FROM HMMS TO SEGMENT MODELS - A UNIFIED VIEW OF STOCHASTIC MODELING FOR SPEECH RECOGNITION

In recent years, many alternative models have been proposed to address some of the shortcomings of the hidden Markov model (HMM), which is c urrently the most popular approach to speech recognition, In particula r, a variety of models that could be broadly classified as segment mod els have been described for representing a variable-length sequence of observation vectors in speech recognition applications, Since there a re many aspects in common between these approaches, including the gene ral recognition and training problems, it is useful to consider them i n a unified framework. Thus, the goal of this paper will he to describ e a general stochastic model that encompasses most of the models propo sed in the literature, pointing out similarities of the models in term s of correlation and parameter tying assumptions, and drawing analogie s between segment models and HMM's. In addition, we summarize experime ntal results assessing different modeling assumptions and point out re maining open questions.