Study of Emotion Recognition Based on Surface Electromyography and Improved Least Squares Support Vector Machine

In order to improve human-computer interaction (HCI), computers need to recognize and respond properly to their user’s emotional state. This paper introduces emotional pattern recognition method of Least Squares Support Vector Machine (LS_SVM). The experiment introduces wavelet transform to analyze the Surface Electromyography (EMG) signal, and extracts maximum and minimum of the wavelet coefficients in every level. Then we construct the coefficients as eigenvectors and input them into improved Least Squares Support Vector Machines. The result of experiment shows that recognition rate of four emotional signals (joy, anger, sadness and pleasure) are all more than 80%. The results of experiment also show that the wavelet coefficients as the eigenvector can be effective characterization of EMG. The experimental results demonstrate that compared with classical L_M BP neural network and RBF neural network, LS_SVM has a better recognition rate for emotional pattern recognition.