Hidden Markov Model Representation Using Probabilistic Neural Network

Abstract

Hidden Markov Model (HMM) is a statistical network used in knowledge representation in many applications. In the same system/application, we could have several HMMs that might have similar or different visible and invisible attributes. In this paper, a theoretical framework proposal based on Probabilistic Neural Network (PNN) concept to represent all HMMs in a given system in one structure is presented. This representation framework will help in making the system provides results for cases that are not well formulated or provided in any of the given set of HMMs. The general idea of PNN has been adopted in this research to represent HMMs as patterns but the computation and representation are different. Our PNN will have at least two layers. The first layer is input layer, the second is the pattern and output layer, but we might have more than one pattern and output layer and the third layer is the sum and output layer. The proposed approach has been applied on three cases, one HMM, hierarchical HMM and independent related HMMs.