A novel divergence measure for hidden Markov models

In this paper, a novel divergence measure for hidden Markov models (HMMs) is introduced. The widely used distance measure between two HMMs is the Kullback-Leibler divergence (KLD). The Monte-Carlo method is usually applied to calculate the KLD, whose computational complexity is prohibitive in practical applications. Numerical examples show that the proposed divergence measure closely approximates the KLD with a saving of hundreds of times in computational complexity.