CAJ | 학술논문

脑电信号的时间和频率间隔选择对脑机接口的分类性能具有重要的影响。针对多类运动想象脑机接口系统，提出一个新的基于时间段和频带联合选择的分类算法。该算法首先使用滑动窗将运动想象产生的脑电信号在时域和频域进行分割，然后在每一对截取的时间段和频段，使用多类共空域模式算法提取脑电特征信号，最后使用k－最近邻算法对特征信号进行分类。交叉验证的分类识别率作为最优时间段和频带的选择标准。使用一个四类数据集对这个分类算法的性能进行了测试。与现有的3个典型算法比较，这个算法取得了最高的平均分类正确率，证实了这个基于时间段和频带联合选择的分类算法的有效性。
뇌전신호적시간화빈솔간격선택대뇌궤접구적분류성능구유중요적영향。침대다류운동상상뇌궤접구계통，제출일개신적기우시간단화빈대연합선택적분류산법。해산법수선사용활동창장운동상상산생적뇌전신호재시역화빈역진행분할，연후재매일대절취적시간단화빈단，사용다류공공역모식산법제취뇌전특정신호，최후사용k－최근린산법대특정신호진행분류。교차험증적분류식별솔작위최우시간단화빈대적선택표준。사용일개사류수거집대저개분류산법적성능진행료측시。여현유적3개전형산법비교，저개산법취득료최고적평균분류정학솔，증실료저개기우시간단화빈대연합선택적분류산법적유효성。
The time and frequency intervals of EEG signals have significant influence on the classification performance of brain-computer interfaces(BCIs).Based on joint selection of time segment and frequency band,a new classification algorithm was proposed to classify multiclass motor imagery BCI systems in this paper.A sliding window was firstly used to segment the EEG signals in time and frequency domain.Then the multiclass common spatial pattern(CSP)was utilized for feature extraction in each pair of time and fre-quency segment.Finally we employed the k-nearest neighbor(KNN)to classify the extracted feature sig-nals.Cross-validated classification accuracy was used for the selection criterion of optimal time and frequen-cy segment.The performance of the algorithm was tested and compared with those of three typical algo-rithms by a classification experiment on a four-class data set.The results suggested that the algorithm could achieve a highest accuracy rate,validating the effectiveness of the j oint selection of time segment and frequency band based classification algorithm.