How many stars are there? Find out in


[Archive: 7 March 1998]

Nowhere to hide

With electronic eyes, ears and noses,
nothing will be secret from the robocops
and the spies of the 21st century

IF YOU WANT to see the future of surveillance, take a trip into the world of millimetre waves and the video cameras that are sensitive to them. People are stripped of their clothes and become featureless, luminous humanoids. Silhouetted against their bodies and suspended as if by magic, hang coins, buckles, pens and keys. Cars are dark and sinister, although their hot radiator grills are bright. Only the steel in reinforced concrete shows up, so buildings look more like cages of copper pipes and electricity cables than homes and offices.

There is little privacy in the world revealed by the millimetre-wave camera. Sitting rooms, bedrooms, bathrooms and lavatories all merge into one open space. People sit in their cages watching warm boxes, others sleep while suspended a few centimetres above the floor. The images make an apt metaphor for the brave new digital world of electronic surveillance. It is a world where there will be nowhere to hide, nor anywhere to hide anything. There are already devices under development that will see through walls and strip-search suspects from a distance, looking under their clothes and inside their bodies. Individuals may be identified by their unique smells and tracked down, or 'recognised' electronically, even before they have had time to complete a crime. And thanks to cheap digital video cameras and powerful new search algorithms, individuals will be tracked by computers. There will be no anonymity even in the once welcoming crowds.

Millimetre waves sit in the electromagnetic spectrum between the infrared and microwaves. They are emitted by anything that contains water, especially if it is warm. The human body is an excellent source, and it stands out like a beacon at these frequencies. 'Millimetre waves are incredibly useful,' says Steve Bohrer, an electrical engineer at Millitech, a company in Massachusetts that has spent 10 years developing cameras that can spot them. They pass straight through any nonconducting material-this includes almost all clothing and most types of building material, he says. Metals are poor emitters, while dielectric materials such as plastics, ceramics, plastic explosives and powdered drugs lie somewhere in between. The amount of millimetre wave radiation that any of these materials emits also depends on their temperature.

Like visible and infrared light, millimetre waves can be focused to form an image. 'It's not difficult. All you need is a plastic lens,' explains Bohrer. What is more difficult is detecting this image once it has been formed.

In a conventional video camera, light is turned into an electrical signal by sensors known as charge coupled devices (CCDs). An array of CCDs at the camera's focal plane generates the electronic signal that creates a picture on a monitor. The more sensors in the array, the finer the camera's resolution, so an important factor is the size of each CCD and hence the number that can be crammed into the array. Most home video cameras have thousands.

But CCDs cannot be used to pick up millimetre waves. Instead radio antennas are needed detect the radiation, and the tiny currents induced in the antennas must be amplified into a clear electronic signal. Such devices tend to be relatively bulky. The trick that Millitech has pulled off is to develop a way of making antennas only two or three millimetres across and then assembling 256 of them into a two-dimensional array. At a distance of a metre or so the camera can resolve objects a few millimetres across. Like a conventional video camera, the array takes 30 pictures per second. The result is a real-time, moving image of the world in millimetre waves.

Millitech's two prototype cameras can easily spot concealed metal knives and guns against the bright background of the bearer's body. If the temperature of an object is known, it is also possible to identify the material it is made from, by assessing the brightness of the image. Since most concealed objects and packages are kept close to the body, their temperature can be estimated fairly accurately. Using this technique, crystalline substances such as sugar or powdered drugs can be identified clearly, although it is not possible to distinguish between the two. By 1997, Millitech expects to have a millimetre-wave camera on sale for around $10 000, and a rugged portable version for $80 000.

To complement its camera, Millitech is developing computer software that will scan images produced by the camera and alert a human operator if it spots something suspicious. Fooling the computer will not be easy. 'I suppose you could hide a gun inside a hot water bottle filled with water at body temperature,' suggests Bohrer. 'But the system would still pick up the rubber bottle.' The only way to beat it will be to hide objects inside the body.

The Millitech camera could be a voyeur's delight. At close range, males and females can be easily distinguished, says Bohrer. The intensity of the millimetre waves that human flesh emits depends on its temperature. A man's genitals are slightly cooler than the rest of the body, so they appear darker. Bohrer says that the computerised scanning system will safeguard people's privacy by doing away with the need for routine human surveillance. While millimetre-wave cameras ruffle through pockets and clothes, microwave imaging devices will look inside the human body for contraband hidden in even more intimate places (see 'The Pocket Radar Revolution', New Scientist, 12 August 1995).

Developed at the Lawrence Livermore National Laboratory in California, these devices comprise two tiny radars that broadcast microwave impulses and listen for the reflections. Each radar controls its range by briefly opening its 'ears' a short time after each pulse is broadcast, ignoring echoes from nearer or more distant objects. This effectively creates a shell around the radar, which typically has a radius of a few metres. By using two radars with shells that overlap, and varying the size of each shell, it is possible to map an entire volume of space.

Such devices have enormous potential for surveillance. They could make strip-searches a thing of the past, for example. To a trained operator, foreign objects inside the body will show up as easily as those outside. And because microwaves pass easily through walls and doors, similar systems could be used to mount a discreet surveillance operation. The radar set should cost only a few dollars to assemble. It is already little bigger than a bar of soap, and could be made much smaller. Only the computing power to assemble the image will be expensive and bulky.

Further down the electromagnetic spectrum, the military electronics manufacturer Raytheon has come up with a way of detecting metal objects using radio waves with wavelengths of a few metres. The idea is that the radiation will excite electrons in a metal object which will reradiate energy as they settle back into a lower energy state. The intensity and duration of this secondary radiation will depend on the size and shape of the object. Raytheon hopes to build a detector that bathes people with radio waves, and matches the secondary signal that this produces against a library of radio signatures from objects such as handguns and knives. For the moment, however, it is not clear how accurately this can be done.

'Artificial sniffers' that will identify us by our smell are another technology that is appearing on the horizon. One day such scent sensors could be as common as the video cameras that have sprouted over the past few years in city streets, over road junctions, and inside shopping centres, airports and railway stations. 'If you build new shopping mall, you could have sniffers all over the place,' predicts George Dodd, a researcher with the Highlands Scientific Research Group at the Craig Dunain Hospital in Inverness, and the acknowledged father of the electronic nose. The sniffers could monitor how often an individual visits a store, identify known shoplifters as they enter and alert security staff if necessary. 'You could detect intruders in an office, even identify them,' Dodd claims.

Police forces in Holland, Germany and Hungary have extensive databanks of human smells taken with swabs from crime scenes, which they use to set sniffer dogs on the track of the culprit. With electronic noses, however, everybody's smell could be stored on computer, says Dodd. In future, smell could be used as evidence of a person's identity much as fingerprints and DNA tests are today.

The human body produces thousands of smelly chemicals, but electronic noses will have to be able to home in on the few hundred that give individuals their unique scent. Dogs do this effortlessly, and can identify the owner up to three years later, says Barbara Sommerville, senior research associate at the University of Cambridge department of clinical veterinary medicine. By painstaking trial and error she has worked out which chemicals are important and which should be ignored. 'The compounds we look for are acids with a low molecular weight, like those found in smelly socks and unwashed clothes,' she says. Armed with these data, it is possible to develop electronic sniffers that look only for these smells.

As well as these unique scents, the body produces a range of smells that depend on the food a person has eaten recently, their state of health, and physiological factors such as their stress levels. Diseases such as lung cancer, schizophrenia, diabetes and duodenal ulcers can all be diagnosed using smell. It could even be used to tell when someone is telling lies. Dodd and John Parker, professor of electronics at the University of Glasgow, are developing an electronic nose that will look for these changing smells. They aim to make it the size of a computer chip, and as cheap to manufacture. Placed in a telephone mouthpiece, the sniffer could analyse the speaker's breath so that doctors could monitor patients without paying a home visit. Because they could indicate the stress of anyone speaking on the phone, sniffer chips planted secretly could provide valuable information-during tense negotiations, for example. 'The technology could easily be turned to spying,' says Dodd.

Virtually all the sniffers now being developed are based on pioneering artificial noses built by Dodd in the 1970s and 1980s. The human nose has about 10 000 sensors, but for his artificial models Dodd has had to make do with arrays of a dozen or so. Each one is coated with a range of electrically conducting polymers that can adsorb selected smelly organic chemicals. The change in electrical conductivity that this causes produces a measurable electronic signal, and the pattern of electronic signals from the different sensors in the array forms a characteristic fingerprint for any given odour. Identifying a smell then becomes a matter of simply comparing this signature with a database of known signatures.

'The noses work, but we don't know how,' says Sommerville. 'It's a real witch's cauldron.' Small variations in the recipe for polymerisation-such as the addition of a minute concentration of metallic ions or a small change in pH-have huge impacts on what the sensor picks up. 'The tedious bit is screening until you pick up and exclude what you want.'

Electronic noses may not be with us just yet, but electronic eyes and ears certainly are, and they are deluging their operators with information. For example, at London's Gatwick Airport a network of over 400 conventional cameras keeps the airport buildings under continuous surveillance. The system can track any individual from the moment they enter the airport to the moment they leave, but for this an operator must sit in the control room, panning, tilting and zooming the cameras. Output from all the cameras is continuously recorded on tape, but recalling these images at a later date requires many hours of laborious work in front of a video suite, combing through the tapes one by one, looking for a particular individual or incident. Computers are little help because conventional cameras generate images in a form that is difficult for a computer to handle.

It is a different story with a new system of 24 digital cameras that monitor the car parks at Gatwick. Every day they capture up to 20 000 images, recording every car and driver entering the car parks, along with the vehicle's registration number. The data are stored on a computer system's disc. Finding out which cars were at a given point at a particular time is then a matter of simply punching in the time and site of interest.

As digital cameras become more common, the possibility opens up of using computers to sift through huge banks of video information. A program called Matchmaker, from the Berkshire company Iterated Systems, can hunt through a database of images such as the car park shots from Gatwick, searching for a particular face. The speed with which it does this depends on the image, its resolution and the power of the computer. For example, on a personal computer, Matchmaker would take several minutes to count the number of aircraft in an aerial photograph of an airport, according to Andrew Crolla, an imaging specialist at Iterated Systems. 'But the computing power is available to do this in real time,' he says. 'It would even be possible to spot the face of a terrorist walking through customs.' In future, programs like Matchmaker will be able to pick an individual out of a crowd and track them as they move about and even hunt through old footage to see where they have been.

Monitoring large numbers of phone conversations for keywords is even easier. 'The watershed for speech processing occurred in the 1980s with the development of fast digital signal processors,' says Nial Donnelly of the Dublin-based company Voice Control Systems. A 10-year-old PC with a 286 chip is powerful enough to monitor two telephone conversations simultaneously for a specific word, he says.

But the success of the technique depends on the word you are looking for. 'Multisyllabic words are great for word spotting,' says Donnelly. 'It would be easy for a computer to scan conversations for words such as 'Semtex', but the word 'bomb' could be confused with 'bum',' he adds. Word-recognition programs start by building up a data file of the way people pronounce the keywords they are looking for. This might be done by digitally recording 1000 people speaking the word. The software breaks the word down into its component sounds, or 'phonemes' and stores them. When scanning a conversation, the computer listens out for the required phonemes in the right order, and starts recording when it hears them.

Producing a reliable transcripts of an entire spoken conversation is still beyond such systems. 'If you talk with a gap between each word then it's easy,' says Donnelly. But when people talk normally, words tend to merge together, and this is too much for a computer system to disentangle. 'It'll be twenty years before we solve this one,' he says.

By Andy Coghlan, Vincent Kiernan and Justin Mullis

Return to main feature page

From New Scientist, 04 November 1995


© Copyright New Scientist, RBI Limited 1998