We are going to refer to the problem of speech recognition under noisy conditions, with medium size vocabularies of isolated words. We describe a method for feature extraction and the scores obtained with a HMM based recognizer. This method uses a combination of others two methods: the Short-time Modified Coherence and the spectral subtraction.
Bibliographic reference. Gomez-Mena, Juan / Santos-Suarez, J. / Garcia-Gomez, Ramón (1991): "A robust feature extraction method for automatic speech recognition in noisy environments", In EUROSPEECH-1991, 1383-1386.