Related Experiment Video
Updated: Jul 7, 2026
06:37
Artificial Intelligence-Based System for Detecting Attention Levels in Students
Published on: December 15, 2023
User adaptive handwriting recognition by self-growing probabilistic decision-based neural networks
1Department of Computer Science and Information Engineering, National Chiao Tung University, Hsin-Chu, Taiwan, R.O.C.
IEEE Transactions on Neural Networks
|February 6, 2008
Summary
User adaptation significantly improves handwriting recognition accuracy, especially with limited training data. Self-growing probabilistic decision-based neural networks (SPDNNs) enable efficient model initialization for new users, boosting performance.
Area of Science:
- Artificial Intelligence
- Machine Learning
- Pattern Recognition
Background:
- User-dependent handwriting recognition systems typically outperform user-independent ones with sufficient data.
- Limited user-specific training data hinders the performance of user-dependent systems.
- Leveraging existing knowledge from multi-user databases can improve performance with minimal training data.
Purpose of the Study:
- To address user adaptation challenges in handwriting recognition systems.
- To develop an effective user adaptation module using self-growing probabilistic decision-based neural networks (SPDNNs).
- To enhance recognition accuracy when user-specific training data is scarce.
Main Methods:
- Developed an SPDNN-based handwriting recognition system.
- Implemented a two-stage recognition structure with a global coarse classifier and a user-independent recognizer.
- Utilized a three-phase training methodology incorporating incremental reinforced and anti-reinforced learning for user adaptation.
Main Results:
- The user adaptation module significantly improved recognition accuracy on a 600-word Chinese character set.
- Average recognition rate increased from 44.2% to 82.4% within five adaptation cycles.
- Recognition performance reached up to 90.2% after ten adaptation cycles.
Conclusions:
- The proposed SPDNN-based user adaptation method effectively enhances handwriting recognition accuracy.
- This approach is particularly beneficial for scenarios with limited user-specific training data.
- The developed system demonstrates a practical solution for personalized handwriting recognition.
