Research on SSVEP Recognition Method Based on Residual Network

Currently, Brain-Computer Interface (BCI) is a research hotspot, playing a crucial role in various fields. The difficulty of research is to identify the obtained Electroencephalography (EEG) signals. There are now a large number of recognition methods, but they all have certain shortcomings. Canonical Correlation Analysis (CCA), which is highly sensitive to noise and has poor performance for frequency or phase dense stimuli. Some researchers also use machine learning for signal recognition, but it requires relatively complex feature extraction. Research results indicate that its accuracy is relatively low. In order to tackle the aforementioned challenges such as low recognition accuracy and complex feature extraction, the paper proposes a deep learning-based method for recognizing steady-state visually evoked potentials (SSVEPs). The neural network employs a residual network (ResNet)for implementation. This method utilizes residual networks to extract features from signal data and complete model training through fully connected operations. Input samples into the model to achieve adaptive feature extraction and end-to-end classification of EEG signals. The recognition method proposed in this article can fully extract signal features, achieve effective recognition of SSVEP signals under short-term visual stimuli, and have high target recognition efficiency. The experimental results indicate that compared to the aforementioned recognition methods, the method proposed in this paper exhibits improvements in both recognition accuracy and efficiency.