Affection Recognition by GSR Signal Use a Improved K-nearest Neighbor Algorithm

Abstract

For afiection recognition problems of GSR (Galvanic Skin Response) signal, this paper presents a method of afiection recognition based on a improved K-nearest neighbor algorithm. In this research, on the flrst step, the group collected 175 participants’ GSR signals on the test of emotional arousal which was acquired in the lab environment; then used the Batterworth low pass fllter to remove GSR signals noise, and extracted the original features of GSR signals at the same time; then recognized the flve feelings of happiness, sadness, fear, anger and calm with the improved K- nearest neighbor classifler. The experimental results show that the method can improves the accuracy of identiflcation indeed.