On study of surface electromyographic signals (sEMG) pattern recognition in rehabilitation equipment and intelligent prosthetic applications,a square-mediation method that extracts the envelope of multi-channel sEMG features is presented,with which the finger gesture recognition rate and accuracy rate is improved.In the process,the sEMG is squared by the finger movement acquisition experiment,and then the envelope was formed through the low-pass filtering.Using the amplitude-multiplication method,the envelope of different types of finger action is used to creat the teacher sample label.With these label,the BP neural network is used to accomplish the recognition and classification of the action.Experimental results show that the average correct rate of finger behavior recognition is 94.93%,including thumb,index finger,middle finger,ring finger,little finger and all finger flexion actions.The average time delay for each action recognition is 50.7 ms.%针对表面肌电信号模式识别在康复器械以及智能假肢中的应用问题,通过平方调解法来提取多通道sEMG特征包络线,以提高手指动作识别速率与正确率;首先将手指动作采集实验获取的表面肌电信号进行平方处理,再经低通滤波形成包络线;利用幅值乘方法对不同的动作类型的包络线进行处理并形成学习用的教师样本标签,最后通过BP神经网络完成动作的识别分类;实验结果显示,屈拇指、屈食指、屈中指、屈无名指、屈小指和屈五指这6种动作的平均识别正确率为94.93%,每次动作识别的平均延时为50.7 ms.
展开▼