Optimizing usability of electric wheelchairs with voice user experience for acceleration wheel rotation design by the kinematics method

Abstract

Individuals with quadriplegia experience total paralysis of all four limbs due to spinal cord injuries, leaving them unable to operate conventional electric wheelchairs that rely on joystick control. Existing alternative interfaces, such as head motion and eye-gaze sensors, are often cost-prohibitive and fail to deliver the maneuverability and accuracy required for daily use. Voice recognition emerges as a practical solution because speech ability is typically retained in quadriplegia, offering a hands-free, intuitive control method. This study proposes an electric wheelchair system integrating voice user experience (VUX), machine learning (ML), and kinematics-based wheel rotation control to address these challenges. Voice commands are processed using natural language processing (NLP) for word recognition and support vector machines (SVM) for amplitude classification to dynamically adjust speed and direction. Forward and inverse kinematics optimize wheel rotation angles, ensuring smooth and precise navigation even in constrained spaces. Experimental results demonstrate 92.82% word recognition accuracy and 94.48% accuracy in frequency and amplitude detection. Functional testing recorded average speeds of 0.343 m/s (no load) and 0.305 m/s (with 60 kg load). Usability testing with 15 quadriplegic users reported 93%.