Label-1

Label-2

Label-3

āļāļēāļĢāļĢู้āļˆāļģāļĨāļēāļĒāļĄืāļ­āđ€āļ‚ีāļĒāļ™āļ āļēāļĐāļēāđ„āļ—āļĒāļ”้āļ§āļĒāđ‚āļ„āļĢāļ‡āļ‚่āļēāļĒāļ›āļĢāļ°āļŠāļēāļ—āđ€āļ—ีāļĒāļĄ
Thai Handwritten Character Recognition by Artificial Neural Networks


āļāļēāļāļˆāļ™āļē āđ€āļĢื่āļ­āļ‡āļ˜āļ™āļēāļ™ุāļĢัāļāļĐ์
āļ“ัāļāļ˜ิāļ”āļē āļĨีāļŠāļĄ
āđ‚āļ­āļŽāļēāļĢิāļ āļŠุāļĢิāļ™āļ•๊āļ°

āļšāļ—āļ„ัāļ”āļĒ่āļ­
āļāļēāļĢāļĢู้āļˆāļģāļĨāļēāļĒāļĄืāļ­āđ€āļ‚ีāļĒāļ™āļ āļēāļĐāļēāđ„āļ—āļĒāļ”้āļ§āļĒāđ‚āļ„āļĢāļ‡āļ‚่āļēāļĒāļ›āļĢāļ°āļŠāļēāļ—āđ€āļ—ีāļĒāļĄ āđ„āļ”้āđ€āļ็āļšāļĢāļ§āļšāļĢāļ§āļĄāļ‚้āļ­āļĄูāļĨāļ•ัāļ§āļ­ัāļāļĐāļĢāļĨāļēāļĒāļĄืāļ­āđ€āļ‚ีāļĒāļ™āļ āļēāļĐāļēāđ„āļ—āļĒāļˆāļģāļ™āļ§āļ™ 100 āļŠุāļ”āļ•่āļ­āļ•ัāļ§āļ­ัāļāļĐāļĢ āļˆāļēāļāļœู้āđ€āļ‚ีāļĒāļ™āļˆāļģāļ™āļ§āļ™ 10 āļ„āļ™ āđ€āļžื่āļ­āļ™āļģāļĄāļēāđ€āļ‚้āļēāļŠู่āļāļĢāļ°āļšāļ§āļ™āļāļēāļĢāļ—āļēāļ‡āļāļēāļĢāļ§ิāļˆัāļĒ āļ›āļĢāļ°āļāļ­āļšāļ”้āļ§āļĒ āļāļēāļĢāļ›āļĢāļ°āļĄāļ§āļĨāļœāļĨāļ āļēāļžāđ€āļšื้āļ­āļ‡āļ•้āļ™ āļāļēāļĢāļŦāļēāļ„ุāļ“āļĨัāļāļĐāļ“āļ°āļžิāđ€āļĻāļĐāļ‚āļ­āļ‡āļĢูāļ›āļ āļēāļžāļĨāļēāļĒāļĄืāļ­āđ€āļ‚ีāļĒāļ™āļ āļēāļĐāļēāđ„āļ—āļĒ āđāļĨāļ°āļāļēāļĢāļĢู้āļˆāļģāļ”้āļ§āļĒāđ‚āļ„āļĢāļ‡āļ‚่āļēāļĒāļ›āļĢāļ°āļŠāļēāļ—āđ€āļ—ีāļĒāļĄāđāļšāļšāđāļžāļĢ่āļĒ้āļ­āļ™āļāļĨัāļš āđ‚āļ”āļĒāļ—ี่āļāļĢāļ°āļšāļ§āļ™āļāļēāļĢāļŦāļēāļ„ุāļ“āļĨัāļāļĐāļ“āļ°āļžิāđ€āļĻāļĐāđ€āļ›็āļ™āļāļĢāļ°āļšāļ§āļ™āļāļēāļĢāļ—ี่āļŠāļģāļ„ัāļ āđ€āļ™ื่āļ­āļ‡āļˆāļēāļāļŠ่āļ§āļĒāđ€āļžิ่āļĄāļ›āļĢāļ°āļŠิāļ—āļ˜ิāļ āļēāļžāđƒāļ™āļāļēāļĢāļĢู้āļˆāļģ āļ‡āļēāļ™āļ§ิāļˆัāļĒāļ‰āļšัāļšāļ™ี้āđ„āļ”้āļ™āļģāđ€āļŠāļ™āļ­āļ„ุāļ“āļĨัāļāļĐāļ“āļ°āļžิāđ€āļĻāļĐ 7 āļ§ิāļ˜ีāļ„ืāļ­ āļāļēāļĢāļŦāļēāļ„āļ§āļēāļĄāļŦāļ™āļēāđāļ™่āļ™ āļāļēāļĢāļŦāļēāļˆุāļ”āļŠิ้āļ™āļŠุāļ”āļ‚āļ­āļ‡āļ•ัāļ§āļ­ัāļāļĐāļĢ āļāļēāļĢāļŦāļēāļ•āļģāđāļŦāļ™่āļ‡āļŦัāļ§āļ‚āļ­āļ‡āļ•ัāļ§āļ­ัāļāļĐāļĢ āļāļēāļĢāļŦāļēāļĢāļŦัāļŠāļĨูāļāđ‚āļ‹่ āļāļēāļĢāļŦāļēāđ€āļŠ้āļ™āļŠāļĄāļĄุāļ•ิāđƒāļ™āđāļ™āļ§āļ™āļ­āļ™āđāļĨāļ°āđāļ™āļ§āļ•ั้āļ‡ āļāļēāļĢāļŦāļēāļ—ิāļĻāļ—āļēāļ‡ āđāļĨāļ°āļāļēāļĢāļŠāđāļāļ™āđƒāļ™āđāļ™āļ§āļ™āļ­āļ™āđāļĨāļ°āđāļ™āļ§āļ•ั้āļ‡ āļ„ุāļ“āļĨัāļāļĐāļ“āļ°āļžิāđ€āļĻāļĐāļ—ั้āļ‡ 7 āļ§ิāļ˜ีāļ—ี่āđƒāļŠ้āļัāļšāļ‚้āļ­āļĄูāļĨāļ•ัāļ§āļ­ัāļāļĐāļĢāļĨāļēāļĒāļĄืāļ­āđ€āļ‚ีāļĒāļ™āļ āļēāļĐāļēāđ„āļ—āļĒāđƒāļ™āļ‡āļēāļ™āļ§ิāļˆัāļĒāļ™ี้āļ›āļĢāļ°āļāļ­āļšāļ”้āļ§āļĒ 235 āļ„ุāļ“āļĨัāļāļĐāļ“āļ° āļŠ่āļ‡āļœāļĨāđƒāļŦ้
āļ›āļĢāļ°āļŠิāļ—āļ˜ิāļ āļēāļžāļāļēāļĢāļĢู้āļˆāļģāļĨāļēāļĒāļĄืāļ­āđ€āļ‚ีāļĒāļ™āļ āļēāļĐāļēāđ„āļ—āļĒāļ”้āļ§āļĒāđ‚āļ„āļĢāļ‡āļ‚่āļēāļĒāļ›āļĢāļ°āļŠāļēāļ—āđ€āļ—ีāļĒāļĄāļ–ูāļāļ•้āļ­āļ‡āļ„ิāļ”āđ€āļ›็āļ™āļĢ้āļ­āļĒāļĨāļ° 87

āļ„āļģāļŠāļģāļ„ัāļ: āļāļēāļĢāļĢู้āļˆāļģāļĢูāļ›āđāļšāļš, āļāļēāļĢāļĢู้āļˆāļģāļ•ัāļ§āļ­ัāļāļĐāļĢ, āđ‚āļ„āļĢāļ‡āļ‚่āļēāļĒāļ›āļĢāļ°āļŠāļēāļ—āđ€āļ—ีāļĒāļĄ, āļāļēāļĢāļ•ัāļ”āđāļĒāļāļ•ัāļ§āļ­ัāļāļĐāļĢ, āļ„ุāļ“āļĨัāļāļĐāļ“āļ°āļžิāđ€āļĻāļĐāļ‚āļ­āļ‡āļ•ัāļ§āļ­ัāļāļĐāļĢāļĨāļēāļĒāļĄืāļ­āđ€āļ‚ีāļĒāļ™āļ āļēāļĐāļēāđ„āļ—āļĒ

Abstract
The purpose of this research is to present the Thai handwritten character recognition by artificial neural networks. 100 collections of Thai handwritten characters were collected from 10 writers. The data are then analyzed using pre-processing feature extraction and recognition by back-propagation neural network. However, the feature extraction method is aimed to identify the intensity of black pixel, end of character, head of character, chain code, cross horizontal and vertical line, mark direction, and horizontal and vertical scanning. From the feature extraction, it was found that there are 235 features. Seven feature extraction methods provide the efficacy of Thai handwritten recognition. In addition, back-propagation neural network provide the Thai handwritten recognition accuracy of 87%

keywords: Pattern Recognition, Character Recognition, Artificial Neural Network, Character Segmentation, Feature Extraction of Thai Handwritten

Presentation

Thai Handwritten Character Recognition
Grey Image
Binary Image
Noise Reduction
Line Segmentation
Character Segmentation
Thinning
Feature Extraction
Artificial Neural Networks
Evaluation
Character Recognition Evaluated
Optimal Hidden Node
OCR Program GUI




Abstract

Feature extraction techniques can be important in character recognition, because they can enhance the efficacy
of recognition in comparison to featureless or pixel-based approaches. This study aims to investigate the novel feature extraction technique called the hotspot technique in order to use it for representing handwritten characters and digits. In the hotspot technique, the distance values between the closest black pixels and the hotspots in each direction are used as representation for a character. The hotspot technique is applied to three data sets including Thai handwritten characters (65 classes), Bangla numeric (10 classes), and MNIST (10 classes). The hotspot technique consists of two parameters including the number of hotspots and the number of chain code directions. The data sets are then classified by the k-Nearest Neighbors algorithm using the Euclidean distance as function for computing distances between data points. In this study, the classification rates obtained from the hotspot, mark direction, and direction of chain code techniques are compared. The results revealed that the hotspot technique provides the largest average classification rates.

keywords
Handwritten Character Recognition, Feature Extraction, k-Nearest Neighbors, Classification

Conference Sitevilamoura, Algarve, Portugal
 
āđ€āļ—āļ„āļ™āļīāļ„āļāļēāļĢāļ™āļģāđ€āļŠāļ™āļ­
āļ§ีāļ”ีāđ‚āļ­āđ€āļĢื่āļ­āļ‡ "āđ€āļ—āļ„āļ™ิāļ„āļāļēāļĢāļ™āļģāđ€āļŠāļ™āļ­" āļšāļĢāļĢāļĒāļēāļĒāļžิāđ€āļĻāļĐāđ‚āļ”āļĒāļ­āļēāļˆāļēāļĢāļĒ์ āļŠāļĢāļĢāļŠāļĢีāļ āļĢāļ“์ āđ€āļ§āļŽุāļ§āļ™āļēāļĢัāļāļĐ์ āļ­āļēāļˆāļēāļĢāļĒ์āļ›āļĢāļ°āļˆāļģāļŠāļēāļ‚āļēāļ§ิāļŠāļēāļāļēāļĢāļŠื่āļ­āļŠāļēāļĢāļĄāļ§āļĨāļŠāļ™ āļ„āļ“āļ°āļ§ิāļ—āļĒāļēāļāļēāļĢāļŠāļēāļĢāļŠāļ™āđ€āļ—āļĻ āļĄāļŦāļēāļ§ิāļ—āļĒāļēāļĨัāļĒāļĄāļŦāļēāļŠāļēāļĢāļ„āļēāļĄ


āļŠāļēāļĄāļēāļĢāļ–āļĢัāļšāļŠāļĄāļ§ีāļ”ีāđ‚āļ­āļ—ั้āļ‡āļŦāļĄāļ”āļˆāļģāļ™āļ§āļ™ 10 āļ§ีāļ”ีāđ‚āļ­ āļ”ัāļ‡āļˆāļēāļāļĨิāļ‡āļ์āļ”ัāļ‡āļ•่āļ­āđ„āļ›āļ™ี้
http://www.youtube.com/playlist?p=PLAFFEDD35FAEA9B56
āļāļēāļĢāļŠāļĢ้āļēāļ‡āļŠื่āļ­āļāļēāļĢāļŠāļ­āļ™āļ”้āļ§āļĒāđ‚āļ›āļĢāđāļāļĢāļĄ Authorware 7.0

āļŠāļēāļĄāļēāļĢāļ–āđ€āļ‚้āļēāļŠāļĄāļ§ีāļ”ีāđ‚āļ­ "āļāļēāļĢāļŠāļĢ้āļēāļ‡āļŠื่่āļ­āļāļēāļĢāļŠāļ­āļ™āļ”้āļ§āļĒāđ‚āļ›āļĢāđāļāļĢāļĄ Authorware 7.0" āļ—ั้āļ‡āļŦāļĄāļ”āđ„āļ”้āļˆāļēāļāļĨิāļ‡āļ์āļ•่āļ­āđ„āļ›āļ™ี้ http://www.youtube.com/playlist?p=PL45EF991C74E81B21

āļ•ัāļ§āļ­āļĒ่āļēāļ‡āļāļēāļĢāđ€āļ‚ีāļĒāļ™ ASP.NET āļ”้āļ§āļĒāđƒāļŠ้ Visual Studio 2008 āđ€āļ›็āļ™āđ€āļ„āļĢื่āļ­āļ‡āļĄืāļ­ āđāļĨāļ°āđƒāļŠ้āļ āļēāļĐāļē C# āđƒāļ™āļāļēāļĢāđ€āļ‚ีāļĒāļ™āđ‚āļ›āļĢāđāļāļĢāļĄ āđ‚āļ”āļĒāđ€āļŠื่āļ­āļĄāļ•่āļ­āļัāļšāļāļēāļ™āļ‚้āļ­āļĄูāļĨāļˆāļēāļ SQL Server

āļŠāļēāļĄāļēāļĢāļ–āđ€āļ‚้āļēāļŠāļĄāļ§ีāļ”ีāđ‚āļ­āļ—ั้āļ‡āļŦāļĄāļ”āļ—ี่āđ€āļี่āļĒāļ§āļัāļš Visual studio & web application with c# āđ„āļ”้āļˆāļēāļāļĨิāļ‡āļ์āļ•่āļ­āđ„āļ›āļ™ี้ http://www.youtube.com/playlist?p=PL04AC36FE5A17B55E