摘要
采集人体表面肌电信号并进行有效模式识别,是多功能假肢研究的关键技术之一.本文研究肌电信号(EMG)的频域信息,以提取有效特征识别不同的手腕部动作.实验采集了8种不同的手腕动作的EMG信号.分别提取了EMG信号的时间序列模型(AR)、时域统计量(DT)、功率谱估计(PSD)、短时傅里叶变换(STFT)和互功率谱(CPSD)5种不同的特征向量,并采用支持向量机(SVM)对特征进行分类.通过特征分类正确率的分析比较,发现提取的信号频率特征能够有效地提高动作的识别率.
Extracting the body surface EMG signals EMG and acting the effective pattern recognition is one of the key technologies in the multi-function prosthetics research. In this paper, the frequency domain infor- mation of EMG is studied to extract effective features and identify the different wrist action. The EMG signal data of 8 different wrist actions are collected. Then, five different feature vectors of EMG signals are extract- ed, including the time series model (AR), the time-domain statistics (DT), the power spectrum estimation (PSI)), the short time Fourier transform (STFT) and the cross-power spectrum (CPSD), and classified us- ing the support vector machine (SVM). With analyzing and comparing the feature classification accuracy, it is found that the extracted frequency characteristics of the signals can effectively improve the recognition rate on the movements.
出处
《测试技术学报》
2011年第4期346-350,共5页
Journal of Test and Measurement Technology
基金
常州科技局立项科技型中小企业技术创新计划(CN20090051)
江苏技术师范学院青年科研基金(KYY08044)资助
关键词
多功能假肢
肌电信号
频域特征
支持向量机
分类识别方法
multi-functional prosthetics
EMG
frequency characteristics
SVM
the classification and recog-nition methods