基于支持向量机多分类器的运动想象电位识别

Recognition of motor imagery based on support vector machine multi-classifier

提出一种基于支持向量机多分类器的运动想象电位识别方法。首先通过neuroscan软件进行脑信号的脑地貌图分析,根据地貌图在不同任务下的脑区优势变化利用小波提取相应脑区的特定频率段信号。再通过小波包提取其能量特征,得到时域、频频域和空域相结合的特征序列。最后利用支持向量机多分类器对想象左手、右手、脚或者想象左手、脚、舌头的脑信号进行识别,并取得了较好的结果。

The method to recognize the electroencephalogram （EEG） signal of the motor imagery base on the support vector machine （SVM） multi-classifier is presented in this paper. Firstly, the physiognomy picture of EEG is obtained through neuroscan software and we get the frequency sect based on the physiognomy picture of EEG using the wavelet. Then we make use of wavelet packets to get the energy feature of EEG and form the feature sequence in time, frequency and space domain. Finally, SVM Multi-classifier is applied to classify the imaginary movements of left hand, right hand and feet as well as left hand, feet and tongue, with a high accuracy.