Search

Recognition following long delays is superior for highly attractive and highly unattractive faces (cf. medium-attractive faces). In the current work, we investigated participants’ ability to recreate from memory faces of low-, medium- and high physical attractiveness. In Experiment 1, participants constructed composites of familiar (celebrity) faces using the holistic EvoFIT system. When controlling for other variables that may influence face recognition (memorability, familiarity, likeability and\ud age), correct naming and ratings of likeness were superior for composites of low attractiveness targets. Experiment 2 replicated this design using the feature based PRO-fit system, revealing superiority (by composite naming and ratings of\ud likeness) for medium attractiveness. In Experiment 3, participants constructed composites of unfamiliar faces after a forensically-relevant delay of 1 day. Using ratings of likeness as a measure of composite effectiveness, these same effects were observed for EvoFIT and PRO-fit. The work demonstrates the importance of attractiveness for method of composite face construction.

Facial identification based on a comparison with a photo-ID is the most standard way to prove identity at security controls. Two experiments are performed controlling for the presence or absence of within-contour facial periphery (masking) and its substitution for an average periphery image (averaging), measuring matching accuracy, reaction time and signal-detection measures d’ an C. Experiment 1 compared face matching for the original periphery in the pair, its masking and its averaging using a sample of frontal image pairs. Experiment 2 compared matches in the average condition with and without an apparent gap around internal features. Results show that masking of facial periphery had a detrimental effect on unfamiliar face matching accuracy accompanied by an increased tendency to positive responses, while averaging composites yielded no difference as compared to the original pair. No differences were found for matching in conditions with or without an apparent gap. The results suggest that face periphery contributes to unfamiliar face matching accuracy through a holistic process which is disrupted when focusing exclusively on the innermost features; the effect is dependent on the global structure of the face image and not related to low-level details. These results should be considered in the context of improving unfamiliar face matching.

Witnesses may construct a composite face of a perpetrator using a computerised interface. Police practitioners guide witnesses through this unusual process, the goal being to produce an identifiable image. However, any changes a perpetrator makes to their external facial-features may interfere with this process. In Experiment 1, participants constructed a composite using a holistic interface one day after target encoding. Target faces were unaltered, or had altered external-features: (i) changed hair, (ii) external-features removed or (iii) naturally-concealed external-features (hair, ears, face-shape occluded by a hooded top). These manipulations produced composites with more error-prone internal-features: participants' familiar with a target's unaltered appearance less often provided a correct name. Experiment 2 applied external-feature alterations to composites of unaltered targets; although whole-face composites contained less error-prone internal-features, identification was impaired. Experiment 3 replicated negative effects of changing target hair on construction and tested a practical solution: selectively concealing hair and eyes improved identification. Practitioner Summary: The research indicates that when a target identity disguises or changes hair, this can lead to a witness (or victim) constructing a composite that is less readily identified. We assess a practical method to overcome this forensic issue. Abbreviation: GEE: Generalized Estimating Equations.

In recent years there has been growing interest in the identification of people with superior face recognition skills, for both theoretical and applied investigations. These individuals have mostly been identified via their performance on a single attempt at a tightly controlled test of face memory—the long form of the Cambridge Face Memory Test (CFMT+). The consistency of their skills over a range of tests, particularly those replicating more applied policing scenarios, has yet to be examined systematically. The current investigation screened 200 people who believed they have superior face recognition skills, using the CFMT+ and three new, more applied tests (measuring face memory, face matching and composite-face identification in a crowd). Of the sample, 59.5% showed at least some consistency in superior face recognition performance, although only five individuals outperformed controls on overall indices of target-present and target-absent trials. Only one participant outperformed controls on the Crowds test, suggesting that some applied face recognition tasks require very specific skills. In conclusion, future screening protocols need to be suitably thorough to test for consistency in performance, and to allow different types of superior performer to be detected from the outset. Screening for optimal performers may sometimes need to directly replicate the task in question, taking into account target-present and target-absent performance. Self-selection alone is not a reliable means of identifying those at the top end of the face recognition spectrum. Electronic supplementary material The online version of this article (10.1186/s41235-018-0116-5) contains supplementary material, which is available to authorized users.

Purpose – The purpose of this paper is to investigate the potential susceptibility of eyewitness memory to the presence of extraneous background speech that comprises a description consistent with, or at odds with, a target face. Design/methodology/approach – A between-participants design was deployed whereby participants viewed an unfamiliar target face in the presence of quiet, or extraneous to-be-ignored speech comprising a verbal description that was either congruent or incongruent with the target face. After a short distractor task, participants were asked to describe the target face and construct a composite of the face using PRO-fit software. Further participants rated the likeness of the composites to the target. Findings – Recall of correct facial descriptors was facilitated by congruent to-be-ignored speech and inhibited by incongruent to-be-ignored speech compared to quiet. Moreover, incorrect facial descriptors were reported more often in the incongruent speech condition compared with the congruent speech and quiet conditions. Composites constructed after exposure to incongruent speech were rated as worse likenesses to the target than those created after exposure to congruent speech and quiet. Whether congruent speech facilitated or impaired composite construction was found to depend on the distinctiveness of the target face. Practical implications – The results suggest that the nature of to-be-ignored background speech has powerful effects on the accuracy of information verbally reported from having witnessed a face. Incongruent speech appears to disrupt the recognition processes that underpin face construction while congruent speech may have facilitative or detrimental effects on this process, depending on the distinctiveness of the target face. Originality/value – This is one of the first studies to demonstrate that extraneous speech can produce adverse effects on the recall and recognition of complex visual information: in this case, the appearance of a human face.

Victims of and witnesses to crime are asked to describe an offender using cognitive interviewing techniques (CI), before constructing a visual likeness of the face. The aim of the current experiment was to investigate whether composite construction using the EvoFIT holistic system would benefit from the parallel use of three enhancement techniques. The study manipulated the type of interview used to elicit a face description (CI vs. holistic cognitive interview, or H-CI) and trialled a visualisation technique for selecting faces (no visualisation vs. visualisation of external features). Also included was a new construction procedure for EvoFIT that requested constructors to focus on the region around the eyes when making face selections. Based on past research, it was anticipated that both the H-CI and external-feature visualisation would promote construction of a more identifiable composite (compared to when each technique was not used). Rather unexpectedly, the results revealed that neither technique improved correct naming of composites, yet an interaction was observed: visualisation of external features led to a benefit that approached significance when used in conjunction with the H-CI (cf. CI). However, when no external-feature visualisation was used, composites were better named following the more usual CI (cf. H-CI) protocol. Results are promising for the new method of face selection, which was used by all participants (focusing on the eye region). Indeed, in the baseline condition, where this was the only `enhancement' method used, naming of composites was 55% correct. Arguably, focusing on the character of the face during the H-CI may instate a processing style that is not well aligned with the new procedure of selecting faces according to the eye region. To overcome this misalignment of processing stages, we propose to ask constructors to focus on this region of the face during the H-CI itself. Implications for theory and police practice are discussed.

When perceiving faces under normal circumstances, the focus of attention is likely to be on the upper (e.g., hair, eyes) than lower (e.g., mouth, chin) facial half [1]. If such a bias were to extend to face construction, then it may hinder the effectiveness of forensic evidence collected from witnesses and victims of crime. In Experiment 1, participants constructed a single face using the EvoFIT holistic (police) composite system 24 hours after having seen an unfamiliar target identity. When constructing the face, participants were asked to select items from face arrays based on the whole face, or for upper and lower facial halves separately; these faces were presented in arrays either intact or horizontally-misaligned [2], and with external features (hair, ears, neck) present or absent. More-identifiable composites were predicted from (i) selection of separate facial halves (cf. currently-used whole-face selection), (ii) presence of horizontal misalignment and (iii) absence of external features. Experiment 2 used the same basic design but participants were requested to select for (i) upper-face half during evolution, (ii) the same as (i) but also for subsequent adjustments of the face, and (iii) overall face (Control). The composites constructed in both experiments were named and rated for likeness. Experiment 1 unexpectedly revealed that the Control group produced the highest-named composites. In Experiment 2, upper face selection during the evolution stage produced more-effective composites. In terms of practical implications, for the EvoFIT composite system, and potentially for other holistic systems, witnesses should be instructed to select faces for the upper facial half during evolution, to maximise subsequent identification of their composites.

Purpose – Research suggests that memory for unfamiliar faces is pictorial in nature, with recognition negatively affected by changes to image-specific information such as head pose, lighting and facial expression. Further, within-person variation causes some images to resemble a subject more than others. Here, the purpose of this paper is to explore the impact of target-image choice on face construction using a modern evolving type of composite system, EvoFIT. Design/methodology/approach – Participants saw an unfamiliar target identity and then created a single composite of it the following day with EvoFIT by repeatedly selecting from arrays of faces with “breeding”, to “evolve” a face. Targets were images that had been previously categorised as low, medium or high likeness, or a face prototype comprising averaged photographs of the same individual. Findings – Identification of composites of low likeness targets was inferior but increased as a significant linear trend from low to medium to high likeness. Also, identification scores decreased when targets changed by pose and expression, but not by lighting. Similarly, composite identification from prototypes was more accurate than those from low likeness targets, providing some support that image averages generally produce more robust memory traces. Practical implications – The results emphasise the potential importance of matching a target's pose and expression at face construction; also, for obtaining image-specific information for construction of facial-composite images, a result that would appear to be useful to developers and researchers of composite software. Originality/value – This current project is the first of its kind to formally explore the potential impact of pictorial properties of a target face on identifiability of faces created from memory. The design followed forensic practices as far as is practicable, to allow good generalisation of results.

Facial composite systems help eyewitnesses to show the appearance of criminals. However, likenesses created by unfamiliar witnesses will not be completely accurate, and people familiar with the target can find them difficult to identify. Faces are processed holistically; we explore whether this impairs identification of inaccurate composite images and whether recognition can be improved. In Experiment 1 (n = 64) an imaging technique was used to make composites of celebrity faces more accurate and identification was contrasted with the original composite images. Corrected composites were better recognized, confirming that errors in production of the likenesses impair identification. The influence of holistic face processing was explored by misaligning the top and bottom parts of the composites (cf. Young, Hellawell, & Hay, 1984). Misalignment impaired recognition of corrected composites but identification of the original, inaccurate composites significantly improved. This effect was replicated with facial composites of non-celebrities in Experiment 2 (n = 57). We conclude that, like real faces, facial composites are processed holistically: recognition is impaired because unlike real faces, composites contain inaccuracies and holistic face processing makes it difficult to perceive identifiable features. This effect was consistent across composites of celebrities and composites of people who are personally familiar. Our findings suggest that identification of forensic facial composites can be enhanced by presenting composites in a misaligned format.

We investigate the effect of target familiarity on the construction of facial composites, as used by law enforcement to locate criminal suspects. Two popular software construction methods were investigated. Participants were shown a target face that was either familiar or unfamiliar to them and constructed a composite of it from memory using a typical 'feature' system, involving selection of individual facial features, or one of the newer 'holistic' types, involving repeated selection and breeding from arrays of whole faces. This study found that composites constructed of a familiar face were named more successfully than composites of an unfamiliar face; also, naming of composites of internal and external features was equivalent for construction of unfamiliar targets, but internal features were better named than the external features for familiar targets. These findings applied to both systems, although benefit emerged for the holistic type due to more accurate construction of internal features and evidence for a whole-face advantage. STATEMENT OF RELEVANCE: This work is of relevance to practitioners who construct facial composites with witnesses to and victims of crime, as well as for software designers to help them improve the effectiveness of their composite systems.

The police often ask witnesses and victims to construct a facial composite of people they have seen committing a crime. Research has suggested, however, that these images are of poor quality due to memory decay, the face construction method used and the presence of the external facial features (hair, ears, face shape). In the current work, the role of the most important external feature, hair, was explored. Participants were shown an unfamiliar target face and constructed a composite of it 24 hours later using a modern recognition‐based system and one of three types of hair: exact, similar or poor match to the target. Better‐matching hair was found to help participants construct a more identifiable central region of the face (eyes, brows, nose, mouth). The results highlight the importance of accurate hair at the start of the construction session. Implications for police practice are discussed.

Witnesses and victims of serious crime are often required to construct a facial composite from their memory, a visual likeness of a suspect’s face. The traditional method is for them to select individual facial features to build a face, but often these images are of poor quality. We have developed a new method whereby witnesses repeatedly select instances from an array of complete faces and a composite is evolved over time by searching a face model built using Principal Components Analysis. While past research suggests that the new approach is superior, performance is far from ideal. In the current research, face models are built which match a witness’s description of a target. It is found that such ‘tailored’ models promote better quality composites, presumably due to a more effective search, and also that smaller models may be even better. The work has implications for researchers who are using statistical modelling techniques for recognising faces.

Witness and victims of serious crime are normally requested to construct a facial composite of a suspect's face. While modern systems for constructing composites have been evaluated extensively in the U.K., this is not the case in the U.S. In the current work, two popular computerized systems in the US, FACES and Identikit 2000, were evaluated against a 'reference' system, PRO-fit, where performance is established. In experiment 1, witnesses constructed a composite with both PRO-fit and FACES using a realistic procedure. The resulting composites were very poorly named, but the PRO-fit emerged best in 'cued' naming and two supplementary measures: composite sorting; and likeness ratings. In experiment 2, PRO-fit was compared with Identikit 2000, a sketch-like feature system. Spontaneous naming was again very poor, but both cued naming and sorting suggested that the systems were similar. The results support previous findings that modern systems do not produce identifiable composites.

Facial caricatures exaggerate the distinctive features of a face and may elevate the recognition of a familiar face. We investigate whether the recognition of facial composites, or pictures of criminal faces, could be similarly enhanced. In this study, participants first estimated the degree of caricature necessary to make composites most identifiable. Contrary to expectation, an anti-caricature was found to be best, presumably as this tended to reduce the appearance of errors. In support of this explanation, more positive caricature estimates were assigned to morphed composites: representations which tend to contain less overall error. In addition, anti-caricaturing reduced identification for morphed composites but enhanced identification for individual composites. While such improvements were too small to be of value to law enforcement, a sizeable naming benefit was observed when presenting a range of caricature states, which appeared to capitalise on individual differences in the internal representation of familiar faces.

Three experiments are reported that compare the quality of external with internal regions within a set of facial composites using two matching-type tasks. Composites are constructed with the aim of triggering recognition from people familiar with the targets, and past research suggests internal face features dominate representations of familiar faces in memory. However the experiments reported here show that the internal regions of composites are very poorly matched against the faces they purport to represent, while external feature regions alone were matched almost as well as complete composites. In Experiments 1 and 2 the composites used were constructed by participant-witnesses who were unfamiliar with the targets and therefore were predicted to demonstrate a bias towards the external parts of a face. In Experiment 3 we compared witnesses who were familiar or unfamiliar with the target items, but for both groups the external features were much better reproduced in the composites, suggesting it is the process of composite construction itself which is responsible for the poverty of the internal features. Practical implications of these results are discussed.

Purpose – The purpose of this paper is to investigate whether the presence of a whole-face context during facial composite production facilitates construction of facial composite images. Design/methodology/approach – In Experiment 1, constructors viewed a celebrity face and then developed a facial composite using PRO-fit software in one of two conditions: either the full-face was visible while facial features were selected, or only the feature currently being selected. The composites were named by different participants. The authors then replicated the study using a more forensically valid procedure: in Experiment 2 non-football fans viewed an image of a premiership footballer and 24 hours later constructed a composite of the face with a trained software operator. The resulting composites were named by football fans. Findings – In both studies, the presence of the facial context promoted more identifiable facial composites. Research limitations/implications – Current composite software was deployed in a conventional and unconventional way to demonstrate the importance of facial context. Practical implications – Results confirm that composite software should have the whole-face context visible to witnesses throughout construction. Although some software systems do this, there remain others that present features in isolation and these findings show that these systems are unlikely to be optimal. Originality/value – This is the first study to demonstrate the importance of a full-face context for the construction of facial composite images. Results are valuable to police forces and developers of composite software.

Eyewitnesses are often invited to construct a facial composite, an image created of the person they saw commit a crime that is used by law enforcement to locate criminal suspects. In the current paper, the effectiveness of composite images was investigated from traditional feature systems (E-FIT and PRO-fit), where participants (face constructors) selected individual features to build the face, and a more recent holistic system (EvoFIT), where they ‘evolved' a composite by repeatedly selecting from arrays of complete faces. Further participants attempted to name these composites when seen as an unaltered image, or when blurred, rotated, linearly stretched or converted to a photographic negative. All of the manipulations tested reduced correct naming of the composites overall except (i) for a low level of blur, for which naming improved for holistic composites but reduced for feature composites, and (ii) for 100% linear stretch, for which a substantial naming advantage was observed. Results also indicated that both featural (facial elements) and configural (feature spacing) information was useful for recognition in both types of composite system, but highly-detailed information was more accurate in the feature-based than the holistic method. The naming advantage of linear stretch was replicated using a forensically more-practical procedure with observers viewing an unaltered ¬composite sideways. The work is valuable to police practitioners and designers of facial-composite systems.

Eyewitnesses often construct a “composite” face of a person they saw commit a crime, a picture that police use to identify suspects. We described a technique (Frowd, Bruce, Ross, McIntyre, & Hancock, 2007) based on facial caricature to facilitate recognition of these images: Correct naming substantially improves when composites are seen with progressive positive caricature, where distinctive information is enhanced, and then with progressive negative caricature, the opposite. Over the course of four experiments, the underpinnings of this mechanism were explored. Positive-caricature levels were found to be largely responsible for improving naming of composites, with some benefit from negative-caricature levels. Also, different frame-presentation orders (forward, reverse, random, repeated) facilitated equivalent naming benefit relative to static composites. Overall, the data indicate that composites are usually constructed as negative caricatures.

Research has indicated that traditional methods for accessing facial memories usually yield unidentifiable images. Recent research, however, has made important improvements in this area to the witness interview, method used for constructing the face and recognition of finished composites. Here, we investigated whether three of these improvements would produce even-more recognisable images when used in conjunction with each other. The techniques are holistic in nature: they involve processes which operate on an entire face. Forty participants first inspected an unfamiliar target face. Nominally 24 hours later, they were interviewed using a standard type of cognitive interview (CI) to recall the appearance of the target, or an enhanced ‘holistic’ interview where the CI was followed by procedures for focussing on the target’s character. Participants then constructed a composite using EvoFIT, a recognition-type system that requires repeatedly selecting items from face arrays, with ‘breeding’, to ‘evolve’ a composite. They either saw faces in these arrays with blurred external features, or an enhanced method where these faces were presented with masked external features. Then, further participants attempted to name the composites, first by looking at the face front-on, the normal method, and then for a second time by looking at the face side-on, which research demonstrates facilitates recognition. All techniques improved correct naming on their own, but together promoted highly-recognisable composites with mean naming at 74% correct. The implication is that these techniques, if used together by practitioners should substantially increase the detection of suspects using this forensic method of person identification.

Clay or computerised facial reconstructions are often presented to the public for recognition without information about the external features of a head (hair, ears and neck) as these are thought to be potentially distracting or even misleading. In this context, the mechanisms of face recognition are poorly understood, but existing research using photographs of familiar faces suggests that external features play an important role for recognition, and external features may even be necessary for recognition to occur at all. The current research aimed to determine the contribution that external features make to the recognition of a familiar face rendered in 3D. It will also determine whether the inclusion or exclusion of external features from a reconstruction is likely to be beneficial. Volunteers were asked to name images of 3D faces of people known to them, presented as either full face 3D surface scans, or where the internal or external features had been removed. As was expected, a clear correlation was found between information presented in the scans and the recognition rate, with participants correctly naming full face images most often and images of external features least often. Logistic regression analysis demonstrated a significant linear trend in recognition rate in the order of external features, internal features and full face. Incorrect naming also increased linearly, indicating that participants were more likely to offer a name (correct or otherwise) when more useful facial information was provided.

Recognition memory for unfamiliar faces is facilitated when contextual cues (e.g. head pose, background environment, hair and clothing) are consistent between study and test. By contrast, inconsistencies in external features, especially hair, promote errors in unfamiliar face-matching tasks. For the construction of facial composites, as carried out by witnesses and victims of crime, the role of external features (hair, ears and neck) is less clear, although research does suggest their involvement. Here, over three experiments, we investigate the impact of external features for recovering facial memories using a modern, recognition-based composite system, EvoFIT. Participant-constructors inspected an unfamiliar target face and, one day later, repeatedly selected items from arrays of whole faces, with ‘breeding’, to ‘evolve’ a composite with EvoFIT; further participants (evaluators) named the resulting composites. In Experiment 1, the important internal-features (eyes, brows, nose and mouth) were constructed more identifiably when the visual presence of external features was decreased by Gaussian blur during construction: higher blur yielded more identifiable internal-features. In Experiment 2, increasing the visible extent of external features (to match the target’s) in the presented face-arrays also improved internal-features quality, although less so than when external features were masked throughout construction. Experiment 3 demonstrated that masking external-features promoted substantially more identifiable images than using the previous method of blurring external-features. Overall, the research indicates that external features are a distractive rather than a beneficial cue for face construction; the results also provide a much better method to construct composites, one that should dramatically increase identification of offenders.

The development of facial identification technologies is of vital importance for security. We demonstrate that psychology can ensure innovation produces applications with effective human outcomes. Experiments 1a-c show spectacle removal imaging techniques that enhance automated face matching, can seriously impair accuracy if employed within a human interface. Experiment 2 shows how understanding and working with human perceptual strategies can optimize identification of facial composites.

Constructing the face of a criminal from the selection of individual facial parts is a hard task. We have been working on a new system called EvoFIT that involves the selection and breeding of complete faces. The approach is theoretically better founded and produces more identifiable composites than those from a traditional 'feature' system. In the current paper, we explored three new methods of presenting faces to a person using EvoFIT. A better quality face was evolved if (1) the external parts of the face were subjected to a Gaussian blur, allowing a user to focus on the important inner region of the face, (2) if the faces were simplified to make them sketch-like in appearance, and (3) if users took longer in deciding which faces to select for breeding. Taken together, these approaches would appear to make a marked improvement in the ability to evolve an identifiable likeness of a target.

We have developed a novel software program called 'Predict Your Child' that, given photographs of potential parent faces, generates plausible looking children. The parent photographs are imported into a PCA-based model of facial appearance to give a set of face parameters that can be mixed together to produce offspring faces. The program is intended for entertainment and has been used commercially, with customers emailing photographs of parents to be bred. In this paper, we describe the system, developed from an evolutionary facial composite system called EvoFIT, outline some of the problems encountered and present some performance data.

Witnesses and victims of serious crime are normally requested to construct a picture of the criminal’s face. These pictures are known as facial composites and are typically produced by a witness recalling details of the face and then selecting individual facial features: hair, eyes, nose, mouth, etc. While composites remain an important tool for the apprehension of criminals, research has suggested that, even under favorable conditions, they are rarely recognized. In the current chapter, we present a new method called EvoFIT whereby users select complete faces and a composite is “evolved” using a Genetic Algorithm. While considerable development was required to tune the new approach, research indicates that EvoFIT now produces more identifiable composites than those produced from the traditional “feature” systems. Novel applications of the technology are also discussed.

Witnesses and victims of serious crime are often required to construct a facial composite, a visual likeness of a suspect's face. The traditional method is for them to select individual facial features to build a face, but often these images are of poor quality. We have developed a new method whereby witnesses repeatedly select instances from an array of complete faces and a composite is evolved over time by searching a face model built using PCA. While past research suggests that the new approach is superior, performance is far from ideal. In the current research, face models are built which match a witness's description of a target. It is found that such 'tailored' models promote better quality composites, presumably due to a more effective search, and also that smaller models may be even better. The work has implications for researchers who are using statistical modelling techniques for recognising faces.

This paper examines two facial composite systems that present multiple faces during construction to more closely resemble natural face processing. A 'parallel' version of PRO-fit was evaluated, which presents facial features in sets of six or twelve, and EvoFIT, a system in development, which contains a holistic face model and an evolutionary interface. The PRO-fit parallel interface turned out not to be quite as good as the 'serial' version as it appeared to interfere with holistic face processing. Composites from EvoFIT were named almost three times better than PRO-fit, but a benefit emerged under feature encoding, suggesting that recall has a greater role for EvoFIT than was previously thought. In general, an advantage was found for feature encoding, replicating a previous finding in this area, and also for a novel 'holistic' interview.

Facial composites are pictures of human faces. These are normally constructed by victims and witnesses of crime who describe a suspect's face and then select individual facial features. Unfortunately, research has shown that composites constructed in this way are not often recognised. In contrast, we are quite good at recognizing complete faces, even if the face is unfamiliar and only seen briefly. This more natural way of processing faces is at the heart of a new composite system called EvoFIT. With this computer program, witnesses are presented with sets of complete faces for selection and a composite is 'evolved' over time. The current work augments EvoFIT by developing a set of psychologically useful scales - such as facial weight, masculinity, and age - that allow EvoFIT faces to be manipulated. These holistic dimensions were implemented by increasing the size and variability of the underlying face model and by obtaining perceptual ratings so that the space could be suitably vectorised. The result of three evaluations suggested that the new dimensions were operating appropriately. Index Terms—facial composite, holistic, witness, crime, EvoFIT

Facial composites are typically constructed by witnesses to crime by describing a suspect's face and then selecting facial features from a kit of parts. Unfortunately, when produced in this way, composites are very poorly identified. In contrast, there is mounting evidence that other, more recognition-based approaches can produce a much better likeness of a suspect. With the EvoFIT system, for example, witnesses are presented with sets of complete faces and a composite is 'evolved' through a process of selection and breeding. The current work serves to augment EvoFIT by developing a set of psychologically useful 'knobs' that allow faces to be manipulated along dimensions such as facial weight, masculinity, and age. These holistic dimensions were implemented by increasing the size and variability of the underlying face model and obtaining perceptual ratings so that the space could be suitably vectorised. Two evaluations suggested that the new dimensions were operating appropriately.

Facial composites produced using traditional feature-based systems are notoriously hard to recognise. We have been developing a new more recognition-based system called EvoFIT that is performing better than other computerised approaches. In the current work, potential ways of improving performance even further were explored. It was found that asking the same person to construct two composites of a target face was successful in improving target identification. The data also found that composites constructed second were as identifiable as those constructed first, suggesting that the system does not appear to be interfering with a user's memory of a target face. The work also indicated that switching from a monochrome to a colour face model produced a slight decrement in performance. Lastly, the work replicated a previous finding that constructing a composite of a distinctive face produces a more identifiable rendition than a composite of a more average-looking face.

Publisher Summary This chapter describes a system that allows the evolutionary generation of near-photographic quality face images. The major anticipated use for such a system would be, allowing eyewitnesses to a crime to construct a likeness of the suspect. Current methods for doing this center on photofit-like composite systems. The user is presented with a large range of possible alternatives for major features, such as nose, mouth, and eyes and attempts to reconstruct a face piecemeal. The original photofit does not work very well, probably because it breaks the face into components in a way that we do not usually do when viewing a face. More recent systems, such as CD-fit, are far more flexible, allowing the position, size, and shape of individual features to be altered interactively on a computer. However, the essential problem of being a composite system remains. If you change, for example, the nose, our perception of the eyes and mouth may well change: we perceive faces holistically. The resultant images also look rather synthetic. The system described here produces face images by recombining principal components (eigenfaces), which are inherently global in nature and produce more photographic-like images.