Kinematic Integration Network With Enhanced Temporal Intelligence and Quality-Driven Attention for Precise Joint Angle Prediction in Exoskeleton-Based Gait Analysis

Exoskeleton robots offer transformative potential in aiding the rehabilitation of patients with lower limb motor dysfunction, where precise and real-time prediction of knee joint angles is critical. Despite advances in deep learning for motion prediction, existing models struggle with balancing accuracy and real-time performance, particularly in wearable applications. This study introduces KINETIQA (Kinematic Integration Network with Enhanced Temporal Intelligence and Quality-Driven Attention), a novel predictive framework that advances the state of the art in knee angle prediction, demonstrating significant improvements in accuracy and clinical utility. The core innovation of KINETIQA lies in its Kinematic Integration Network (KIN), an architecture that seamlessly integrates advanced techniques, including Transformers, temporal convolutions, and multi-headed attention mechanisms. This integration enables KINETIQA to effectively capture complex temporal dynamics of knee movement across both short and extended time horizons. Specifically, KINETIQA achieves an MAE of 0.605 at the 20 ms horizon, improving over Transformer (0.705) by 14.2%, and maintains superior performance at 100 ms with an MAE of 1.174, representing a 19.8% reduction compared to Transformer’s 1.464 baseline. Beyond its predictive capabilities, KINETIQA employs rigorous feature selection and validation protocols to ensure clinical-grade precision, delivering actionable insights for personalized rehabilitation, injury prevention, and movement analysis. The model’s design improves prediction accuracy and enhances interpretability, facilitating practical adoption in clinical settings. Additionally, KINETIQA maintains real-time compatibility, achieving an inference latency of 22.469 ms in the 100 ms forecasting case. This performance remains well within the 50 ms threshold typically required for safe and responsive exoskeleton control. These findings underscore KINETIQA’s potential in robotic-assisted rehabilitation, offering a meaningful advancement toward reliable and effective patient support systems.