基于AR参数模型与聚类分析的肌电信号模式识别方法

The Method of Surface EMG Pattern Recognition Based on AR Parameter Model and Clustering Analysis

肌电信号是与神经肌肉活动有关的生物电的体现,肌电信号的模式识别是肌电应用的基础。利用现代功率谱估计中的参数模型法,对从掌长肌、肱桡肌、尺侧腕屈肌和肱二头肌采集的4路表面肌电信号建立AR参数模型,并提取其AR模型参数作为信号的特征,构造特征矢量,提供给基于距离测度的Mahalanobis距离分类器进行模式分类,能够成功地识别出握拳、展拳、腕内旋、腕外旋、屈腕、伸腕、前臂内旋、前臂外旋8种动作模式。实验表明,该方法识别率高、鲁棒性好,为肌电等非平稳生物电信号的模式识别提供了一种新方法。

Surface electromyography （SEMG） is a bio-electrical manifestation related to neuromuscular activation. Electromyography pattern recognition plays an important role in SEMG application. A new SEMG pattern recognition method based on AR parameter model and clustering analysis is proposed. Four channel SEMG signals from corresponding muscles （palmaris longus, brachioradialis, flexor carpi ulnaris, biceps brachii） are picked up and analyzed with AR model parameter. Using AR model parameters as signal characteristics, an eigenvector is composed and inputted to the Mahalanobis distance classifier to identify different movement patterns. Eight movement patterns （hand grasps, hand extension, wrist circumrotates entad, wrist cireumrotates forth, wrist bends, wrist spreads, forearm cireumrotates entad, forearm circumrotates forth） are successfully identified. Experiments show that the proposed method performs very well and the recognition result is robust. It is believed that this method can be straightforwardly expanded to other nonstationary bioelectric signals pattern recognition study.