An Efficient and Robust Gender Classification System

P. Rai,P. Khanna
DOI: https://doi.org/10.1109/CICN.2015.58
2015-12-01
Abstract:This manuscript present a gender classification technique uses the (2D)2PCA, Gabor filter, and SVM for classifying the gender from occluded and non occluded face images. Present work explores two approaches, i.e., Fusion at feature level, and fusion at classifier level. Experimental result shows that, both the proposed approaches give an acceptable result on non-occluded face image database. For occlude face images the approaches render accuracies in range of 90% to 86% for various occlusion conditions. Proposed schemes for gender classification are evaluated using a number of well-known face image databases including LFW, FERET, AR, FEI A and FEI B. The analysis also reveals that lower part of face contain more useful information as compared to eyes for gender classification. The experimental analysis also show the effectiveness of proposed method against various variation in face images.
What problem does this paper attempt to address?