Eye-Tracking ROI Sampling for Low-Power Gaze Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Artificial reality systems face challenges in efficiently tracking objects and minimizing power consumption due to continuous high-resolution image capture, which burdens computing resources and drains battery life.
Innovation Solution
Implementing region of interest sampling and retrieval, a two-step process involving an image controller to identify and retrieve high-resolution regions of interest from a subset of photodetectors, reducing the number of accessed photodetectors and conserving computing resources and power.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If continuous high-resolution image capture is performed to track objects of interest, then tracking accuracy is improved, but power consumption and computing resource usage increase
Solution Approach 1:
The patent divides the image sensor's photodetectors into multiple groups, where only a subset of groups (those containing objects of interest) are actively read out at high resolution. This segmentation allows the system to maintain tracking accuracy for specific regions while reducing overall power consumption by leaving other photodetector groups in a low-power state.
Solution Approach 2:
The patent applies different quality levels to different regions of the image sensor. Regions containing objects of interest are read out at full high resolution, while other regions are either downsampled or not read out at all. This local quality differentiation maintains tracking precision where needed while minimizing energy consumption in irrelevant areas.
2Measurement precision
If continuous high-resolution image capture is performed to track objects of interest, then tracking accuracy is improved, but computing resource burden increases
Solution Approach 1:
The patent extracts and processes only the relevant portions of the image data that contain objects of interest. By identifying which photodetector groups correspond to tracked objects and reading out only those groups at high resolution, the system reduces the volume of data requiring computational processing, thereby lowering the computing resource burden while maintaining tracking accuracy.
Solution Approach 2:
The patent performs partial action by reading out only a subset of photodetector groups rather than the entire sensor array. This partial readout approach provides sufficient data for accurate object tracking without the excessive computational burden of processing complete high-resolution images from all photodetectors.
3Loss of information
If full-resolution image data is continuously captured, then complete image information is obtained, but power consumption increases
Solution Approach 1:
The patent performs preliminary identification of objects of interest before full high-resolution readout. By first identifying which photodetector groups contain tracked objects and then selectively reading out only those groups, the system ensures that complete image information is obtained for relevant regions while avoiding the energy cost of reading out unnecessary regions.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach reduces power consumption and computing resources while maintaining accurate object tracking, such as eye and gaze estimation, by focusing on specific regions of interest within captured images.
Implementation Method 1
an image sensor comprising photodetectors (e.g., photodiodes), a full-resolution image (e.g., a face of a user) described by pixel data captured by the photodetectors
Data Source
AI summary
An imaging device is configured to perform eye tracking. The imaging device comprises an image sensor and an image controller. The image sensor is configured to capture an image of a portion of a user's face. The image controller is configured to: in a first-access step, read out a set of sample pixels, from the image sensor to main memory; identify a pixel of interest from the set of sample pixels and a location of the pixel of interest; in a second-access step, read out remaining pixels in a photodetector group of interest corresponding to the location of the pixel of interest, from the image sensor to the main memory; generate a high-resolution region of interest by combining the pixel of interest and the remaining pixels in the photodetector group; and perform gaze estimation using the full resolution region of interest.