Language Identification via Stroke Count for OCR Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current character recognition methods are time-consuming, often taking several minutes to process a 15-page document, as they analyze each written character individually without efficiently identifying the language first.
Innovation Solution
A method and apparatus that quickly identify the language of a document by determining the number of strokes per character, then utilize a specifically tailored recognition engine for that language, reducing processing time by using separate engines for different languages.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If character recognition is performed by analyzing each written character individually without language identification, then the recognition process is straightforward, but the processing time is excessively long (up to a minute for a 15-page document)
Solution Approach 1:
The patent applies preliminary action by performing language identification before character recognition. The system first determines the language of the document using statistical methods (such as n-gram analysis or hidden Markov models), then selects and applies the appropriate language-specific recognition model. This preliminary step avoids using incorrect models and reduces overall processing time, directly addressing the time loss issue.
Solution Approach 2:
The patent segments the character recognition process into distinct phases: language identification phase and language-specific recognition phase. By dividing the monolithic recognition process into these segments, the system can optimize each phase independently and avoid the inefficiency of applying a single generic recognition approach to all characters, thereby reducing total processing time.
2Productivity
If a single character recognition engine is used for all languages, then the system is simple to implement, but the processing time increases significantly when handling multiple languages
Solution Approach 1:
The patent implements universality by creating a language identification module that serves multiple functions: it identifies the language of incoming documents, selects appropriate recognition engines, and manages the deployment of multiple language-specific models. This universal module coordinates the use of multiple specialized engines without requiring each engine to be independently controlled, thus improving productivity while managing complexity.
Solution Approach 2:
The patent applies parameter changes by dynamically selecting different recognition engine parameters (such as character sets, encoding schemes, and statistical models) based on the identified language. Instead of using a fixed recognition configuration, the system adjusts its parameters to match the document language, significantly improving recognition speed and accuracy for multiple languages.
3Loss of time
If language identification is performed before character recognition, then processing time is reduced, but the system complexity increases due to additional language detection mechanisms
Solution Approach 1:
The patent introduces an intermediary language identification module that acts as a mediator between the document input and the character recognition engines. This intermediary performs lightweight statistical analysis (such as character frequency counting or n-gram matching) to determine language probability, then routes the document to the appropriate recognition engine. The intermediary approach reduces overall processing time by preventing mismatched recognition attempts while adding only moderate complexity through well-established statistical methods.
Data Source
AI summary
A method and apparatus for identifying a language used in a document based on a number of strokes per character is provided herein. Once identified, character recognition may take place based on the language identified. In one embodiment, a character recognition engine is utilized for character recognition, wherein the character recognition engine is specifically tailored to the identified language.