Information Processor With Pupil-Size Learning for Fatigue Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting user fatigue and drowsiness in information terminals require dedicated devices and are not capable of real-time, accurate, and efficient detection.
Innovation Solution
An information processor equipped with an imaging unit and an arithmetic unit performing machine learning to detect and analyze pupil size changes over time, utilizing neural networks for inference and learning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a dedicated device is used to detect eye fatigue by pupil diameter changes, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The information terminal's existing camera is made to serve multiple functions: its primary function for capturing images/video is combined with a secondary function for detecting pupil diameter changes to assess eye fatigue. This eliminates the need for dedicated detection devices while maintaining measurement capability through software-based image analysis of the existing camera feed.
Solution Approach 2:
Instead of using a dedicated optical device to directly measure pupil diameter, the system creates a visual copy of the pupil through standard camera imaging. The pupil diameter is then measured by analyzing this copied image representation, allowing indirect measurement through a simpler imaging device rather than requiring specialized measurement equipment.
2Productivity
If real-time fatigue detection is implemented, then productivity is improved, but use of energy increases
Solution Approach 1:
The system performs partial image processing by focusing only on specific regions of interest (the eye and pupil areas) rather than analyzing the entire image frame. This selective processing approach reduces computational load and energy consumption while still achieving real-time detection capability by concentrating resources on the critical fatigue-indicating features.
Solution Approach 2:
Instead of continuous full-frame analysis, the system implements periodic detection at specific time intervals or triggered by certain conditions (such as detected eye presence). This periodic sampling approach maintains real-time monitoring capability while significantly reducing overall processing requirements and energy consumption compared to continuous analysis.
3Measurement precision
If machine learning is used to presume fatigue with high accuracy, then measurement precision is improved, but loss of time increases
Solution Approach 1:
The system performs preliminary detection using simplified algorithms that quickly assess basic pupil diameter changes and fatigue indicators. This preliminary action provides immediate feedback while more sophisticated machine learning analysis operates in parallel or subsequently, allowing high-accuracy fatigue presumption without significant time delay by having the simpler detection ready in advance.
Solution Approach 2:
The system implements a multi-stage detection process where obvious fatigue cases are identified and processed quickly through streamlined pathways, skipping unnecessary complex analysis steps for clear-cut situations. This allows the system to rush through simple cases efficiently while reserving full machine learning analysis for ambiguous or complex cases requiring higher precision.
Data Source
AI summary
An information processing system is provided. The information processing system includes first and second information processors. The first information processor obtains a first moving image including a face and detecting an eye from two or more first images among the first moving image. The first information processor detects a pupil from the eye detected from the first image, calculating its size, and performing learning using a change over time in the size of the pupil. The second information processor obtains a second moving image including a face and detecting an eye from two or more second images among the second moving image. The second information processor detects a pupil from the eye detected from the second image, calculating its size, and performing inference on the change over time in the size of the pupil on the basis of the result of the learning performed by the first information processor.