基于块稀疏贝叶斯学习肌电信号特征提取_丁帅.pdfVIP

35 12 Vol. 35 No. 12 2014 12 Chinese Journal of Scientific Instrument Dec. 2014 基于块稀疏贝叶斯学习的肌电信号特征提取 丁 帅, 王 亮 (北京航空航天大学 北京 100191) 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 sEMG 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 sEMG 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。 关键词: 表面肌电; 特征提取; 非平稳信号; 块稀疏贝叶斯学习 中图分类号: TP391.4 TH39 文献标识码: A 国家标准学科分类代码: 410.4 Feature extraction of surface electromyography based on block sparse Bayesian learning Ding Shuai, Wang Liang (Beihang University, Beijing 100191, China) Abstract: Aiming at the multi-movement pattern recognition problem of upper limb, the feature extraction method of surface electromyography (sEMG) is studied to obtain stronger distinguishing features. A new method based on Block Sparse Bayesian Learning (BSBL) is proposed to capture the movement information contained in sEMG, which is used to classify gross hand and elaborate finger movements. The features, namely decomposition coefficients (DC), extracted with the proposed method can represent the nonlinear characteristics of sEMG. In the pattern recognition experiment on classifying the gross hand and elaborate finger movements of eight participants, DC feature combining with Willison amplitude (WAMP) feature can effectively reflect the divers