Neural Network Pattern Recognizer Using Positional Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current pattern recognition technologies require extensive research and development to create specific recognizers for each application, making them labor-intensive and costly, and lack the ability to process images in a direct and efficient manner like text processing.
Innovation Solution
An image-based pattern recognizer using a neural network with positional coding, allowing for the creation of generic pattern recognizers that can be customized by users to recognize various patterns without explicit references to the problem or patterns, employing observers and pattern filters to process images and generate ranked lists of coordinates for feature recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a classic pattern recognition approach with explicit routines and variables is used for each specific recognition task, then the recognition function can be implemented, but long periods of research and development are required and the process is labor intensive
Solution Approach 1:
The patent implements a universal pattern recognition system using neural networks that can be applied to multiple recognition tasks (OCR, ASR, face recognition, gesture recognition) without requiring separate development for each task. The neural network architecture with observers and pattern filters provides a generic framework that adapts to different recognition problems, eliminating the need for task-specific explicit routines and variables.
Solution Approach 2:
The patent replaces the mechanical approach of explicitly programming routines and variables for each recognition task with a neural network system that learns patterns automatically. The neural network substitutes the manual configuration process with automated learning, where the system adapts to different recognition tasks through training data rather than requiring hand-crafted algorithms for each task.
2Reliability
If a classic pattern recognition approach with explicit routines and variables is used for each specific recognition task, then the recognition function can be implemented, but the cost of resulting applications increases
Solution Approach 1:
The universal neural network framework allows a single system to handle multiple recognition tasks, reducing the need to develop and maintain separate recognition systems for each application. This consolidation lowers development costs and makes the system more economical to implement across different domains.
Solution Approach 2:
By replacing manual programming of explicit routines with automated neural network learning, the patent reduces the labor-intensive nature of developing recognition systems. The automated learning process lowers development costs by eliminating the need for extensive manual configuration and tuning for each recognition task.
3Productivity
If text processing methods are used for image-based pattern recognition, then direct and efficient processing can be achieved, but current technologies lack the capability to process images in the same way
Solution Approach 1:
The patent substitutes traditional image processing methods with a neural network system that processes images in a manner analogous to text processing. The neural network with observers and pattern filters enables operations similar to text indexing and search, allowing efficient processing and retrieval of image patterns without being constrained by conventional image analysis techniques.
Solution Approach 2:
The neural network framework provides universal processing capabilities that can handle both text and image data using the same underlying mechanisms. This allows text-like operations (indexing, searching, pattern matching) to be applied to images, making the system versatile across different data types while maintaining high processing efficiency.
Data Source
AI summary
An image-based pattern recognizer and a method and apparatus for making such a pattern recognizer are disclosed. By employing positional coding, the meaning of any feature present in an image can be defined implicitly in space. The pattern recognizer can be a neural network including a plurality of stages of observers. The observers are configured to cooperate to identify the presence of features in the input image and to recognize a pattern in the input image based on the features. Each of the observers includes a plurality of neurons. The input image includes a plurality of units, and each of the observers is configured to generate a separate output set that includes zero or more coordinates of such units.