In recent years, brain-computer interface (BCI) is expected to solve the physiological and psychological needs of patients with motor dysfunction with great individual differences. However, the classification method based on feature extraction requires a lot of prior knowledge when extracting data features and lacks a good measurement standard, which makes the development of BCI. In particular, the development of a multiclassification brain-computer interface is facing a bottleneck. To avoid the blindness and complexity of electroencephalogram (EEG) feature extraction, the deep learning method is applied to the automatic feature extraction of EEG signals. It is necessary to design a classification model with strong robustness and high accuracy for EEG signals. Based on the research and implementation of a BCI system based on a convolutional neural network, this article aims to design a brain-computer interface system that can automatically extract features of EEG signals and classify EEG signals accurately. It can avoid the blindness and time-consuming problems caused by the machine learning method based on feature extraction of EEG data due to the lack of a large amount of prior knowledge.