An Fpga Implementation Of Intelligent Visual Based Fall Detection

Peng Shen Ong, Yoong Choon Chang, Chee‐Pun Ooi, Ettikan Kandasamy Karuppiah, Shahirina Mohd Tahir

Falling has been one of the major concerns and threats to the independence of the elderly in their daily lives. With the worldwide significant growth of the aging population, it is essential to have a promising solution of fall detection which is able to operate at high accuracy in real-time and supports large scale implementation using multiple cameras. Field Programmable Gate Array (FPGA) is a highly promising tool to be used as a hardware accelerator in many emerging embedded vision based system. Thus, it is the main objective of this paper to present an FPGA-based solution of visual based fall detection to meet stringent real-time requirements with high accuracy. The hardware architecture of visual based fall detection which utilizes the pixel locality to reduce memory accesses is proposed. By exploiting the parallel and pipeline architecture of FPGA, our hardware implementation of visual based fall detection using FGPA is able to achieve a performance of 60fps for a series of video analytical functions at VGA resolutions (640x480). The results of this work show that FPGA has great potentials and impacts in enabling large scale vision system in the future healthcare industry due to its flexibility and scalability.