Toward Recognizing Two Emotion States from ECG Signals

Abstract

Emotion recognition based on physiological signals which can reflect people’s real emotion correctly is more robust and objective than any other ways, so it has a bright prospect of research and applications. This paper may firstly carry out the work of feature extraction for electrocardiogram (ECG) obtained from 391 subjects containing two emotion states (joy, sad) by the method of discrete wavelet transform (DWT). Then feature selection could be performed using the method on the combination of Particle Swarm Optimization (PSO) and KNN classifier. Eventually, the optimal feature subset could be found and the total recognition rate reached 84.45%. Experiment and simulation results showed that it is feasible and efficiency that using PSO and KNN to recognize emotion states by physiological signals.