Optimization of weights in ELM for face recognition

Since last many decades, identification of the human beings on the basis of their face images is getting attraction in the domain of recognizing any pattern. In the proposed approach, a novel algorithm is utilized for the recognition of face images. Extreme learning machine (ELM), which is a non-iterative algorithm, is taken into consideration by computing their input weights with the help of genetic algorithm (GA). PCA is utilized here as the dimension reduction tool. Various face databases like ORL, Georgia Tech and YALE are tested over the proposed approach and it can be observed from the percentage error rates for all the databases that improvement through this proposed algorithm is achieved. The results are also compared with the standard extreme learning machine algorithm to show the improvement by the proposed approach compared to the standard ELM.

Automated summary

This paper presents a new algorithm for face image recognition using extreme learning machine, genetic algorithm, and PCA, achieving improvement in recognition accuracy across various face databases compared to standard methods.