Face Recognition Based on Intensified Firefly Algorithm and Extreme Learning Machine

Recognition accuracy is still remaining an issue in face recognition in which it can be affected by factors such as illuminations and dissimilarities in appearances. The paper presents a new approach for the recognition of facial images, using Intensified Firefly Algorithm (IFA) and extreme machine classifier. This approach employs multi linear sparse principle component analysis, which receives sparsity from Sparse Principle Component Analysis and gains a series of sparse projections which can be efficiently computed by sequence of procedures using sparse regression method. Firefly algorithm (FA) works based on shining arrangements and performance of fireflies and uses to choose the input and output weights and biases of hidden layer of Extreme Learning Machine (ELM) which provides integrated solutions to generalized feed forward networks. The proposed technique has been applied to YALE, JAFFE and ORL databases and the experimental results reveal excellent recognition accuracy 99.7%. Extensive experiments with different number of training images produces less complexity of 350 ms for different number of training images. Performance and error comparison with other algorithms for various iterations illustrates the effectiveness of the proposed framework.