The document summarizes research using a transformer model to classify biomedical signals, specifically EEG signals. Key points:
- A transformer model was used to classify EEG signals from epilepsy patients using statistical features extracted from wavelet decompositions of the signals.
- The model was tested on the Bonn and CHB-MIT EEG datasets, achieving accuracy above 97% on binary, tertiary and 5-class problems from the Bonn data and above 95% accuracy for individual patients in the CHB-MIT data.
- The transformer model outperformed previous methods like MLP, SVM, ANN and CNN in classifying the EEG signals, demonstrating its effectiveness for biomedical signal classification tasks.