ROBUST PARAMETERS FOR NOISY SPEECH RECOGNITION

A method of automatic processing of noise-affected speech captures and digitizes speech in the form of at least one digitised signal and extracts several time-based sequences or frames corresponding to the signal, by means of an extraction system. Each frame is decomposed by means of an analysis system into at least two different frequency bands so as to obtain at least two first vectors of representative parameters for each frame, one for each frequency band. The method converts, by means of converter systems, the first vectors of representative parameters into second vectors of parameters substantially insensitive to noise, wherein each converter system (50) is associated with one frequency band and converts the first vector of representative parameters associated with the same frequency band, and wherein a learning of the converter systems is achieved on the basis of a learning corpus which corresponds to a corpus of speech contaminated by noise.