A massive amount of speech and video content is stored on servers run by video-sharing sites, broadcast stations, call centers, e-learning host services, and other such organizations.
Spoken Term Detection (STD), which the enables search for a keyword on the speech/video database is one of the essential technologies for effectively utilizing such information.
Kouichi Katsurada and Tsuneo Nitta at the Toyohashi University of Technology (Toyohashi Tech) have developed a rapid STD method that outputs the most reliable search results in a few milliseconds. The search speed is at least 10 times faster than existing methods.
They researchers use a suffix array as a data structure and apply dynamic time warping on it to realize a high-speed ambiguous search. In addition, they embed keyword division and iterative lengthening search into the search algorithm. These techniques make it possible to detect a keyword on the speech/video database in a very short time.
This technique can be applied to search for a scene from movies, TV/radio programs, lecture videos, as well as being utilized at call centers for finding customer services information from the database.
The Toyohashi team is looking for partners to commercialize this powerful technology.
----------------------------------------------------------------------
Related publication:-
K. Katsurada, S. Teshima, and T. Nitta, Fast Keyword Detection Using Suffix Array, InterSpeech2009, 2147–2150, (2009).
----------------------------------------------------------------------
Further information
Ms. Junko Sugaya and Mr. Masashi Yamaguchi
International Affairs Division
TEL: (+81) 0532-44-2042; FAX: (+81)0532-44-6557
E-mail:[email protected]
About Toyohashi University of Technology
Founded in 1976, Toyohashi University of Technology is a vibrant modern institute with research activities reflecting the modern era of advanced electronics, engineering, and life sciences.
