Phoneme-Based Persian Speech Recognition

Undoubtedly, one of the most important issues in computer science is intelligent speech recognition. In these systems, computers try to detect and respond to the speeches they are listening to, like humans. In this research, presenting of a suitable …

Authors: Saber Malekzadeh

                                                                                                                                                                                                                                                                                                                                                                                                                 STFT                                                                                                                                                                                                                                                                                                                                                                                                                             -  -  -  -                                     Mel-frequency Cepstrum                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                        :                                                                                                                                                                                                                                                                                                                                                                                                                -  -  -         STF T                                   MFCC                                                                                                                                               :             -  -                                                                                                                                                                                                                                                                                               -                                  -        STFT                           -        M FCC                           -                                   -                                    -                                              -                           "  "    "  "                    -              B I DE    BOWED                      -      MFCC         "  "    "  "                      -                                     -      STFT                                   -      STFT                                   -                                        -                                                           -                                     -                                        -                                                        1 - 1 -                                     1 - 2 -                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                               1 - 3 -                                                    [  ]                                                                                                                                                   ]  [                         Moore                                                 [  ]                                                                                                                                                                                                                                                                              periodic  cy clic rate   multirate  random    pulse width modulated        periodic         [  ]                                                                                                                                                                                                         [  ]                                    Harvey Fletcher        [  ]                                              [  ]                                                                 [  ]                                                                                                                                                                                        Raj Reddy                                                                                                                                           [  ]                                                                                                                                             [  ]                                             BBN    I BM     Carnegie Mellon   Stanford         [  ]  [  ]          L eonard E. Baum          Carnegie Mellon    Raj Reddy       James Baker   Janet Bake r                              [  ]        Fred Jelinek     IBM                               [  ]        Raj Reddy    ung Xe Dong H   sphinx                                                                    I BM         MedSpeak                                                                                    Office     [  ]                    [  ]              Windows Vista                           Google               I phone                S iri     Apple                         Micr osoft        Cortana        Google                                 [  ]   1 - 4 -                                                                                                2 - 1 -                        2 - 2 -                                                                                                                                                                                                                                                                                                           [  ]   2 - 3 -   2 - 3 - 1 -                 -                              [  ]   2 - 3 - 2 -                              x(t)    X( f)           󰇛 󰇜      󰇛  󰇜                          [  ]   2 - 3 - 2 - 1 -                                                                                                            [  ]  󰇝󰇛󰇜󰇞󰇛 󰇜   󰇛    󰇜     󰇛  󰇜  󰇛    󰇜         -        ST FT              W    x          2 - 3 - 3 - Mel-frequency Cepstrum                                               Spectrum   Mel                        Mel                                                           [  ]    -      M F C C  2 - 4 -    2 - 4 - 1 -                                                                      [  ]             2 - 4 - 2 -                                                                                                [  ]   2 - 4 - 3 -                               [  ]   2 - 4 - 4 -                                                                                                                                                                                                   [  ]   2 - 5 -    2 - 5 - 1 -                                                                                                                .                                 2 - 5 - 2 -                                                                                      [  ]                                                                                                                   [  ]   2 - 6 -      2 - 6 - 1 -                                                                                                                                                                                                                        [  ]  2 - 6 - 2 -                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                 factors of variation   Convolution   Recurrent                                                                                                                                                                                                                                                                             -      [  ]                                                                      Pooling   Fully Connected                                                                                                                      3 - 1 -                                                          3 - 2 -   3 - 2 - 1 -                                                                          Curvelet   Ridglet                       [  ]   3 - 2 - 2 -                                                                                                                                                                       [  ]   3 - 3 -                3 - 3 - 1 -    1                                                                                                                                              [  ]   3 - 3 - 2 -                                                                                                                                                                                                                                                                                                                                                                              [  ]                                           Perceptron                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                 Warr en Mcc ulloch   Walter Pitts   euron N   Do nald O. Hebb   Rosenblatt   Bernard Widrow   Marcian Hoff   Adaline   Marvin  Minsky   Sey mour Papert   Grasberg                                                                                                                                                                                                                                                                                                                  Alexnet                                                       Resnet  Densnet   Nasnet                                                           3 - 3 - 3 -                                                                 Avalanch   Carpenter Gail   Anderson James   Kohonen Teuvo   Verbose               3 - 4 -    3 - 4 - 1 -                                                                                                                                                     [  ]    -                   3 - 4 - 2 -                                                                           [  ]   3 - 4 - 3 -                                              [  ]                                                                    [  ]   3 - 5 -                                                                                                                                                                                     4 - 1 -                              4 - 2 -    4 - 2 - 1 -                                                                    4 - 2 - 2 -                                                                                                                                                                                                                                                                                                          [  ]                                                                                                                                                                                                                                                                                                                                  -                                                                                                                                                                                                                                                                                                                                   4 - 3 -                                                                                                                                                                                                                                                                                                                                                                                      -                                                                              -                                                                                                                                                                                                              4 - 3 - 1 -                                                                                                                                                                                      -                           "  "    "  "  4 - 3 - 2 -                                                                                                                                                                                                                                                     "  "              "  "  "  "                                                                                  "   "                                 "  "   "  "                                                                                       "    "                            "  "  "  "  "  "  "  "                      "  "                                                                                                      4 - 3 - 3 -                                                                                                                                                                                                                                                                                                            4 - 3 - 4 -                                                                           )  SVM (                                                                                         Support Vector Machin e                                                                                                            SVM                                                                          [  ]                                                                                                                                  4 - 3 - 5 -                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                             4 - 3 - 6 -                                                     MFCC  STFT            MFCC                       MFCC                                                                                                   MFCC                                                MFCC                                                         -   STFT                         MFCC  STFT                                                                                                             Luven              4 - 3 - 7 -                   STFT                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                 4 - 3 - 8 -    STFT  ST FT                                                                                                                                                                                                              STFT                                                                                    STFT                                             SVM                                                                      STFT                                                       STFT                                     STFT ،               ST FT  Formant                 -              b i de    b o w e d                                                                                                                                                                                                                                                                                   4 - 3 - 9 -   MFCC                        MF CC                                         MFCC                                                                                                                                                                                                              "   "             "   "                                                                                                                                                                                                                                                                               -      M F C C         "  "    "  "                                                                                                                 MFCC                                                          MFCC      MF CC               4 - 4 -            STFT                                                                                                    -                                                                                                                                                                                             4 - 5 -                                                            Adobe Audtion               Adaptive Noise Reduction                                                 STFT                  -      STFT                    -      ST F T     4 - 6 -                                                   Persian form English form Persian Example  A   I   ʊ   æ   e    o   P   B   T   D   tʃ   dʒ   K   G   F   V   Kh   S                Z   ʃ   ʒ   M   N   H   L   R   Q   j    -                                                                                                                   Dropout   n Bat chNormalizatio                -                                                                                                                                                                                                                    same           4 - 6 - 1 -                                                               Relu                                       Sequential   adding P   ) rectified linear unit Relu (                                    Relu      Leaky R elu                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                 Softmax                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                   MaxPooling   Perceptron   Neuron   Softmax                                                   Onehot      Categorical Cross Entropy                                                                                                                                                                                                                                                                                        AdaDelta             SGD   Adam                  2                             3                                                        Cross Entropy    Compile    Fit                                                                                                                                                                                      5 - 1 -                              5 - 2 -                                                                                                                             Support F  -score Recall Precision Phoneme Number                                                 .                                                                             .                                                 Avg / Total   -                                                                                                                                                                                                                                                                                                        F        5 - 2 - 1 -                                                                          F                  -                           5 - 2 - 2 -                                                                                                                            recision P   Recall                                                                                                      PCVC    5 - 3 -                                    PC VC               [  ]         [  ]         [  ]           -                                                      6 - 1 -                               6 - 2 -                                                                                         AutoEncoder                GAN                       AC GAN                                                                                                                                                                                                                         End to En d                                                                                                                                                               6 - 3 -                                                                                                                                                                             Miller, R., Low e, J    Boar, R. (  ). The incredible music machine. Quartet Books Ltd      James R. Lewis (  ): The Voice in the Machine: Building Computers T hat Understand Spee ch, International Journal of Human-Computer Interaction,  ,         Moore, Gordon E. (      ). "Cramming more components onto integrated circuits". Electronics. Retrieved           Kuo, B.C.:(  ) Digital Control Systems,  nd edn. Oxford University Press, New York      Sejdić  E.; Djurović  I.; Jiang J. (  ). "Time-frequency feature representation using energy concentration: An overview of recent advances". Digital Signal Processing.   (  ):     doi:  /j.dsp.       Huffman, Larry. "Stokowski, Harvey Fletcher, and the Bell Labs E xperimental Recordings      Juang, B. H.; Rabiner, Lawrence R.(  ) "Automatic speech reco gnition  a brief history of the technology development" (PDF):  . Retrieved   January      Pierce, John R. (  ). "Whither speech recognition   Journal of the Acoustical Society of America.   (  ):  . doi:         Reddy, D. Raj. (  ) " Computer recognition of connected sp eech." The Journal of the Acoustical Society of America   :         Benesty, Jacob; Sondhi, M. M.; Huang, Yiteng (  ). Springer Handbook of Speech Processing. Springer Science   Business Media. ISBN       Blechman, R. O.; Blech man, Nicholas (June  ,  ). "Hello, Hal ". The N ew Yorker. Retrieved   January        Klatt, Dennis H. (  ). " Review of the A RPA speech understanding project". The Journal of the Acoustical Society of America.   (  ):     doi:         James K. Baker.(    Machine-aided La beling of Connected Spe ech   in Working Papers in Speech Recognition XI, Technical Reports, C omputer Science Department, Carnegie-Mellon University, Pittsburgh, PA         Pioneering Speech Recognition". Retrieved   January      Thompson, Terry. "DO-IT" .(  ) University of Washington. Retrieved   June      Froomkin, Dan. " The Computers Are Listening".(  ) The Intercept. Retrieved   June      Haşim Sak, Andrew Senior, Kanishka Rao, Françoise Beaufays and Johan Schalkwyk (September  ): Google voice search: faster and more accurate                 Zölzer, Udo (  ). Digital Audio Signal Processing. John W iley and Sons. ISBN             Larson, David R. (  ). "Wavelet Analysis and Applications ( See: Unitary systems and wavelet sets)". Appl. Numer. Harmon. Anal. Birkhäuser:         Cochran, William T., et al. " W hat is the fast Fourier transform?." Proceedings of the IEEE   :         Okumura, Shuhei. (  ) The short time Fourier transform and local signals. Diss. Carnegie Mellon University      Sahidullah  Md.; Saha , Gou tam (May  ). "Design, analysis and ex perimental evaluation of block based transformat ion in MFCC computation for speaker recognition". Speech Comm unication.   (  ):     doi:  /j.specom.       Boashash, Boualem (  ). Time-frequency signal analysis and processing: a comprehensive reference. Academic Press      Beigi, Homayoon.(  ) Fundamentals of spe aker recognition. Springer Science   Business Media      Graves, Alex, Abdel -rahman Mohamed, and Geoffrey Hinton. (  ) " Speech recognition with deep recurrent neural networks." Acoustics, speech and signal processing (icassp),   ieee international conference on. IEEE      LeCun, Yann, Yoshua Bengio, and Geoffrey Hinton. (  ) "Deep learning." Nature  :         Albelwi, S    Mahmood, A. (  ). A framework for designing the architectures of deep convolutional neural networks. Entropy,  (  ),       Stoica, Petre; Moses, Randolph (   ). Spectral Analysis of Signals (PDF). NJ: Prentice Hall      Rumelhart, David E., Geoffrey E. Hinton.(  ) "Learning Internal Repr esentations by Error Propagation". David E. Rumelhart, James L . McClelland, and the PDP research group. (editors), Parallel distributed processing: Explorations in the mic      MacNeilage, Peter, (  ). The Origin of Speech. Oxford: Oxford University Press         Ostendorf, Mari, and Salim Roukos.(  ) " A stochastic segment model for phoneme-based continuous speech recognition." IEEE Transactions on Acoustics, Speech, and Signal Processing  :         Gupta, Vishwa N., et al.(  ) "Phoneme based spe ech recognition." U.S. Patent No.  .   Feb      Titze, Ingo R. (  ). Principles of Voice Production. Prentice Hall. ISBN          . Archived from the original on           Cortes, C.; Vapnik, V. (  ). "Support-vector networks" . Machine Learning.   (  ):     doi:  /BF                  Nazari, M., Sayadiyan, A    Valiollahzadeh, S. M. (  , April). Spea ker- independent v owel recognition in Persian spee ch. In Information and Communication Technologies: From Theory to Applications,  . ICT TA  .  rd International Conference on (pp.       Sadeghi, V. S    Yaghmaie, K. (  ). Vowel recognition using neural networks. IJCSNS International Journal of Computer Science and Network Security,  (  ),         Tavanaei, A., Manzuri, M. T    Sameti, H. (  , June). Mel-scaled Discrete Wavelet Transform and dynamic features for the Persian phoneme r ecognition. In Artificial Intelligence and Signal Processing (AISP),   International Symposium on (pp.       Abstract Undoubtedly , one of the most important issues in computer science is intelligent speech recog nition. In these sy stems, computers try to detect and respond to the speeches they are listening to, like humans. I n this research, presenting of a suitable method for the diagnosis of Persian phonemes by AI using the signal processing and classification algorithms have tried. For thi s purpose, the STFT algorithm has been used to process the audio signals, as well as to detect and classify the signals processed by the deep artificial neural network. At first, educational samples were provided as two phonological phrases in Persian langua ge and then signal processing operations were performed on them. Then the results for the data training have been given to the artificial deep neural network. At the final stage, the experiment was conducted on new sounds. Keyword s: Phoneme Recognition, Speech Recognition, Deep L earning  Vali-e-Asr Univer sity of Rafsanjan Faculty of Math and Com puter Sciences Phoneme-Base d Persian Speech Rec ognition Report as Thesis of Master of C omputer Science in the fie ld of Artificail Intellig ence Supervisor: Dr. Mohamm ad Hosein Gholizadeh Adviser: Dr. Seyyed Nas er Razavi Saber Malek zadeh September 

Original Paper

Loading high-quality paper...

Comments & Academic Discussion

Loading comments...

Leave a Comment