Handwriting Search Using Ink Stroke Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current optical character recognition (OCR) software is prone to errors when processing human handwriting, making it difficult for users to accurately search and retrieve handwritten documents stored electronically.
Innovation Solution
A method that accepts and analyzes handwriting ink strokes, producing search data to search existing handwriting data, allowing for the display of relevant results on an input and display device, which includes using OCR and additional analysis techniques such as vector coordinate systems and characteristic analysis to improve search accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If current OCR software is used to process handwritten documents, then the documents can be converted to machine text for searching, but the search accuracy deteriorates due to OCR errors
Solution Approach 1:
The patent segments the search process into multiple independent components: (1) extracting visual features from handwritten query and database images, (2) converting to coordinate representations, (3) comparing geometric characteristics, and (4) optionally applying OCR as a supplementary step. This segmentation allows the system to bypass OCR's error-prone text conversion for the primary matching operation, thereby maintaining search accuracy while preserving full-text search capability.
Solution Approach 2:
The patent introduces coordinate representations and visual feature extractions as intermediary forms between the handwritten images and the search matching process. Instead of directly comparing OCR text outputs (which are error-prone), the system uses these intermediary representations that preserve the original visual characteristics of handwriting, enabling accurate comparison without relying on potentially incorrect text conversions.
2Adaptability or versatility
If handwriting is converted to machine text using OCR, then the text becomes searchable, but the conversion process introduces errors that reduce reliability
Solution Approach 1:
The patent implements a dynamic search approach where multiple search methods are available and can be applied in sequence or combination. The system first performs visual feature-based searching on handwritten images, then optionally supplements with OCR-based text searching. This dynamic multi-method approach ensures reliable results by not depending solely on the error-prone OCR conversion, thereby maintaining both versatility and reliability.
Solution Approach 2:
The patent prepares multiple search pathways in advance (visual feature matching, coordinate comparison, and OCR-based text search) so that if one method fails or produces errors, other methods are already available to compensate. This prior cushioning ensures that the system maintains reliable search functionality even when OCR accuracy is poor, as the visual-based methods provide a backup that does not depend on correct text conversion.
3Device complexity
If traditional search methods are used on handwritten documents, then the process is simple, but the ability to locate specific information deteriorates due to lack of text recognition
Solution Approach 1:
The patent replaces the traditional mechanical OCR-based text conversion system with a visual feature extraction and coordinate comparison system. Instead of mechanically converting handwriting to text and then searching the text, the system extracts visual features (lines, curves, intersections) and compares their geometric characteristics directly. This substitution maintains process simplicity while dramatically improving information retrieval capability by working with the visual characteristics of handwriting rather than relying on text conversion.
Data Source
AI summary
One embodiment provides a method, involving: accepting, at an input and display device, handwriting ink strokes; analyzing, using a processor, the handwriting ink strokes; producing search data based on the analyzing; searching, using a processor, existing handwriting data based on the search data; returning, on the input display and device, at least one result based on the searching. Other aspects are described and claimed.