Irish police force pilot facial recognition system to identify suspects
For the past three weeks, Irish police force, the Gardai, has been testing facial recognition software that enables victims of crime or witnesses to more easily identify suspects, according to the Irish Times.
The police force has already seen some success with the EVO-Fit software, achieving up to a 74% success rate in correctly identifying suspects that are wanted by law enforcement agencies.
Developed by Dr Charlie Frowd of the University of Winchester, the mobile system taps into the more subliminal and subtle “recognition memory to help victims and eyewitnesses more effectively identify suspects.
Gardaí said the traditional system of asking witnesses and victims to describe a suspect is often met with limited results, especially if the person was traumatized, or if the crime occurred quickly or in a dark setting.
Using the EVO-Fit system, witnesses and victims are asked to recall their impressions of the suspect’s character or personality in addition to their physical description in order to trigger a person’s recognition memory.
Investigators ask these witnesses and victims to choose photos from a large database that best resemble the suspects they saw.
Based on their selections, the system compiles an initial photo-fit by combining the shapes and textures of the various faces to produce a crude facial image.
The police compare the image against over 60 databases which contain facial images for both male and female suspects spanning all ethnic groups, aged brackets and other features.
The system then repeats the process multiple times before it finally identifies the most accurate depiction of the suspect possible.
There are currently 16 police forces in the UK that are using the EVO-Fit software, said Garda commissioner Noirin O’Sullivan.
The system has greatly improved Humberside Police’s success rate. When the police force switched over from traditional photo fit techniques to the EVO-Fit system, they were able to boost their suspect identification rate from 14% to a whopping 74%.