Emotion feature selection from physiological signals using tabu search

A method of feature selection based on tabu search is proposed for recognizing affective states (joy, anger, sadness, pleasure). After combining with knear neighbour classifier, the tabu search heuristic algorithm achieved good recognition performance in the affective states classification , obtained the robust feature subsets which belong to 4 physiological signals (ECG, EMG, SC, RSP). Simulation showed that it is feasible and effective to recognize emotion using tabu search algorithm.