Smart handwritten notes recognition: AI-powered solutions for learning beyond
Smart handwritten notes recognition: AI-powered solutions for learning beyond
Every day, millions of handwritten notes often record valuable knowledge, critical insights, and critical information. Given the importance of handwritten notes, an effective method to digitize, store, and extract information from these notes would not only greatly enhance the ability to securely preserve and manage information but also unlock a variety of real-world applications. However, handwriting recognition still remains a challenging within Computer Vision (CV) and Artificial Intelligence (AI). Although various studies in Optical Character Recognition (OCR) have achieved remarkable success in recognizing handwritten text at the character level, they still struggle with overlapping characters, inconsistent spacing, and the wide variety of handwriting styles encountered in practice. To address these limitations, this research proposes a Handwritten Word Recognition (HWR) model with a two-stage pipeline, specifically optimized for handwritten English notes. The first stage focuses on accurately detecting the location of each word in an image. The second stage utilizes ResUNet-101 for feature extraction, followed by Bi-LSTM and CTC decoding to recognize entire words. Finally, a post-processing phase leverages Natural Language Processing (NLP) techniques to enhance accuracy further. The proposed method achieves an accuracy of 77% on testing data, demonstrating significant improvements over traditional OCR systems in handling handwritten word. This research not only advances handwritten note digitization but also lays the groundwork for automated information extraction systems, benefiting fields such as education, research, and document archiving.
Handwritten word recognition, Image processing, Optical Character Recognition, Post - Processing
258-273
Pham, Vu Tuyet Anh
77a4d399-13ab-4ac5-9185-9818650c0c23
Le, Tan Duy
9df04978-eb18-4ea8-aef1-54b3fe6f154e
Tuyen, Nguyen Tan Viet
f6e9374c-5174-4446-b4f0-5e6359efc105
Huynh, Kha Tu
5bc0e303-caec-479a-b7ff-76629581f23d
Pham, Vu Tuyet Anh
77a4d399-13ab-4ac5-9185-9818650c0c23
Le, Tan Duy
9df04978-eb18-4ea8-aef1-54b3fe6f154e
Tuyen, Nguyen Tan Viet
f6e9374c-5174-4446-b4f0-5e6359efc105
Huynh, Kha Tu
5bc0e303-caec-479a-b7ff-76629581f23d
Pham, Vu Tuyet Anh, Le, Tan Duy, Tuyen, Nguyen Tan Viet and Huynh, Kha Tu
(2025)
Smart handwritten notes recognition: AI-powered solutions for learning beyond.
Nguyen, Ngoc Thanh, Kozierkiewicz, Adrianna, Dinh Duc Anh, Vu, Nguyen Van, Sinh, Nunez, Manuel, Treur, Jan and Vossen, Gottfried
(eds.)
In Advances in Computational Collective Intelligence - 17th International Conference, ICCCI 2025, Proceedings.
vol. 2747 CCIS,
.
(doi:10.1007/978-3-032-10202-7_18).
Record type:
Conference or Workshop Item
(Paper)
Abstract
Every day, millions of handwritten notes often record valuable knowledge, critical insights, and critical information. Given the importance of handwritten notes, an effective method to digitize, store, and extract information from these notes would not only greatly enhance the ability to securely preserve and manage information but also unlock a variety of real-world applications. However, handwriting recognition still remains a challenging within Computer Vision (CV) and Artificial Intelligence (AI). Although various studies in Optical Character Recognition (OCR) have achieved remarkable success in recognizing handwritten text at the character level, they still struggle with overlapping characters, inconsistent spacing, and the wide variety of handwriting styles encountered in practice. To address these limitations, this research proposes a Handwritten Word Recognition (HWR) model with a two-stage pipeline, specifically optimized for handwritten English notes. The first stage focuses on accurately detecting the location of each word in an image. The second stage utilizes ResUNet-101 for feature extraction, followed by Bi-LSTM and CTC decoding to recognize entire words. Finally, a post-processing phase leverages Natural Language Processing (NLP) techniques to enhance accuracy further. The proposed method achieves an accuracy of 77% on testing data, demonstrating significant improvements over traditional OCR systems in handling handwritten word. This research not only advances handwritten note digitization but also lays the groundwork for automated information extraction systems, benefiting fields such as education, research, and document archiving.
Text
Smart_Handwritten_Notes_Recognition_AI-Powered_Solutions_for_Learning_Beyond-CameraReady
- Accepted Manuscript
More information
e-pub ahead of print date: 8 November 2025
Keywords:
Handwritten word recognition, Image processing, Optical Character Recognition, Post - Processing
Identifiers
Local EPrints ID: 507747
URI: http://eprints.soton.ac.uk/id/eprint/507747
ISSN: 1865-0929
PURE UUID: c087486b-11bf-4991-8fbf-ac17055890ce
Catalogue record
Date deposited: 06 Jan 2026 10:53
Last modified: 08 Jan 2026 03:23
Export record
Altmetrics
Contributors
Author:
Vu Tuyet Anh Pham
Author:
Tan Duy Le
Author:
Nguyen Tan Viet Tuyen
Author:
Kha Tu Huynh
Editor:
Ngoc Thanh Nguyen
Editor:
Adrianna Kozierkiewicz
Editor:
Vu Dinh Duc Anh
Editor:
Sinh Nguyen Van
Editor:
Manuel Nunez
Editor:
Jan Treur
Editor:
Gottfried Vossen
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics