A contactless computer vision system for underwater walking and jogging gait analysis using YOLO-pose and Multi-CNN BiLSTM architecture

Buoyancy-assisted hydrotherapy exercise has been shown to reduce joint loading and accelerate functional recovery. However, conventional marker or sensor-based approaches are costly and impractical for underwater use due to water interference and setup constraints when monitoring recovery progress monitoring. To overcome these challenges, a computer vision-based gait analysis model was trained for jogging sessions in hydrotherapy pools. In this study, 2D coordinates extracted using You Only Look Once (YOLO) 11m-pose served as the model input without noise filtration to validate their robustness. A comparison of hyperparameter optimization algorithms was conducted, with the combination of multivariate tree-structured Parzen estimators (MultiTPE) and Hyperband identified as the optimal approach. Two convolutional bidirectional long short-term memory architectures, i.e., single vs. multiple convolutional layers (CNNs) per pooling were applied and compared in multi-head and single-head regression settings. Result indicated that multi-CNNs per pooling with multi-task learning best exploit inter-parameter correlations. On a 45-sample test set, the model achieved an intraclass correlation coefficient (ICC) with two-way random effects, absolute agreement, single rater model of 0.8999, Pearson’s correlation coefficient (PCC) of 0.9066, mean absolute error (MAE) of 0.0954 s for swing, stance, and stride time, while 3.5141 steps/min for cadence. The developed system thus achieves precise analysis for underwater leg movements.