A Neural Network for Stance Phase detection in smart cane users

Abstract

Persons with disabilities often rely on assistive devices to carry on their Activities of Daily Living. Deploying sensors on these devices may provide continuous valuable knowledge on their state and condition. Canes are among the most frequently used assistive devices, regularly employed for ambulation by persons with pain on lower limbs and also for balance. Load on canes is reportedly a meaningful condition indicator. Ideally, it corresponds to the time cane users support weight on their lower limb (stance phase). However, in reality, this relationship is not straightforward. We present a Multilayer Perceptron to reliably predict the Stance Phase in cane users using a simple support detection module on commercial canes. The system has been successfully tested on five cane users in care facilities in Spain. It has been optimized to run on a low cost microcontroller.