Machine Learning in Biomechanics: Enhancing Human Movement Analysis

Biomechanical analysis of human movements is highly relevant for discovering strategies to prevent injury, treat disease, and enhance performance. In this context, high-dimensional datasets are typically collected using either laboratory-based biomechanical measurement systems or wearable sensors. In recent years, Machine Learning (ML) has become increasingly popular for exploiting the potential of high-dimensional biomechanical data. There are three major ML paradigms: supervised learning, unsupervised learning, and reinforcement learning, with the first two used primarily in biomechanics. In supervised learning, ML models are trained, for example, to classify knee injury status based on muscle activation patterns or to predict knee joint forces using wearable sensor data through regression algorithms. Unsupervised learning in biomechanical applications involves, for example, reducing high-dimensional kinematic data into compact low-dimensional representations or identifying characteristic groups of people, such as individuals with similar gait abnormalities. Reinforcement learning presents, for example, a promising approach to developing controllers for biomechanical models capable of generating physiologically feasible high-dimensional movements. ML-based analysis complements traditional biomechanical analysis well, as both have their own strengths and weaknesses. Overall, ML can support our understanding of human movement biomechanics and optimize movement patterns to prevent injuries and enhance human health and performance.