基于肌电信号与肌肉形变信号的手语识别

针对手语手势识别问题，提出一种基于肌电信号与肌肉形变信号的手语识别架构。首先，设计信号采集系统；然后，采集肌电信号与肌肉形变信号，利用滤波及小波降噪等方法对原始数据进行降噪处理。采用基于能熵比的双门限端点检测法提取信号有效活动段；分别提取肌电信号以及肌肉形变信号特征，将所提取的信号特征融合组成特征向量；最后，采用基于网格搜索的支持向量机识别模型对所采集手语动作进行识别。信号融合后手语识别正确率达到97.2%，相对于仅采用肌电信号的手语识别方法，融入肌肉形变信号后识别率提高9.3%。结果表明，基于肌电信号和肌肉形变信号的手语识别框架对动态手语手势具有良好的识别效果。

Sign language recognition based on electromyogram signal and muscle deformation signal

Abstract: Regarding to sign language and gesture recognitions, a framework for sign language recognition based on electromyogram (EMG) signal and muscle deformation signal is proposed. A signal acquisition system is designed and then is used to collect EMG signal and muscle deformation signal. The noise in original data is removed by filtering and wavelet denoising. The effective active segment of the signal is extracted by the double-threshold endpoint detection method based on energy-to-entropy ratio and the features of EMG signal and muscle deformation signal are extracted. The extracted signal features are fused to form eigenvectors. Finally, the collected sign language motions are identified by support vector machine recognition model based on grid search. The accuracy of the sign language recognition method after signal fusion is improved to 97.2%. Compared with the sign language recognition method which only uses EMG signal, the recognition rate is improved by 9.3% after fusing with muscle deformation signal. The experimental results show that the framework for sign language recognition based on EMG signal and muscle deformation signal has good recognition effect on dynamic sign language and gesture.