基于相关性分析和支持向量机的手部肌电信号动作识别
doi: 10.3724/SP.J.1146.2007.00499
SEMG Movement Pattern Recognition of Hand Based on Correlation Analysis and SVM
-
摘要: 为了有效提取表面肌电信号(SEMG)的特征,该文提出了一种基于相关性分析的改进的特征提取方法。首先用空域相关法对两路SEMG信号进行消噪预处理,然后对处理后的SEMG信号进行四尺度小波变换,并通过相关性分析提取SEMG信号的重要边缘在各尺度上的小波系数,以各尺度上的这些系数的平方和构建六维特征向量输入支持向量机分类器,对手部的多个动作进行分类。实验结果表明,基于相关性分析和小波变换构筑的特征向量结合支持向量机的方法能够以较高识别率区分伸腕、屈腕、展拳、握拳4种动作,能够得到比传统的神经网络分类器更为准确的分类结果。Abstract: In order to extract effectively the feature of SEMG signal, an improved method of feature extraction based on correlation analysis is proposed. Firstly, the paper decreases the noise included in two channel SEMG signals using spatial correlation filtering. Secondly, the paper analyzes SEMG signal after de-noising with 4-scale wavelet transformation and extract wavelet coefficient of the main fringe by arithmetic of correlation analysis. A 6-dimension eigenvector which is constructed with sum of squares of the wavelet coefficient is inputted SVM. The result shows that four movements (wrist spreads, wrist bends, hand extension, hand grasps) are successfully identified by the method of SVM combined with the eigenvector which is constructed at the condition of correlation analysis and wavelet transformation. The more precise classified results can be get than neural network sorter with this method.
-
Graupe D and Cline W K. Functional separation of EMGsignals via ARMA identification methods for prosthesiscontrol purposes[J]. IEEE Trans. on Syst. Man Cybern, 1975,5(3): 252-259.[2]Hudgins B, Philip Parker, and Scott R N. A new strategy formultifunction myoelectric control. IEEE Trans. on BiomedEng, 1993, 40(1): 82-94.[3]Kang Werr-Jun, et al.. The application of cepstral coefficientsand maximum likelihood method in EMG patternrecognition[J].IEEE Trans. on Biomed Eng.1995, 42(8):777-785[4]罗志增, 杨广映. 表面肌电信号的AR 参数模型分析方法[J].传感技术学报, 2003, 16(4): 384-387.[5]Vapnik V N. An overview of statistical learning theory[J].IEEETrans. on Neural Network.1999, 10(5):988-999[6]Witkin A. Scale-space filtering: A new approach tomulti-scale description[J].IEEE International Conference onICASSP8.1984, 9(3):150-153[7]Xu Yansun, et al.. Wavelet transform domain filters: A spatially selective noise filtration technique[J].IEEE Trans. onImage Processing.1994, 3(6):747-758[8]Mallat S and Hwang W L. Singularity detection andprocessing with wavelets[J].IEEE Trans. on Inform. Theory.1992, 38(2):617-643[9]贾雪琴, 王旭等. 基于小波变换和K-L 展开的单通道表面肌电信号识别. 东北大学学报(自然科学版), 2006, 27(8): 859-862.[10]张学工. 关于统计学习理论与支持向量机. 自动化学报, 2000,26(1): 32-42.[11]王崇文, 李见为, 陈为民. 基于HMM 和SVM 的指纹分类方法[J].电子与信息学报.2003, 25(11):1488-1493浏览[12]Weston J and Watkins C. Multi-class support vectormachines[R]. Royal Holloway College. Tech Rep: CSDTR-98-04, 1998.
计量
- 文章访问数: 3817
- HTML全文浏览量: 122
- PDF下载量: 1035
- 被引次数: 0