WASHINGTON — The U.S. National Geospatial-Intelligence Agency has launched a cash prize competition to find new ways to locate where audio and video files were recorded.

As the primary source of geospatial intelligence, the NGA prides itself in its ability to exploit imagery in support of the intelligence community and the military. The agency’s activities cover a wide range, from using satellites to map the Arctic to using artificial intelligence for automatic detection of objects in overhead imagery.

But at least one task remains elusive: Geolocating the source of video and audio recordings.

That’s why the NGA is launching the Soundscapes Competition, which will award cash prizes to entrees who can find new methods for identifying a recording’s location based on ambient sound in the background. Participants will be asked to come up with methods of “identifying, analyzing, and modelling these sound and acoustic scene indicators to uniquely classify audio recordings as originating in one of nine cities," according to the competition website.

The competition was launched Oct. 20 and runs through Nov. 27. Would-be solvers will receive short non-speech audio files, with each file recorded in one of nine cities. Using provided training data, participants will need to create an algorithm that can determine the undisclosed city in which each recording was made.

Participants will first describe their proposed method in a whitepaper. Then they will label each test audio file with the city in which it was made, explaining how confident they are in their determinations.

Over the five-week period, would-be solvers will be able to submit and test their solutions while seeing their ranking on a group leaderboard.

The agency will award up to eight cash prizes, including $27,000 for the top participant. The best solvers will be invited to present a paper on their efforts at a workshop in 2021.

More information can be found on the competition website.

Nathan Strout covers space, unmanned and intelligence systems for C4ISRNET.

More In Intel/GEOINT