Toward a bio-inspired rehabilitation aid: sEMG-CPG approach for online generation of jaw trajectories for a chewing robot. (May 2019)

• Record Type: Journal Article
• Authors: Kalani, Hadi; Moghimi, Sahar; Akbarzadeh, Alireza
• Is Part Of: Biomedical signal processing and control. Volume 51(2019)
• Page Start: 285
• Page End: 295
• Publication Date: 2019-05
• Subjects: Central pattern generator -- Chewing pattern -- Classification -- Rehabilitation robotics -- Surface electromyography; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2019.02.022
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: We combined two biological concepts CPG and sEMG to control a chewing robot. Chewing patterns were generated using several concepts, e.g. inverse kinematics, CPG, feature selection/classification, and sEMG based control. Our proposed masticatory robot was designed to deliver biomimetic exercises and maneuvers that resemble natural masticatory movements. sEMG-based control removed the need for hand maneuvers and instead enabled the operator to guide the training session through realistic masticatory motions. Abstract: The purpose of this study was to develop a bio-inspired masticatory robot that generates real-time trajectories, using surface electromyography signals (sEMG). We employed the central pattern generator (CPG) concept to generate smooth transitions from one chewing pattern to another during an exercise. Online changes in the recreated chewing patterns were provided based on the features extracted from the sEMG of the masticatory muscles of a tele-operator. The proposed method employed several concepts, including kinematics, sEMG feature extraction and selection, classification, and robotic control. First, chewing patterns were recognized by a multiclass support vector machine based on time-domain features extracted from sEMG signals. Next, CPG neurons generated a suitable trajectory for the robot actuators to reproduce the corresponding chewing pattern in the jaw (supposedly mounted on the moving platform of a 6RSS robot). The performance of the proposed approach was examined using a semi-real life chewing scenario. The average recognition rate for all the chewing classes, time windows, trials, and subjects was 86.36% ± 5.2%. Despite the sudden changes in the chewing patterns throughout the experiment, variations in actuator angles during transitions were smooth due to the limit cycle property of the CPG. The proposed method provided a solution for some inherent problems in generating a smooth and continuous trajectory in applications related to rehabilitation robots. This would make the proposed system and methodology feasible for a rehabilitation robot in real life exercise therapy.