Modeling self-Principal Component Analysis for age invariant face recognition

Abstract

Face recognition is a biometric approach which can extract the facial characteristics of a person without his/her cooperation. The face recognition system fails to identify a person after some years because of the age related variations shown on the face. The face recognition system should be able to handle such age related variations and recognise the person irrespective of his age. The aging variations on a face are seen in the form of wrinkles, shape changes etc. The process of aging is highly composite and its effects on face are unpredictable. The aging effects are unique to every individual's face. In this paper we try to model the age variations of face using a self -Principal Component Analysis (PCA) based method. We have compared the recognition rate with the conventional PCA and an improved recognition rate is obtained using our method.

Face Gender Classification using Combination of LPQ-Self PCA

“…Meanwhile, recognizing the impact of age on gender classification performance, there is a suggestion to enhance the classification process by employing the Self-Principal Component Analysis (Self-PCA) method [12]. Self-PCA is a development of PCA that involves creating specific eigenfaces for each class to be identified, resulting in more accurate eigenfaces for each class [13]. The research demonstrated that the application of Self-PCA to facial images for gender classification, especially when considering the age factor, gives better performance than traditional PCA, using Euclidean distance measurement as the classifier.…”

Analysis on Age Invariance Face Recognition Study and Effects of Intrinsic and Extrinsic Factors on Skin Ageing

“…Locale around eyes is utilized as input feature highlight rather than whole face as it is progressively steady piece of face. J. S. Nayak and Nagarathna N. et al [7] proposed this self-PCA based face acknowledgment technique to think about maturing impacts by building subspace at individual dimension. Z. Li and U.…”

“…By arbitrary inspecting preparation set just as element space, different LDA-based classifiers are built and after that joined to create a strong choice by means of a combination rule[7].3.2.1 Densely sampled local feature descriptionThe entire face picture is partitioned into a lot of covering patches and after that chose neighborhood picture descriptors is connected to each fix. Removed highlights from these patches are connected together to frame a component vector with huge dimensionality for further investigation.…”

“…Filter highlight portrayal comprises of two fundamental parts: key point extraction, and highlight descriptors. Thickly test SIFT include descriptors from whole facial picture rather than just at a generally modest number of extricated key focuses[7].…”

“…In this way, it is relied upon to acquire discriminative properties of these neighborhood depiction plans, and moreover have ability in extricating age invariant highlights, for example, dispersion of edge bearing in face. (ii) MFDA has ability to adequately consolidate rich data passed on by thickly tested SIFT and MLBP descriptors, which are integral to some degree[7].…”