HMM topology selection for on-line Thai handwriting recognition

Siriboon, Kritawan; Jirayusakul, A.; Kruatrachue, Boontee

doi:10.1109/cw.2002.1180872

Cited by 6 publications

(7 citation statements)

References 3 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…this approach, the comparison of the recognition results obtained from the proposed method multiple representations and the elastic matching is given. 7 As shown in Fig.6, the directional code for the rotation of Fig. 7…”

Section: -mentioning

confidence: 97%

“…The uncertain group consists of 225 315-noise and real signal that can't be distinguish. The real group 7 has high curvature enough to be certain as a real signal. given as 1 + 2/3 and 4+1/4, respectively.…”

Section: -mentioning

confidence: 99%

“…However a few methods for online Thai character have been proposed [7,8,9]. In this paper, a new method for unconstrained handwritten Thai character recognition is proposed.…”

mentioning

confidence: 99%

See 2 more Smart Citations

Online Unconstrained Handwritten Thai Character Recognition Using Multiple Representations

Bounnady

Kruatrachue

Matsuura

2008

2008 International Symposium on Communications and Information Technologies

View full text Add to dashboard Cite

This paper proposes an unconstrained handwritten counter clockwise direction as the main feature. Even Thai character recognition method using multiple representations.though this feature quite coincides with the way writers learnThe proposed method can recognize different handwritten Thai to write (Fig. 1), it may consist of noisy curves. Usually, the characters having small curve segments with clockwise and noise curve segment can be distinguished from the real curve counter clockwise directions, although the conventional method by the high amount of curvature of the real signal. But when such as the elastic matching method was difficult to recognize the written speed is fast, the curvature is reducing close to the them. As the experimental results, the recognition rate was 97.50% th e shown in fig. the curv is hard to and the recognition time (average) using the PC with 1.4 GHz noisy curve as shown in Fig. 1. These noisy curves are hard to Pentium 4 was 0.01 second. distinguish with normal noise reduction technique such as Fourier expansion or Fourier descriptor due to their similarity I. INTRODUCTIONin frequency and amplitude. In order to successfully use the curvature sequence in the recognition, one may have to Many methods for online character recognition have been identify real curvature from the noisy one.proposed [1]. However a few methods for online Thai character have been proposed [7,8,9]. In this paper, a new method for unconstrained handwritten Thai character recognition is proposed. The example of unconstraint handwritten character is shown in Fig. 1, where characters are written out of proportion and varied in amount of rotation Printed Handwritten (slant) and writing style. Most zone features [1] are less effecti * d . t stheo Mo pro onse some characters Fig. 1. Examples of unconstrained handwritten Thai character with skew effective clue to the out of proportion since some characters s and in proportion.will have big, small or omitted head. The proposed method will be compared with elastic matching which provides a very Dominance curvature good recognition rate and has been widely used since the early [2] and recently handwritten recognition research [3,4].The elastic matching can be used to perform matching the unknown writing with the prototype character and chosen the Distance A is smaller than distance B closest match as the recognized character. The elastic matching find the optimal alignment distance among all the n Fig. 2. The elastic matching of overall shape and missed dominant points of both matched characters by summation distance of curvature. all the matched point along the Viterbi path. This require 0(n2) computational cost proportional to the Viterbi matrix Test Prototype Test Prototype Test Prototype Test Prototype size [5]. Even though the Viterbi search reduce the ] 74 computational cost from exponential to O(n2), the cost still 0 = l | high to perform real time recognition, especially for a lot of i V i1 f J |j |m prototype characters. Another disadvantage of elastic matching...

show abstract

Section: -mentioning

confidence: 97%

Section: -mentioning

confidence: 99%

See 1 more Smart Citation

Online Unconstrained Handwritten Thai Character Recognition Using Multiple Representations

Bounnady

Kruatrachue

Matsuura

2008

2008 International Symposium on Communications and Information Technologies

View full text Add to dashboard Cite

show abstract

“…For Maximum Likelihood (ML) objective function, the topology with the maximum value of such likelihood is considered as the best topology [1]. However, ML function is prone to the over-fitting problem.…”

Section: Literature Reviewsmentioning

confidence: 99%

“…The first algorithm predefines rigidly the connectivity among the states [1]. By doing so, this algorithm estimates only the number of state and the number of Gaussian mixtures per state in some predefined lengths.…”

Section: Literature Reviewsmentioning

confidence: 99%

A Genetic Algorithm-aided Hidden Markov Model Topology Estimation for Phoneme Recognition of Thai Continuous Speech

Bhuriyakorn

Punyabukkana

Suchato

2008

2008 Ninth ACIS International Conference on Software Engineering, Artificial Intelligence, Networking, and Parallel/Distributed

View full text Add to dashboard Cite

The use of Hidden Markov Models (HMM) in many pattern recognition tasks is now very common. Like other pattern recognitions, most Automatic Speech Recognition systems rely on HMM acoustic models. In such systems, recognition performances are significantly affected by their topologies. In this paper, we propose an HMM topology estimation approach for Thai phoneme recognition tasks whose process is divided into 2 stages. First, a set of suitable topologies are constructed by combinations of different objective functions and topology generation methods. Second, a Genetic Algorithm is deployed as the topology selection algorithm which considers global fitness and selects the most suitable topology from the candidates proposed in the previous stage for each phoneme. As a result, the well-trained topology yields a maximum of 4.36% error reduction over predefined left-to-right models. The estimated topologies still work well when the topology estimation was performed on speech utterances whose recording environments differ from the ones recognized.

show abstract

State Machine Induction with Positive and Negative Training for Thai Character Recognition

Kruatrachue

Pantrakarn

Siriboon

2007

2007 International Conference on Communications, Circuits and Systems

View full text Add to dashboard Cite

HMM topology selection for on-line Thai handwriting recognition

Cited by 6 publications

References 3 publications

Online Unconstrained Handwritten Thai Character Recognition Using Multiple Representations

Online Unconstrained Handwritten Thai Character Recognition Using Multiple Representations

A Genetic Algorithm-aided Hidden Markov Model Topology Estimation for Phoneme Recognition of Thai Continuous Speech

State Machine Induction with Positive and Negative Training for Thai Character Recognition

Contact Info

Product

Resources

About