事件相关电位(P300)脑电信号解码的两个问题及其解决方法

Two Problems in Decoding Event-related Potential(P300)EEG Signals and Their Solutions

为提高脑机接口系统中目标分类准确率并保证一定的信息传输速率,本文首先建立了多尺度卷积神经网络模型,然后建立一个通道选择算法,给出针对每个被试的、更有利于分类的通道组合.最后利用第十七届中国研究生数学建模竞赛C题公开数据训练得到面向受试者的P300识别的特定模型.实验结果表明:筛选出特定5位被试者的最优通道,识别平均准确率最高可达72%,平均信息传输速率最高可达35.7bits/min,取得了较好的效果.

In order to improve the accuracy of target classification and ensure a certain rate of information transmission in the brain computer interface system,a multi-scale convolutional neural network model is set up,and then a channel selection algorithm is established to give a channel combination that is more conducive to classification for each subject.Finally,a specific model of tested-oriented P300 recognition is obtained by using the open data of Question C in the 17th China Graduate Mathematical Contest in Modeling.The experimental results show that the average recognition accuracy rate of the optimal channel selected for the specific 5 subjects is up to 72%,and the average information transmission rate is up to 35.7bits/min.