The Research Of Emotion Recognition Based On Galvanic Skin Response Signal

Emotion recognition based on physiological signals is an important research field with extensive research and application prospects. Since internal feelings and emotional changes can be identified through the analysis of physiological signals, so the most reliable way for emotion recognition is extracting ideal physiological data in accord with the real environment. Galvanic Skin Response (GSR) signal varies obviously with the change of emotion states. The skin conductance level is closely related to emotion and attention. However, there is still many limitations in the existing researches based on physiological signal feature, such as hard to find a specific feature to represent a certain emotion, unfavorable identification effect, and poor robustness.Because of the complexity and difficulty for the study of all kinds of physiological signals, the research method is improved and innovated to break through the former limitations, and one single physiological signal is used for emotion recognition. The Galvanic Skin Response (GSR) Signal contains abundant emotional information, so it can reflect more pronounced changes in human emotion states. The paper mainly studies the variations of GSR signal in emotion recognition, and describes the whole experimental method including GSR data acquisition, data preprocessing, feature extraction, feature selection and classification of emotions (happy, surprise, disgust, sadness, angry and fear).Data acquisition is the first step of the whole study. Whether the target emotion is elicited is the key issue to gain corresponding affective data, so data acquisition is as important as feature extraction and selection. The movies with six emotions were presented to 245 subjects whose GSR data were acquired efficiently at the same time without much discomfort from the body surface.According to the characteristics of GSR signals, a large number of original features are extracted from GSR signals. However, not all features make contributes to emotion recognition, so it is necessary to find affective features from them, namely feature selection. Feature selection in emotion recognition is a combinatorial optimization problem thus a NP problem. So an effective intelligent optimization algorithm is advisable to find a satisfying solution to the problem. It has low computational complexity compared with traditional methods (SFS, SBS, Fisher, ANOVA, etc). Tabu Search (TS) is a highly efficient intelligent optimization algorithm, and has a higher global search capability compared with other intelligent optimization algorithms. According to the specific issues in the research, improved Tabu Search Algorithm is adopted to find feature subsets which enable better recognition effects.The classification and recognition results evaluate the feature subsets which are selected by improved TS. The evaluation process is a classification process. As the algorithm of Fisher classifier is fast and efficient, in the feature search process which requires repeatedly calling the classifier, Fisher has its unique advantages. Therefore, the research applied Fisher classifier algorithm is used as evaluation function in the feature search process to enhance the computing speed and effect.According to the above method, the research results verify that using improved TS algorithm with Fisher for emotion recognition based on physiological signals is effective. GSR signal contains large amount of characteristics which can recognize the six emotions, thus a better result is achieved. Most important of all, the mapping relationship between emotion states and some features of GSR signal is found effectively.