In this paper, a binaural beamforming algorithm for hearing aid applications is introduced.The beamforming algorithm is designed to be robust to some error in the estimate of the target speaker direction. The algorithm has two main components: a robust target linearly constrained minimum variance (TLCMV) algorithm based on imposing two constraints around the estimated direction of the target signal, and a post-processor to help with the preservation of binaural cues. The robust TLCMV provides a good level of noise reduction and low level of target distortion under realistic conditions. The post-processor enhances the beamformer abilities to preserve the binaural cues for both diffuse-like background noise and directional interferers (competing speakers), while keeping a good level of noise reduction. The introduced algorithm does not require knowledge or estimation of the directional interferers' directions nor the second-order statistics of noise-only components. The introduced algorithm requires an estimate of the target speaker direction, but it is designed to be robust to some deviation from the estimated direction. Compared with recently proposed state-of-the-art methods, comprehensive evaluations are performed under complex realistic acoustic scenarios generated in both anechoic and mildly reverberant environments, considering a mismatch between estimated and true sources direction of arrival. Mismatch between the anechoic propagation models used for the design of the beamformers and the mildly reverberant propagation models used to generate the simulated directional signals is also considered. The results illustrate the robustness of the proposed algorithm to such mismatches.
A HEARING aid is a common and effective solution to sen- sorineural hearing loss. Despite enormous advances in hearing aid technology, the performance of hearing aids under noisy environments remains one of the most common complaints from hearing aid users [1], [2], and hearing-impaired people face H. As'ad and M. Bouchard are with the School of Electrical Engineering and Computer Science, University of Ottawa, Ottawa, ON K1N 6N5, Canada (e-mail: hasad056@uottawa.ca; martin.bouchard@uottawa.ca).
H. Kamkar-Parsi is with WS Audiology 91058, Erlangen, Germany (e-mail: homayoun.kamkarparsi@sivantos.com).
Digital Object Identifier 10.1109/TASLP.2019.2924321 challenges in understanding and separating speech in noisy environments [1]- [3].
For noise reduction, single channel processing algorithms, which rely on frequency and temporal information of the input signals, have been extensively researched such as in [4], [5]. However, single channel algorithms suffer from several limitations under low-SNR acoustic scenarios, especially for nonstationary noise and multi-talkers conditions. Single channel solutions typically also introduce distortion and do not provide true speech intelligibility improvement. A notable exception is the solution in [6] which has been found to improve speech intelligibility. The solution in [6] is based on deep neural networks and a binary masking of some speech components in the T-F domain. This solution, however, does not preserve naturalness of the target speaker speech (high distortion), which is a concern for its use in hearing aids. It has also not been developed for the case of one or two competing talkers.
As an alternative, microphone array processing (beamforming) has been widely used in modern hearing aids, leading to directionally sensitive hearing aids [7]. Binaural hearing aids have also recently been introduced in the market. Binaural hearing aids have a hearing aid device at each ear, each possibly equipped with multiple microphones, and the devices are capable to transmit signals or information from one side to the other through a “binaural wireless link”. Microphone arrays can provide good noise reduction with low distortion, and the use of additional microphones and different microphone geometry in binaural hearing aids can lead to further improvements in the directional response, compared to monaural single-sided beamforming. However, even binaural hearing aids have still not achieved the required robustness in case of real-life complex environments [8]. The performance of binaural beamformers can be significantly affected by a mismatch or an error between the target source propagation model assumed for the beamformer design and the actual physical target source propagation [9], [10]. This includes errors in the estimated target direction of arrival (DOA) used in the beamformer algorithms, i.e., target DOA mismatch. This kind of mismatch can be generated from imperfect target DOA estimation schemes, from small head movements of the hearing aid user, and from multipath propagation. To address this problem, several acoustic beamforming methods robust to the mismatch in target propagation models have been introduced in the literature [11]- [21], and some of these solutions are not specifically for binaural hearing aids.
Unfortunately, most of the previous work rely on sophisticated Voice Activity Detection (VAD), speech presence probability estimation, and/or SNR estimation. These can become difficult to measure in complicated multi-talker reverberant environments, with speakers having variable activity patterns. An interesting solution for hearing aids based on inequality constrained optimization has been proposed in [22] and discussed in [23], to increase the robustness to target DOA mismatch. However, since this design uses extra constraints for directional sources to increase robustness to DOA mismatch, this can lead to low degrees of freedom available for residual noise reduction (e.g., low number of adaptive “nulls”) in case of limited number of available microphones signals. In addition, it requires an estimation of the DOA for the directional interferer sources.
All the beamforming designs in [11]- [21] were not designed to preserve the binaural cues of the residual directional interferers and diffuse-like noise in the binaural output signals. Several binaural beamforming solutions have been introduced to preserve some of the binaural cues of these components, while also preserving the target signal and achieving a good noise reduction level. Under some assumptions (e.g., accurate direction of arrival estimates), binaural beamforming processing such as the second and third methods in [24] can provide directional noise reduction and preserve the binaural cues of the target signal and the directional interferers, depending on the number of available microphones. However, this binaural beamforming is not designed to preserve the binaural cues of the diffuse-like
This content is AI-processed based on open access ArXiv data.
