A Neuro-Fuzzy Based Sensing Approach for the Classification of Emulated Postural Instability

Assistive Technology helps to assess the daily living of frail people and may have a strategic role detect prevent falls. In this paper, task classifying different classes postural sway behaviors has been addressed by developing Neuro-Fuzzy inference approach that is robust against noise. The proposed classifies four namely Stable Standing, Antero-Posterior, Medio-Lateral Unstable. strategy exploits data generated wearable sensor node, be positioned on user chest. A dedicated experimental set-up realized emulate dynamics generate dataset. Two novel indices robustness system proposed. first index measure residuals between predicted expected status, which equally weights estimations with respect classes. second metric reliability index, allows for assessing degree trust each estimation performed inference. Results obtained demonstrate suitability methodology, showing capability almost 100% correctly classify patterns among allowed classes, indexes 97.56% 98.50% training test patterns, respectively. Also, classification algorithm noisy demonstrated.