基于拼写器的脑电信号分析和判别模型研究

Research on EEG Signal Analysis and Discriminative Model for Speller

针对P300拼写器中信息传输和目标分类的性能问题,提出了脑功能网络和多模型融合的脑电信号处理方式.首先,对原始信号进行独立成分分析、最佳滤波频带选择及数据分段等预处理操作去除信号中的伪迹噪声.其次,针对原始脑电数据冗余问题,引入了脑功能网络模型对通道敏感度进行分析,仅选择部分通道数据,在不降低分类准确率的情况下,提高信息传输速率,减少目标字符识别所需轮次.最后,针对脑电信号分类问题,提出了基于随机森林和支持向量机的融合模型,通过引入权重和投票机制,根据多个模型投票的结果选择目标字符,提高分类准确率.实验表明,研究结果可以为基于拼写器的脑电信号分析提供一定的理论支持.

In this paper,we propose a brain function network and multi-model fusion for EEG signal processing to address the performance problems of information transmission and target classification in P300 speller.First,the original signal is pre-processed with independent component analysis,optimal filtering band selection and data segmentation to remove artifact noise from the signal.Secondly,to address the redundancy problem of the original EEG data,a brain functional network model is introduced to analyze the channel sensitivity and select only some of the channel data to improve the information transmission rate and reduce the number of rounds required for target character recognition without reducing the classification accuracy.Finally,a fusion model based on random forest and support vector machine is proposed for the EEG signal classification problem.By using a weighting and voting mechanism,the target characters are selected based on the results of multiple model voting to improve the classification accuracy.The experiments show that the results of this paper can provide theoretical support for the analysis of EEG signals for speller.