Analyses of Face Recognition Rate Using Holistic Methods

The paper focuses on the analysis of holistic facial recognition rates obtained by changing the parameters related to the number of components and the number of images for the same subject in the training set. It has been observed that, regardless of the method used, choosing a small number of samples for the training set does not lead to acceptable results. The recognition rate for using the PCA algorithm is directly influenced by the number of samples and by the size of the feature vector. Thus, if all the features were retained and the number of images in the training set was high, 95% rates were obtained. Similarly, in the case of the extraction of vectors of variable dimensions with DCT and the use of a neural network for classification, the obtained rates were not satisfactory for a small number of samples in the training set.