Mixture splitting technique and temporal control in a HMM-based recognition system
Claude Montacié, M.-J. Caratay, Clément Barras · 2002
Studies various techniques to improve the performance and to reduce the computation cost and the required memory of a speech recognition system based on hidden Markov models (HMMs). For the efficiency of the system, we first study the optimization of the number of HMM parameters according to training data. We experiment with the temporal control of the phonetic transitions on a lexical decoding task with a significant 5% improvement. Finally, a preliminary method for dynamically selecting a sublexicon is studied in order to reduce the lexical decoding cost.