The Design And Implementation Of Emotion Recognition System From Multiple Physiological Signals Based On Artificial Neural Network

Emotion is a psychological and physiological phenomenon of human, which exists in people’s lives on all sides, and plays an irreplaceable role. Emotions are an important bridge for communication between people, and also closely related with human health. People in various fields such as psychology, medicine, information science, rehabilitation science and so on, are studying emotion. Emotion recognition has become an important research methods and means in the people’s study of emotions. Emotion recognition method generally includes in two types:physiological and non-physiological signal recognition. There are a large number of experiments and studies confirm the correlation between physiological signals and their emotional state, many researchers analysis with human’s physiological signal, then use them to identify person’s emotional state and achieved good experimental results. Because of the physiological signal recognition method’s good accuracy and objectivity, it’s high research valued and has been very widely used.This paper aims to design an emotion recognition system based on physiological signals. The system includes a measurement platform and the emotion recognition platform based on physiological signals. The system measures the human body’s physiological signals, and get physiological signal feature extraction, ultimately analyzed the characteristic parameters through the emotion recognition platform to identify the person’s emotional state. In this study, I studied the internal and international researches on physiological signals of emotion recognition, and selected the emotional state type, the emotion-induced physiological solution and the type of physiological signals for the present study used in the emotion recognition. After this, I designed a physiological signal measurement platform, including the hardware design and the PC software design which deal with the measurement of physiological signal feature extraction. The platform can be simultaneously collected by the human skin electrical pulse signal, the respiration signal, and the skin temperature signal for real-time processing and extracted characteristic parameters12required for emotion recognition. Then I designed a platform for emotion recognition based on BP Artificial Neural Networks model, after training the emotion recognition platform could identify the person’s emotional state based on12parameters of physiological signals. Finally, I confirmed the feasibility and practicality of the study design emotion recognition system through a study of validation experiments. By the final experimental verification, the system samples’average recognition rate reached73.3%with the individual differences, the sample which got through to individual differences in processing average recognition rate reached86.7%.