Akademik Veri Yönetim Sistemi
INTERNATIONAL JOURNAL OF INNOVATIVE COMPUTING INFORMATION AND CONTROL, cilt.9, sa.5, ss.1819-1834, 2013 (ESCI)
The main purpose of this paper is to propose a novel framework based on the separate classification of each partition of a face using the Common Vector Approach (CVA) and the combination of the classification results with the Majority Voting technique to classify test images. First of all, the thirteen different partitions (forehead, two eyebrows, two eyes, two cheeks, nose, mouth, chap and three spaces between some partitions) that effectively represent a face are extracted. When the partition vectors are applied to our proposed framework, higher recognition rates are obtained when compared with those obtained by the PCA, D-LDA and CVA methods for the image vectors obtained by stacking the partition vectors. Although the SVM classifier is superior to all other classifiers in terms of recognition performance, our proposed framework is an appropriate choice when the recognition results, processing time and memory requirement issues are all taken into account.