Gazette - Picture this

QUESTIONS & ANSWERS

Software gives eyewitness facial recollection a new look

Ever heard of EFIT-V? It’s the latest generation in facial imaging software, and it can create accurate facial composites with minimal verbal input from eyewitnesses. Det/Cst Clifford Clark, a witness interviewing specialist, chaired the United Kingdom E-FIT user group from 2004 to 2008. Now a member of the New Zealand Police, Clark speaks with Gazette writer Caroline Ross about how EFIT-V is changing the landscape of facial imaging.

Describing a suspect’s face in detail is a difficult task for most eyewitnesses, says Det/Cst Clifford Clark of the New Zealand Police, pictured here with a facial imaging system that relies more on visual — not verbal — input.

First, what is facial imaging software?

Facial imaging software is used to generate the face (of a suspected offender) from the memory of an eyewitness. In essence, it’s about getting the eyewitness’s memory of the offender’s face into a tangible form that everyone else can see.

Why is EFIT-V so “leading edge”?

EFIT-V treats the suspect’s face as a whole, rather than a collection of separate facial features that the witness must identify and change one by one. After the witness selects the suspect’s gender, ethnic appearance, hair style and general face shape, the software generates nine faces. The witness then chooses the face that is “most similar” (to his or her memory of the suspect), and the software creates a new generation of faces based on the chosen face. With each successive generation (of faces), the program reduces the variation between each of the faces, so they become more similar to each other — and to the witness’s memory of the offender.

How does this “whole-face” approach differ from other facial imaging techniques?

The majority of software packages use a system that builds a face by combining photographs of (individual) features, which are placed together to make up the whole face, then changed or corrected feature by feature — for example by selecting a nose from a database of images. FBI sketch artists (also) rely on the witness looking for each feature one by one from an album of many faces, thus the initial selection of features is in isolation.

What makes the whole-face approach more effective than other approaches?

Psychology has taught us that humans are good at recognizing whole faces but not so good at recognizing individual features, particularly from faces that are unfamiliar. If you saw a face for only a moment, it would be very difficult to pick out that unfamiliar person’s eyes from hundreds of examples, yet you might be able to recognize that person’s whole face in a crowd.

How much verbal input must a witness provide for EFIT-V to be effective?

EFIT-V is largely visually driven rather than verbally directed. Previous systems relied on a full cognitive interview (with the witness) to help develop a description of the offender prior to developing the composite face on the computer. It would appear that EFIT-V may do as well with no interview, or with a shortened version. The software provides an opportunity (for witnesses) to see (multiple) faces, pick the closest image, and/or reject images that are least like the offender. It thus plays to our inherent capacity for recognition rather than requiring us to negotiate limited descriptions, which are easily misinterpreted.

What are the advantages for law enforcement?

EFIT-V does seem to be much faster than conventional methods (with regard to arriving at a usable facial composite). There is also evidence that EFIT-V can be effective with witnesses who are less able to communicate, for example children, people who speak a different language, or people who suffer from cognitive impairment. Witnesses can point to images rather than describe what they want.

Any final comments?

It’s important to remember that a facial composite is only an image from a witness’s memory, often from a fleeting glance. It will never be a photograph of the offender. Facial composites help us solve crimes by helping us find suspects, but it is still good quality detective work that will prove or disprove any suspect’s involvement in an offence.