基于FastICA和卷积神经网络的脑电信号分类算法

EEG signal classification algorithm based on FastICA and Convolution Neural Network

为了改善传统脑电信号分类时间长、精度不够准确且分类难度较大的问题,利用脑电传感器(Mind Wave传感器)及Real Term软件从串口抓取数据获取脑电波TGAM数据包,并对采集的脑电信号数据进行分解计算处理,得到各个波段数据,使用基于负熵的独立分量分析的固定点算法(FastICA)提取脑电信号特征,并用深度学习分类算法对脑电信号进行分类。传统机器学习算法不能准确分类复杂的脑电信号,运用卷积神经网络(Convolutional Neural Network,CNN)提取数据进行训练,构建分类器,实现了对脑电信号更高效更准确的分类。实验结果表明,与Fisher线性判别、BP神经网络、朴素贝叶斯相比,此算法可以更准确地区分是否清醒的状态,对脑电信号分类的研究具有重大意义。

In order to improve the long classification time,low accuracy and difficult operation for the traditional EEG signal classification,the brainwave sensor（ Mind Wave sensor） and Real Term software are used to capture the brainwave TGAM data packet from the serial port so as to obtain all band data by decomposition computation on the sampled EEG. FastICA is adopted to extract the EEG signal characteristics and the EEG signal is classified by the depth learning classification algorithm. Convolutional Neural Network（ CNN） is used to train the data,and the classifier is built to realize the accurate classification for EEG signal. The experimental results show that the proposed algorithm can distinguish more clearly from the Fisher linear discriminant,BP neural network and Naive Bayesian,and it is of great significance to study the classification of EEG.