SISTEM PENGENALAN TULISAN TANGAN PADA DOKUMEN YANG MENGANDUNG SINGKATAN TIDAK LAZIM MENGGUNAKAN SKRIPSI JARINGAN SARAF TIRUAN
Handwritten Character Recognition System (HCR) is classified into groups, offline and online. Offline HCR aims to recognize static image after handwritting is performed. As opposite, online HCR recognize image while handwriting is being performed. Offline HCR has more difficult challenges than onlin...
Saved in:
| Main Author: | |
|---|---|
| Format: | Theses and Dissertations NonPeerReviewed |
| Language: | English Indonesian |
| Published: |
2018
|
| Subjects: | |
| Online Access: | http://repository.unair.ac.id/80194/1/ST.SI.05-19%20Kar%20s%20abstract.pdf http://repository.unair.ac.id/80194/2/ST.SI.05-19%20Kar%20s.pdf http://repository.unair.ac.id/80194/ http://lib.unair.ac.id |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Universitas Airlangga |
| Language: | English Indonesian |
| Summary: | Handwritten Character Recognition System (HCR) is classified into groups, offline and online. Offline HCR aims to recognize static image after handwritting is performed. As opposite, online HCR recognize image while handwriting is being performed. Offline HCR has more difficult challenges than online character recognition due to insufficient temporal information such as number and direction of strokes, ink pressure, unpredictable and high variation of handwriting. These difficulties cause low accuracy achieved.
The aim of this research is to detect existence of uncommon abbreviation on documents written in Bahasa Indonesia. Several steps conducted to detect uncommon abbreviations: collecting handwritten samples, performing image processing, neural network training. From classficiation process, there are two accuracy indicators used. Character accuracy by class and uncommon abbreviation detected. Character accuracy by class achieved is 60.47 % and for uncommon abbreviation detected accuracy is 27.89 %. Recognition performance achieved by using artificial neural network is better than other research using K-Nearest Neighbor where accuracy achieved is 46 %. |
|---|