Modeling biological individuality using machine learning: A study on human gait

• DOI: 10.1016/j.csbj.2023.06.009
• Publication Date: 2023-06-16T01:06:29Z
• Authors: Fabian Horst; Djordje Slijepcevic; Marvin Simak; Brian Horsak; Wolfgang Immanuel...; Matthias Zeppelzauer
• Keywords: ddc:796; ddc:610; Human gait recognition; 796 Sport; 610 Medizin; 610; 796 Athletic and outdoor sports and games; 796; Ground reaction forces; 610 Medical sciences; Biomechanics; Layer-wise relevance propagation; Explainable artificial intelligence; Force-based gait recognition; TP248.13-248.65; Biotechnology; Research Article

Human gait is a complex and unique biological process that can offer valuable insights into an individual's health well-being. In this work, we leverage machine learning-based approach to model individual signatures identify factors contributing inter-individual variability in patterns. We provide comprehensive analysis of individuality by (1) demonstrating the uniqueness large-scale dataset (2) highlighting characteristics are most distinctive each individual. utilized data from three publicly available datasets comprising 5368 bilateral ground reaction force recordings during level overground walking 671 distinct healthy individuals. Our results show individuals be identified with prediction accuracy 99.3% using signals all components, only 10 out 1342 our test being misclassified. This indicates combination components provides more accurate representation signature. The highest was achieved (linear) Support Vector Machines (99.3%), followed Random Forests (98.7%), Convolutional Neural Networks (95.8%), Decision Trees (82.8%). proposed powerful tool better understand has potential applications personalized healthcare, clinical diagnosis, therapeutic interventions.