M. Wester, Judith M. Kessens, H. Strik (2002)
Automatic Speech Recognition for MUMIS
Deliverable D-5.1-T20 of MUMIS (Multimedia Indexing and Searching Environment),
Project ref. no. IST-1999-10651, 28 February 2002, 26 pages.
Security (distribution level) : Project internal
This report describes the status of the automatic speech recognition
tools developed for MUMIS data. Two languages have been studied so
far, Dutch and English. The data comprises the commentaries that
accompany TV broadcasts of football matches (Euro-2000). The speech
can be described as spontaneous. The recordings are extremely noisy as
they contain a great deal of background noise in the form of noise
produced by the crowd, the referee etc. This noise greatly increases
the difficulty of the speech recognition task.