In the brain-computer interface ( BCI),the feature extraction and classification of electroencephalogram ( EEG) can be a-chieved by massive study of the multilayer feedforward neural network. But the BP neural network based on error back propagation con-verges slowly,and has low efficiency in training,limited accuracy in classification. To solve these problems,the quick and stable Leven-berg-Marquardt algorithm is adopted in this article instead of the BP algorithm to train the neural network. The MATLAB simulation ex-periment about classifying the EEG signals of the motor imagery of left hand and right hand uses the Graz data set B from the BCI com-petition 2008. The simulation results show that the accuracy rate of this algorithm is 87. 1%,which is superior to 78. 2% of the BP algo-rithm,and it converges better as well. This technology provides an effective way for EEG classification.% 在脑机接口(BCI)中,脑电信号(EEG)的特征提取和分类识别可以通过多层前馈神经网络的大量学习来实现,但是基于误差反向传播的BP神经网络标准算法收敛速度慢,在训练中效率不高,分类正确率也很有限.针对这些问题,文中提出使用一种快速稳定的Levenberg-Marquardt算法来代替BP算法进行神经网络的学习训练,并利用BCI 2008竞赛的Graz数据集B进行了对左右手想象运动脑电信号分类的MATLAB仿真实验.该方法使得脑电信号分类的正确率达到87.1%,比BP算法的正确率78.2%要高,并且具有更好的收敛性.该算法为脑电信号的分类提供了有效的手段.
展开▼