Automatic video understanding

A new system will detect and understand text in English and foreign language video to provide timely and actionable intelligence.


Cambridge, MA-based BBN Technologies has been awarded a contract valued at $0.9 million under the Video Analysis and Content Extraction (VACE) program sponsored by the Disruptive Technology Organization (formerly Advanced Research and Development Activity, or ARDA).


Under the contract, BBN will integrate several speech and language technologies in a single interface that will allow analysts to extract meaning from video files that contains text.


This videotext understanding system will incorporate multilingual videotext detection, tracking and recognition technology, and videotext understanding technology.


Video often contains videotext that is useful for analysts and decision makers. For example, graffiti sprayed on a wall can give analysts an indication of local political sentiment, text on signs may help them identify the location where a video was made, and captions on foreign news video can provide valuable information that is not included in the speech track.


The VACE program’s goal is to develop technologies that enable better use to be made of the large body of data available from these kinds of video sources, as well as those from meetings, conferences, and surveillance and reconnaissance, regardless of the language of the text.


“News, information, and intelligence come from so many different sources now, and being able to understand and distill them rapidly is critical,” said Tad Elmer, President and CEO, BBN Technologies. “The work we’re doing on the VACE program will provide advanced tools to find information for critical decision making quickly.”