AMI Large vocabulary continuous speech recognizer
AMI Rozpoznávač spojité řeči s velkým slovníkem
required - licence fee
The AMI speech recognition system is being developed in cooperation of partners of EU-sponsored project AMI: University of Sheffield, IDIAP, University of Edinburgh, BUT and TNO Netherlands. It is based on Hidden Markov models. Brno group has contributed novel methods of acoustic modeling (discriminative training and linear transforms) to the recognizer. The system was successfully tested in NIST Meeting Transcription Evaluations in 2005.