21-08-2012, 10:52 AM
A Hybrid HMM/ANN System for Recognizing Unconstrained Offline Handwritten Text Lines
[attachment=32244]
ABSTRACT
Recently, several papers have proposed pseudo dynamic methods for automatic handwritten signature verification. Each of these papers uses texture measures of the gray level signature strokes. This paper explores the usefulness of local binary pattern (LBP) and local directional pattern (LDP) texture measures to discriminate off-line signatures. A comparison between several texture normalizations is made so as to look for reducing pen dependence. The experiments conducted with MCYT off-line and GPDS960Graysignature corpuses show that LDPs are more useful than LBPs for automatic verification of static signatures. Additionally, the results show that the LDP codes of the contour are more discriminating than the LDPs of the stroke interior, although their combination at score level improves the overall scheme performance. The results are obtained by modeling the signatures with a Support Vector Machine (SVM) trained with genuine samples and random forgeries, while random and simulated forgeries have been used for testing it.
INTRODUCTION
OFFLINE handwritten text recognition is one of the most active areas of research in computer science and it is inherently difficult because of the high variability of writing styles. High recognition rates are achieved in character recognition and isolated word recognition, but we are still far from achieving high-performance recognition systems for unconstrained offline handwritten texts [1], [2], [3], [4], [5], [6], [7]. Automatic handwriting recognition systems normally include several preprocessing steps to reduce variation in the handwritten texts as much as possible and, at the same time, to preserve information that is relevant for recognition. There is no general solution to preprocessing of offline handwritten text lines, but it typically relies on slope and slant correction and normalization of the size of the characters. With the slope correction, the handwritten word is horizontally rotated such that the lower baseline is aligned to the horizontal axis of the image. Slant is the clockwise angle between the vertical direction and the direction of the vertical text strokes. Slant correction transforms the word into an upright position. Ideally, the removal of slope and slant results in a word image that is independent of these factors. Finally, size normalization tries to make the system invariant to the character size and to reduce the empty background areas caused by the ascenders and descenders of some letters.
EXISTING SYSTEM
Our existing system handwritten character recognition using Modified Direction Feature (MDF), it is nothing but a system which recognize a hand written character Modified Direction Feature (MDF) generated encouraging results, reaching an accuracy of 81.58%.
PROPOSED SYSTEM
Our proposed system is Off-line Signature Verification using the Enhanced Modified Direction Feature and Neural-based Classification in which we are using MDF with signature images. Specifically, a number of features have been combined with MDF, to capture and investigated various structural and geometric properties of the signatures to perform verification or identification of a signature, several steps must be performed. After preprocessing all signatures from the database by converting them to portable bitmap (PBM) format, their boundaries are extracted to facilitate the extraction of features using MDF .Verification experiments are performed with classifiers We are using Radial Basis Function (RBF) which is a classifier which gives an accuracy level of 91.21%
TECHNICAL FEASIBILITY
This study is carried out to check the technical feasibility, that is, the technical requirements of the system. Any system developed must not have a high demand on the available technical resources. This will lead to high demands on the available technical resources. This will lead to high demands being placed on the client. The developed system must have a modest requirement, as only minimal or null changes are required for implementing this system.
ECONOMICAL FEASIBILITY
This study is carried out to check the economic impact that the system will have on the organization. The amount of fund that the company can pour into the research and development of the system is limited. The expenditures must be justified. Thus the developed system as well within the budget and this was achieved because most of the technologies used are freely available. Only the customized products had to be purchased.
OPERATIONAL FEASIBILITY
The aspect of study is to check the level of acceptance of the system by the user. This includes the process of training the user to use the system efficiently. The user must not feel threatened by the system, instead must accept it as a necessity. The level of acceptance by the users solely depends on the methods that are employed to educate the user about the system and to make him familiar with it. His level of confidence must be raised so that he is also able to make some constructive criticism, which is welcomed, as he is the final user of the system.
ANALYSIS MODEL
The model that is basically being followed is the WATER FALL MODEL, which states that the phases are organized in a linear order. First of all the feasibility study is done. Once that part is over the requirement analysis and project planning begins. If system exists one and modification and addition of new module is needed, analysis of present system can be used as basic model.
The design starts after the requirement analysis is complete and the coding begins after the design is complete. Once the programming is completed, the testing is done. In this model the sequence of activities performed in a software development project are: -
Requirement Analysis, Project Planning, System design, Detail design, Coding, Unit testing, System integration & testing
CONCLUSION
In this paper, we have presented a hybrid HMM/ANN system for recognizing unconstrained offline handwritten text lines. The key features of the recognition system are the novel approach to preprocessing and recognition, which are both based on ANNs. The preprocessing is based on
using MLPs: . to clean and enhance the images, . to automatically classify local extrema in order to correct the slope and to normalize the size of the text lines images, and . to perform a nonuniform slant correction. The recognition is based on hybrid optical HMM/ANN models, where an MLP is used to estimate the emission probabilities. The main property of ANNs which is useful for preprocessing tasks is their ability to learn complex nonlinear input-output relationships from examples. Used for regression, an MLP can learn the appropriate filter from examples. We have exploited this property to clean and enhance the text images. Used for classification, MLPs can be used to determine the membership of interest points from the image to the reference lines, which is useful for slope correction and size normalization, and to locally detect slant in a text image. This preprocessing behaved favorably when compared to other preprocessing techniques. We tested our HMM and HMM/ANNsystems, performing the same experiments that those presented here, but by using more classical techniques to correct slope, slant, and size normalization [10]. We obtained a 54.3 percent and 29.8 percent test WER, respectively, which represent a percentual decrease of 29 percent and 25 percent when compared to the test results from Table 6.
[attachment=32244]
ABSTRACT
Recently, several papers have proposed pseudo dynamic methods for automatic handwritten signature verification. Each of these papers uses texture measures of the gray level signature strokes. This paper explores the usefulness of local binary pattern (LBP) and local directional pattern (LDP) texture measures to discriminate off-line signatures. A comparison between several texture normalizations is made so as to look for reducing pen dependence. The experiments conducted with MCYT off-line and GPDS960Graysignature corpuses show that LDPs are more useful than LBPs for automatic verification of static signatures. Additionally, the results show that the LDP codes of the contour are more discriminating than the LDPs of the stroke interior, although their combination at score level improves the overall scheme performance. The results are obtained by modeling the signatures with a Support Vector Machine (SVM) trained with genuine samples and random forgeries, while random and simulated forgeries have been used for testing it.
INTRODUCTION
OFFLINE handwritten text recognition is one of the most active areas of research in computer science and it is inherently difficult because of the high variability of writing styles. High recognition rates are achieved in character recognition and isolated word recognition, but we are still far from achieving high-performance recognition systems for unconstrained offline handwritten texts [1], [2], [3], [4], [5], [6], [7]. Automatic handwriting recognition systems normally include several preprocessing steps to reduce variation in the handwritten texts as much as possible and, at the same time, to preserve information that is relevant for recognition. There is no general solution to preprocessing of offline handwritten text lines, but it typically relies on slope and slant correction and normalization of the size of the characters. With the slope correction, the handwritten word is horizontally rotated such that the lower baseline is aligned to the horizontal axis of the image. Slant is the clockwise angle between the vertical direction and the direction of the vertical text strokes. Slant correction transforms the word into an upright position. Ideally, the removal of slope and slant results in a word image that is independent of these factors. Finally, size normalization tries to make the system invariant to the character size and to reduce the empty background areas caused by the ascenders and descenders of some letters.
EXISTING SYSTEM
Our existing system handwritten character recognition using Modified Direction Feature (MDF), it is nothing but a system which recognize a hand written character Modified Direction Feature (MDF) generated encouraging results, reaching an accuracy of 81.58%.
PROPOSED SYSTEM
Our proposed system is Off-line Signature Verification using the Enhanced Modified Direction Feature and Neural-based Classification in which we are using MDF with signature images. Specifically, a number of features have been combined with MDF, to capture and investigated various structural and geometric properties of the signatures to perform verification or identification of a signature, several steps must be performed. After preprocessing all signatures from the database by converting them to portable bitmap (PBM) format, their boundaries are extracted to facilitate the extraction of features using MDF .Verification experiments are performed with classifiers We are using Radial Basis Function (RBF) which is a classifier which gives an accuracy level of 91.21%
TECHNICAL FEASIBILITY
This study is carried out to check the technical feasibility, that is, the technical requirements of the system. Any system developed must not have a high demand on the available technical resources. This will lead to high demands on the available technical resources. This will lead to high demands being placed on the client. The developed system must have a modest requirement, as only minimal or null changes are required for implementing this system.
ECONOMICAL FEASIBILITY
This study is carried out to check the economic impact that the system will have on the organization. The amount of fund that the company can pour into the research and development of the system is limited. The expenditures must be justified. Thus the developed system as well within the budget and this was achieved because most of the technologies used are freely available. Only the customized products had to be purchased.
OPERATIONAL FEASIBILITY
The aspect of study is to check the level of acceptance of the system by the user. This includes the process of training the user to use the system efficiently. The user must not feel threatened by the system, instead must accept it as a necessity. The level of acceptance by the users solely depends on the methods that are employed to educate the user about the system and to make him familiar with it. His level of confidence must be raised so that he is also able to make some constructive criticism, which is welcomed, as he is the final user of the system.
ANALYSIS MODEL
The model that is basically being followed is the WATER FALL MODEL, which states that the phases are organized in a linear order. First of all the feasibility study is done. Once that part is over the requirement analysis and project planning begins. If system exists one and modification and addition of new module is needed, analysis of present system can be used as basic model.
The design starts after the requirement analysis is complete and the coding begins after the design is complete. Once the programming is completed, the testing is done. In this model the sequence of activities performed in a software development project are: -
Requirement Analysis, Project Planning, System design, Detail design, Coding, Unit testing, System integration & testing
CONCLUSION
In this paper, we have presented a hybrid HMM/ANN system for recognizing unconstrained offline handwritten text lines. The key features of the recognition system are the novel approach to preprocessing and recognition, which are both based on ANNs. The preprocessing is based on
using MLPs: . to clean and enhance the images, . to automatically classify local extrema in order to correct the slope and to normalize the size of the text lines images, and . to perform a nonuniform slant correction. The recognition is based on hybrid optical HMM/ANN models, where an MLP is used to estimate the emission probabilities. The main property of ANNs which is useful for preprocessing tasks is their ability to learn complex nonlinear input-output relationships from examples. Used for regression, an MLP can learn the appropriate filter from examples. We have exploited this property to clean and enhance the text images. Used for classification, MLPs can be used to determine the membership of interest points from the image to the reference lines, which is useful for slope correction and size normalization, and to locally detect slant in a text image. This preprocessing behaved favorably when compared to other preprocessing techniques. We tested our HMM and HMM/ANNsystems, performing the same experiments that those presented here, but by using more classical techniques to correct slope, slant, and size normalization [10]. We obtained a 54.3 percent and 29.8 percent test WER, respectively, which represent a percentual decrease of 29 percent and 25 percent when compared to the test results from Table 6.