【摘要】 在表面肌电信号(electromyography,EMG)中,各类动作的识别是一个重要研究方向。本文采用独立分量分析independent component analysis,ICA)对肌电信号进行处理,消除各动作信号之间的相互线性耦合叠加,并采用信号的小波熵作为特征向量进行模式识别。试验表明,在对信号进行先期ICA处理以后,动作模式的识别效果较好。此方法也可应用于其他生理信号的识别分类。更多还原

【Abstract】 For the electromyography(EMG) processing,the motions recognition is a hot domain.The independent component analysis(ICA) is used to pre-process and decouple the EMG signals.The wavelet entropy of the EMG signals are executed as the characterization vectors.Using the characterization vectors as the input and training the neural networks,we can get an excellent classifier with good performance.The result shows that the motions pattern recognition with signals pre-processed by ICA is better than that before decoupling the EMG.更多还原