Extracting age features and age detection from facial skin images (part 1)

Abstract

The project objective is to implement, experiment and analyse the performance of PCA using face skin image patches to classify young and old persons. For the project, 40 photographs of faces of old persons and 40 from young persons were taken and used as the database for this project. From these, 160 eye patches were extracted from the database (i.e. 2 eye patches from each photograph) to undergo PCA for dimension reduction and other necessary mathematical comparisons, to generate age detection results. The 160 eye patches were grouped into Training Set and Testing Set whereby each contains 80 randomly selected eye patches, 40 from the old category and 40 from the young category. The programming software, Matlab, is used to generate the codes for this project. PCA is performed to obtain the eigenvalues and eigenvectors from the Training Set for dimension reduction. The Nearest Neighbour Rule (NNR) is applied as the classifier for testing. Lastly, experiments were then performed and recognition results were obtained to determine the effectiveness of the given methodology. A breakdown on the results generated is analyzed and ways to improve age detection accuracy are proposed.