Driver Drowsiness Detection Using Segmented Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current vehicle camera systems struggle to detect drowsiness in drivers effectively, as it requires analyzing multiple images over a duration, which is computationally intensive and may not provide real-time monitoring.
Innovation Solution
An apparatus with a camera and processing unit featuring a first neural network model to identify driver poses and a second neural network model to process feature information over time, determining drowsiness based on metrics like blink rate, eye closure, and driving participation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple images capturing drowsiness behavior are analyzed to detect drowsiness, then detection accuracy is improved, but processing power and computational resources are significantly consumed
Solution Approach 1:
The patent divides the drowsiness detection task into two separate neural network models: a first model that processes images to extract feature information (such as eye closure state, head pose), and a second model that processes the extracted features to determine drowsiness. This segmentation allows each model to be specialized and more efficient, reducing the overall computational burden compared to using a single comprehensive model that analyzes all image data directly.
2Measurement precision
If multiple images capturing drowsiness behavior are analyzed to detect drowsiness, then detection accuracy is improved, but real-time monitoring capability deteriorates
Solution Approach 1:
The first neural network model performs preliminary processing by extracting key feature information from images before the second model makes the final drowsiness determination. This preliminary extraction of essential features (eye closure, head pose, blink detection) from multiple images allows the system to maintain high detection accuracy while reducing the computational load for the final decision, thereby enabling real-time monitoring capability.
3Measurement precision
If feature information is processed from multiple images over a duration, then drowsiness detection accuracy is improved, but device complexity increases
Solution Approach 1:
The first neural network model extracts only the essential feature information from multiple images (such as eye closure state, head pose, blink characteristics) and passes these extracted features to the second model. This extraction process separates the complex image processing task from the decision-making task, allowing the system to achieve high detection accuracy through multi-image analysis while keeping the overall device complexity manageable by focusing computational effort on key features rather than processing all image data in full detail.
Data Source
AI summary
An apparatus includes: a camera configured to view a driver of a vehicle; and a processing unit comprising a first model and a second model; wherein the first model is configured to process the images of the driver and to output feature information based on the images, and wherein the second model is configured to receive the feature information from the first model, and to generate an output that indicates whether the driver is drowsy or not.