基于MTACNet网络的运动想象脑电分类

Motor imagination EEG classification based on MTACNet

为了更好地利用到脑电信号中的相关特征,改善运动想象脑电的分类性能,构建了一种基于混合特征和并行多尺度TCN模块的多层卷积网络(MTACNet)。首先,搭建基于混合特征的多层卷积神经网络,并在其中嵌入高效通道注意力机制,选取PReLU作为激活函数,以提取脑电信号中的时域和空域信息;然后对TCN模块进行改进,构建并行多尺度时域特征提取模块,接入多层卷积网络,进一步挖掘不同时间尺度的特征信息。在公开数据集BCI_IV_2a和自采数据集SCU_MI_EEG上进行测试,平均分类准确率分别为86.15%、77.10%,标准差分别为9.17%、13.58%。并且针对自采数据集,设计了一种融合多频域脑电信号进行三通道输入的预处理方法,经过预处理后使平均分类准确率提升了3.29%。实验结果表明,与其他方法相比,本文所构建的分类网络取得了较为不错的分类效果,所设计的预处理方法能够降低复杂环境和无关干扰因素对分类结果的影响。

In order to make better use of the relevant features in EEG signals and improve the classification performance of motor imagery EEG,a multi-layer convolutional network(MTACNet)based on mixed features and parallel multi-scale TCN modules was constructed.First,build a multi-layer convolutional neural network based on mixed features,and embed an efficient channel attention mechanism in it,and select PReLU as the activation function to extract the temporal and spatial information in the EEG signal;then improve the TCN module,build a parallel multi-scale timedomain feature extraction module,connect to a multi-layer convolutional network,and further mine feature information at different time scales.Tested on the public dataset BCI_IV_2aand the self-collected dataset SCU_MI_EEG,the average classification accuracy rates are 86.15%,77.10%,and the standard deviations are 9.17%,13.58%,respectively.And for the self-collected data set,apreprocessing method was designed to fuse multi-frequency domain EEG signals for three-channel input.After preprocessing,the average classification accuracy rate increased by 3.29%.The experimental results show that:Compared with other methods,the classification network constructed in this paper has achieved relatively good classification results,and the designed preprocessing method can reduce the impact of complex environments and irrelevant interference factors on the classification results.