Drowsy Driving Detection Using Multi-Source Sensor Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems face challenges in accurately detecting driver drowsiness due to limitations in image analysis, facial landmark detection, and computing device constraints, which can lead to inaccurate evaluations and inefficiencies.
Innovation Solution
The proposed system integrates inward and outward camera feeds to analyze driver behavior and facial expressions, using machine learning models to calculate a drowsiness scale index, thereby overcoming previous limitations and enhancing accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If dashboard camera footage is used for facial landmark detection, then the system can monitor driver behavior, but the accuracy of facial landmark detection does not meet anticipated standards
Solution Approach 1:
The patent combines multiple data sources including dashboard camera footage, wearable device sensors (accelerometers, gyroscopes, heart rate monitors), and vehicle telematics data to create a comprehensive driver state assessment system. This multi-source integration compensates for the limitations of individual facial landmark detection and provides more reliable drowsiness evaluation through corroborating evidence from multiple independent measurement systems.
2Measurement precision
If image analysis is used to evaluate driver fatigue, then the system can detect drowsiness signs, but establishing a definitive chronological sequence from visual data poses inherent difficulties
Solution Approach 1:
The patent introduces wearable devices and vehicle sensors as intermediary systems that provide timestamped physiological and behavioral data independent of image analysis. These intermediaries capture driver state information continuously with precise timing, allowing the system to establish chronological sequences of drowsiness events without relying solely on visual data processing, thereby reducing time loss while maintaining detection accuracy.
3Measurement precision
If comprehensive image analysis is performed to assess driver drowsiness, then detection accuracy improves, but constraints on available computing devices regarding storage and processing capacities impact efficiency
Solution Approach 1:
The patent extracts and separates processing tasks between the vehicle's computing system and wearable devices. The wearable devices perform initial processing of physiological data locally, filtering and pre-processing information before transmission to the vehicle system. This extraction of computational burden from the central vehicle computer reduces storage and processing demands on the main system while maintaining comprehensive analysis capabilities through distributed intelligence.
Data Source
AI summary
Techniques are presented for detecting when drivers drive while drowsy. In some implementations, a drowsiness model is trained with data associated with inward videos and outward videos captured during a trip. The inward videos capture the inside of the cabin with the driver, and the outward videos capture the view in front of the vehicle in the direction of travel. Further, a device at the vehicle periodically calculates a drowsiness scale index value that indicates the level of drowsiness of the driver. Calculating the drowsiness scale index value includes obtaining a set of inward frames from the inward videos; for each inward frame, creating a face image by cropping the inward frame; obtaining a set of outward frames from the outward videos; calculating inward embeddings of the face images and outward embeddings of the outward frames; and calculating, by the drowsiness model, the drowsiness scale index value.