Automatic word recognition based on second-order hidden Markov models

Abstract

We propose an extension of the Viterbi algorithm that makes second-order hidden Markov models computationally efficient. A comparative study between first-order (HMM1's) and second-order Markov models (HMM2's) is carried out. Experimental results show that HMM2's provide a better state occupancy modeling and, alone, have performances comparable with HMM1's plus postprocessing.

Using Markov Models to Mine Temporal and Spatial Data

“…We have shown in (Mari et al, 1997) that an HMM2 can be estimated following the same way. The estimation is an iterative process starting with an initial model and a corpus of sequences of observations that the HMM2 must fit even when the insertions, deletions and substitutions of observations occur in the sequences.…”

Double Layer Architectures for Automatic Speech Recognition Using HMM

“…This assumption is motivated by the existence of efficient, tractable algorithms for model estimation and recognition. To overcome the drawbacks of regular HMMs regarding segment duration modelling and trajectory (frame correlation) modelling, some authors have proposed a new class of models in which the underlying state sequence is a second-order Markov chain (HMM2) (Mari et al, 1997). These models show better state occupancy modelling, at the cost of higher computational complexity.…”

Hidden Markov Models for Real-Time Estimation of Corn Progress Stages Using MODIS and Meteorological Data

“…High-order stage transition dependency would result in good modeling of stage duration [42]. Mari et al [43] carried out a comparative study between first-and second-order HMMs on automatic word recognition. Seifert et al [44,45] utilized high-order HMM to improve modeling of spatial dependencies between chromosomal regions.…”