Finger Movement Estimation From Surface EMG Using Mamba Architecture
Year: 2026
Type: conference-paper
Abstract: With advances in virtual reality (VR), human machine interfaces (HMIs), prosthesis and rehabilitation, and many other applications, the need for kinematics tracking has become more popular in recent years. To this day surface electromyography (sEMG) is still the most common method having the advantage of being non-invasive, yet alternatives that match this profile have been considered and studied such as, force myography (FMG) that measures changes in muscle volume to determine the activity of said tissue, or A-mode ultrasound (A-M US) that sends an ultrasound pulse which penetrates the skin, measuring the time that it takes for the echoes to return. This study explores the possibility of using Mamba architecture in order to reconstruct finger movement using data gathered from sEMG sensors and joint angles from a dedicated data glove provided by the authors of Ninapro DB8 dataset. The resulting model is a regressor capable of estimating fingers movement given a sequence of sEMG datapoints.
Institution: University of Craiova
Language: English
Subfield: Biomedical Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7166861435
