基于肌电信号层级分类的手部动作识别方法

目的 利用肌电信号对手部动作进行识别,是控制现代康复假手的关键,其中使用少量电极识别出较多手势又是一难点。为更加充分利用所获得的肌电信息,本文提出一种层级分类方法。方法 首先提出一种基于层级分类的手部肌电信号动作识别方法,该方法首先根据被分类对象的多侧面属性,利用肌电积分值作为特征值,并通过线性判别函数实施预分类;其次建立肌电信号的自回归模型,将模型系数作为特征值,将人工神经网络作为分类器进行细分类;最后进行了对比实验论证。结果 实验结果表明,可以利用2个表面肌电电极以较高的识别率识别出8个常用手部动作。结论 该方法能够以较少的肌电电极识别出较多的动作,比未采用分层方法具有更好的分类效果。

A method of hand movement pattern recognition based on sEMG hierarchical classification

Objective Recognization of hand movement patterns by using electromyography （EMG） signal is the key to controlling modern rehabilitation prosthetic hand.It can promote the common progress of biotechnology and mechatronics technology.To recognize more hand movement patterns by fewer electrodes is one of the difficulties.In order to make the best of EMG information,the article proposes a method of hierarchical classification.Methods Firstly,a method of recognizing the hand movement patterns based on hierarchical classification is proposed.The method utilizes the multiple characteristics of the classified objects.Integrate EMG is used for the feature value and the linear discriminant function actualizes presorting.Then,an autoregressive （AR） model is established and its parameters are used for the features.Artificial neural network can be used to classify finely.Finally,comparative experiments are carried out to verify.Results The experiments show that the eight common hand movements can be recognized effectively by two surface EMG electrodes.Conclusions The method indicates that the more movement patterns can be recognized by fewer electrodes and it performs better than the method without hierarchical classification.