A Fall Warning System for the Elderly with Multiple Nodes Based on DTW Algorithm

Abstract

Aiming at the problem of real-time fall warning for the elderly when walking, this paper improves the traditional dynamic time warping (DTW) algorithm for speech recognition, processes the gait data, and designs a multi node wear system with fall warning function. There are three nodes in the system, which are located in the waist and two ankles. The acceleration data of these three nodes are collected by the motion sensors of the elderly when they walk. Moreover, the system will use DTW algorithm to calculate the difference between the waist acceleration data of two gait cycles as the index of human stability, and calculate the difference between the left and right ankle acceleration data of one gait cycle as the index of human balance, and then use k-nearest neighbor (KNN) algorithm to realize the recognition function. When the elderly walk unsteadily, it will give an alarm, so as to help the elderly avoid walking unsteadily Fall. The experimental results show that the system can effectively identify the elderly gait instability which is the prelude to fall, and the accuracy rate is 95 %.