Recognition of Epilepsy in GTCs Based on Wrist Signals

This paper focuses on the GTCS seizure detection in epilepsy, based on the patient’s wrist measurement acceleration, skin conductance, and heart rate to study and extract the corresponding features, the extracted features are clustered into two groups to reduce the feature dimension of each model, and thus reduce the amount of computation, and the machine learning algorithms in the linear kernel SVM, XGBoost were used for training, the two groups of models obtained were The combination of the two groups of models obtained is carried out for practical validation, and the algorithm with the best validation effect is selected. The experimental results show that SVM and XGBoost perform well on practical validation, with 10 grand mal seizures, all of which are identified, and only two false alarms, which are not inferior to the detection based on EEG signals, and the wrist-based signals are more helpful for developing wearable devices, which are more suitable for epilepsy detection and identification.