Surface electromyography onset-based processing for control of a robotic arm

Upper-limb amputation can have a significant impact on a patient's quality of life, including limited independence. While myoelectric prostheses present the opportunity for more natural and intuitive control when compared to body-powered devices, their effectiveness can be limited by the challenges that come with surface EMG data collection. This study proposes a simplified EMG post-processing approach for converting muscle activation into robotic arm movements. Surface EMG data was collected sequentially from ten locations on the upper limb of a single subject that was instructed to perform four predefined motion sequences designed to cover common upper limb motions. The signals were processed via a custom pipeline that includes filtering and onset detection using NeuroKit2. The muscle activations detected were then mapped to commands for a Baxter robotic arm, which was used as a proxy for a prosthetic device. The resulting robotic motions were compared to the human subject's via the use of joint angle measurements and the time to complete each sequence. The results show that the proposed pipeline can reproduce the general structure and range of human motion, with the robot achieving joint movements similar to those of the participant. However, the robot suffered from a longer time required to complete each sequence of movements. This study shows that a relatively simple post-process EMG-based control scheme can capture multi-step human motion while remaining transparent. However, improvements in speed and real-time performance will be necessary for practical use in prosthetic devices.