Jiachen Wang

Deep learning has been widely used in wearable sensors to improve accuracy in gait analysis. However, these deep learning models typically focus on single tasks, either in gait parameter estimation or gait phase detection. This study presents a novel multi-task learning framework for regression (i.e., gait cycle percentage prediction) and classification (i.e., gait phase prediction) tasks in pathological gait analysis using wearable sensors. Our framework employs a Multi-gate Mixture-of-Experts (MMoE) architecture to achieve soft parameter-sharing, integrating expert networks, cross-expert attention mechanisms, and dynamic routing to balance shared and task-specific representations. To reduce computational burden in wearable applications, we compare lightweight model configurations that optimize expert count and feature dimensionality. Model performance has been validated on a public dataset consisting of 158 frail older adults, demonstrating that our framework significantly outperforms single-task learning and hard parameter-sharing baselines, achieving an accuracy of 97.56% and a Mean Absolute Error (MAE) of 0.0397. Notably, the most compact lightweight configuration reduces the parameter count by nearly 98% (from 2.118 million to 0.0469 million), achieving an accuracy of 96.47% and a MAE of 0.0549. Attention mechanisms significantly enhance performance across all configurations, with improvements ranging from 17.9% to 30.4%. These findings validate the potential of lightweight multi-task approaches for real-time gait assessment, offering promising applications for clinical evaluation and rehabilitation monitoring in geriatric populations.

Wearable sensors are widely used to assess spatiotemporal gait parameters and their variability, which are critical for fall risk prediction. However, the impact of gait analysis accuracy and precision on fall risk prediction remains unexplored. This study collected gait data from 95 older adults using instrumented footwear on an instrumented walkway which is recognized as a system with gold standards during the 6-minute walking test. Participants were classified into fallers and non-fallers based on retrospective fall history (falls in the 6 months prior to completing the experiment), prospective fall occurrence (falls in the subsequent 6 months after completing the experiment), and a combination of both. Gait parameters and their variability were estimated using three algorithms: the conventional foot displacement method and two support vector regression (SVR) techniques. These features were used to develop fall risk prediction models with four machine learning classifiers: logistic regression, decision tree, support vector machine, and artificial neural network. Our findings demonstrate that the accuracy and precision of gait analysis algorithms significantly influences the estimation of gait parameters and their variability, directly impacting fall risk prediction performance. Using a support vector classifier, the area under the receiver operating characteristic curve (AUC) values for predicting retrospective falls, prospective falls, and either fall type increased from 0.79, 0.84, and 0.77 (conventional method) to 0.85, 0.89, and 0.83 (SVR). These findings show the importance of refining gait analysis accuracy and precision in future studies that aim to use wearable sensors for fall risk assessment in older adults.