The Face Recognition System By Using The Radial Basis Function Neural Network (RBFNN)

Siti Dhamirah 'Izzati, Damni (2011) The Face Recognition System By Using The Radial Basis Function Neural Network (RBFNN). Project Report. UTeM, Melaka, Malaysia. (Submitted)

	PDF (24 pages) The_Face_Recognition_System_By_Using_The_Radial_Basis_Function_Neural_Network_(RBFNN)_-_24_pages.pdf - Submitted Version Download (450kB)
	PDF (Full Text) The_Face_Recognition_System_By_Using_The_Radial_Basis_Function_Neural_Network_(RBFNN)_-_Full_Text.pdf - Submitted Version Restricted to Registered users only Download (3MB)

Abstract

The demand in the ubiquitous applications of the face recognition system gives the inspiration to establish this project with the application of one of the artificial neural network, Radial Basis Function Neural Network (RBFNN). This report leads to the development of the face recognition system by using the Radial Basis Function Neural Network (RBFNN). It is also aims to analyze the characteristic for the training and testing of RBFNN fit with the system so that it can recognize the subject successfully. However, it limits to the MATLAB Graphical User Interface (GUI) as its interface and uses the grey scale input image with the format „jpeg‟ for the image only. Therefore the methodology applies the MATLAB GUI in order to develop the system and also the adaption with the match score of the test subject is very crucial to analyze the outputs from the system. This determines its performance whereby the experiments conducted revealed that the range of 25 to 30 is the best spread value for the system with the higher pixels also gives better performance. There also another experiment which shows the comparison of the genuine match and impostor match. The outputs verified that the higher score of genuine match and lower impostor match is the best combination so that the system will recognize enrollee consistently and precisely. Lastly, conclusion for this project is that its security, relevancy and cheaper that contributes its marketability value for industry. For the recommendation, it is suggested to apply it in multimodal biometric system, or using new advanced algorithm such as Scale-Invariant Feature Transform (SIFT). The upgrade of the system from verification to identification mode also should be included for the future research.

Actions (login required)

View Item