Citation:
Rabiner, Lawrence R. "A tutorial on hidden Markov models and selected applications in speech recognition." Proceedings of the IEEE 77.2 (1989): 257-286.
Summary:
This paper is a landmark article on Hidden Markov Models, and is cited by most scientists who use HMMs, and it presents a comprehensive review on HMMs an their applications. The core aspects include the construction of a basic Marov model, additional parameters required to construct a Hidden Markov Model, along with forward and backward chaining algorithms. The paper also deals with the three main problems that are solved using HMMs.
Discussion:
The paper begin with explaining probability primitives and basic Markov models. It then discusses the characterization of the basic parameters required to build a HMM, i..e No. of states, No. of observation, apriori probability, transitional probability and emission probability.
The paper begin with explaining probability primitives and basic Markov models. It then discusses the characterization of the basic parameters required to build a HMM, i..e No. of states, No. of observation, apriori probability, transitional probability and emission probability.
The 3 main uses of the HMM are :
1. Given an observation sequence and a model, what is the probability of that observation sequence? (likelihood of those events occurring)
2. Given a set of observations and a model, what is the most likely sequence of states under which the observations hold?
3. How can the model parameters be adjusted to increase likelihood of a desired observation.
The paper the delves into the application of the Viterbi algorithm for forward and backward chaining. The math details are avoided in this post.
No comments:
Post a Comment