Hybrid Model using Subspace Distribution Clustering Hidden Markov Models and Semi-Continuous Hidden Markov Models for Embedded Speech Recognizers

摘要

Today's state-of-the-art speech recognition systems typically use continuous density hidden Markov models with mixture of Gaussian distributions. Such speech recognition systems have problems; they require too much memory to run, and are too slow for large vocabulary applications. Two approaches are proposed for the design of compact acoustic models, namely, subspace distribution clustering hidden Markov models and semi-continuous hidden Markov models. However, these models require also large memory to acquire high recognition accuracy. In this paper, we propose a new hybrid model using subspace distribution clustering hidden Markov model and semi-continuous hidden Markov model with the aim of achieving much more compact acoustic models.