Human upper limb motions recognition for stroke rehabilitation with smartphone sensors

EasyChair Preprint 920

Abstract

In order to improve the effective of rehabilitation training for stoke elders and provide accurate rehabilitation guidance for therapist, we construct a neural network based on Multi-Layer Perceptron(MLP) to recognize five upper limb motions basing on mobile phone sensors. Five rehabilitation movements of upper limb are chosen to be recognized include hand horizontal, hand turn left and right, hand scroll down, elbow flexion. In this experiment, the raw data are collected from the three-dimensional data of accelerometer of smartphone. After preprocessing and feature extraction of the data, the neural network can identify the five motions and some combined motions. Through the experiment, this system has a nicely performance with 98.2% accuracy.

Keyphrases: Rehabilitation, Smartphone, Upper Limb Motion Recognition, neural network

@booklet{EasyChair:920,
  author    = {Chen Xiaochen and Han Ping and Xu Wenchao and Yang Yanqin and Zhu Xuanming},
  title     = {Human upper limb motions recognition for stroke rehabilitation with smartphone sensors },
  howpublished = {EasyChair Preprint 920},
  year      = {EasyChair, 2019}}