Full-body pose reconstruction and correction in virtual reality for rehabilitation training

• DOI: 10.3389/fnins.2024.1388742
• Publication Date: 2024-04-04T04:49:18Z
• Authors: Xiaokun Dai; Zhen Zhang; Shuting Zhao; Xueli Liu; Xinrong Chen
• Keywords: Multilayer Perceptron (MLP); 0202 electrical engineering, electronic engineering, information engineering; rehabilitation training; virtual reality; deep learning; Neurosciences. Biological psychiatry. Neuropsychiatry; full-body pose reconstruction; 02 engineering and technology; RC321-571; Neuroscience

Existing statistical data indicates that an increasing number of people now require rehabilitation to restore compromised physical mobility. During the process, therapists evaluate and guide movements patients, aiding them in a more effective recovery preventing secondary injuries. However, immutability mobility expensive price training hinder some patients from timely access rehabilitation. Utilizing virtual reality for might offer potential alleviation these issues. prevalent pose reconstruction algorithms primarily rely on images, limiting their applicability reality. Furthermore, existing evaluation correction methods field focus providing clinical metrics doctors, failed efficient movement guidance. In this paper, reality-based method is proposed. The sparse motion signals devices, specifically head-mounted displays hand controllers, used reconstruct full body poses. Subsequently, reconstructed poses standard are fed into natural language processing model, which contrasts difference between two provides guidance form language. Quantitative qualitative results indicate proposed can accurately real-time. By referencing poses, model generates professional text. This approach facilitates training, reducing cost enhancing efficiency self-rehabilitation training.