Date: 11/28/05 (Algorithms)    Keywords: no keywords

does anyone here know much about acoustic tracking algorithms? i'd like to put an implementation on a robot as part of a distributed surveillance project. ultimately, i'd like the following functionality:

• the robots automagically deploy themselves in some optimal manner (where optimal has yet to be defined.
• the robots go into a semi-sleep mode, powering down as much of their electronics as possible
• as the robots are in their semi-sleep mode, they have a set of n acoustic sensors and attached DSPs listening for unusual sounds. the sensors are arranged in some convenient pattern
• when a robot hears an unusual sound, it wakes up the other robots.
• the robots figure out who is nearest to the sound. that robot is dispatched to track the source of the sound. the remaining robots reconfigure themselves to maintain an optimal configuration.

what i'm hoping for is some suggestions on how to attack this problem. i know how to wake the robots up and compute the initial bearing and range guestimate. what i don't know how to do, however, is actively follow a sound. there are several papers (such as work done by rama chellapa at umd) on the more general subject of fusing acoustic and video sensing for tracking, but they are far more computationally demanding than what my platform will be able to handle. any suggestions on a reasonably simple approach to solving this will be greatly appreciated.

Source: http://www.livejournal.com/community/algorithms/66879.html