Irregular Gait Detection using Wearable Sensors

Abstract

This paper presents a personalized system for detecting irregular gait parameters that may lead to a fall. Accurate detection of gait irregularities may be used to deliver targeted feedback for improving gait patterns and thereby reducing the risk of a fall. The proposed system uses Inertia Measurement Units(IMUs), proximity (PR) and infrared (IR) sensors. We separate the system into two distinct components. The first component is used to detect the current gait phase of the wearer based on the incoming sensor data. The second component combines the sensor data with the label produced by the first component used to classify the gait as regular or irregular in a manner that may potentially lead to a fall. The system can identify the occurrence of three distinct gait irregularities that may lead to a fall: small step width, low foot clearance and excessive trunk sway. For this, we use an Adaptive Neuro-Fuzzy Inference System (ANFIS). The system was trained on three healthy subjects to evaluate its ability to identify irregular gait. Results show that the system can provide real time results with an accuracy equal or greater to similar systems in the existing literature.