DRIVER EMOTION RECOGNITION USING CNN AND ATTENTION MECHANISM

Abstract

Driver emotions
significantly influence
road safety, as emotional distractions can impair cognitive functions, leading to delayed
reactions and poor decision-making. Traditional methods for emotion recognition involve
physiological signals such as EEG or heart rate monitoring, which are often intrusive and
impractical for real-world applications. In contrast, vision-based deep learning techniques
offer a non-intrusive and scalable approach. This paper presents a novel Convolutional
Neural Network (CNN) and Attention Mechanism-based model for real-time driver emotion
recognition using facial expressions. The CNN extracts deep spatial features from facial
images, while the attention mechanism enhances focus on emotion-relevant facial regions
such as the eyes and mouth. The proposed system is evaluated on the Driver Emotion Facial
Expression (DEFE) dataset, demonstrating high accuracy (90.2%) and robustness under
varying lighting conditions.
The model is implemented in a real-time driver monitoring system, capable of detecting
emotional states such as anger, drowsiness, and stress. The integration of an attention
mechanism improves performance by highlighting critical facial regions, ensuring better
generalization across diverse driver profiles. The proposed approach outperforms
conventional CNN models and serves as a potential component for Advanced DriverAssistance Systems (ADAS), contributing to enhanced road safety and accident prevention.
Future work will explore multi-modal approaches combining facial recognition with speech
and physiological signals for an even more comprehensive emotion detection system