Face Classification under Different Kernel Function Compared to KNN Classifier

In this paper, we present a comparative study between Daubechies-DCT approach, Discrete Cosine Transform (DCT) and Histograms of Oriented Gradient (HOG) under different kind of kernel function. We obtain Daubechies-DCT by fusing the DCT features and Daubechies features. The implementation of HOG achieved by dividing the face image into small connected regions, named cells, and for each cell compiling a histogram of gradient directions. We use the fusion of DCT and Daubechies wavelets then, HOG method to process face classification focused on SVM (Support Vector Machine) and KNN (K Nearest Neighbors) classifiers. The fusion features are inputted into SVM and KNN classifiers. Results show that the HOG with SVM-Rbf kernel function achieves the highest performance in terms of the detection rate which we obtained 96.5%. We present experimental results applied on MIT face database to demonstrate the effectiveness of a novel comparative study in terms of accuracy and running time.