ARM Based Security System Using Linear Discriminant Analysis

This paper presents an ARM Based Security system using facial recognition Techniques. Also a comparison analysis between Principal Component Analysis (PCA), Independent Component Analysis (ICA) and Linear Discriminant Analysis (LDA) algorithms for facial recognition implemented in ARM Cortex M4 processor is done. A training database is created with images of all the authenticated users. The image of the user, whose identity is to be authorized, is captured using a webcam and the ARM microcontroller processes the algorithms to convert the images into vectors and components which are then compared with the images existing in the training database. Results are displayed in a LCD. To compare the performance of the algorithms ORL image database is used. The performance parameters used for comparison are Recognition rate and Cumulative Match score (CMS). The experimental results indicate that LDA algorithm outperforms in terms of recognition accuracy and CMS and gives best results under different illumination conditions, various expressions and poses. It can also be observed that the execution time decreases drastically when executed in ARM Cortex M4 microcontroller compared to execution in MATLAB.