文章摘要

文章信息/Info

Title:: Analysis on the characteristics of upper limb muscle fatigue of athletes based on surface EEG

Author(s):: ZHAO Zhiwei1; WANG Meiling2; 1.Department of Physical Education,Inner Mongolia Medical University,Hohhot 010110,China;2.Inner Mongolia Medical University, School of Pharmacy, Hohhot 010110

Keywords:: Energy characteristics; EEG signal; Signal energy characteristics; Upper limb movement; Sports intention; Intelligent recognition

摘要:: 传统的对运动员上肢运动意图识别方法，没有对采集获得的大量脑电信号进行平滑滤波，存在较多毛刺干扰，导致识别准确率和识别率不高。我们提出一种基于能量特征的脑电信号上肢运动意图智能识别方法，采用快速傅里叶变换方法对采集获得的运动障碍患者脑电信号进行频率分析，获得患者脑电信号中的μ波和β波频率分布规律，找到脑电信号噪声所在频段；并采用Daubechies小波将患者脑电信号进行3阶分解，将患者脑电信号中低频部分的小波系数进行归零处理后，再进行脑电信号重构，即可消除低频脑电信号中的噪声干扰；在此基础上，采用小波包系数分析患者脑电能量，实现患者脑电信号能量特征提取；基于脑电信号能量特征，采用马氏距离判别方法对上肢运动意图进行智能识别。实验结果显示，所提方法能够去除原始脑电信号中的“毛刺”干扰，平均识别率结果为88.6%，识别准确率和识别率较高。

Abstract:: The traditional recognition method of athletes’ upper limb motion intention does not smooth filtering a large number of EEG signals collected, and there are many burrs, which leads to low recognition accuracy and recognition rate. An intelligent recognition method of upper limb motion intention based on energy characteristics was proposed. The frequency of EEG signals collected from motor disorder patients was analyzed by using fast Fourier transform (FFT) method. The frequency distribution of μ wave and β wave in EEG were obtained, and the frequency band of EEG noise was found, and the low frequency part of EEG was decomposed by Daubechies wavelet. After the wavelet coefficients were returned to zero and the EEG signals were reconstructed, the noise interference in the low-frequency EEG signals could be eliminated; on this basis, the wavelet packet coefficients were used to analyze the EEG energy of the patients, and the EEG signal energy characteristics were extracted; on the basis of the EEG signal energy characteristics, the upper limb movement was detected by Mahalanobis distance discrimination method. The experimental results show that the proposed method can eliminate the "burr" interference in the original EEG signals, and the average recognition rate is 88.6%. It has high recognition accuracy and recognition rate.