IEEE International Conference on Acoustics Speech and Signal Processing · 1993 · 19 citations · 4 references
ComputingComputer ScienceLinguisticsComputational LinguisticsMachine LearningPhoneticsConnected SpeechHidden Markov ModelsPattern RecognitionSpeech RecognitionAdult Speech RecognitionError Rate AnalysisError Rate MetricsMaximum Mutual InformationMutual Information EstimationSpeaker-independent Speech RecognitionSpeech Recognition AccuracyTraining
The authors describe developments by the speech research group at CRIM (Centre de Recherche Informatique de Montreal), in the field of speaker-independent connected digit recognition, using hidden Markov models (HMMs) trained with maximum mutual information estimation (MMIE). The experiments described were all performed on the complete adult portion of the TIDIGITS corpus. Techniques that made it possible to improve greatly the recognition rate are described. New results include a 0.28% word error rate and a 0.84% string error rate with two models per digit (one for male and one for female speakers) using context-dependent discrete HMMs.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">></ETX>
4