1 Introduction

Fall detection uses three types of sensors: wearable, ambient and visual cameras ^[1-4]. In the last decade, many researchers have proposed some methods for fall detection ^[5-9]. Among wearable sensors, the accelerometer is the most effective sensor for fall detection. Computer vision uses image processing to analyze human body postures and realize fall detection. This method is non-invasive, does not need to wear equipment, and is convenient for the elderly. Video detection includes image capture, storage in the loop buffer and feature extraction. The visual detection provides the most extensive application prospect for fall detection.