摘要
基于经验模态分解(EMD)理论,提出一种左右手运动想象脑电信号分析方法。首先利用时间窗对脑电信号数据进行划分,对每段数据通过经验模态分解法将其分解为一组固有模态函数IMF,提取主要信号所在的IMF层去除信号中的噪声。对含有主要信号的几层IMF进行Hilbert变换,得到瞬时频率与对应的瞬时幅值。再提取左右手想象的特定频段mu节律和beta节律的能量信号作为特征,分别利用支持向量机(SVM)和Fisher进行了分类比较。对EMD和小波包在去噪和特征提取进行了比较。结果表明,EMD是一种很有效的去噪方法,经过EMD分解后提取的能量信号在区分左右手想象上更具有优势,识别率高。
Based on Empirical Mode Decomposition(EMD) theory, We present a method to analysis the electroencephalogram (EEG)signal of right and left motor imagery. First, EEG signals were divided into several segments using time scale. EMD method decomposes every EEG signal segment into a group of intrinsic mode functions(IMFs) ,using the IMF to denoise. According to the Hilbert transform, we can obtain the instantaneous frequency and instantaneous amplitude. In order to recognize the EEG signal of right and left motor imagery using SVM and Fisher, it is important to obtain the energy of the special frequency mu and beta. Meantime ,we compare EMD with wavelet to denoise and to recognize. The results show that EMD is a useful method to denoise, and EMD is a efficiency method to analysis the electroencephalogram (EEG)signal of right and left motor imagery. In addition, the proposed method obtains a high recognition rate .
出处
《生物医学工程学进展》
CAS
2009年第3期125-130,共6页
Progress in Biomedical Engineering
基金
国家自然科学基金项目(60775033
60674089)
上海市浦江人才计划项目(07PJ14031)
上海市重点学科项目(B504)
关键词
EMD
固有模态函数
小波
支持向量机
Empirical mode decomposition , Intrinsic mode function , Wavelet , SVM