基于Bayes决策理论的表面肌电信号模式分类的研究

通过对采集的四通道表面肌电信号进行分析，对其建立AR（Autoregressive）参数模型，提取AR模型参数构建特征矢量。根据实际表面肌电（SEMG）信号的随机性特征，提出了一种采用Bayes决策理论对肌电信号的AR模型参数特征进行分类的新方法，并运用最小错误率Bayes分类器，很好地实现了对前臂八种动作表面肌电信号的模式分类。平均识别率为99．125％。此外，还提出采用动态聚类中心的方法对其进行了改进，使其平均识别率提高到99．5％。研究表明，采用Bayes分类器对肌电信号的AR模型参数特征进行分类，是一种有效的处理手段，并可直接应用到其它具有随机性特征的生理电信号的模式分类中。

Study of Surface EMG Pattern Classification Based on Bayes Decision Technique

Four channel surface electromyographic(SEMG) signals from four corresponding muscles(palmaris longus,brachioradialis,flexor carpi ulnaris,biceps brachii) are analyzed and an autoregressive(AR) parameter model is constructed.Using AR model parameters as signal characteristics,an eigenvector is constructed.To SEMG randomicity characteristic,a novel SEMG pattern classification method,which is based on Bayes decision technique,is proposed.Eight forearm movement patterns are successfully identified by Bayes decision technique and the average identifying ratio is 99.125%.Moreover eight forearm movement patterns average identifying ratio achieves 99.5% after Bayes decision technique is improved on by using floating clustering center.Experiments show that it is an effective method for SEMG pattern classification,which can be straightforwardly expanded to other randomicity bioelectric signals pattern classification study.