Development of a Modified Local Binary Pattern-Gabor Wavelet Transform Aging Invariant Face Recognition System

EasyChair Preprint 5001

Abstract

Most implementations of the currently existing facial recognition schemes lack adequate provision for age variations in individuals which make them less effective for practical usage. This is because most execution of such age-invariant facial identification systems are based on global feature extraction techniques which lack strong clarification ability and it is computationally inefficient.

This work presents an attempt to review various techniques adopted for face recognition which include Principal Component Analysis-(PCA), Histogram of Gradients-(HOG), Local Binary Pattern-(LBP), Gabor Wavelet Transform-(GWT), Particle Swarm Optimized Technique (PSO) and Support-Vector Machine (SVM) for particle extraction, particle selection and classification. In this paper, a hybrid local feature extraction technique was developed based on Local Binary-Pattern and Gabor-Wavelet Transform (LBP-GWT)) for a more efficient age-invariant feature extraction with strong discrimination ability.

The performance of the developed technique was compared with some state of the art feature extraction procedures including HOG which produced FAR of 21, FRR of 27, RA of 86.92% and RT of 124.533s. LBP yielded FAR of 18, FRR of 32, RA of 84.75% and RT of 101.221s. Furthermore, GWT produced FAR of 12, FRR of 26, RA of 88.41% and RT of 112.692s. PCA produced FAR of 22, FRR of 38, RA of 81.71% and RT of 151.421s. However, LBP-GWT yielded FAR of 6, FRR of 15, RA of 93.6 and RT of 81.667s.

The results obtained indicate that the developed hybrid LBP-GWT outperforms HOG, PCA, LBP and GWT as far as RA, RT, FAR and FRR.

Keyphrases: Aging-Invariance, Gabor Wavelet Transform, Local Binary Pattern, face recognition, system

@booklet{EasyChair:5001,
  author    = {Ayodele Oloyede and Temitayo Fagbola and Stephen Olabiyisi and Elijah Omidiora and John Oladosu},
  title     = {Development of a Modified Local Binary Pattern-Gabor Wavelet Transform Aging Invariant Face Recognition System},
  howpublished = {EasyChair Preprint 5001},
  year      = {EasyChair, 2021}}