The Rehabilitation Exoskeletons Based on Intention Recognition

Abstract

With the global aging population and the increasing number of stroke and gout patients, the number of individuals with mobility impairments is rising. Rehabilitation exoskeleton, as an emerging human-machine fusion assistive technology, has been widely used in recovery training for patients with neurological injuries. Motion intention recognition is the core component for achieving human-machine interaction, with its accuracy directly determining the safety and effectiveness of rehabilitation training. Driven by advancements in artificial intelligence and multimodal sensing technologies, single-modality motion intention recognition is evolving toward multimodal fusion. This paper systematically reviews the development of intent recognition for rehabilitation exoskeletons, highlighting the limitations of single-modality approaches such as electromyography and electroencephalography. Consequently, researchers have shifted focus to multimodal intent recognition. The current state of multimodal intent recognition is presented, encompassing sensory modalities, feature extraction and fusion methods, and representative application cases. It also summarizes the current difficulties and challenges in development, along with potential future trends, aiming to outline developmental trajectories and provide reference and insights for subsequent research.