Gaze Area Retrieval in Head-Worn AI for Low-Latency Responses
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current eye-tracking technologies in head-worn devices are costly due to the use of multiple cameras and LEDs, leading to high power consumption and processing overhead, and struggle with accurately interpreting complex scenes captured by wide field-of-view cameras, resulting in delayed responses and suboptimal user experiences.
Innovation Solution
A head-wearable device with coplanar illumination sources and cameras, combined with machine-learning algorithms, isolates the user's gaze area from the field-of-view image and processes it using a multi-modal AI to provide targeted responses.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple cameras and LEDs are used for eye tracking, then tracking accuracy is improved, but device cost and power consumption increase
Solution Approach 1:
The patent segments the eye tracking function by using a single camera positioned at a specific angle rather than multiple cameras. The coplanar configuration with a single camera at 45 degrees to the optical axis allows the system to achieve sufficient tracking accuracy without the power consumption and cost of multiple cameras and LEDs
Solution Approach 2:
The patent extracts the essential function of eye tracking by removing unnecessary components (multiple cameras, rings of LEDs) and retaining only the critical element (single coplanar camera). This extraction maintains tracking capability while significantly reducing power consumption and device complexity
2Area of stationary object
If wide field-of-view cameras are used, then scene coverage is improved, but processing complexity and latency increase
Solution Approach 1:
The patent extracts only the relevant portion of the image (gaze area) from the full field-of-view image for processing. By isolating and processing only the gaze area rather than the entire scene, the system reduces processing complexity and latency while maintaining awareness of the overall scene context
Solution Approach 2:
The patent segments the image processing task by focusing computation only on the gaze area identified by eye tracking, rather than processing the entire wide field-of-view image. This segmentation approach reduces processing load and latency while preserving scene understanding
3Measurement precision
If high refresh rate rings of LEDs and cameras are used, then tracking precision is improved, but power draw increases
Solution Approach 1:
The patent removes the high-power rings of LEDs and multiple cameras while retaining sufficient tracking precision through a single coplanar camera positioned at 45 degrees to the optical axis. This extraction maintains measurement precision with significantly reduced power draw
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
Enhances user interaction by reducing latency and improving response relevance through focused gaze-based image processing, optimizing power usage and reducing complexity in head-worn devices.
Implementation Method 1
two groups of illumination sources configured to illuminate the respective eye of the user
Data Source
AI summary
A method of providing a response to a user based on a field-of-view and a gaze of the user is described. A head-wearable device is communicatively coupled to a non-transitory, computer-readable storage medium including executable instructions that, when executed by one or more processors cause the one or more processors to perform the method. The method includes, causing the one or more cameras of the head-wearable device to capture an image of a field-of-view of the user and causing an eye-tracking device of the head-wearable device to determine a gaze of the user. The method further includes, in response to a capture command, isolating a gaze area of the image from a remainder of the image based on the gaze of the user and identifying, using a machine-learning algorithm, an object in the gaze area. The method further includes generating a response, using another machine-learning algorithm, based on the object.