Deep Neural Network Approach for Pose, Illumination, and Occlusion Invariant Driver Emotion Detection.

Area of Science:

• Intelligent Transportation Systems
• Computer Vision
• Affective Computing

Background:

• Driver emotion monitoring is crucial for advanced driver assistance systems (ADAS) to ensure vehicle safety.
• Challenges in driver emotion detection include pose variations, illumination changes, and occlusions.
• Accurate monitoring of a driver's mental status is vital for preventing road accidents.

Purpose of the Study:

• To propose two novel machine learning and deep neural network approaches for robust driver emotion monitoring.
• To overcome challenges in detecting driver emotions caused by variations in pose, illumination, and occlusions.
• To enhance the safety and reliability of intelligent vehicles through effective driver state assessment.

Main Methods:

• Development of two distinct approaches utilizing machine learning and deep neural networks.
• Implementation of algorithms designed to handle pose variations, diverse illumination conditions, and occlusions.
• Validation of the proposed methods on multiple benchmark datasets (CK+, FER 2013, KDEF, KMU-FED).

Main Results:

• The first approach achieved accuracies of 93.41% (CK+), 83.68% (FER 2013), 98.47% (KDEF), and 98.18% (KMU-FED).
• The second approach demonstrated improved accuracies of 96.15% (CK+), 84.58% (FER 2013), 99.18% (KDEF), and 99.09% (KMU-FED).
• Both approaches outperformed existing state-of-the-art methods in driver emotion detection.

Conclusions:

• The proposed machine learning and deep neural network methods offer a significant advancement in driver emotion monitoring.
• These novel approaches effectively address common challenges, leading to more reliable driver state assessment.
• The enhanced accuracy contributes to the development of safer and more intelligent vehicles.