New Proposed Fusion between DCT for Feature Extraction and NSVC for Face Classification

Feature extraction is an interactive and iterative analysis process of a large dataset of raw data in order to extract meaningful knowledge. In this article, we present a strong descriptor based on the Discrete Cosine Transform (DCT), we show that the new DCT-based Neighboring Support Vector Classifier (DCT-NSVC) provides a better results compared to other algorithms for supervised classification. Experiments on our real dataset named BOSS, show that the accuracy of classification has reached 99%. The application of DCT-NSVC on MIT-CBCL dataset confirms the performance of the proposed approach.