Deformable Expression Detection Using Gradient Sign Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current electronic devices lack efficient methods for detecting deformable expressions, such as facial emotions or gestures, which limits their functionality in applications like human-machine interfaces and image capture systems.
Innovation Solution
A method and apparatus for deformable expression detection using a combination of directional gradient components, where signs of these components are coded into values to classify expressions like smiling, blinking, or anger, employing machine learning algorithms like SVM, KNN, and boosting algorithms, with feedback mechanisms to adjust classification errors and decision thresholds.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional expression detection methods are used, then detection accuracy may be maintained, but computational complexity increases and processing efficiency decreases
Solution Approach 1:
The patent extracts only the essential sign information from gradient components rather than processing complete gradient magnitude and direction data. By taking out only the critical sign features (positive/negative directions) and discarding redundant magnitude information, the system achieves simpler computation while maintaining expression detection accuracy
Solution Approach 2:
The patent transforms the continuous gradient information into discrete sign-based coded values (e.g., 0, 1, 2, 3 representing different direction combinations). This parameter transformation from continuous to discrete space reduces computational burden while preserving the essential directional information needed for expression detection
2Measurement precision
If comprehensive image processing is performed, then expression detection accuracy improves, but processing time increases
Solution Approach 1:
The patent performs preliminary preprocessing operations (alignment, cropping, scaling, histogram equalization) before the main expression detection. By preparing the image in advance with these standardized operations, the subsequent gradient-based detection can proceed efficiently without needing to perform complex real-time adjustments during detection
Solution Approach 2:
The patent extracts only the sign information from gradient components rather than processing complete gradient magnitude and direction data. By taking out only the critical sign features (positive/negative directions) and discarding redundant magnitude information, the system achieves simpler computation while maintaining expression detection accuracy
3Device complexity
If simple coding methods are used, then computational complexity is reduced, but information loss increases
Solution Approach 1:
The patent converts the potential harm of information loss into a benefit by demonstrating that gradient magnitude information is actually redundant for expression detection. The sign information alone suffices to capture the essential directional patterns of facial expressions, turning what could be seen as an information deficiency into a computational advantage
Data Source
AI summary
A method for deformable expression detection is disclosed. For each pixel in a preprocessed image, a sign of a first directional gradient component and a sign of a second directional gradient component are combined to produce a combined sign. Each combined sign is coded into a coded value. An expression in an input image is detected based on the coded values.