Human memory is a notoriously unreliable thing that can be easily influenced. That’s good news for criminals and bad news for law enforcement agencies that often rely on eyewitnesses to provide a description of a criminal. Around the world, law enforcement agencies employ sketch artists to piece together faces in a process similar to assembling a Mr. Potato Head toy. The witness describes key features, such as hair length, nose size or sharpness of the chin, and the artist combines them to create a likeness. Research into psychology suggests that this kind of method doesn’t take into account how the memory actually works, so researchers have developed new software that helps witnesses recreate and recognize suspects using principles borrowed from the fields of optics and genetics.
Developed by Dr Christopher Solomon of the University of Kent in Canterbury, England, the new software, called the EFIT-V system, generates its own faces that progressively evolve to match the witness’ memories. The witness starts with a general description such as "I remember a young white male with dark hair." Nine different computer-generated faces that roughly fit the description are generated, and the witness identifies the best and worst matches. The software uses the best fit as a template to automatically generate nine new faces with slightly tweaked features, based on what it learned from the rejected faces.
"Over a number of generations, the computer can learn what face you're looking for," says Solomon.
The mathematics underlying the software is borrowed from Solomon's experience using optics to image turbulence in the atmosphere in the 1990s. Solomon says, "I then realized that the same technique could be applied to human faces, which in many respects are mathematically similar to turbulent wavefronts" – I know mine is, especially first thing in the morning.
The software integrates this approach with an interactive genetic algorithm that progressively changes the features based on principles borrowed from evolution. Characteristics such as nose size and chin sharpness are represented as mathematical genes that mutate. As the features change, the witness' selections guide the evolution of the face.
One advantage of this technique, says Solomon, is that it can be used on witnesses who can't recall details about a suspect, but say that they would remember the face if they saw it again. Traditionally, police sketch artists cannot work with these people. By tapping into recognition instead of recall, "the EFIT-V system proved to be quite effective even when witnesses say they can't describe a person," says Solomon.
The software is being used by approximately 15 police departments in the United Kingdom, and by a half dozen European countries, including France and Switzerland. In field trials conducted by the Derbyshire police force, it led to twice as many identifications of suspects as traditional methods. The software has now started to make its way to the United States, where it being used by researchers in university settings and, in the future, Solomon hopes to partner with a suitable U.S. company to market the technology to police departments.
Solomon will present the software at the Optical Society of America's (OSA) Annual Meeting, Frontiers in Optics, which will take place Oct. 11-15 in San Jose, California.