CAJ | 학술논문

在对运动想象电位进行模式识别时，需要对原始信号进行滤波以提取信号中区分度较高的成分用于分类，而在噪声较为严重时，滤波方法会导致有用信息的丢失，从而降低分类正确率。针对该问题，本文提出节律成分提取（ Rhythmic Component Extraction， RCE）与共空子空间分解（The Common Spatial Subspace Decomposition ， CSSD）相结合的特征提取方法，对提取的特征信息使用线性分类器进行分类。采用该方法对2003年国际BCI竞赛数据进行识别，测试数据的分类正确率达到87．23％，比使用传统空间滤波方法进行特征提取时的分类正确率提高了6．8％，表明该方法可有效地应用于左右手运动想象电位的识别。
재대운동상상전위진행모식식별시，수요대원시신호진행려파이제취신호중구분도교고적성분용우분류，이재조성교위엄중시，려파방법회도치유용신식적주실，종이강저분류정학솔。침대해문제，본문제출절률성분제취（ Rhythmic Component Extraction， RCE）여공공자공간분해（The Common Spatial Subspace Decomposition ， CSSD）상결합적특정제취방법，대제취적특정신식사용선성분류기진행분류。채용해방법대2003년국제BCI경새수거진행식별，측시수거적분류정학솔체도87．23％，비사용전통공간려파방법진행특정제취시적분류정학솔제고료6．8％，표명해방법가유효지응용우좌우수운동상상전위적식별。
The original signal usually needs filtering to extract the components with higher degree of distinction before the motor imagery potential classification, which results in loss of the useful information under the serious noise interference , as the result, the classification correct rate will be degraded .To solve this problem , hereby a rhythmic component extraction ( RCE) combined with common spatial subspace decomposition (CSSD) for feature extraction while using the linear classifier was proposed .The i-dentification accuracy reached 87.22%by using the method for the international BCI Competition 2003 data, classification accu-racy improved by 6.8%compared with traditional methods of spatial filtering , which show that the method can be effectively ap-plied to the identification of left and right hand motor imagery potential .