基于表面肌电信号的性别差异性手势识别

Recognition of hand gestures with gender differences based on surface electromyographic signals

对于同一手势不同性别的表面肌电信号差异性较大。为了减小差异性,提出滑动平均能量与能量补偿相结合的方法。本实验共采集10种手势动作的表面肌电信号;利用滑动平均能量对活动段进行检测,并对女性的动作段进行能量补偿;小波包分解采用Db4、Bior3.2、Haar、Sys8、Dmey这5种小波函数提取特征;最后并通过粒子群优化支持向量机进行分类。结果分析表明,能量补偿增大了特征的辨识度,减小了性别差异性,提高了手势识别率。

For the same gestures,there are some gender differences in surface electromyographic signals(s EMG).Herein a method combining sliding average energy and energy compensation is proposed to reduce the gender differences for the recognition of hand gestures.The sEMG of 10 hand gestures are collected.The active segment is detected by sliding average energy,and then the energy of female motion segment is compensated.Five wavelet functions of wavelet packet decomposition,namely Db4,Bior3.2,Haar,Sys8 and Dmey,were used to extract features.Finally,the obtained data are classified and recognized by particle swarm optimization-support vector machine.The results show that energy compensation improves the identification of features,reduces gender differences,and increases the recognition rate of hand gestures.