Hand Recognition in Head-Mounted Displays Using Pixel Difference Thresholds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing augmented reality (AR) systems in head-mounted displays and image processing devices face challenges in accurately recognizing user hands due to varying illumination conditions and individual skin tone differences, leading to decreased recognition accuracy and increased complexity in defining model colors.
Innovation Solution
An image processing device that calculates differences in pixel values within predetermined thresholds to recognize hand shapes, allowing for comparison with stored contour shapes without relying on specific model colors, thereby improving recognition accuracy and versatility across different illumination conditions and user skin tones.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If model color is defined in advance as a fixed reference for hand recognition, then recognition process is simplified, but recognition accuracy decreases under varying illumination conditions and different skin tones
Solution Approach 1:
The patent applies dynamics by transitioning from a static model color approach to a dynamic difference calculation approach. Instead of using a fixed model color that cannot adapt to changing conditions, the system calculates color differences between adjacent pixels in real-time, allowing the recognition process to adapt dynamically to varying illumination conditions and different skin tones while maintaining operational simplicity
Solution Approach 2:
The patent changes the recognition parameter from absolute color matching to relative color difference calculation. By calculating the difference between adjacent pixel colors and comparing it to a threshold, the system maintains simplicity while achieving accuracy across diverse conditions. The parameter change from fixed model color to dynamic difference value effectively resolves the contradiction between simplicity and accuracy
2Measurement precision
If model color is newly defined according to each user's hand color, then recognition accuracy is improved, but convenience deteriorates and system complexity increases
Solution Approach 1:
The patent implements self-service by enabling the system to automatically adapt to each user's hand color without requiring manual intervention. The difference calculation method inherently adapts to individual skin tones by comparing adjacent pixel colors, eliminating the need for users to participate in model color definition while maintaining high recognition accuracy for each individual
3Reliability
If color matching with margin is used to secure recognition accuracy, then recognition is more robust, but versatility across different skin tones decreases
Solution Approach 1:
The patent achieves universality by creating a recognition method that works across all skin tones without requiring skin-tone-specific parameters. The difference calculation approach is universally applicable because it relies on the relative color relationship between adjacent pixels rather than absolute color values, making the system equally effective for all users regardless of their skin tone
Data Source
AI summary
When a hand of the user is recognized in an image pickup region of a camera, a head mounted display stores a contour shape of the hand which would be imaged by the camera in advance. In addition, the head mounted display receives an input of image data per pixel included in the camera, calculates a difference between colors of adjacent pixels represented by the image data, sets a set of image data having the same color system, where the calculated difference is within a predetermined threshold, as a group, and captures a contour of a region of the data. Next, the head mounted display compares the captured contour to a contour shape of the hand which is stored in advance to allow the user to recognize the hand of the user in the image pickup region.