一种能识别情感状态的遗传算法及应用

A genetic algorithm capable of affective states identification and its application

将离散的情感状态的识别看成模式识别问题,运用组合优化的思想,从4种生理信号心电(ECG)、肌电(EMG)、皮肤电(SC)、呼吸信号(RSP)中抽取情感特征,将遗传算法和最近邻算法相结合构造出一种新的方法,以最近邻为准则,遗传算法寻优,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.仿真实验表明,用肌电信号(EMG)来识别4种情感状态效果要好于用其他3种生理信号,按唤醒度轴方向识别情感状态较按效价轴方向识别情感状态效果好.

Regarding the discrete emotion recognition as a mode identification problem and employing the idea of integrated optimization,a new genetic algorithm was constructed by means of extraction of affective feature from four physiological signals： such as electrocardiogram（ECG）,electromyogram（EMG）,skin current（SC）,respiration space（RSP）.In this method,the genetic algorithm was integrated with the nearest neighbor algorithm and the nearest neighbor was taken as the criteria to search the optimal feature subset which represents exactly the relevant affective states,i.e.joy,anger,sadness,and pleasure.The simulation experiment demonstrated that the identification performance with EMG signal was better than that of other three physiological signals,and it also turned out that,for four emotions,it was easier to distinguish emotions along the arousal axes than the valence axes.