基于DSP的三自由度肌电假手实时控制方法

 利用安置在拇长屈肌，指深屈肌和指伸肌上的3个电极所测得的肌电信号，采用所提出的新的模式分类器，可以实现基于DSP的三自由度假手手指运动的实时控制。该分类器采用自回归(AR)参数模型和样本熵的方法构造特征矢量，经过由弹性反向传播(RP)算法构建的3层前馈神经网络的分类，能够成功地分辨出拇指、食指和中指的弯曲与伸展运动，平均识别率可以达到91%以上。实验结果表明，该分类器具有很高的辨识能力，同时由于其较小的计算量，也为嵌入式的多自由度肌电假手控制提供了一种新的控制方法。

Real-time Control Method for 3-DOF EMG Prosthetic Hand Based on DSP

A novel classifier is presented in this paper.It can achieve the finger motion real-time control of a 3-DOF prosthetic hand by measuring the surface electromyography(EMG)signals.The signals are measured by three electrodes which are mounted on the flexor pollicis longus,flexor digitorum profundus and extensor digitorum.The methods of autoregressive(AR)model and sample entropy are used to construct the feature vectors for the classifier.Using these feature vectors,the flexions and extensions of the thumb,the index finger and the middle finger can be successfully identified via three-layer feedforward neural network trained by resilient backpropagation(RP)algorithm.The average correct rate is above 91%.Furthermore,the experimental results show that the classifier has high recognition capability and lower calculation,and also present a new control method for the embedded EMG control of the multi-DOF prosthetic hand.