基于ANOVA和BP神经网络的最优肌电信号测量位置选择

EMG measurement position optimization based on ANOVA and BP neural network

基于肌电信号的手部动作识别中,肌电信号测量位置的选择直接关系到动作识别的准确率.本文以使用最少的肌电传感器和获得较高的动作识别率为目标,提出一种基于ANOVA(方差分析)和BP神经网络的肌电信号测量位置优选方法.使用4个肌电传感器采集受试者做出指定动作时的肌电信号,提取肌电信号的时域特征,并按测量位置组合构成15个不同的样本进行BP神经网络的训练和测试.采用单因素ANOVA分析测量位置对动作识别结果影响的显著性,采用Tukey HSD将测量位置进行归类,并从动作识别率最高的子集中选择测量位置最少但识别准确率最高的测量位置组合作为最优的肌电信号测量位置.实验结果表明,测量位置对动作识别的结果具有显著的影响,随着测量位置数的增加,动作识别准确率呈上升趋势,最优的测量位置组合为P1+P3+P4,其动作识别准确率为94.6%.

The locations of electromyography(EMG)measurements are directly related to the accuracy of motion recognitionin hand gesture recognition based on EMG signals.This study proposes an EMG measurement position optimization strategy based on ANOVA and back propagation(BP)neural network to obtain the best motion recognition with the fewest EMG sensors.Four EMG sensors are used to capture the EMG signals when the subjects perform specific hand gestures.Feature data extracted from the raw EMG signals are combined into 15 different vectors according to different measurement position combinations.These 15 feature vectors are used to train and test the BP neural network.Single factor analysis of variance(ANOVA)is employed to analyze the significance of the influence of the measured position on themotion recognition.Tukey s honest significant differencetest is adopted to classify the position combinations.The position combinations are divided into several subsets.In the subset with the highest recognition rate,the position combination with the least measurement position and the highest recognition accuracy is considered to be the optimized measurement position.The experimental results show that the measurement position has a significant impact on the results of motion recognition.The accuracy of motion recognition shows an upward trend with the increase in measurement position.The optimal combination of measurement position is P1+P3+P4,and the accuracy of motion recognition is 94.6%.