Head-Eye Tracking Probability Maps for Area-of-Interest Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current gaze-tracking systems are inflexible, computationally inefficient, and prone to measurement errors, limiting their effectiveness in dynamic environments with varying users, and existing methods for improving accuracy through calibration are time-consuming and user-specific.
Innovation Solution
A method and system that utilize a spatial probability map in the head coordinate system to determine areas-of-interest by integrating head pose and eye-gaze data, allowing for flexible and efficient area-of-interest determination without extensive user-specific calibration, using a head-eye coordination concept to handle uncertainties and adapt to individual peripheral vision capabilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If individual calibration is performed to improve measurement accuracy, then measurement precision improves, but time consumption increases and flexibility decreases
Solution Approach 1:
The patent applies preliminary action by pre-calculating and storing probability density values for multiple predefined gaze directions in a lookup table during system initialization or manufacturing. This allows the system to quickly retrieve and combine pre-computed probability distributions during runtime without performing real-time complex calculations or requiring individual user calibration, thus resolving the contradiction between measurement precision and time consumption.
2Measurement precision
If remote cameras are used for eye monitoring, then measurement capability is achieved, but field-of-view is limited to a narrow range
Solution Approach 1:
The patent applies dimensionality change by transforming the limited two-dimensional camera field-of-view into a comprehensive three-dimensional probability map around the user's head. By representing gaze directions as points on a sphere or partial sphere and computing probability density values in three-dimensional space, the system effectively expands the monitoring coverage beyond the camera's narrow field-of-view to encompass a much wider angular range.
3Area of stationary object
If head-pose estimation is used to enlarge angular field-of-view, then coverage area increases, but measurement accuracy becomes coarse
Solution Approach 1:
The patent applies merging by combining head-pose estimation data with eye-gaze tracking data to create a fused probability map. The system integrates the coarse angular information from head pose with the finer eye-gaze measurements, merging multiple data sources to achieve both wide coverage and improved accuracy that neither method could achieve alone.
4Adaptability or versatility
If individual calibration is performed for each user, then adaptability to individual capabilities improves, but device complexity and inflexibility increase
Solution Approach 1:
The patent applies universality by creating a user-independent probability map system that can be applied to any user without individual calibration. The predefined gaze directions and probability density values serve multiple users universally, allowing the system to adapt to individual capabilities through the general probabilistic model while avoiding the complexity of individual calibration procedures for each user.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
A method and a system for determining an area-of-interest in a head-eye-tracking application use head pose determination and/or eye gaze tracking for determining the area-of-interest in applications such as human assistance systems or human-robot collaboration. The method determines the area-of-interest of a person based on at least one of head pose or eye gaze, and obtains at least one of left eye gaze direction data for a left eye of the person, right eye gaze direction data for a right eye of the person, and head pose data for a head of the person. A probability map in the head coordinate system is calculated based on the obtained at least one of left eye gaze direction data, right gaze direction data and head pose data. The area-of-interest is determined based on the calculated probability map by generating at least one candidate area-of-interest in the head coordinate system, by calculating for the at least one candidate area-of-interest a probability value based on the probability map, and by selecting the candidate area-of-interest as the determined area-of-interest, in case the calculated probability value is equal to or exceeds a threshold, for example a predetermined threshold value.