Driver Attention Estimation Using Probabilistic Gaze Ray Projection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing driver monitoring systems face challenges in accurately estimating attention states due to noisy and uncertain raw gaze and head pose data, leading to errors in determining where a driver's attention is projected.
Innovation Solution
A method and system that utilize an imaging camera and light sources to capture images, generate visual attention rays, project them onto a digital scene representation with predefined regions of interest, determine intersections, apply weightings, and estimate attention states based on these intersections, incorporating confidence measures and historical behavior.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If raw gaze and head pose data are used directly for attention state estimation, then the system is simple and fast, but the estimation accuracy is low due to noisy and uncertain data
Solution Approach 1:
The patent introduces an intermediary processing layer between raw gaze/head pose data and attention state estimation. This layer includes confidence measure calculation, angular distribution generation, and ray projection onto scene geometry, which filters and refines the noisy raw data before final estimation, thereby improving accuracy without requiring fundamental system redesign
Solution Approach 2:
The system implements feedback through confidence measures that are calculated from the raw data quality and used to adjust the angular distribution of visual attention rays. This feedback mechanism allows the system to adapt its estimation process based on data reliability, improving accuracy while maintaining computational efficiency
2Reliability
If multiple visual attention rays with angular distribution are generated and projected onto scene regions, then the reliability of attention state determination is improved, but the computational complexity increases
Solution Approach 1:
The patent segments the attention estimation process into distinct computational stages: confidence measure calculation, angular distribution generation, ray projection onto predefined scene regions, and intersection determination. This segmentation allows each stage to be optimized independently and enables parallel processing of multiple rays across different region intersections, reducing overall computational burden while maintaining reliability
Solution Approach 2:
The system generates multiple visual attention rays with angular distribution around the primary gaze direction, which is more computationally intensive than using a single ray. However, this excessive action (generating more rays than strictly necessary) improves reliability by accounting for uncertainty and variability in gaze measurement, with the computational cost managed through efficient projection algorithms
Data Source
Figure 1
Figure 2
Figure 3
AI summary
System for estimating an attention state of a subject within a scene, including: an imaging camera to capture digital images of the subject and one or more light sources configured to illuminate the subject during a period in which the digital images are captured; a processor configured to process the captured images and generate subject attention data of the subject; processor to: (i) generate a primary visual attention ray of the subject from current subject attention data; (ii) generate a distribution of visual attention rays having an origin common to the primary visual attention ray; (iii) project the visual attention rays onto a digital representation of the scene having a plurality of predefined regions of interest; (iv) determine an intersection of the projected visual attention rays with one or more of the regions of interest; and (v) based on the intersection, estimate an attention state of the subject's attention.