基于块稀疏贝叶斯学习的肌电信号特征提取

发布时间：2018-01-31 13:33

本文关键词： 表面肌电特征提取非平稳信号块稀疏贝叶斯学习　出处：《仪器仪表学报》2014年12期 　论文类型：期刊论文

【摘要】：针对上肢多种运动的模式识别问题,对表面肌电的特征提取方法进行了研究,旨在取得更有鉴别能力的特征。采用基于块稀疏贝叶斯学习的方法提取s EMG信号中蕴含的运动信息,用于上肢动作中的整手以及精细的手指动作的识别。基于块稀疏贝叶斯学习的方法提取到的分解系数特征DC能表征s EMG信号的非线性特征。在对8位受试者6类整手以及精细的手指动作的识别实验中,将DC特征结合Willison幅值特征能更有效地反映出6类动作之间的差异,平均动作识别成功率可达92.9%。实验表明,所提出的特征提取方案可行。与传统的特征提取方法相比,此方案进一步提高了多种动作模式的识别精度。
[Abstract]:Aiming at the pattern recognition of multiple movements of upper limbs, the feature extraction method of surface electromyography is studied. In order to obtain more discriminant features, a block sparse Bayesian learning method is used to extract the motion information contained in s EMG signal. It is used to recognize the whole hand and fine finger movement in upper limb movement. The decomposition coefficient feature DC which is extracted from block sparse Bayesian learning method can represent s. The nonlinear characteristics of EMG signal. In the experiment of recognition of 6 kinds of whole hand and fine finger movement of 8 subjects. DC features combined with Willison amplitude features can more effectively reflect the differences between the six types of actions, and the average success rate of motion recognition can reach 92.9%. Compared with the traditional feature extraction method, the proposed scheme can further improve the recognition accuracy of various action patterns.
【作者单位】：北京航空航天大学;
【分类号】：TN911.7
【正文快照】： 1引言人体自主运动时,神经肌肉系统活动会发放出电信号。从人体浅表骨骼肌的表面经生物电极记录到的此类电信号就是表面肌电信号(surface electromyography,s EMG)。s EMG信号能够反映人体神经、肌肉的活动状态。提取s EMG的信号特征可用于控制或诊断,因此s EMG信号在人机交互

【参考文献】