Surveillance: ‘Behaviour recognition systems’…

MORE THAN A RELIABLE GUIDE?

TECHNOLOGICAL EXPERTS believe that recent advances in technology and within information technology allow them to detect hostile intentions before events may occur. But, unless deemed perfect, that might not be such a good thing.

Firstly, we should consider why a need might arise in developing an intelligent computerised surveillance system that might meet with such objectives. For people who are engaged monitoring people’s movements through the use of surveillance cameras, the task is bound to be deemed tedious. Even if a person employed on security tasks is concentrating, identifying suspicious behaviour is generally very difficult. Consider, for example, a nondescript man who descends to a subway platform several times over the course of a few days without getting on a train. Is that suspicious behaviour? It could be. Is the average security guard going to actually notice? Maybe not. An example, then – albeit fictional – of why experts would like to develop fully integrated systems in assessing people’s future actions.

The perceived need for such systems is stimulating the development of computerised devices that can both recognise people and objects, whilst also detecting suspicious or unorthodox forms of behaviour.

For the moment, much of this technology remains in laboratories. However, Cybernet Systems, a firm based in Ann Arbor (Michigan) in America, is working with the United States Army Research Laboratory and believes that behaviour-recognition systems are getting very good. Some are already being deployed at security checkpoints.

Human gaits, for instance, can provide a lot of information about possible intentions. The use of special object-recognition software allows the locking-on to particular features of a video recording (such as the movement of a person’s knees or elbow joints) and then following individual people around. Correlating these movements with consequences, such as the throwing of an incendiary, allows the development of computer models that link posture and consequences reasonably reliably. Such a system can, for example, identify a person in a crowd who is carrying a concealed package with the weight of a large explosives belt. The US army plans to deploy the system at military checkpoints, on vehicles and at embassy perimeters.

Some intelligent surveillance systems are even able to go beyond this. Instead of merely learning what a threat looks like, they are being adapted to learn the context in which behaviour might probably be threatening. That people linger in places such as bus stops, for example, is normal. Loitering in a stairwell, however, is a much rarer occurrence that may warrant introspective examination by human security staff.

The systems concept is also being developed and tested for the United States Air Force Research laboratory: it uses a network of cameras in tracking people identified as suspicious. It might be used, too, in tracking pedestrians who have left a package on the ground, continuing to walk through town.

As object and motion-recognition technology improves, researchers are starting to focus on facial expressions and what they might reveal. The ‘Human Factors Division’ of America’s Department of Homeland Security (DHS) is running what it calls “Project Hostile Intent”. Such a systems device seeks to scrutinise and extrapolate fleeting ‘micro-expressions’ that might easily be missed by the human eyes. Flashes last for less than a tenth of a second and involve just a small portion of the face.

Terrorists are often trained in the technique of concealing emotions; micro-expressions, though, are largely involuntary human actions. This actually aids the interest of researchers because any conscious attempt in suppressing facial expressions will accentuate human micro-expressions. Disturbingly, this is being referred to as “micro-facial leakage”.

 

IN TOTAL, there are some 40 micro-expressions. The DHS whilst refusing to describe them in detail does seem a little odd, as they have been studied and documented for many years by civilian research companies. However, it is worth pointing-out that signals which might be deemed to reveal hostile intent change with context. If, for instance, many travellers in an airport-screening line are running late, a good example that would most certainly raise expressions of anguish – such as raised cheeks and eyebrows, or lowered lips and gaze – the system would create signals that raise less cause for concern.

Proponents in developing this sort of technology argue that it avoids the contentious issue of racial profiling, because only behaviour is studied. However, that hardly resolves the issue because cultures and races express themselves in different ways. The University of Arizona which conducts research for the US Defence Department and who also employ experts on automated behaviour-recognition says that system improvements need to be tailored in dealing with cultural input. Passengers from repressive countries, for example, who may already be under suspicion because of their origins, are typically known to display signs of increasing levels of anxiety. This might be revealed through rigid body movements when nearing security exits. But, that might result in a high incidence of false-positives. It is understood that a software package (known as ‘Agent 99′) is being finely tuned that takes into account the body movements of people from differing cultures.

FAST, or Future Attributable Screening Technology, is another programme that is run by the Human Factors Division, and is being developed as a complement to Project Hostile Intent. An array of sensors, set at a distance of a couple of metres, measures variables like skin temperature, blood-flow patterns, perspiration, and heart and breathing rates. In a series of inaugural tests that involved 140 volunteers, the system detected about 80% of those who had been asked to try and deceive it by being hostile or by attempting to smuggle a weapon through it.

The researchers are playing down the significance of the testing statistics, and few additional details are available. However, it is believed that FAST – which only began 16 months ago – will improve yet further in terms of accuracy due to extra sensors that are being developed that will detect eye movements and even body odours, both of which can provide clues to emotional states.

Civil liberty groups argue that FAST amounts to a forced medical examination with too many innocents likely becoming entangled in intrusive questioning. In terms of technological evolution this resembles something very similar to the workings of polygraphs (lie detectors) – an electrical device that registers involuntary physical processes such as pulse rate and perspiration – in measuring physiological correlates of stress. Polygraphs have had a patchy and controversial history, accusing nervous innocents while acquitting practiced liars.

 

© Mark Dowe 2008: all rights protected

mark.dowe@googlemail.com

Copyright is the currency by which information may be exchanged in certain instances. If you are unsure of your rights relating to digital communications in partial or complete form you should seek independent legal advice.

This site is governed by the Laws of Scotland.

Outstanding political and current affairs coverage