Design of Wearable Physiological Signal Measuring Device for Elder's Emotional Response Detection

• Main Authors: Lin, Yu-Sian, 林育賢
• Other Authors: Hou, Chun-Ju
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/33n4nb

碩士 === 南臺科技大學 === 電機工程系 === 107 === With the increasing number of elderly people, Taiwan has officially become an aged society. The degeneration of physical and mental function due to aging, can cause social isolation and emotional distress in elderly people. In daily life, there are many stressors. Good amount of stress contributes to physical and mental development. On the other hand, excessive stress can cause autonomic nervous disorders, decreased immunity, and negative emotions. According to the literature [16] [18], the individual emotional response has a significant impact on the autonomic nervous system, and will affect the normal functions of the human body. It is found that the amygdala and prefrontal cortex are the neural basis for emotion and cognition, emphasizing the relationship between emotion and cognition. The clinical evaluation of emotional status is done through the assessment of the emotional states of the patient after being examined by the psychologist in view of their subjective and objective responses. Recently, many scholars use physiological signals for emotional assessment and feedback because they can reflect the individual’s actual physiological responses and mental states. The purpose of this study is to compare the changes in the three physiological signals of the elderly while playing the interactive game and during rest. By designing a wearable physiological signal measurement system, the physiological signal is measured through the front-end sensors. The analog circuit filters and amplifies the signals before the digital circuit performs signal sampling and data encoding. Then the data is transmitted to the computer via the Bluetooth module. The transmitted data will undergo physiological signal processing and feature extraction. After the aforementioned procedures, statistical analysis, feature selection, and emotion recognition is executed. The objectives of this study are: (1) Three-channel physiological signal analog circuit design: ECG, PPG, EEG; (2) Digital circuit design: analog/digital conversion, data encoding, serial transmission; (3) Signal processing and analysis: pre-processing, HRV time-domain analysis, HRV frequency-domain analysis, EEG wavelet analysis, feature extraction; (4) System calibration and experimental design; (5) Statistical analysis, feature selection, emotion recognition. The results of this study include: (1) system calibration: the system frequency response and sensitivity test results met the design requirements; (2) signal processing and feature analysis: 38 features have been successfully extracted after pre-processing and feature analysis of three signals (ECG:16, PPG:18, EEG:4); (3) experimental test results: elderly people respond differently to physiological signals under stress and non-stress stimuli, HRV features reflect the autonomic activity, and EEG features reflect the degree of relaxation and concentration of the individual. The classification results showed that the accuracy was 93.7%, and the cross-validation accuracy was 92.1%. This indicates that this study is indeed feasible for emotion recognition of elderly people.