ETRW on Speech Processing in Adverse Conditions Cannes-Mandelieu, France November 10-13, 1992

Isolated Word Recognition In Noisy Environments

J. Ortega, J. Alvarez, E. López, L. Hernández

ETSI Telecomunicación, Universidad Politécnica de Madrid, Spain

In this paper, the problem of isolated word recognition in noisy environments is presented. First, the problem is solved combining the information provided by Vector Quantizers (VQs) and Hidden Markov Models (HMMs) adapted to different noise situations. And after, introducing spectral substraction and mapping between vector quantizers with and without noise as a computationally efficient technique for robust word spotting speech recognition. The results provided show high recognition rates of speech in noise, solving the classical drawbacks of previous systems.

Full Paper

Bibliographic reference.  Ortega, J. / Alvarez, J. / López, E. / Hernández, L. (1992): "Isolated word recognition in noisy environments", In SPAC-1992, 215-218.