AI Recognition of Human Drawings via Order Parameter Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current AI systems face limitations in accuracy and processing time when recognizing user-drawn inputs, such as handwriting and images, due to the variability and disorder in human-drawn data, which can lead to misinterpretation and inefficiency.
Innovation Solution
The method involves computing an order parameter of user-drawn images by applying a Fourier transform and an idealized modulation transfer function, followed by an inverse Fourier transform to create a modified image, which is then used to extract a numerical value quantifying the degree of order. This value is fed into an AI program to improve recognition accuracy and reduce processing time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If AI training methods are used to recognize user-drawn inputs, then recognition capability is improved, but processing time and training time increase
Solution Approach 1:
The patent applies preliminary action by pre-processing the input image through Fourier transform and MTF filtering before AI recognition. This transforms the image into frequency domain representation, extracting essential features in advance. The order parameter calculation is also performed beforehand, providing pre-computed metrics that guide the AI recognition process, thereby reducing the computational burden during actual recognition and decreasing processing time.
Solution Approach 2:
The patent extracts the order parameter as a separate quantitative feature from the input image using Fourier analysis. This extracted parameter independently characterizes the disorder level of the drawing, separating this critical information from the full image data. The AI system then uses this extracted order parameter alongside the image data, improving recognition efficiency by focusing on key features rather than processing all pixel information equally.
2Measurement precision
If AI training methods are used to recognize user-drawn inputs, then recognition capability is improved, but training time increases
Solution Approach 1:
The patent changes parameters by transforming the image from spatial domain to frequency domain using Fourier transform. This parameter transformation reveals the order/disorder characteristics of the drawing that are not apparent in the original image. The MTF filtering further modifies the frequency parameters to enhance relevant features. These parameter changes provide the AI training process with enriched feature representations, improving accuracy while the systematic approach keeps training time manageable.
3Device complexity
If traditional image recognition is used without order parameter, then processing is simpler, but recognition accuracy decreases due to variability in human-drawn data
Solution Approach 1:
The patent introduces the order parameter as an intermediary between the raw image data and the AI recognition system. This intermediary quantitative measure of disorder acts as a mediator that bridges the gap between variable human-drawn inputs and the AI model's recognition requirements. The order parameter provides an additional layer of information that helps the AI system distinguish between intentional variations in drawing and actual errors, improving accuracy without overwhelming the system with excessive complexity.
Data Source
AI summary
A method of interpreting human-drawn images includes modifying a human-drawn image. A numerical value corresponding to an order parameter squared (S2) is extracted from the modified image. An artificial intelligence (AI) program characterizes the human-drawn image utilizing the human-drawn image and the numerical value of the order parameter. The disclosure further includes systems, computer readable media, programs capable of the same.