基于PNN的手势识别

Gesture recognition based on PNN

为了实现人体手势姿态识别的目标,选用氯化银(Ag Cl)贴片电极作为信号传感端,通过采集前臂表面肌电(SEMG)信号,经信号放大、滤波等前期处理,再经活动段检测、降噪等信号处理后,提取伸食指、握拳、伸腕、屈腕4种手势的均方根值和积分EMG值作为特征向量,送入概率神经网络(PNN)中进行训练识别,实现人体手势识别。实验结果表明:PNN对前臂SEMG信号的模式识别的正确率可达到97.62%,将PNN应用于手势识别系统具有可行性。

In order to achieve the goal of gesture recognition,AgC1 patch electrude is used as signal sensing end. By collecting forearm surface electromyography（SEMG） signal, through signal amplification, filtering and other pre- processing,and by active segment detection, noise reduction and other signal processing, the root mean square values and integral EMG values of four kinds of gesture which are stretched forefinger, fist, wrist extension and wrist flcxion are extracted as eigenvector, and send to probabilistic neural network （PNN） for training and to realize the identification of gestures. The experimental results show that PNN can achieve accuracy of 97. 625 % for pattern recognition on forearm SEMG signals,and it is feasible to apply PNN to gesture recognition system.