CCTV cameras that listen as well as watch

News by Richard Thurston

University researchers are developing artificial intelligence software which can be loaded onto CCTV cameras to enable them to locate incidents by noise alone

CCTV cameras could soon be capable of reacting to noise, university researchers say.

The addition of artificial intelligence software would enable CCTV cameras to automatically rotate to focus on incidents, rather than waiting for a human to spot the incident and move the camera.

If the camera 'hears' shouting or the sound of breaking glass, for example, it would pan round to focus on the incident, potentially filming precious seconds that would otherwise be missed.

The researchers, from the University of Portsmouth, hope it could slash response times and revolutionise the way crimes are caught on camera. It could also save police officers from having to view hours of footage by flagging up key incidents.

Revealing his research to SC Magazine today, Dr David Brown, head of the University's Institute of Industrial Research, said he had won three years of funding from the Engineering and Physical Sciences Research Council (EPSRC) to undertake the project.

"It's a very fast, real-time method of identifying sounds," he said.

His proposed system means attaching two or three directional microphones to each CCTV camera, with the addition of specially-designed articial intelligence software.

When a CCTV camera hears a suspicious sound, it uses the directional attributes of the microphones to locate the source of the problem, and swivels round in just 300milliseconds. It all produces an alert for the CCTV operator.

The camera could also make use of motion analysis, which flags up actions like individuals running away from a crowd or rapidly moving their arms.

Brown said the system has worked well in a lab environment, but that further testing would be required before it is deployed in practice. He added that the software can 'learn' the signs of an incident to more accurately respond to situations.

Brown said the system might later be developed to recognise specific words which indicate the start of an incident.

The Institute of Industrial Research has been studying CCTV since 2001. As well as sound and motion detection, its work has included improving the quality of the grainy images CCTV often produces.

Separately, the university is preparing a report for the Home Office to help it produce guidelines for the police on how to review CCTV. The report will focus on the ergonomic, attentional, perceptional and decision-making issues involved with reviewing CCTV.

Find this article useful?

Get more great articles like this in your inbox every lunchtime

Video and interviews