According tot he New Scientist Dutch police use surveillance cameras with microphones to estimate anger levels and anticipate violence.
The NS report says:
“Microphones attached to the cameras feed the sound signals to software that can detect voices that are aggressive in tone. “Aggressive people tend to tense their larynx, and the sound made by their vocal cords is distorted,” says Peter van Hengel of developer Sound Intelligence, a spin-off of the University of Groningen. This means that high frequency vowel sounds span a broader frequency range. “A truly aggressive voice is very hard to imitate,” he says. In a trial earlier this year, police made three arrests after being alerted by the system.”
Compare this earlier post about attempts to estimate behaviour from movement.
In each case the system is used to scan mounds of video footage and flag interesting sections for police to observe. Looked at from one level, they are just a search engine. However the accuracy of their algorithms has to be critical.
Sound Intelligences own press releases suggest that the system is used to switch cameras on when they spot aggressive behaviour – ie the cameras can then remain off – with surveillance footage ungathered – at other times. If so, thats a good idea, though I cant imagine the British authorities missing any opportunity to film us all.