A Two-Stream Approach to Fall Detection With MobileVGG

Qing Han,Haoyu Zhao,Ruikang Liu,Ke Zuo,Xiang Zhou,Hao Cui,Weidong Min

doi:10.1109/access.2019.2962778

Abstract

The existing deep learning methods for human fall detection have difficulties to distinguish falls from similar daily activities such as lying down because of not using the 3D network. Meanwhile, they are not suitable for mobile devices because they are heavyweight methods and consume a large number of memories. In order to alleviate these problems, a two-stream approach to fall detection with the MobileVGG is proposed in this paper. One stream is based on the motion characteristics of the human body for detection of falls, while the other is an improved lightweight VGG network, named the MobileVGG, put forward in the paper. The MobileVGG is constructed as a lightweight network model through replacing the traditional convolution with a simplified and efficient combination of point convolution, depth convolution and point convolution. The residual connection between layers is designed to overcome the gradient disappeared and the obstruction of gradient reflux in the deep model. The experimental results show that the proposed two-stream lightweight fall classification model outperforms the existing methods in distinguishing falls from similar daily activities such as lying and reducing the occupied memory. Therefore, it is suitable for mobile devices.

Highlights

A fall refers to a person’s sudden, involuntary, unintended position change, such as falling onto the ground or onto a lower place
The main contributions of this paper are as follows: a) A two-stream approach to fall detection with the MobileVGG is proposed. It outperforms the existing methods in distinguishing falls from similar daily activities such as lying and reducing the occupied memory. It is suitable for mobile devices. b) The MobileVGG is put forward to accelerate the training process
In this paper, a two-stream approach to fall detection with the MobileVGG is proposed. It first uses a stream based on motion characteristics to extract the video frames of possible fall behavior, and can effectively distinguish the fall from similar daily activities such as lying down

Summary

Introduction

A fall refers to a person’s sudden, involuntary, unintended position change, such as falling onto the ground or onto a lower place. Falling is the number one killer of the accidental injury of the aged [1]–[3]. With the development of the aging society, an increasing number of old people live alone and are not taken care of. Falls become a severe issue in the care of the elderly. As many old people have chronic diseases, even minor falls may threaten their health and life. Some old people luckily survived from falls but rely on others’ care in daily life and need medical aids for walking. The development of automatic fall detection has become an urgent need for protecting vulnerable people, especially the old. It has become a hot research topic

Objectives

Methods

Findings

Conclusion