c|net reports (along with many other sources) on a company with a very appropo name, Sound Intelligence, that has introduced a CCTV system product that includes an audio-based automatic aggression detector called SIgard. One or more microphones are co-located with a CCTV camera. A software-based algorithm monitors the incoming audio stream and "detects" aggressive human speech. When the system makes a detection, it alerts the CCTV operator staff to attract their attention to that particular camera and location. Only at that point may the CCTV staff monitor the audio for that camera and not before, according to the company.
The company is based in the Netherlands and has deployed systems there. They are now entering the market in the UK.
This technology was briefly mentioned in an interesting debate on the BBC's Moral Maze radio program this week. The topic was the "surveillance society" and the program is well worth a listen for those interested in the general topic. I should note that I detected quite a few inaccuracies in the debate, so don't take everything in it at face value.