Distance-Based Binary Classifier for Handwritten Word Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current handwriting recognition technologies face challenges in accurately classifying handwritten words when characters overlap or intersect, requiring effective character segmentation algorithms that perform well under practical conditions.
Innovation Solution
A computer-implemented method using a distance-based binary classifier that imports a known answer, constructs a reference dictionary of handwritten characters, and applies a modified tangent distance process to classify handwritten responses by independently altering each character in the reference dictionary to minimize distance with the handwritten response, thereby classifying the response as correct or incorrect.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If character segmentation algorithms are applied to handle overlapping characters, then handwriting recognition capability is improved, but segmentation accuracy deteriorates under practical conditions
Solution Approach 1:
The patent extracts the classification task from traditional character segmentation approaches. Instead of attempting to segment and recognize individual characters, the system directly extracts and classifies complete handwritten words using a distance-based binary classifier that compares handwritten words against reference dictionaries, thereby avoiding the segmentation accuracy problem entirely while maintaining recognition capability
Solution Approach 2:
The patent changes the fundamental parameter of comparison from character-level features to word-level features. By using a distance-based classifier that computes distances between handwritten words and reference words (rather than segmenting and matching characters), the system achieves robustness to overlapping characters while improving overall recognition accuracy
2Ease of operation
If traditional handwriting recognition methods are used, then character-level processing is achieved, but performance deteriorates when characters overlap or intercept
Solution Approach 1:
The patent merges the processing unit from individual characters to complete words. By treating the entire handwritten word as a single unit for classification (rather than processing characters separately), the system maintains ease of operation while significantly improving reliability in handling overlapping and intersecting characters through word-level distance-based comparison
3Measurement precision
If robust handwriting recognition technology is implemented, then accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent segments the reference dictionary into multiple subsets based on different distance thresholds and classification criteria. This segmentation allows the system to process only the most relevant reference words for each classification task, reducing computational complexity while maintaining high accuracy through targeted comparisons rather than exhaustive searches
Data Source
AI summary
This disclosure provides methods and systems to classify handwritten words associated with an answer to a question. According to an exemplary embodiment of this disclosure, a reference dictionary is constructed as a function of a known answer to the question, the reference dictionary including handwritten characters which are used for comparison purposes to determine if the handwritten answer if correct or not.