A linear regression model for estimation of isometric wrist joint torques with varying arm configurations using EMG signals

With an aging population, facilitation of an independent lifestyle for the elderly and those with neuromuscular conditions is a prominent field of research. Several robotic devices have been developed that aid patients and seniors regain their autonomy by augmenting forces exerted by the device wearer. A majority of these devices are controlled using torque estimation models based on electromyography (EMG) signals. However, the accuracy of these models deteriorates with alteration of limb configurations that render them unreliable for practical applications. This paper proposes and demonstrated the viability of a model to overcome this problem. We used experimental EMG and torque data gathered from eleven volunteers at various arm configurations to showcase estimation accuracy preservation under varying configurations achievable using the proposed model.