Amrutha Sethuram

Aspects of Age Variation in Facial Morphology Affecting Biometrics

There is a great degree of variation that occurs in the face during aging of adults that will affect outcomes for face- based biometric systems. These changes have been studied sporadically in anthropological research and have received even less attention in biometric-related literature. This paper presents a summary of recent research that spans biometric, forensic, and anthropologic literature in an attempt to unify a variety of findings related to adult aging, discusses work in synthesizing images of aged faces and its relation to the anthropological findings, and also presents work demonstrating a need for further study in this area to improve face-based biometric techniques. Only age-invariant or aging-aware methods will have the best success in longer stretches of time.

Improvements and Performance Evaluation Concerning Synthetic Age Progression and Face Recognition Affected by Adult Aging

Aging of the face degrades the performance of face recognition algorithms. This paper presents recent work in synthetic age progression as well as performance comparisons for modern face recognition systems. Two top-performing, commercial systems along with a traditional PCA-based face recognizer are compared. It is shown that the commercial systems perform better than the baseline PCA algorithm, but their performance still deteriorates on an aged data-set. It is also shown that the use of our aging model improves the rank-one accuracy in these systems.

Improvements in Active Appearance Model based synthetic age progression for adult aging

Normal adult aging in the face can drastically affect performance of face recognition systems. Synthetically generating age-progressed or age-regressed images for aiding recognizers is one method of improving the robustness of face-based biometrics. These synthetic age progressions may also aid human law enforcement and other applications. There has been wide interest in these techniques in recent years, and the use of Active Appearance Models (AAMs) for synthetic age progression has been shown to be a promising approach but has not yet been demonstrated on a large human population with wide variation. This paper presents improvements in AAM-based age progression that generate significantly improved visual results, taking into account a much wider gender, age, and ethnic range than published to date for age progression techniques.

A hierarchical approach to facial aging

Active Appearance Models (AAMs) have been used as a promising tool in the field of synthetic age progression. However, they are yet to be demonstrated on a large human population with wide variation. This paper presents a novel AAM-based hierarchical approach to facial aging. This work is motivated from studies in medical and anthropological literature on classification of human faces based on gender, ethnic and age groups. The proposed hierarchical model approach is a ethnicity and gender specific aging paradigm. Specifically, the Caucasian (European descent) and African American ethnic groups are considered. This work will further show that using individual hierarchical models generate better age-progressed synthetic images when compared to a general model approach. The results are evaluated by visual perception of the intended age group and preservation of identity. Also, a quantitative evaluation was performed using FaceVACS, a commercial face recognition system, as a surrogate measure. Higher match scores for synthetic images generated by hierarchical models when compared to those generated by a general model suggests the efficiency of the proposed hierarchical model approach.

Establishing a test set and initial comparisons for quantitatively evaluating synthetic age progression for adult aging

Generating accurately age-progressed images has application in improving the robustness of face-based biometrics, aiding human law enforcement, and serving other applications. There has been a growing interest in this topic, but age-progression techniques and particularly means to evaluate their effectiveness quantitatively are still underdeveloped. This paper presents work in establishing a standard dataset over which techniques may be evaluated as well as some initial evaluation work. Firstly, a photographic test set that has several representative images of the same individuals across many years has been assembled in an effort to provide a comparison standard. By using photos of an individual at a younger age when the matching photos of that person at an older age are available, age-progression methods may be quantitatively compared. Secondly, several initial variations of age-progression methods are compared over this dataset.

Implications of Adult Facial Aging on Biometrics

1.1 Statement of the problem Single features of the human face, facial components, as well as the human face taken as a whole may be viewed as a biometric tool for purposes of individual human identification. Using computer technology, automated face recognition (FR) systems have been created to match individual faces from print and digital photographs and video to faces of the same individual whose image is stored among many others in a computer database. One of the primary problems that arise with FR systems is how to contend with the passage of time and the resultant effects of facial aging. Enrolled, i.e. gallery, images become difficult to match against a query, i.e. probe, image as the time span between the gallery and the probe increases. Several researchers have highlighted this temporal performance degradation (Pentland & Choudhury, 2000; Phillips et al., 2000; Ricanek et al., 2006) .To meet this challenge, two approaches have been taken: (1) research what is known about adult age-related craniofacial morphological changes to better understand how to (2) synthetically age individuals from a facial image, or rather, develop workable artificial age progression techniques to anticipate how an individual’s facemay appear years down the road. In addition to probing the question of how a person’s face may age across the adult lifespan, other facial appearance questions have arisen, such as how might a person’s face change simply over the course of one day? Herein we present an overview of face aging, its implications on biometrics, specifically FR systems, and provide an example of our work via relaying results of experiments designed to measure facial age changes and artificially synthesize facial images, all in an effort to improve FR systems and ensure that facial features, when used as a biometric tool, are reliable, consistent, and accurate with replicable results.

An improved rendering technique for active-appearance-model-based automated age progression

Age progression is the process of creating images that suggest how a person may appear in a certain amount of time based on the effects of the aging process. Traditionally these images have been created manually by forensic artists who use both art and science to guide how representations appear, whether drawn or photo-manipulated. Automated age-progression seeks to use algorithmic methods to create accurate images of how the individual in a photo could appear after aging effects. It is still a fairly young area of research, but one promising technique suggested so far has been to use parametrically driven face models such as Active Appearance Models to modify the face appearance in an image based on a data-driven model of face aging. These can be successful but tend to suffer from reconstructed texture artifacts.

Gender classification using the profile

Gender classification has attracted a lot of attention in computer vision and pattern recognition. In this paper, we propose a gender classification method. First, we present a robust profile extraction algorithm; Second, we implement Principal Components Analysis (PCA) and Independent Components Analysis (ICA) to extract discriminative features from profile to estimate the face gender via SVM. Our experimental results on Bosphorus 3D face database show that our proposed method works well.

Facial landmarking: comparing automatic landmarking methods with applications in soft biometrics

Registration is a critical step in computer-based image analysis. In this work we examine the effects of registration in face-based soft-biometrics. This form of soft-biometrics, better termed as facial analytics, takes an image containing a face and returns attributes of that face. In this work, the attributes of focus are gender and race. Automatic generation of facial analytics relies on accurate registration. Hence, this work evaluates three techniques for dense registration, namely AAM, Stacked ASM and CLM. Further, we evaluate the influence of facial landmark mis-localization, resulting from these techniques, on gender classification and race determination. To the best of our knowledge, such an evaluation of landmark mis-localization on soft biometrics, has not been conducted. We further demonstrate an effective system for gender and race classification based on dense landmarking and multi-factored principle components analysis. The system performs well against a multi-age face dataset for both gender and race classification.

Extremely dense face registration: Comparing automatic landmarking algorithms for general and ethno-gender models

Registration is a very important step in object recognition. Accurate detection of the eye centers, eye corners, mouth and nose are critical for face recognition and more broadly, for face processing. In this work, we have evaluated three techniques, namely AAM, Stacked ASM and CLM, for automatic detection of landmarks under the problem of extremely dense registration scheme for the face. Further we compare the efficacy of these techniques for the general case and for the specific case based on ethnicity and gender. It is shown that the performance of STASM and CLM are comparable and better than AAM. It is also shown that, in general, models trained on ethno-gender groups perform better than the models trained on general exemplars.