A new technique developed by scientists at Manchester Metropolitan University that interprets facial gestures could be the most accurate 'lie detector' yet discovered. Using just a laptop and a camera, the method has demonstrated significantly better results than the traditional polygraph lie detectors used by the CIA and is much simpler to use, needing no experts to operate it and no physical contact with the suspect.
The 'Silent Talker' the system uses artificial intelligence to detect and analyse thousands of micro-gestures, many of which go unnoticed by the naked eye. The system reads stress, deception, tiredness and other traits to analyse non-verbal behaviour, which makes up 93% of all human communication. Tiny involuntary movements that are considered impossible to fake are detected by Silent Talker, which then analyses incongruities to identify when someone is lying. Project director Dr Zuhair Bandar in the University's Department of Computing and Mathematics said: "We have looked at systems across the world and are convinced that this is the most sophisticated".
"A breakthrough in the UK is all the more significant given the large budgets made available for security in the USA post September 11.
"In phased tests to date, the system is accurate in over 80% of cases, compared to 70% for the best other current systems.In one test, researchers asked volunteers to steal '10 each from a box and asked them to deny knowledge of the theft when later questioned. To compare the different reactions, another group who had seen but not stolen the money, was also quizzed about the missing money. During questioning, a secret camera filmed the interviewees faces. It correctly identified the liars in 80% of cases.Banks and insurance companies have already expressed interest in the system, which has manifold applications from job interviews and police investigations to airport security and counter-terrorism. It also has applications in medicine and could aid the diagnosis and monitoring of depressions, schizophrenia and psychopathy. The team is currently working with psychologists from MMU and the Universityof Liverpool, and has secured a patent application for its system.
