脑电信号的混沌分析和小波包变换特征提取算法  

EEG feature extraction algorithm based on chaos analysis and wavelet packet transform

作　　者：罗志增[1] 李亚飞[1] 孟明[1] 孙曜[1] 

机构地区：[1]杭州电子科技大学机器人研究所,杭州310018

出　　处：《仪器仪表学报》2011年第1期33-39,共7页Chinese Journal of Scientific Instrument

基　　金：国家863项目(No.2008AA04Z212);国家自然科学基金(No.60705010;No.60874102);浙江省教育厅科研项目(Y200909010)资助

摘　　要：针对脑电(EEG)信号的手部动作模式信息处理,提出一种混沌分析和小波包变换相结合的特征提取方法。用眼动辅助来采集手部动作时的脑电信号,对采集的C3、C4、P3和P4脑电信号消噪后分别用混沌分析和小波包变换的方法进行特征提取,前者提取混沌特征的最大Lyapunov指数和关联维数,组成8维向量;后者提取脑电信号的4种特征节律波,分别计算其相对能量,组成16维向量;最后把两种方法提取的向量组成24维特征向量,输入SVM分类器,实现基于EEG信号的手部动作模式的识别。对不同个体上翻、下翻、展拳、握拳4种手部动作的识别实验表明,平均识别率均在80%以上,明显优于其他方法识别的结果。In order to recognize hand movement based on EEG,a feature extraction method is presented,which combines chaos analysis and wavelet packet transform.Firstly,the eye-motion assisted hand-motion C3,C4,P3 and P4 EEG signals are recorded;after de-noising,the EEG features are extracted using the above mentioned two methods.Chaos analysis method is used to extract the maximum Lyapunov exponents and correlation dimensions,which make up an 8 dimension vector.Wavelet packet transform is used to extract the 4 feature rhythm waves;their relative energies are calculated and a 16 dimension vector is built.Then the two resultant vectors are combined to form a 24 dimension feature vector,which is inputted into an SVM classifier to recognize the hand motions.Experimental results indicate that the proposed method can discriminate four hand motion patterns in different individuals（including hand up and down,hand opening and closing） with mean correct recognition rates all above 80%,which is obviously better than those using other methods.

关 键 词：脑电信号 手部动作识别 混沌分析 小波包变换 SVM分类器 

分 类 号：TP212.13[自动化与计算机技术—检测技术与自动化装置]