A Multiple Classifier System with Learning based Boosting for Clear and Occluded Color Face Identification
Dhananjoy Bhakta · American Journal Of Advanced Computing · 2014
Abstract — An automatic color face identification technique with improved identification accuracy has been designed where skin color feature are employed as an additional feature. In this system the most important facial features are extracted and face recognition is performed by individual extracted features. These decisions are combined then using learning based Boosting ensembles method. Each extracted features are used to train the system separately to learn the weights iteratively which finally boost the overall system accuracy. Experiments performed over a varied test dataset produce promising results in terms of accuracy, precision, recall and F-score for both clear as well as occluded color face images. The average accuracy of the overall system is 98.8 % for clear faces and 79.31 % for occluded faces. At the same time the performance evaluation time of our system is moderately low.