Extracting Emotional Features from ECG by Using Wavelet Transform

One key element of emotion recognition is to extract emotional features effectively from physiological signals. In this paper, a wavelet transform based feature extraction is proposed to recognize emotions through ECG (Electrocardiogram) signals. Four emotional data sets collected on the same day from one subject are decomposed by DWT (Discrete Wavelet Transform) and 84 statistic values of wavelet coefficients are selected as the emotional features according to their amplitude relations. Furthermore, in order to eliminate the negative impacts of material, time and environment, these selected features are normalized with respect to the emotional mode ´Pleasure´. The initial results show that, with the normalized features, the best correct-classification ratio of joy and sadness reaches 92%.