中文摘要：

针对脑机接口研究中运动想象脑电信号的模式识别问题,提出了一种基于离散小波变换和AR双谱的特征提取方法。该方法首先利用Daubechies类小波函数对二路脑电信号进行3层分解,抽取小波系数的均值、能量均值、均方差三个特征;然后,采用5阶AR模型进行双谱估计,抽取双谱切片特征;最后,将这两类特征进行组合后使用马氏距离线性判别进行分类。利用BCI2003竞赛的标准数据,该方法使得EEG的识别正确率达到92.86%,与竞赛的最好结果（89.29%）相比提高了3.57%,为BCI研究中脑电信号的模式识别提供了有效的手段。

英文摘要：

Aiming at the issue of motor imagery electroencephalography（EEG） pattern recognition in the research of brain-computer interface（BCI）,a novel feature extraction method based on discrete wavelet transform（DWT） and autoregressive（AR） bispectrum is proposed.Firstly,two-channel EEG signals are decomposed to three levels using Daubechies wavelet function.Secondly,the mean,average power and standard deviation of the wavelet coefficients are computed.Thirdly,bispectrum is estimated using fifth-order AR model and diagonal slice characteristic of the bispectrum is extracted.Finally,the above two kinds of features are combined,and linear discriminant analysis（LDA） based on Mahalanobis distance is utilized to classify the combined feature.This method is applied to the standard dataset of BCI Competition 2003,and experimental results show that the recognition rate reaches 92.86%,which is 3.57% higher than the best result（89.29%） of the competition.This technology provides an effective approach to EEG pattern recognition in BCI research.