基于径向基网络的表面肌电信号处理

利用AR模型对实验所采集到的原始二通道表面肌电信号（SEMG）加以分析，提取AR系数作为特征值，将其作为训练样本输入到RBF神经网络进行训练，用此网络对前臂的伸臂和曲臂两种运动模式的表面肌电信号进行模式分类。实验表明，基于径向基函数RBF神经网络分类准确率比BP神经网络更高，具有较强的鲁棒性和自适应能力，可以有效识别肌肉的单动作模式。

Surface Electromyography Analytical Method based on the Method of RBF Neural Network

This paper provides a new feature extraction method to extract the eigenvectors(AR coefficient)from the original two-channel surface electromyography(SEMG)signals.We use the AR coefficients as the input to train RBF Neural Network and to classify the two movement patterns(flexor carpi,extensor carpi)of the forearm.The experiment shows that RBF Neural Network has a more high discrimination rate,robustness and adaptability compared with BP Neural Network.Muscle's action pattern also can be effectively recognized by this method.