Fast Online Decoding of Motor Tasks with Single sEMG Electrode in Lower Limb Amputees

Abstract

The quality of life of lower limb amputees strongly depends on the performance of their prosthesis. Active prostheses controlled by prosthesis sensors can participate to the movement and improve the walking performance of the amputees. However, a promising control mechanism involves the use of electromyography (EMG) to decode motor intentions. This approach could timely inform the prosthesis about the steps that the patient is going to perform much earlier compared to the feedback given by sensors. Here, we investigate whether an EMG-based algorithm is able to detect the motor intentions of transfemoral amputees. Subjects with a transfemoral amputation performed different motor tasks (e.g., ground level walking, climbing up/down stairs), while we recorded the EMG signals from surface electrodes placed on the subject’s stump. Our decoding algorithm achieved 100% motion intention discrimination. Such perfect decoding was achieved usually after less than 100 ms from the onset of the movement, thus ensuring that the information about the next step could be transmitted to the active prostheses with a sufficient advance to achieve its proper control. These results showed not only the feasibility of EMG-based online decoding of motor intentions, but also that perfect decoding can be achieved online with as little as one recording site, ensuring a minimum discomfort and encumbrance of the whole system.

Authors thank Scuola Superiore Sant’Anna, Pisa, and Centro Protesi INAIL, Budrio. This study was funded by MOTU (Protesi Robotica di Arto Inferiore con Smart Socket ed Interfaccia Bidirezionale per Amputati di Arto Inferiore), INAIL PPR-AI 1/2. A special thanks to Dr. Pericle Randi.