Combining Structure and Parameter Adaptation of HMMs for Printed Text Recognition

Abstract

We present two algorithms that extend existing HMM parameter adaptation algorithms (MAP and MLLR) by adapting the HMM structure. This improvement relies on a smart combination of MAP and MLLR with a structure optimization procedure. Our algorithms are semi-supervised: to adapt a given HMM model on new data, they require little labeled data for parameter adaptation and a moderate amount of unlabeled data to estimate the criteria used for HMM structure optimization. Structure optimization is based on state splitting and state merging operations and proceeds so as to optimize either the likelihood or a heuristic criterion. Our algorithms are successfully applied to the recognition of printed characters by adapting the HMM character models of a polyfont printed text recognizer to new fonts. Our experiments involve a total of 1,120,000 real and 3,100,000 synthetic character images and concern a set of 89 HMM models. A comparison of our results with those of state-of-the-art adaptation algorithms (MAP and MLLR) shows a significant increase in the accuracy of character recognition.

Open-vocabulary recognition of machine-printed Arabic text using hidden Markov models

“…Ait-Mohand et al [9] recently presented an interesting study of mixed-font text recognition using HMMs. The main contribution of the study was related to HMM model length adaptation techniques that were integrated with HMM data adaptation techniques, such as maximum likelihood linear regression (MLLR) and maximum a posteriori (MAP) techniques.…”

“…Another approach for addressing mixed fonts was proposed by Ait-Mohand et al [9]. They proposed HMM adaptation techniques in which the adaptation was performed using the HMM data and the model length (number of states).…”

“…This approach proved to be more effective than the commonly employed approach for recognizing text using a recognizer that was trained on text samples from multiple fonts [7], [8]. Our approach overcomes the common limitations of other techniques, such as the need for labeled samples of the text images in the font to be recognized and the assumption of data isogeny, i.e., the text lines to be recognized are obtained from only one font at a time [9]. This paper is organized as follows: The next section presents related studies on printed text recognition, which primarily focus on printed Arabic text recognition.…”

Language Identification in Document Images

“…We use the LITIS OCR based on HMM with variable state number, described in [8]. Since the language is unknown during recognition, this OCR is a language free version working at the character level (without any language model nor dictionary).…”

“…The textual content of each line is decoded using Viterbi decoding without contextual resources as it is the case for standard recognizer (no dictionary, no language model used). A detailed description of the recognition engine is given in [8].…”