I摘 要脑机接口是人工智能的一项重要内容,也是一种全新的大脑与计算机通讯的方式,具有改变人机交互方式的巨大价值。近年来对于脑电信号的研究越来越成熟,许多方法被应用于对运动想象中的脑电信号进行分类。然而卷积神经网络作为一门新兴起的技术,在该领域上应用还比较少。本文探究了卷积神经网络在运动想象脑电信号分类上的应用。主要使用脑机接口竞赛 IV 2b 数据集,探索运动想象脑电信号分类任务。首先查阅了近年来的众多脑电信号分类模型,对于传统机器学习的方法和深度学习的方法进行了分析,然后在实用卷积神经网络的基础上,加入深度可分离卷积构建了一个轻量级的卷积神经网络,使用该模型将相关运动想象信息从脑电信号中提取出来,并由此对脑电信号中的时频域特征进行进一步的理解, 最后给出了模型进一步改进的思路。本文研究表明,在对该数据集的运动想象脑电信号做二分类时,本文所提出的卷积神经网络模型分类准确率为 92.9%,信息传输率能达到 27.03 比特每分钟,取得了较好的效果。因此卷积神经网络能够对脑电信号的时频域特征做到很好的解码和分类。关键词:脑机接口;卷积神经网络;脑电信号;解码 IIABSTRACTThe brain-computer interface is an important content of artificial intelligence, and also a brand-new way of brain-computer communication, which has great value in changing the way of human-computer interaction. In recent years, research on EEG signals has become more and more mature, and many methods have been applied to classify EEG signals in motor imagination. However, as a new technology, convolutional neural network(CNN) is still relatively few in this application. This article explores the application of convolutional neural networks in the classification of motor imaging EEG signals. Mainly use the BCI competition IV 2b data set to explore the task of motor imaging EEG signal classification. First I consulted many EEG signal classification models in recent years, then I analyzed the traditional machine learning method and deep learning method. In addition, on the basis of pra...