Earticle

Research on early facial composite systems has often demonstrated their poor capability to produce a good likeness of a target face. Photofit composites have been shown to produce poorer identification than simply furnishing a description of a target face. The current study revisits this comparison with composites produced using a holistic based composite system, EFIT-V. The effectiveness of verbal descriptions of faces was compared with facial composites in two laboratory tasks. Prior to the empirical work, eight ‘witnesses’ viewed an unfamiliar face and then provided a verbal description and created a facial composite. In Experiment 1 participants evaluated the relationship between a target face and three combinations of witness information (description alone, composite alone and description and composite combined). Providing a description was rated as more useful than providing a composite. In addition, presenting both a composite and description together was rated as less useful than presenting a description on its own. In Experiment 2 participants were supplied with the same types of information while they attempted to choose the target from an array of faces. Better performance was achieved from a description than from a composite or a description and composite combined. Analysis of participants’ confidence in their decisions showed it was higher when prompted by descriptions, irrespective of whether the decisions were correct or not. The implications of these findings for the utility of facial composites are discussed.

Remote protein homology detection has been widely used as a part of the analysis of protein structure and function. In this study, the good quality of protein feature vectors is the main aspect to detect remote protein homology; as it will assist discriminative classifier model to discriminate all the proteins into homologue or non-homologue members precisely. In order for the protein feature vectors to be characterized as having good quality, the feature vectors must contain high protein structural similarity information and are represented in low dimension which is free from any contaminated data. In this study, the contaminated data which originates from protein dataset was investigated. This contaminated data may prevent remote protein homology detection framework to produce the best representation of high protein structural similarity information in order to detect the homology of proteins. To reduce the contaminated data and extract high protein structural similarity information, some research has been done on the extraction of protein feature vectors and protein similarity. The extraction of protein feature vectors of good quality is believed could assist in getting better result for remote protein homology detection. Where, the good quality of protein feature vectors containing the useful protein similarity information and represent in low dimension will be used to identify protein family precisely by discriminative classifier model. Referring to this factor, a method which combines Protein Substring Scoring (PSS) and Pairwise Protein Substring Alignment (PPSA) from sequence comparison model, chi-square and Singular Value Decomposition (SVD) from generative model, and Support Vector Machine (SVM) as discriminative classifier model is introduced.

This paper aims to review the technique of Photometric Stereo (PS), with specific application to face recognition. PS is a method to rapidly estimate the three-dimensional geometry of a face (or any other Lambertian-like object) using several images with an identical viewpoint but varied illumination directions. The contributions of this paper are to (1) summarise the pros and cons of PS compared to alternative methods; (2) cover the theory of PS, in particular with respect to the related method of shape-from-shading; (3) outline some of the key extensions of PS to help overcome its weaknesses; and (4) discuss an application of PS for a practical and complete face recognition system.

Using physiological or behavioral characteristics to identify humans has been in use for quite some time now. Many wildlife animals also show distinctive natural body marks that can be used to identify them individually. Scientists in conservation research often use this approach but the process is manual and can be slow and error prone. This paper reports on an investigation to use biometric techniques for the identification of an important endangered species – The Great Crested Newt. The paper reports on novel techniques for extraction of the belly patterns of these animals as a source of biometric information. Features and classification techniques used for their automatic recognition are presented. The proposed approach is tested on a database of newts under investigation by conservationists. Preliminary studies are also reported on the ageing effects when belly images are compared over a number of years. The results suggest that such biometric techniques may be suitable for developing effective and flexible identification of wildlife in the field.

There is mounting evidence to suggest that the external features of a person face—shape, ears and, in particular, hair—exert a detrimental effect on the construction of a facial composite. The effect was first demonstrated for EvoFIT, a software system whereby constructors repeatedly select whole faces from arrays of alternatives, with ‘breeding’, to ‘evolve’ a face. In research by Frowd and Hepton (2009), volunteers saw a target face and, 24 hours later, were interviewed to describe the face in detail and then used EvoFIT in one of two ways: constructors saw face arrays containing hair that was either similar-to or exactly-matched hair on a target face. The study found that using exactly-matching hair promoted much-more identifiable composites than using similar hair. More recent research, however, has found that system performance is improved following use of a novel interview given to constructors. This holistic-cognitive interview prompts constructors to recall the target face in detail and then make seven personality-type judgments about it, with the aim of improving their face-recognition ability and thereby produce a better-quality composite. In the current research, we carried out a partial replication of Frowd and Hepton using the holistic-cognitive interview. It was found that identification of composites constructed in this way did not differ significantly by type of hair, and so the enhanced interview appears to mask inaccuracies in presented hair, promoting more identifiable images. Theoretical implications of the research are discussed for EvoFIT along with other system developments that have focused on the potential influence of hair.

Facial composite systems may be used by police to help a witness to a crime create a likeness of the perpetrator. Evaluation of new facial composite systems in the laboratory allows a measure of experimental control, but lacks the emotional impact of a real crime. As a step towards a more realistic level of stress for our participant witnesses, we presented target face images while they were engaged in playing an action thriller computer game. The quality of the composites they subsequently produced was compared with that of a second ‘onlooker’ participant, who merely observed the game and had the same view of the target face. Heart rate monitoring confirmed that the players were more stressed than the onlookers while the recognition rate of the onlooker composites was twice as good. We conclude that the method holds some promise as a method for composite system evaluation.