A Synergy-Coherence Integration Method for Real-Time Knee Contact Force Estimation under Multiple Gait Patterns

Accurate real-time estimation of medial knee contact force (KCF) is essential for knee osteoarthritis research and rehabilitation assessment. To overcome the limitations of existing methods such as numerous sensors requirements and insufficient robustness, this study proposes a novel real-time medial KCF estimation framework that integrates muscle synergies with intermuscular coherence, enabling high-accuracy estimation using five EMG sensors across five gait patterns. First, muscle synergy extraction is employed to select five key muscles for dimensionality reduction. Second, intermuscular coherence is computed to characterize neuromuscular coordination under different gait conditions. These complementary temporal and frequency domain features are subsequently integrated and processed through a dual-branch LSTM network for contact force prediction. Evaluation on a public dataset demonstrated strong performance (R2= 0.87 ± 0.07) across five gait conditions, exceeding baseline methods, while leave-one-gait-out testing further confirmed the method's superior generalization and robustness. Real-world validation with 12 participants, combined with EMG-informed musculoskeletal modeling, demonstrated that the real-time peak force estimation errors were no more than 0.18 BW (body-weight) during walking tests across five gait patterns. This study provides a practical and robust solution for real-time knee force monitoring with fewer sensors across multiple gait conditions, offering significant potential for applications in motor function assessment, personalized rehabilitation interventions, and integration with intelligent assistive devices.