Human activity recognition based on wrist PPG via the ensemble method
Omair Rashed Abdulwareth Almanifi,
Ismail Mohd Khairuddin,
Mohd Azraai Mohd Razman,
Rabiu Muazu Musa,
Anwar P.P. Abdul Majeed
Affiliations
Omair Rashed Abdulwareth Almanifi
Innovative Manufacturing, Mechatronics and Sports Laboratory, Faculty of Manufacturing and Mechatronic Engineering Technology, Universiti Malaysia Pahang, Pahang Darul Makmur, Malaysia
Ismail Mohd Khairuddin
Innovative Manufacturing, Mechatronics and Sports Laboratory, Faculty of Manufacturing and Mechatronic Engineering Technology, Universiti Malaysia Pahang, Pahang Darul Makmur, Malaysia
Mohd Azraai Mohd Razman
Innovative Manufacturing, Mechatronics and Sports Laboratory, Faculty of Manufacturing and Mechatronic Engineering Technology, Universiti Malaysia Pahang, Pahang Darul Makmur, Malaysia
Rabiu Muazu Musa
Centre for Fundamental and Continuing Education, Department of Credited Co-curriculum, Universiti Malaysia Terengganu, Terengganu, Malaysia
Anwar P.P. Abdul Majeed
Innovative Manufacturing, Mechatronics and Sports Laboratory, Faculty of Manufacturing and Mechatronic Engineering Technology, Universiti Malaysia Pahang, Pahang Darul Makmur, Malaysia; School of Robotics, XJTLU Entrepreneur College (Taicang) Xi’an Jiaotong-Liverpool University, Suzhou. 215123, P. R. China; Centre for Software Development and Integrated Computing, Universiti Malaysia Pahang, Gambang, Pahang Darul Makmur, Malaysia; Faculty of Engineering, Technology and Built Environment, UCSI University, Kuala Lumpur Campus, Kuala Lumpur, Malaysia; EUREKA Robotics Centre, Cardiff School of Technologies, Cardiff Metropolitan University, Cardiff, UK; Corresponding author at: School of Robotics, XJTLU Entrepreneur College (Taicang) Xi’an Jiaotong-Liverpool University, Suzhou. 215123, P. R. China.
Human activity recognition via Electrocardiography (ECG) and Photoplethysmography (PPG) is extensively researched. While ECG requires less filtering and is less prone to disturbance and artifacts, nonetheless, PPG is cheaper and widely available in smart devices, making it a desired alternative. In this study, we explore the employment of the ensemble method with several pre-trained machine learning models namely Resnet50V2, MobileNetV2, and Xception for the classification of wrist PPG data of human activity, in comparison to its ECG counterpart. The study produced promising results with a test classification accuracy of 88.91% and 94.28% for PPG and ECG, respectively.
