Research on Stroke Rehabilitation Technology Based on Brain-Computer Interface with Deep Learning

Stroke has become the leading cause of disability in China and the second leading cause of death globally. Brain-Computer Interface (BCI) technology does not rely on the brains normal output pathways but instead forms a connection between the brain and external machines based on the collection, output, transformation, and decoding of EEG signals. This provides a new solution for the field of neural rehabilitation. This paper focuses on the application of neural network models, particularly deep learning models (DLMs), in brain signal decoding. In the area of motor intent decoding for BCI, research on various models based on convolutional neural networks (CNNs) has addressed issues such as the accuracy of motor imagery EEG signal decoding. In the field of speech communication BCI decoding, deep learning contributes to the development of BCI spellers, especially in steady-state visual evoked potential (SSVEP) decoding and other speech-related signal decoding. By using various DLMs and optimization strategies, it improves decoding efficiency, accuracy, generalization ability, and robustness.