Abstract: The importance of face recognition system in our world today is enormous. Issues that the existing face recognition systems had to solve emanated from in consistence in facial patterns which do not conform to the traditional facial patterns. Existing face recognition models which employed the use of genetic algorithm for modeling CNN have the problem of slow convergence and local minimal entrapment. Firefly algorithm (FA) has been found to produce consistent and better performance in terms of time and optimality than other algorithm. Firefly (FA) is therefore applied for modeling CNN face recognition system. Three models were used, FFA-CNN, CNN1, and CNN2. Where FFA-CNN is a CNN model designed to obtain optimized parameter using FFA as the optimizer, CNN1 and CNN2 are CNN model designed by Random model parameters. For each model a total number of 694 sample facial images which is about (70%) of total dataset were used for training and 299 sample of facial images which is about (30%) of total dataset were used for testing the trained system. 3 experiments were carried out. The result shows that FFA-CNN is 100% accurate while CNN1 and CNN2 are 30.10% and 81.61% respectively. With this excellent result, FFA-CNN model develop in this research work can be recommended for use in face recognition system. This research has contributed immensely to knowledge by developing an algorithm that improved the performance of CNN model for face recognition.

Keywords: Convolutional Neural Network, Face Recognition System, Firefly Algorithm, Optimized Parameter.


PDF | DOI: 10.17148/IJARCCE.2022.11603

Open chat
Chat with IJARCCE