A Hybrid Model for Driver Emotion Detection Using Feature Fusion Approach.

Area of Science:

• Artificial Intelligence
• Computer Vision
• Human-Computer Interaction

Background:

• Driver emotions significantly impact driving behavior and road safety.
• Advanced Driver Assistance Systems (ADAS) in the automotive industry utilize AI for driver support.
• Continuous monitoring of driver emotions can predict behavior and prevent accidents.

Purpose of the Study:

• To develop a novel hybrid network architecture for predicting driver emotions.
• To enhance road safety by understanding the link between emotions and driving behavior.
• To achieve accurate emotion prediction under various conditions like different poses, occlusions, and illumination.

Main Methods:

• A hybrid network combining a deep neural network and a support vector machine was developed.
• Fusion of Gabor and Local Binary Pattern (LBP) features was used for emotion determination.
• A Support Vector Machine (SVM) classifier integrated with a Convolutional Neural Network (CNN) was employed for classification.

Main Results:

• The proposed model achieved high accuracy across multiple datasets: 84.41% (FER 2013), 95.05% (CK+), 98.57% (KDEF), and 98.64% (KMU-FED).
• The model demonstrated effectiveness in predicting emotions under challenging conditions.
• The hybrid approach proved superior in emotion recognition tasks.

Conclusions:

• The developed hybrid deep learning model shows significant promise for real-time driver emotion recognition.
• This technology can be integrated into ADAS to improve driver monitoring and road safety.
• Accurate emotion detection is a key step towards proactive accident prevention systems.