The classification of EEG-based wink signals: A CWT-Transfer Learning pipeline
Jothi Letchumy Mahendra Kumar,
Mamunur Rashid,
Rabiu Muazu Musa,
Mohd Azraai Mohd Razman,
Norizam Sulaiman,
Rozita Jailani,
Anwar P.P. Abdul Majeed
Affiliations
Jothi Letchumy Mahendra Kumar
Innovative Manufacturing, Mechatronics and Sports Laboratory, Faculty of Manufacturing and Mechatronics Engineering Technology, Universiti Malaysia Pahang (UMP), 26600 Pekan, Pahang Darul Makmur, Malaysia
Mamunur Rashid
Faculty of Electrical and Electronics Engineering Technology, Universiti Malaysia Pahang, 26600 Pahang, Malaysia
Rabiu Muazu Musa
Centre for Fundamental and Liberal Education, Universiti Malaysia Terengganu (UMT), 21030 Kuala Nerus, Terengganu Darul Iman, Malaysia
Mohd Azraai Mohd Razman
Innovative Manufacturing, Mechatronics and Sports Laboratory, Faculty of Manufacturing and Mechatronics Engineering Technology, Universiti Malaysia Pahang (UMP), 26600 Pekan, Pahang Darul Makmur, Malaysia
Norizam Sulaiman
Faculty of Electrical and Electronics Engineering Technology, Universiti Malaysia Pahang, 26600 Pahang, Malaysia
Rozita Jailani
Faculty of Electrical Engineering, Universiti Teknologi MARA (UiTM), 40450 Shah Alam, Selangor Darul Ehsan, Malaysia
Anwar P.P. Abdul Majeed
Innovative Manufacturing, Mechatronics and Sports Laboratory, Faculty of Manufacturing and Mechatronics Engineering Technology, Universiti Malaysia Pahang (UMP), 26600 Pekan, Pahang Darul Makmur, Malaysia; Centre for Software Development & Integrated Computing, Universiti Malaysia Pahang, 26600, Malaysia; Corresponding author at: Innovative Manufacturing, Mechatronics and Sports Laboratory, Faculty of Manufacturing and Mechatronics Engineering Technology, Universiti Malaysia Pahang (UMP), 26600 Pekan, Pahang Darul Makmur, Malaysia.
Brain–Computer Interface technology plays a vital role in facilitating post-stroke patients’ ability to carry out their daily activities of living. The extraction of features and the classification of electroencephalogram (EEG) signals are pertinent parts in enabling such a system. This research investigates the efficacy of Transfer Learning models namely ResNet50 V2, ResNet101 V2, and ResNet152 V2 in extracting features from CWT converted wink-based EEG signals, prior to its classification via a fine-tuned Support Vector Machine (SVM) classifier. It was shown that ResNet152 V2-SVM pipeline could achieve an excellent accuracy on all train, test and validation datasets.
