Statistical Online Character Recognition via Gabor Feature Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing online character recognition (OLCR) methods, particularly structural and statistical-structural models, lack adaptability across different character sets and languages, making them inefficient for recognizing handwritten characters in languages with large character sets like Korean and Japanese.
Innovation Solution
A statistical OLCR system that generates patterns using pre-identified character samples through preprocessing, feature extraction, and statistical training, allowing for flexible recognition across different character sets without complete redesign, utilizing techniques like k-means clustering and Gabor filters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If structural recognition methods are used for online character recognition, then recognition accuracy can be achieved for specific character sets, but adaptability across different languages and character sets deteriorates
Solution Approach 1:
The patent changes the fundamental parameters of character representation from structural descriptions to statistical feature vectors. Instead of using fixed structural models that must be redesigned for each language, the system extracts statistical features (directional patterns, spatial relationships) that can be universally applied across different character sets and languages, thereby improving adaptability while maintaining recognition accuracy.
Solution Approach 2:
The patent creates a universal recognition system by extracting statistical features that are common across different character sets. The same feature extraction and classification algorithms can be applied to recognize characters in Korean, Japanese, English, or other languages without requiring language-specific structural models, achieving multi-lingual adaptability.
2Measurement precision
If statistical-structural models are used, then recognition accuracy improves, but device complexity and adaptability deteriorate due to language-specific model requirements
Solution Approach 1:
The patent extracts the essential statistical features from character images and represents them as compact feature vectors. By taking out only the critical statistical characteristics (directional information, spatial patterns) and discarding unnecessary structural details, the system achieves simplified representation that reduces model complexity while maintaining high recognition accuracy across different languages.
Solution Approach 2:
The patent transforms the complexity from language-specific structural models to a unified statistical feature space. By changing the representation parameters to statistical features that are language-invariant, the system reduces the complexity of adapting models to different languages while preserving recognition accuracy.
3Productivity
If structural analysis methods are used, then recognition can be performed on handwritten characters, but flexibility for new character sets and languages deteriorates
Solution Approach 1:
The patent changes the approach from analyzing fixed structural patterns to extracting statistical features that describe the probabilistic characteristics of handwritten characters. This parameter change enables the system to handle new character sets and languages flexibly, as the statistical features can be extracted from any character set without requiring pre-defined structural models.
Solution Approach 2:
The patent enables the system to automatically adapt to new character sets by extracting statistical features from training data of the new language or character set. The system serves itself by learning the statistical characteristics of new languages through feature extraction and classification, without requiring manual redesign of the recognition system.
Data Source
AI summary
A statistical system and method for generating patterns and performing online handwriting recognition based on those patterns. A plurality of predetermined patterns may be generated by performing feature extraction operations on one or more character samples utilizing a Gabor filter. An online handwritten character may be acquired. The online handwritten character may be pre-processed. One or more feature extraction operations, utilizing a Gabor filter, may be performed on the online handwritten character to produce a feature vector. One or more patterns may be generated, using a statistical algorithm, for the online handwritten character, based on the feature vector. The online handwritten character may be statistically classified based on a comparison between the one or more patterns generated for the online handwritten character and the plurality of predetermined patterns.