基于深度神经网络的sEMG手势识别研究

为了提高表面肌电信号（sEMG）手势识别算法的准确性，并解决人为提取大量特征具有局限性的问题，提出了一种基于深度神经网络的手势识别方法。将MYO臂环采集到的8通道sEMG数据，采用活动段分割的方法探测到有效动作；设计出一种融合卷积神经网络（CNN）和长短时记忆（LSTM）网络的神经网络；实验的结果表明手势识别准确率为91.6%，验证了提出的方案高效可行。

Research on Gesture Recognition of sEMG Based on Deep Neural Network

In order to improve the accuracy of sEMG gesture recognition algorithm and solve the limitation caused by extracting a large number of features artificially, this paper proposes a gesture recognition method based on deep neural network. Firstly, it uses an active segment segmentation method on 8 channel sEMG data which is collected by MYO armband to detect effective actions. Then, it designs a neural network which combines Convolutional Neural Network（CNN） and Long-Short Term Memory network（LSTM）. The result shows that the accuracy of gesture recognition reaches 91.6% and the proposed method is proved to be efficient and feasible.