Human thermal comfort measurement plays a critical role in giving feedback signals for building energy efficiency. A non-invasive measuring method based on subtleness magnification and deep learning (NIDL) was designed to achieve a comfortable, energy efficient built environment. The method relies on skin feature data, e.g., subtle motion and texture variation, and a 315-layer deep neural network for constructing the relationship between skin features and skin temperature. A physiological experiment was conducted for collecting feature data (1.44 million) and algorithm validation. The non-invasive measurement algorithm based on a partly-personalized saturation temperature model (NIPST) was used for algorithm performance comparisons. The results show that the mean error and median error of the NIDL are 0.4834 Celsius and 0.3464 Celsius which is equivalent to accuracy improvements of 16.28% and 4.28%, respectively.
Real-time thermal comfort perception for occupants plays important roles in human-oriented smart buildings and their energy efficiency. 21% of the global energy consumption is due to energy requirements of commercial and residential buildings [1]. In many countries and regions with rapid urbanization, building energy consumption is expected to increase at an annual rate of 32% [1]. 50% of building energy consumption is related to heating, ventilation and air conditioning (HVAC) system [2]. Feedback signals from thermal comfort perception can be used to effectively control and optimize HVAC energy consumption. Since the 1970s, many methods, including questionnaire surveys, environmental measurements, and physiological measurements (invasive and semi-invasive methods), have been explored to measure human thermal comfort. However, due to (1) inter-individual differences, (2) intra-individual variances, and (3) subtle skin variations (that make it difficult to access skin temperature through computer vision), there has been no breakthrough in thermal comfort perception through computer vision-based techniques until now. The drawbacks of current methods can be summarized as follows: (1) lack of big data validation, (2) lack of practical application possibilities for accurate non-invasive techniques, and (3) lack of adequate consideration of inter-individual and intra-individual differences over time, including subtle skin variations. Instead of human oriented design considering individual perception of indoor climate, buildings are regulated to provide constant and standardized climate comfort. Because different occupants have different subjective feelings toward the same indoor environment, the constant indoor environment parameters cannot meet individual needs in a smart building and optimize energy efficiency.
Human thermal comfort is a subjective feeling that depends on how the human body interacts with the environment [3]. For overcoming the drawbacks described above, a non-invasive measurement method of thermal comfort based on subtleness magnification and deep learning (NIDL) was explored and is described in this paper. The subtleness magnification algorithm adopted is Euler Video Magnification (EVM). Using this NILD method, subtle skin variation was first magnified by the EVM algorithm, and a region of interest (ROI) is selected. A deep neural network with 315 layers was optimized and used for extracting skin image features, according to features of human thermal comfort, and a regression relationship between skin image and skin temperature was constructed. A dataset, containing 1.44 million frames, was collected from a physiological experiment and was used for algorithm validation.
The main contribution of this paper can be summarized as follows:
(1) The proposed method makes non-invasive measuring technology for human thermal comfort practically possible.
(2) It is the first time that deep learning is used for skin temperature measurement using EVM combined with deep learning for feature extraction and relationship construction.
(3) It is the first time that a large image-based dataset (1.44 million frames) for human thermal comfort is Non-invasive thermal comfort perception based on subtleness magnification and deep learning for energy efficiency constructed. 16 subjects were invited for a physiological experiment to collect the data set and it was used for algorithm validation. The rest of this paper is organized as follows. Section 2 introduces related work. In section 3, the research method, including the physiological experiment to collect image-based data and algorithm, are introduced. Algorithm validation results are presented in sections 4 and 5, and conclusions in section 6.
Based on Fanger’s theory of thermal comfort [3], the thermal comfort environment is defined by ASHRAE and ISO (No. 7730) as “at least 80% of building occupants are psychologically satisfied with the temperature range of thermal environment [4,5]”. As mentioned earlier, human perception thermal comfort varies intra-individually as well as inter-individually. Tracking these variances has traditionally involved three types of methods (2.1-2.3 below).
Based on an offline or online questionnaire, the thermal preference of an occupant is collected and used as a basis for environment parameter regulation [6,7]. According to [4] and [5], the questionnaire is a subjective assessment which can reflect the occupant’s psychological state and thermal comfort level. However, the questionnaire method relies on the continuous and frequent participation of building occupants, therefore the operability is weak and the efficiency is low [8].
For the environmental measurement method, different sensors measure different indoor environmental parameters, including temperature, humidity, and airflow. Based on supervision models, the relationship between the environmental parameters and occupant thermal comfort is constructed to determine
This content is AI-processed based on open access ArXiv data.
