改进BP神经网络的EMG手指运动识别

在基于肌电信号(EMG)手指运动的模式识别中,稳定性和识别率是两个主要问题,为此提出了一种新的EMG模式识别算法。该算法采用现代信号处理理论中的AR模型和改进的BP神经网络相结合的算法,有效的解决了BP网络识别中落入局部极值问题。进行试验,将提取到的特征值输入MATLAB建立一个改进多层BP神经网络,识别三个不同类型的手指运动。实验表明,改进BP算法较传统BP算法获得了更高的识别精度,达到94%左右。

Optimized BP Neural Networks for EMG Finger Movement Recognition

In the pattern recognition of Finger movement based on electromyography （EMG）, the Stability and Recognition rate are both the problem. The paper proposes a new method of pattern recognition of EMG signal. The method combination of the algorithm using BP neural network AR model and the improvement of modern signal pro-cessing in the theory of the algorithm, can effectively solve the problem of BP network into local extremum recogni-tion. To make the classification of the eigenvalues of the EMG, these eigenvalues have been inputted to the MAT-LAB to build up a improved multilayer BP neural networks. For the recognition of three different kinds of finger mo-tion＆#39;s EMG signals, the experiments show that the improved BP algorithm, to obtain higher recognition accuracy than the traditional BP algorithm, to around 94%.