A GENERAL METHOD FOR COMBINING ACOUSTIC FEATURES IN AN AUTOMATIC SPEECH RECOGNITION SYSTEM

A general method for the use of different types of fea- tures in Automatic Speech Recognition (ASR) systems is presented. A gaussian mixture model (GMM) is ob- tained in a reference acoustic space. A specific fea- ture combination or selection is associated to each gaus- sian of the mixture and used for computing symbol pos- terior probabilities. Symbols can refer to phonemes, phonemes in context or states of a Hidden Markov Model (HMM). Experimental results are presented of applications to phoneme and word rescoring after ver- ification. Two corpora were used, one with small vocab- ularies in Italian and Spanish and one with very large vocabulary in French.