3D On-line Handwriting Recognition System for Wearable Devices

다국어 초록 (Multilingual Abstract)

In this paper, we propose a on-line handwriting recognition for wearable computing devices. Most of the existing handwriting recognition require elaborate preprocessing techniques and training processes. We present a online handwriting recognition method using patterns of primitives. The basic idea of our approach is to extract primitives from the handwritten character by using the information of writing direction. The experimental result shows that the classifier with simplified preprocessing and training stage allows handwriting recognition in an efficient way for the writer independent case. We also present a 3D handwriting recognition system for an input interface of wearable devices. To track the motion of the user’s finger, we developed a input system using an accelerometer and a gyroscope. Motion-based feature extraction enables characters well recognized no matter the characters are inclined or overlapped in three dimensional space.

목차 (Table of Contents)

• 1. Introduction
• 2. Related Works
• 3. Handwriting recognition based on the patterns of structured primitives
• 3.1 On-line handwriting recognition
• 3.2 Structural primitives
• 3.3. Primitive extraction
• 3.4 Primitive matching
• 3.5 Primitive pattern-based classification
• 3.6 Experimental results
• 4. 3D handwriting recognition system
• 4.1 3D handwriting recognition challenges
• 4.2 Hardware system
• 4.3 Pre-processing
• 4.4 Feature extraction
• 4.5 Character classes
• 4.6 Classification
• 4.7 Experimental results
• 5. Conclusion
• 6. References