Real-Time Emotion Detection Using CNN-LSTM with Attention Mechanisms for Enhanced Accuracy in Video Streams

This study examines the effectiveness of a hybrid deep learning model that combines Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) networks, and attention mechanisms for real-time emotion detection in video streams. Utilizing the FER2013 dataset, comprising over 35,000 grayscale images of seven facial expressions, the model achieves a classification accuracy of 94.8%. The CNN extracts spatial features from video frames, while the LSTM captures temporal dependencies across sequences, enhanced by an attention mechanism that prioritizes emotionally salient facial regions. Control variables such as data augmentation, normalization, and frame resolution are incorporated to optimize performance. Findings indicate superior performance compared to baseline CNN and LSTM models, with implications for mental health monitoring and human-computer interaction. The study contributes to affective computing by demonstrating robust methodologies for emotion recognition under varying conditions.