传统的机器学习方法中脑电信号通常需要经过繁琐的预处理和特征工程才能进行解码。如何构建一个能够快速、可靠地解码运动想象脑电信号的端到端深度学习网络，成为当前运动想象脑电信号解码研究的关键问题。因此，在结合卷积内滑动窗口裁剪策略（sliding window cropping strategy，SWCS）和时间卷积网络（temporal convolutional network，TCN）的基础上，提出一种新的卷积内SWCS的时间卷积网络，并使用该网络对运动想象脑电信号进行识别研究。该网络利用二维卷积提取脑电信号的浅层特征，使用卷积内SWCS将时间序列划分为多个时间窗口，然后将二维卷积提取的脑电信号浅层特征输送到TCN网络中提取时间序列中更高级的时间特征。在第IV届脑机接口竞赛的数据集上的分类结果表明，卷积内SWCS的时间卷积网络的分类效果优秀。

In traditional machine learning methods， Electroencephalogram （EEG） signals typically require extensive preprocessing and feature engineering before decoding. An end-to-end deep learning network is crucial for quickly and accurately decoding motor imagery EEG signals. This system must interpret complex patterns within EEG data to enable effective applications. Therefore， a new temporal convolutional network with intra-convolutional sliding window cropping strategy （SWCS） is proposed. This network uses 2D convolution to extract shallow features of EEG signals and intra-convolutional SWCS to divide the time series into multiple time windows. These shallow features are then fed into the temporal convolutional network （TCN） to extract more advanced temporal features. The classification results on dataset of the IVth Brain-Computer Interface Competition demonstrate that this approach yields excellent classification results.