Abstract
With the increasing rehabilitation needs of individuals with autism spectrum disorder (ASD) and the development of virtual reality (VR) systems in rehabilitation, VR training systems based on embodied interaction have received growing research attention. However, evaluating rehabilitation-related changes during long-term VR training and improving the use of multi-source feedback remain important challenges. This study aims to explore the longitudinal effects of a VR training and assessment system based on embodied interaction and multi-source interactive feedback for children with ASD. The system integrates brain function evaluation using functional near-infrared spectroscopy (fNIRS), behavioral pattern analysis using Kinect, VR task performance records, and subjective questionnaires. Guided by human-centered design principles, a one-month longitudinal study was conducted in a hospital rehabilitation setting with nine children with ASD. Participants completed 20 VR training sessions, and data were collected before the first training session, after the tenth session, and after the twentieth session. The results provide preliminary support for the feasibility of the system and show positive changes in several rehabilitation-related indicators, including task performance, subjective user experience indicators, selected brain function measures, and upper limb movement patterns. The multi-source feedback approach provides complementary objective information for interpreting rehabilitation-related changes. These findings offer references for applying VR technology in ASD rehabilitation and for designing longitudinal training cycles in future studies.