基于时频特征优化选择的运动想象脑电信号分类研究

Research on Classification Method Based on Time-frequency Feature Selection for Motor Imagery EEG

针对运动想象脑电信号在频域和时域方面普遍存在的个体差异,提出利用滑动窗技术对运动想象过程的频率段和时间段进行分解,得到各种频率段和时间段的组合。在每种组合下,分别进行三分类运动想象脑电数据的分类实验,以分类正确率为标准,找出最佳频率段和时间段,并将其应用于运动想象脑电信号的个性化特征分析中,以进一步提高BCI系统的分类正确率。

Because individual differences generally existed in the motor imagery EEG signals in aspects of frequency and timedomains, the method of using sliding window to decompose the frequency range and time range of motor image ERD/ERSphenomenon was proposed to get various combinations of frequency and time. In each combination, three classifications of motorimagery EEG experiments were made on different subjects. The best combination of frequency and time of each subject has beenfound out, and the personalized characteristic analysis of motor imagery EEG signals has been realized. The research has furtherimproved BCI system爷s correct classification rate.