A Deep Learning-Based Gyroscopic Sensors for Real-Time Gait Deviation Detection

This study introduces a cutting-edge real-time gait deviation detection system that integrates biomechanics sensory acquisition through gyroscopic sensors and advanced deep learning via the combination of MobileNetV2 and a bi-directional long-term memory model. The system evaluates angular misalignments in critical joints (hip, knee, ankle) along with speed and gait pattern analysis, offering a comprehensive dual-approach solution. In clinical trials involving 30 patients, the system achieved a 12% accuracy in detecting hemiplegic, diplegic, and standard gait patterns, highlighting challenges in model optimization. While the system's performance currently lags behind traditional diagnostic methods, it shows promise for further refinement. The system offers a low-cost, scalable, and potentially highly accurate solution that addresses critical gaps in primary healthcare, particularly in resource-limited environments. Integrating biomechanics data with deep learning represents a significant advancement in gait analysis, and continued development enhances diagnostic precision and enables continuous monitoring of rehabilitation progress. Implementing a bi-directional LSTM and MobileNetV2 into gyroscopic sensors revolutionizes the detection and treatment of gait disorders, offering significant benefits for clinical practice and research. The findings underscore the importance of individualized, joint-specific interventions for effective gait rehabilitation and highlight the system's role in advancing gait diagnostics.