Evaluation of Human Balance for Detection of Fall-risk using Non-linear Postural Dynamics and Ensemble Model

Clinical assessment of human balance relies on standardized tests, but their subjectivity leads to low specificity and sensitivity in predicting fall-risk. Biomechanical measures show limited ability to determine fall severity due to complex balance characteristics. This research used Recurrence Quantification Analysis (RQA) to study non-linear postural dynamics with COP trajectories from static posturography. Using an open-access dataset of 163 individuals (116F, 47M), RQA measures were extracted. The proposed two-level ensemble model achieved 80% accuracy in detecting fall severity (fit time: 54.02s), 85% in fall-risk evaluation, and 91.05% in assessing age-related balance characteristics. Anterior-posterior COP fluctuations were more pronounced than medio-lateral, and higher RQA values indicated greater postural stability. Using non-linear dynamics and ensemble modeling fall-risk can be detected with high accuracy.