Publication: Smart lock system integrating face recognition technology on FPGA
Loading...
Date
2023-08
Authors
Law, Xin Yi
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
FPGA-based smart lock systems are becoming increasingly popular in the market due to variation of technologies employed. The use of biometric techniques
such as face recognition in smart lock systems has become more widespread since it outperformed other technologies due to their high convenience and high security to the end users. Although there were many successful recognition algorithms developed in face recognition field, there is still room for improvement to solve the problem of demographic factors such as illuminations and variations of skin tone in face recognition. In this study, a face recognition system using LBPH algorithm with chi-square test is proposed to be developed and integrated on an FPGA-based smart lock system with password-based lock mechanism as backup authentication method. The face recognition system and the password-based lock mechanism are successfully developed on Altera DE1-SoC board. The performance evaluation involved testing the system on nine users with different skin tones under the same lighting conditions. The results showed a recognition accuracy of 96.67% for light skin tone, 86.67% for medium skin tone, and 83.33% for dark skin tone. This indicates that variations in skin tone can affect recognition accuracy, even when considering lighting conditions. Additionally, the estimated power consumption of the face recognition system ranged
from 1590.41mW to 1620.43mW. This low power consumption demonstrates the potential for significant power savings through configuration and optimization of the hardware design on the FPGA.