Research On Affective Recognition And Regulation Based On GSR Signals

In this thesis, a method for affective recognition and regulation based on GSR signals is proposed to construct a human-computer interaction system with the function of affective support. User’s affective real-time state including affective category and affective intensity is determined by analyzing the GSR signals. Among of them, the affective category is recognized by BP neural network and optimized by genetic algorithm. Furthermore, the affective intensity is evaluated by using curve fitting. When it is necessary to regulate affection, an appropriate regulation material will be presented to the user automatically in terms of the user’s affective state, resulting in human-computer affective interaction.The main work and conclusions are shown as follows:(1) Formulate the GSR acquisition scheme and establish the affective physiological signal database. The GSR data were collected to establish an affective physiological signal database when the subjects were in the state of angry, fear, grief and happy. The acquired GSR data were preprocessed, and 28 statistical features were extracted as the primitive character set for the affective recognition.(2) BP neural network optimized by genetic algorithm is used for affective recognition. The genetic algorithm is used to optimize the connection weights and thresholds of the neural network, thus resulting in a faster network convergence speed and a higher recognition rate. The obtained result shows that, the recognition rate is 8% higher compared with no optimization. On this basis, the real-time affective recognition is implemented, and the recognition rate is 81.3%.(3) Evaluation of affective intensity. The functional relation between affective intensity and GSR feature is established by fitting the feature values and the affective intensity values of each affective event. According to this functional relation, the affective intensity can be calculated in real time. The evaluation of affective intensity has two functions:the subject’s affective state is determined by the affective category and affective intensity; the moderating effect of materials is evaluated according to the affective intensity when the subject views the regulation materials.(4) Establish an affective regulation material library. According to the characteristics of affection regulation which is based on human-computer interaction, experimental design team select 30 materials from a large number of related clips to create an affective regulation material library. At the same time, the material library settings updated interface, users can choose new materials to suit their personal taste.(5) Construct an affective recognition and regulation system based on GSR signals. This system consists of three main parts:affective recognition subsystem, affective intensity evaluation subsystem and material selection subsystem. The system uses the physiological signal as the input, therefore the result of affective recognition is more objective and accurate. At the same time, the system selects different regulation material to different user to achieve a personalized affective regulation. To conclude, the system is a human-computer interaction system with the function of affective support.