Performance Of Levenberg-Marquardt Algorithm In Face Recognition

Norjuliana , Abdul Hadi (2009) Performance Of Levenberg-Marquardt Algorithm In Face Recognition. Project Report. UTeM, Melaka,Malaysia. (Submitted)

	PDF (24 Pages) Performance_Of_Levenberg-Marquardt_Algorithm_In_Face_Recognition_Norjuliana_Binti_Abdul_Hadi_TA1650.N67_2009_-_24_Pages.pdf - Submitted Version Download (302kB)
	PDF (Full Text) Performance_Of_Levenberg-Marquardt_Algorithm_In_Face_Recognition_Norjuliana_Binti_Abdul_Hadi_TA1650.N67_2009.pdf - Submitted Version Restricted to Registered users only Download (2MB)

Abstract

This thesis investigates the performance of Levenberg-Marquardt Algorithm in face recognition. This project aim to analyze the performance of the algorithm in face recognition system. The Levenberg-Marquardt algorithm is an iterative technique that locates the minimum of function that is expressed as the sum of square of nonlinear functions. It has become a standard technique for nonlinear least-square problems and can be thought of as a combination of steepest descent and the Gauss-Newton method. This project has focused on using the neural network approach and image processing. The Matlab software has been chosen as programming software because it has an image processing toolbox robot vision and neural network toolbox. The Matlab’s m-files describe a neural network which will identify if an image contains a human face. For this project, 10 difference pictures (only face) from each 10 peoples will add to database. By using Matlab software, related programs will be saving in m-files. The neural network will train for learning and collecting data process. Then, neural network must be testing for recognizing and collecting data. To train the neural network, only 5 from 10 pictures for each people will be used while the other 5 pictures will be used to test the performance of the neural netwok. Time and performance for training and testing process must be recorded to analyze the data. With respect to database search system, Levenberg-Marquardt Algorithm is not so good in terms of training time. Minimization of the memory of image might be a good solution. Moreover the processing time is also influenced by the number of images trained. If the number of images more, so time taken for training process will be more than before. The objectives of this project have been achieved. For Levenberg-Marquardt algorithm, it can recognize 45 of 50 tested images, which mean 90% of the testing image can be recognized.

Actions (login required)

View Item