基于交叉运算的人工蜂群优化BP神经网络的脑电信号分类

Classification of Electroencephalography Based on BP Neural Network Optimized By Crossover Operation of Artificial Bee Colonies

为了提高脑电信号的分类准确率,提出一种基于人工蜂群算法和BP神经网络的分类方法。针对反向传播(BP)神经网络存在全局搜索能力差、对初始权重敏感和人工蜂群算法的搜索公式精于探索但疏于开发等问题,采用全局搜索因子来增强人工蜂群算法的开发能力,再加入交叉运算来解决人工蜂群算法的全局搜索。采用改进的算法来优化BP神经网络对初始权重敏感的问题,进而实现对脑电信号的分类。实验结果表明,所提算法对脑电信号的分类准确率更高,分类准确率达到91.5%,而且可以加快收敛速度。

In order to improve the classification accuracy of EEG signals,a classification method based on an artificial bee colony algorithm and back propagation(BP)neural network is implemented.In order to improve the poor global search abilities and sensitivity to initial weights of BP neural networks,the global search factor is used to enhance an artificial bee colony algorithm search formula,which is proficient in exploration but required further development.A crossover operation is used to improve the global search capacity of the artificial bee colony algorithm.This enhanced algorithm is further used to optimize the sensitivity of the BP neural network to initial weights,enabling classification of EEG signals.The experiment results show that the proposed algorithm produces a highly accurate EEG signal classification of 91.5%with an accelerated convergence speed.