时频分析在语言想像脑机接口中的应用  

Application of time-frequency analysis in speech imagery based brain-computer interfaces

作　　者：王力[1] 张雄[1] 仲雪飞[1] 樊兆雯[1] 张玉[1] 孙瀚[1] 

机构地区：[1]东南大学电子科学与工程学院,南京210096

出　　处：《东南大学学报（自然科学版）》2014年第6期1126-1130,共5页Journal of Southeast University：Natural Science Edition

基　　金：国家重点基础研究发展计划(973计划)资助项目(2010CB327705);国家高技术研究发展计划(863计划)资助项目(2012AA03A302);中央高校基本科研业务费专项资金资助项目(CXLC12_0095)

摘　　要：为了分析语言想像所诱导的脑电信号的时间和频率特性,利用事件相关谱扰动(ERSP)对信号进行时频分析.首先,经ERSP确定能量波动的频率范围;然后,根据该频率范围计算信号的事件相关(去)同步(ERD/ERS),并利用共空间模式和支持向量机分别对单次实验数据的特征值进行提取和分类.对8位被试的试验结果分析表明,被试间的频率范围具有显著的差异,其中4位被试的频率范围含有α波,3位被试含有α波和β波,1位被试的脑电信号在默读汉字时无明显变化.2个汉字默读时的脑电信号可产生相似的ERD/ERS.优化频率范围后针对这2个汉字的平均分类正确率分别提高了2.25%和1.39%.时频分析能更好地显示脑电信号的能量变化率,并能改善语言想像脑机接口的性能.To analyze time and frequency characteristics of electroencephalography signals induced by speech imagery,event-related spectral perturbation （ERSP）is used in the time-frequency analy-sis of signals.First,the frequency range of the energy fluctuation can be determined by ERSP. Then,according to the frequency range,event-related （de）synchronization （ERD/ERS）of the sig-nals is calculated,and the eigenvalues of each experimental data are extracted and classified by com-mon spatial pattern and support vector machine,respectively.The results of eight subjects show that the frequency ranges are obviously different among subjects.The ranges of four subjects include αrhythm and those of three subjects include αand βrhythms.The electroencephalography signals of one subject do not evidently change when he reads a Chinese character in mind.The similar ERD/ERS of electroencephalography signals may produce while reading two Chinese characters in mind. After optimizing the frequency ranges,the average classification accuracy for the two characters are improved by 2. 25% and 1. 39%,respectively.The energy gradient of electroencephalography sig-nals can be exhibited better by time-frequency analysis,and the performance of speech imagery based brain-computer interfaces can be also improved.

关 键 词：语言想像 脑电信号 时频分析 脑机接口 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]