 This paper proposes a new method called multiscale convolutional neural network MHCNN. It uses a convolutional neural network to classify EEG signals from different frequency bands. This method has been shown to have higher accuracy than traditional methods such as CNN and multiscale convolutional neural networks. Additionally, the authors use a dense network to connect the multiscale convolutional neural network, which reduces the number of parameters and improves the feature propagation. The results show that the combination of the theta-beta-2-gamma band has the best classification effect. This article was authored by Dongwen, Roli, Houtang, and others.