Driver Monitoring Alarm Logic Using AI Voting to Cut False Alerts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing driver monitoring systems (DMS) face issues with false alarms, inaccurate detection due to varying driver positions and occlusions, and installation challenges, leading to reduced performance and reliability.
Innovation Solution
A method utilizing a deep learning model to analyze driver images, incorporating voting algorithms to enhance anomaly detection accuracy by considering previous prediction results and thresholds, and determining alarms based on continuous anomaly patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a separated alarm related algorithm is not added to a model used in the DMS, then the model complexity is reduced, but the alarm accuracy deteriorates causing false alarms during driving
Solution Approach 1:
The alarm detection system is segmented into multiple independent algorithms: a primary detection algorithm for initial anomaly identification, and separate alarm-related algorithms for secondary verification. This segmentation allows each algorithm to specialize in specific detection tasks, improving overall alarm accuracy while maintaining manageable model complexity through modular architecture.
Solution Approach 2:
An intermediary verification mechanism is introduced between the primary detection algorithm and the final alarm output. The alarm-related algorithms act as intermediaries that cross-check primary detection results, filtering out false positives before generating final alarms. This intermediary layer enhances reliability without requiring complete redesign of the base model.
2Measurement precision
If a sensitive driver detection algorithm is used, then the detection precision is improved, but the false alarm frequency increases
Solution Approach 1:
The system applies partial verification actions to detection results. Instead of fully verifying every detected anomaly, the system selectively applies alarm-related algorithms based on confidence levels and anomaly types. This partial verification approach maintains high detection precision while reducing false alarm frequency by avoiding over-verification of low-risk cases.
Solution Approach 2:
A feedback mechanism is implemented where alarm-related algorithms provide corrective feedback to the primary detection algorithm. When false positives are identified through secondary verification, the system adjusts detection thresholds and parameters, creating a closed-loop system that continuously optimizes the balance between detection precision and false alarm frequency.
3Measurement precision
If the DMS product is customized for each vehicle installation location, then the detection accuracy is improved, but the ease of manufacture deteriorates
Solution Approach 1:
The DMS product incorporates dynamic calibration capabilities that automatically adapt to different installation locations and vehicle types. Rather than requiring manual customization for each installation, the system dynamically adjusts detection parameters, camera angles, and sensor configurations based on real-time environmental feedback, maintaining high detection accuracy while simplifying the manufacturing and installation process.
Solution Approach 2:
The system implements self-service calibration features that automatically optimize detection accuracy for each installation location without requiring manual customization. The DMS performs self-adjustment of detection parameters, focal points, and alarm thresholds based on initial scanning and environmental assessment, eliminating the need for complex customization while maintaining high detection precision.
Data Source
AI summary
Disclosed is a method for providing an alarm based on a behavior of a driver in a driver monitoring system. The method includes receiving a first image corresponding to the driver within a vehicle, generating first model output information which indicates possibility that an anomaly which compromises safety of the driver exists in the first image, from the first image, by using an artificial intelligence model. The method includes generating a first anomaly primary prediction result indicating whether the anomaly exists in the first image by comparing the first model output information with a first threshold, generating a first anomaly secondary prediction result by performing a first voting using the first anomaly primary prediction result and determining an anomaly alarm corresponding to the first image by performing a second voting using the first anomaly secondary prediction result.