将情感识别中的特征选择看成组合优化问题,从四种生理信号EMG、ECG、RSP、SC中抽取统计特征,将参数可调的遗传算法和K-近邻算法相结合尝试找出最能"代表"某一情感状态joy、anger、sadness、pleasure的最优情感特征组合模式.仿真表明,该方法是有效的.%The feature selection in emotion recognition is regarded as the combinational optimisation problem, statistical features are extracted from four physiological signals: ECG, EMG, SC and RSP, the genetic algorithm with adjustable parameters is integrated with the Knearest neighbour algorithm for trying to find out optimal emotional feature combination model which represents exactly the relevant emotional states: joy, anger, sadness and pleasure. Simulation shows that the method is effective.
展开▼