Lower limb movement state pattern recognition based on EEG-EMG signals

Assessing the lower limb motor states of stroke patients based on biosignals is very important in the field of medical rehabilitation, and the importance of finding effective physiological signal indicators and processing methods for patient rehabilitation training and evaluation is self-evident. In this paper, a CNN-SVM model is constructed based on the CNN classifier to classify the three motion states of the lower limbs of the subjects, and the method is verified in the WAY-EEG-GAL multimodal open dataset to have a better classification effect, and the experimental data are used to verify the effectiveness of the model classification. The results show that the CNN-SVM method proposed in this paper outperforms the CNN classification model for all three classifications on both the WAY-EEG-GAL dataset and the experimental data, with average accuracies of 86.60% and 95.43%, respectively. This study provides a theoretical basis for combining EEG and EMG signals to establish a BCI-based method for lower limb exercise rehabilitation.