An FPGA-based Fall Detection System Using Millimeter-wave Radar and Convolutional Neural Network

Abstract

Falls have become the second leading cause of unintentional injury death in the elderly. Timely detection of falls in the elderly can avoid greater injuries. Therefore, it is increasingly important to study fall detection systems. Based on this background, a fall detection system is designed in this paper. A fall detection system is usually divided into two parts: data acquisition and feature recognition. Due to its excellent confidentiality and applicability, millimeter-wave radar is used for data acquisition in this paper. Usually, the data collected by millimeter-wave radar is processed in a computer, and feedback is given to fall or not. But the data processing software in the computer brings great convenience to the system at the same time, it also limits the stability and portability of the system and hinders the practical application of the system. For this, the system introduces FPGA (Field Programmable Gate Array) to replace the computer for all the data processing including short-time Fourier transform and CNN (Convolutional Neural Network) feature recognition. Experimental results show that the system can detect falls accurately in simple cases, which preliminarily proved the feasibility of FPGA application in the fall detection system. The FPGA-based fall detection system can realize miniaturization and low price through tape-out after further improvement, and promote the practical application of the fall detection system.