Stress-detecting cameras great for pre-crime identification

TECHNOLOGY

18 AUGUST 14  by LIAT CLARK

http://www.wired.co.uk/news/archive/2014-08/18/stress-surveillance

A team of electronic engineers is developing tiny surveillance cameras that pick up stress signals in people. They speculate the design could be used to spot anxious would-be criminals in airports and other high-target areas.

Chen Tong from the Southwest University in Chongqing told the South China Morning Post that the system -- which picks up levels of blood oxygenation -- will find out where anxiety is bubbling under the surface.

Referring to a number of attacks that have taken place in China in the past year, Tong said of the perpetrators: "They all looked and behaved as ordinary people but their level of mental stress must have been extremely high before they launched their attacks. Our technology can detect such people, so law enforcement officers can take precautions and prevent these tragedies."

According to the SCMP, the technique uses hyperspectral imaging, which gathers data across the electromagnetic spectrum down to pixel level. It is designed to pick up unusual dispersions of oxygen in the blood across areas like the face, which might be an indicator of high stress levels. According to the Southwest University researcher, tests have so far indicated the system can differentiate between stress and normal physical activities that might heighten oxygen levels in the blood. It sounds as though Tong envisions law enforcement using some kind of wearable to visually identify suspects using metre readings.

There are, of course, infinite issues here in terms of personal data storage and collection. Any wearables in existence today would not have the processing power to analyse a crowd and deliver results in real time. It means that personal data and images would need to be transferred to a central storage system for processing over Wi-Fi, potentially leaving that data further open to interception. 

Tong told the SCMP: "The suspected terrorists could hopefully be filtered out by non-contact devices and then be checked more intensively with contact devices."

DON'T MISS

UK police hope to catch suspects with facial recognition tech

It's obviously an incredibly controversial technology to start strapping to law enforcement. Stress levels are raised for a variety of reasons -- can technology ever tell us if the reason behind that stress is nefarious? The chances are slim. In which case, the system would only be suitable for doing blanket sweeps that would need follow-up investigations, leaving it open to plenty of abuse. SCMP's story focuses on the technology being suitable for picking out "suicide bombers". In reality, hopefully, such cases would be few and far between. It's the far between, when a system like this could be mass sweeping the public, that is of concern.

Peter Yuen from Cranfield University's Centre for Electronics Warfare in the UK has been investigating similar technology since 2008, with support from the Ministry of Defence's Counter Terrorism Centre. He combines skin temperature (visible using thermal imaging) with blood oxygenation (through hyperspectral imaging) and heart rate data to detect emotional stress in people. He told Wired.co.uk Tong was in fact one of his students from 2009-2012, and he believes the technology being presented is "not that mature".

"I must say that his work has serious issues that need further research to confirm the practical usefulness of the technology," Yuen told us. "We are in progress to patent this technology for security and law and order applications."

Yuen's application has been tested in 100 subjects and achieved around 80 percent accuracy when it comes to picking out stressed individuals. The team has attained 50 percent accuracy for quantifying the level of stress. Even his trial, however, has not been put to the test in a real world, outdoors environment -- only in the lab.

"The accuracies for measuring the heart rate and the blood oxygenation using imaging technique are not great: both are needed to get to within 5 percent error but we have only achieved 10-20 percent, which certainly will need more work!"

Yuen is only now able to discuss his finds -- the investigations were classified up until 2013. Given the lift on that restricted status, he and his team are now looking into applications in fields as disparate as biomedical, law and order, education, training and various other civil applications.

"We are hoping to team up with the Gloucester hospital, Leicester University and ISIS to pull through this  technique for civil applications in five years. We have been talking to a couple of investor firms but there is no formal commitment yet."

Of its potential for infringing on the public's privacy, Yuen comments: "This is a 'mood' detection technique and therefore can see people's mood whether they like it or not. True there will be some privacy intrusion and we will need to think of a method to restrict it.