Distributed Sensor Module for Eye-Tracking with Down-Sampled Image Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
AR/VR devices face challenges in minimizing power consumption and size while maintaining accurate object tracking, as existing machine learning-based object tracking methods require significant power and do not provide sufficient accuracy.
Innovation Solution
A self-sufficient sensor module that detects features from down-sampled images using a machine-learning model, reducing the need to read every segment of the original images, thereby lowering power consumption and improving accuracy by computing the gaze based on corresponding segments in high-resolution images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If machine learning-based object tracking is used, then tracking capability is improved, but power consumption increases
Solution Approach 1:
The patent divides the image processing into segments: first processing a down-sampled version to detect regions of interest, then selectively processing only those specific segments from the full-resolution image. This segmentation approach enables machine learning-based tracking while reducing overall power consumption by limiting full-resolution processing to only necessary regions.
2Measurement precision
If full-resolution images are processed for accurate feature detection, then measurement precision is improved, but use of energy increases
Solution Approach 1:
The patent extracts only the necessary segments from full-resolution images based on region of interest detection from down-sampled images. By taking out and processing only the relevant portions rather than entire high-resolution images, the system maintains feature detection accuracy while significantly reducing power consumption.
3Measurement precision
If every segment in original images is read for feature detection, then measurement precision is improved, but loss of time increases
Solution Approach 1:
The patent performs preliminary processing on down-sampled images to detect regions of interest before processing full-resolution images. This preliminary action identifies which segments require detailed analysis, eliminating the need to read and process every segment from original images, thus reducing processing time while maintaining accuracy.
4Use of energy by moving object
If down-sampled images are used for feature detection, then use of energy is reduced, but measurement precision deteriorates
Solution Approach 1:
The patent applies different processing qualities to different parts of the image: down-sampled images are used for initial region detection where high precision is less critical, while full-resolution images are used for detailed feature analysis in identified regions of interest. This local quality differentiation maintains overall measurement precision while reducing power consumption.
Data Source
AI summary
In one embodiment, a method for eye-tracking comprises capturing images of a user using one or more cameras, the captured images of the user depicting at least an eye of the user, storing the captured images of the user in a storage device, reading, from the storage device, a down-sampled version of the captured images of the user, detecting one or more first segments in the down-sampled version of the captured images by processing the down-sampled version of the captured images using a machine-learning model, the one or more first segments comprising features of the eye of the user, reading, from the storage device, one or more second segments in the captured images corresponding to the one or more first segments in the down-sampled version of the captured images, and computing a gaze of the user based on the one or more second segments in the captured images.