Effect of Short-Term Artificial Intelligence-Based Real-Time Biofeedback Training on Squat Performance

Purpose:The purpose of this study was to evaluate the effects of artificial intelligence (AI)-based real-time biofeedback on squatting performance and lower limb kinematics.Methods: Thirty healthy adults were randomly assigned to either an experimental or a control group.After receiving verbal instructions on squat performance, both groups performed two sessions of 15 squat repetitions; the experimental group trained with AI-based real-time biofeedback using Real PT (Korea Body Information Co.), while the control group received no feedback.Participants were measured at four time points (pre-training, mid-training, post-training, and 3-day retention) using Movement Competency Screen (MCS) scores for the squat movement, as well as lower extremity moments and thoracic angle.Results: A significant grouptime interaction was found for MCS scores, with the AI-biofeedback group showing significant improvement over time, while the control group did not.No significant interactions were found for hip knee, or ankle joint moments and thoracic angle.Subgroup analysis indicated that participants who improved performance accuracy over training sessions showed significantly greater MCS score increase across all test sessions compared to baseline.Conclusion: Short-term AI-based real-time biofeedback squat training demonstrated significantly greater changes in movement competency over time, compared to no feedback, supporting its potential application in various sports training settings.Further studies involving in longer period intervention may be necessary to observe biomechanical changes.