Intent-Synchronized Continuous Forearm Prosthesis Control using Skin Deformation and IMU Sensors

Abstract

Conventional surface electromyography (sEMG)based prostheses often rely on discrete pattern classification, which forces users into unnatural, disjointed, and sequential joint actuation. To address this limitation, this study proposes a novel, intuitive control system for forearm prostheses that enables simultaneous, continuous multi-degree-of-freedom (DoF) articulation. Our approach fuses data from two distinct modalities: a custom near-infrared skin deformation sensor that captures muscle morphology, and an inertial measurement unit (IMU) that measures limb orientation to compensate for postural artifacts. By feeding these physical cross-sectional deformation features into an Extra-Trees Regress (ETR) model, the system accurately estimates continuous 2D wrist and finger joint angles. Offline evaluations with able-bodied subjects yielded high estimation accuracy, particularly regarding wrist kinematics. Furthermore, we implemented the system on a prototype robotic hand and evaluated it through a clinical cylinder manipulation task requiring precise, simultaneous joint adjustments. Compared to conventional sEMG-based methods, the proposed system significantly reduced task completion times and substantially improved intuitive controllability by eliminating sequential control delays. While a performance gap remains compared to intact hands-primarily due to hardware constraints such as the absence of tactile feedback-this multimodal sensing framework establishes a robust and practical foundation for next-generation, intent-driven prosthetic control.