Offline Text Recognition Using Hidden Markov Model

Abstract

In this paper, we have proposed a methodology of implementing a system model based on Hidden Markov Model (HMM) that can effectively recognize digital textual material. The idea behind the model relies on the ease of implementing HMMs to predict the succeeding character depending on the observable of the present. A similar concept is then applied as a whole for complete word detection and recognition. The model is termed as H and relies on heavily pre-processed images of digital textual data-set. The training phase depends heavily on the vocabulary fed to the system in image format and a series of textual characters, sentences and non-sentimental phrases in text format. Evaluation of the model is expressed in terms of the likelihood of occurrence of testing data. The evaluation result is maintained as the final criterion for the model’s ability to filter text from noisy text images. The applications of the project lie in the noise removal from text, clarification of text, scaling the model to operate on a huge amount of textual data and the scope of the project is limitless in image processing and natural language processing.