Driver Attention Estimation Using Top-Down and Bottom-Up Cues

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing driver state determination devices struggle to accurately estimate a driver's attention function degraded state, which is crucial for improving estimation accuracy, especially in cases where the driver's focus is concentrated on a single point or when the distance to a preceding vehicle is unsafe, leading to difficulties in assessing the driver's mental state effectively.

Innovation Solution

A driver state estimation device that utilizes both top-down and bottom-up attention source amounts to estimate the driver's state, incorporating sightline detection, point of interest identification, and saliency distribution analysis to derive index values in specific areas where each attention type predominates, thereby enhancing estimation accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If the driver state is estimated based on only one of the top-down attention source amount or bottom-up attention source amount, then the device complexity is reduced, but the estimation accuracy of the driver state deteriorates

Engineering Contradiction:
Improveestimation accuracyVSAvoiddevice complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent divides the driver's attention system into two distinct components: top-down attention (goal-directed, voluntary attention) and bottom-up attention (stimulus-driven, involuntary attention). Each component is measured separately using different metrics (first index for top-down, second index for bottom-up), allowing the system to capture multiple dimensions of driver attention without requiring a single complex measurement system.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent transitions from unidimensional attention measurement (single index) to bidimensional attention measurement (two independent indexes). By evaluating both top-down and bottom-up attention source amounts separately, the system creates a more comprehensive attention assessment model that captures the multifaceted nature of driver attention, thereby improving estimation accuracy while maintaining manageable device complexity through modular measurement approaches.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Ease of operation

If the driver state is estimated based on sightline concentration at one point and distance to preceding vehicle, then the ease of operation is improved, but the ability to detect attention function degraded state deteriorates

Engineering Contradiction:
Improveease of operationVSAvoiddetection capability
Core Design Contradiction:
Ease of operationVSReliability

Solution Approach 1:

The patent segments the attention assessment into two independent measurement systems: one for top-down attention (first index based on sightline focus on specific points) and one for bottom-up attention (second index based on environmental saliency). This segmentation allows the system to maintain simple operational procedures while simultaneously capturing comprehensive attention information that includes detection of attention function degradation.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent adds a new dimension to attention measurement by introducing bottom-up attention assessment alongside the existing top-down attention measurement. This bidimensional approach enables the system to detect attention function degraded states that may not be apparent through single-point sightline analysis alone, while preserving the ease of operation through established measurement methods.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

Data Source

PatentEP3868622B1Driver state estimation device and movable body
Publication Date: 2023.11.08 MAZDA MOTOR CORP
  • EP3868622B1 patent drawingFigure 1
  • EP3868622B1 patent drawingFigure 2
  • EP3868622B1 patent drawingFigure 3

AI summary

A first measurement section measures a first index value that is correlated with an amount of an attention source allocated to top-down attention of a driver in a first area where the top-down attention of the driver is predominant. A second measurement section measures a second index value that is correlated with an amount of the attention source allocated to bottom-up attention of the driver in a second area where the bottom-up attention of the driver is predominant. An estimation section estimates the driver state including an attention function degraded state of the driver on the basis of the first index value in the first area and the second index value in the second area.