CNN-LSTM based emotion recognition using Chebyshev moment and K-fold validation with multi-library SVM

Area of Science:

• Computer Science
• Artificial Intelligence
• Psychology

Background:

• Human emotions and facial expressions lack a direct, consistent correlation.
• Facial expressions are influenced by psychological factors, affecting their display.
• Machine Learning and Neural Networks are increasingly used to analyze complex human behaviors.

Purpose of the Study:

• To analyze driver facial expressions for mood and emotional state detection.
• To enhance road safety by understanding driver emotions.
• To propose and evaluate a novel hybrid model for accurate emotion recognition.

Main Methods:

• Development of a hybrid Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) model.
• Integration of RESNET152 CNN with a Multi-Library Support Vector Machine (SVM) for classification.
• Utilization of Chebyshev moments for enhanced feature extraction and K-fold validation for performance evaluation.

Main Results:

• The proposed hybrid CNN-LSTM model demonstrated superior performance over traditional methods.
• The model achieved high accuracy in classifying driver emotional states.
• Chebyshev moments proved effective for feature extraction in this context.

Conclusions:

• The hybrid CNN-LSTM model offers a promising approach for real-time driver emotion recognition.
• Accurate emotion detection can significantly contribute to improving road safety.
• Further research can explore more complex emotional states and diverse driving conditions.