Korean Handwritten Text Recognition
Historical handwriting transcription using CNN-based text detection and Bi-LSTM sequence recognition
This project focused on extracting and transcribing historical Korean handwriting from document images. It employed a two-stage workflow: first localizing text regions and then recognizing the character sequences in the detected regions.
Research period: December 2021 to February 2022
Funder: Small and Medium Business Technology Information Promotion Agency, South Korea
Role: Project Lead
Research objectives
- Detect handwritten text regions in scanned Korean documents.
- Transcribe detected historical handwriting into modern Korean text.
Contributions
- Annotated text regions and prepared bounding-box labels for detector training.
- Used a CNN-based approach for text localization.
- Applied bidirectional long short-term memory (Bi-LSTM) sequence modeling to recognize detected text.
- Developed an integrated detection-to-transcription workflow.
Methods
Document image analysis, text detection, convolutional neural networks, Bi-LSTM, optical character recognition (OCR), sequence modeling.