Feature extraction based on joint variational modal decomposition and continuous time wavelet transform for TENS therapy via EEG classification

• Published in: 2024 IEEE Consumer Life Tech (ICLT) pp. 1 - 4
• Main Authors: Zhang, Jinpei, Wang, Meilin, Ling, Bingo Wing-Kuen
• Format: Conference Proceeding
• Language: English
• Published: IEEE 11.12.2024
• Subjects: Accuracy; Brain modeling; classification of EEGs; Continuous wavelet transforms; Diseases; Electroencephalography; Feature extraction; feature extraction based on variational modal decomposition and continuous wavelet transform; insert Transcutaneous electrical nerve stimulation therapy; Medical treatment; Numerical simulation; Random forests; Testing
• DOI: 10.1109/ICLT63507.2024.11038701

The cognitive ability is closely related to the brain disease and the brain disease is a current health problem with the great concern. The studies have shown that the transcutaneous electrical nerve stimulation (TENS) can indirectly modulate the brain activity by stimulating the superficial nerves, so the TENS based EEG studies are of great importance. The purpose of this paper is to investigate whether TENS therapy can improve subjects' cognitive performance in mathematical games. In order to effectively differentiate EEG before and after TENS therapy, this paper proposes a feature extraction method based on variational modal decomposition (VMD) and continuous time wavelet transform (CWT) to improve the accuracy of EEG classification. First, the EEG is denoised. Then, VMD was used to decompose the signal into multiple intrinsic modal functions (IMFs). Next, CWT decomposition was applied to some IMFs and the frequency component with the highest energy was selected for feature extraction. Finally, the EEG before and after TENS treatment were classified using the random forest model. By comparing with the existing methods, it is found that the method proposed in this paper can effectively improve the accuracy of EEG classification, and the AUC values of ROC curves are also improved to different degrees. Among them, the highest accuracy of classification can be improved by 33.62% over the original, and the highest AUC can reach 0.97.