Sixth European Conference on Speech Communication and Technology
(EUROSPEECH'99)

Budapest, Hungary
September 5-9, 1999

Reviving Discrete HMMs: the Myth About the Superiority of Continuous HMMs

Vassilis Digalakis, Stavros Tsakalidis, Leonardo Neumeyer (1)

Dept. of Electronics and Computer Engineering, Technical University of Crete, Hania, Greece
(1) SRI International, Menlo Park, CA, USA

Despite what is generally believed, we have recently shown that discrete-distribution HMMs can outperform continuous-density HMMs at significantly faster decoding speeds. Recognition performance and decoding speed of the discrete HMMs are improved by using product-code Vector Quantization (VQ) and mixtures of discrete distributions. In this paper, we present efficient training and decoding algorithms for the discrete-mixture HMMs (DMHMMs). Our experimental results show that the high-level of recognition accuracy of continuous mixture-density HMMs (CDHMMs) can be maintained at significantly faster decoding speeds.


Full Paper (PDF)   Gnu-Zipped Postscript

Bibliographic reference.  Digalakis, Vassilis / Tsakalidis, Stavros / Neumeyer, Leonardo (1999): "Reviving discrete HMMs: the myth about the superiority of continuous HMMs", In EUROSPEECH'99, 2463-2466.