基于表面肌电信号的腰背动作识别新方法

A New Method for Recognition of Lumbar and Back Movements Based on Surface EMG Signals

为提高基于表面肌电信号的人体腰背动作识别率,提出一种基于小波包能量与改进NARX神经网络的分类识别新方法。利用小波包变换对动作部位进行表面肌电信号特征提取,并采用改进NARX神经网络进行分类识别。选取8名实验者分别在扭腰、弯腰、侧弯腰3种动作下进行表面肌电信号数据采集,选择db4小波包函数对信号进行6层分解,得到第6层64个频带的小波包分解系数,代表各个动作信息的特征向量,作为改进NARX神经网络的输入进行分类识别。对照实验组中,改进NARX神经网络的识别率较高,总体识别率达到96.7%。实验结果表明,利用该识别方法对腰部动作进行分类识别,分类准确,且识别率更高。

In order to improve the recognition rate of human low back motion pattern based on surface EMG,this paper proposes a new classification and recognition method based on wavelet packet energy and improved NARX neural network.The wavelet transform is used to extract the surface EMG signal features of the action part,and the improved NARX neural network is used for classification and identification.Eight experimental subjects were selected to perform surface EMG signal data acquisition under the lumbar motion of twisting,bending and side bending.The db4 wavelet packet function was used to decompose the signal in 6 layers to obtain the decom⁃position coefficient of the 64 wavelet band in the 6th layer.The coefficients represent the feature vector of each action information are used as an input to the improved NARX neural network for classification and recognition.In the control group,the improved NARX neural network has a higher recognition rate,and the overall recognition rate reaches 96.7%.The experimental results show that the waist movement is classified and recognized by this recognition method with accurate classification and higher recognition rate.