Drowsiness Detection via Face Tracking and Sensor Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Modern vehicle event recorders face difficulties in detecting risky operator behavior, such as drowsiness, from raw sensor data, which can lead to anomalous events like accidents.
Innovation Solution
A system that combines face tracking data with sensor data to determine a degree of drowsiness, using a processor to capture and warn the driver if the drowsiness level exceeds certain thresholds, and records data for review, employing cameras mounted on the rear-view mirror or dashboard to monitor driver behavior.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If only sensor data is used to detect drowsiness, then the system complexity is reduced, but the detection precision deteriorates
Solution Approach 1:
The patent combines face tracking data from cameras with sensor data (accelerometer, gyroscope, vehicle state sensors) to create a fused data stream for drowsiness detection. This merging of multiple data sources improves detection precision by cross-validating indicators across different modalities while sharing processing infrastructure to manage system complexity.
Solution Approach 2:
The system uses a unified processing framework that handles multiple types of data (visual, inertial, vehicle state) through common algorithms and thresholds. The same processor and memory infrastructure serves multiple sensing functions, reducing overall system complexity while maintaining high detection precision through multi-source data integration.
2Measurement precision
If face tracking data is combined with sensor data, then the drowsiness detection accuracy is improved, but the device complexity increases
Solution Approach 1:
The patent segments the drowsiness detection process into distinct modules: face tracking module, sensor data acquisition module, data fusion module, and threshold evaluation module. Each module processes specific data types independently before integration, reducing processing complexity while maintaining high detection accuracy through specialized handling of each data source.
Solution Approach 2:
The system introduces a data fusion intermediary layer that bridges face tracking data and sensor data. This intermediary processes and normalizes both data streams before combining them for final drowsiness determination, managing the complexity of multi-source integration while preserving the accuracy benefits of data combination.
3Reliability
If multiple thresholds are used for drowsiness detection, then the reliability of warning system is improved, but the difficulty of detecting and measuring increases
Solution Approach 1:
The patent implements segmented threshold evaluation with distinct thresholds for different drowsiness indicators (eye closure, head position, vehicle state). Each indicator has its own threshold criteria, allowing independent evaluation and reducing the complexity of evaluating a single complex multi-threshold condition while improving reliability through multiple independent checks.
Data Source
AI summary
A system for event triggering comprises an interface and a processor. An interface configured to receive a face tracking data and receive a sensor data. The processor configured to determine a degree of drowsiness based at least in part on the face tracking data and the sensor data; in the event that the degree of drowsiness is greater than a first threshold, capture data; and in the event that the degree of drowsiness is greater than a second threshold, provide a warning.