A Hybrid Framework Combining CNN, LSTM, and Transfer Learning for Emotion Recognition

View Full Paper

Ask AI

Bookmark

Authors

Discussion

This study demonstrates improved accuracy in emotion recognition on the CK+ dataset, highlighting the benefits of a hybrid CNN-LSTM approach.

The hybrid cnn-lstm model achieved superior performance in emotion recognition with an accuracy of 92.30% on the ck+ dataset.
ResNet-50 reached the highest accuracy of 76.85% on the fer2013 dataset, showcasing effective architecture performance.
Preprocessing included grayscale conversion and pixel normalization to enhance the image input for the models.
Findings imply a trade-off between computational efficiency and recognition accuracy, key for real-time applications.

Sarvakar et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark

This study demonstrates improved accuracy in emotion recognition on the CK+ dataset, highlighting the benefits of a hybrid CNN-LSTM approach.