基于LVQ的自发脑电信号的分类研究

Research on spontaneous EEG signal classification based on LVQ

为了识别在不同思维状态下的自发脑电(EEG)信号 ,本文用6阶自回归(AR)模型表示EEG信号 ,用学习矢量量化(LVQ)神经网络作分类器 ,分别用LVQ1和LVQ2.1算法对网络进行训练 ,并对分类结果进行测试 ,比较了网络选择不同参数时对分类正确率的影响。研究表明 :竞争层神经元数目直接影响了正确率 ,当选择最佳参数值时分类正确率为62 %～83 % ,因人而异。

The paper is to identify EEG signals of different mental tasks with sixth-order autoregressive,AR(6)coeffi-cients derived from raw,unfiltered EEG signals and LVQ network as classifier.The VLQ network is trained with LVQ1and LVQ2.1algorithms and compared the classification accuracy with different parameters.It shows that the competi-tive layer neuron numbers affect the result more than others.62%-83%classification accuracy is reached with optimal parameters.