Offline Handwriting Recognition Using Invariant Moments and Curve Let Transform with Combined SVM-HMM Classifier
| Autor: | A. Khatri, B. Nagaria, P. Kumawat |
|---|---|
| Rok vydání: | 2013 |
| Předmět: |
Biometrics
Computer science business.industry Speech recognition Feature vector ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION Computer Science::Computation and Language (Computational Linguistics and Natural Language and Speech Processing) Pattern recognition Facial recognition system Support vector machine ComputingMethodologies_PATTERNRECOGNITION Computer Science::Sound Handwriting Handwriting recognition Computer Science::Computer Vision and Pattern Recognition Artificial intelligence Hidden Markov model business Classifier (UML) |
| Zdroj: | 2013 International Conference on Communication Systems and Network Technologies. |
| DOI: | 10.1109/csnt.2013.40 |
| Popis: | Offline Handwriting recognition is considered as important research field in the filed of forensic and biometric applications. It finds significance in fields like graphology which exploits the physiological behavior of the person based on the handwriting. There are several algorithms for Handwriting recognition. However none of the techniques is yet proved to be satisfactory especially for large number of classes. This is due to the fact that handwriting is a pattern which differs from instance to instance of the same writer. Hence HMM is most preferred technique in this domain. It is due to the fact the HMM produces good result for large number of statistical patterns. However, the performance of the system depends entirely on the feature vectors. Unlike the cases of usual patter recognition like face recognition, a user's training and test sample may vary. Hence recognition of the same is tough. Therefore in this work we propose a novel technique for offline handwriting recognition based on Invariant Moments and curve let transform. Curve let transform and Invariant moments are used predominantly for character recognition problem and hence are more suitable for the work. Further we compare the performance of HMM based technique with SVM based technique and found that for some patterns, the efficiency of SVM classifier is better than that of HMM and performance of HMM is better than HMM in some cases. Hence we develop a combined classifier and prove that the system performs better than both independent HMM and SVM classifier. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|