Facial Emotion Identification Using Convolution Neural Networks: Enhancing Human-Computer Interaction

Facial Emotion identification uses technology to classify the various emotions of humans. Human-Computer Interaction is an emerging field that uses deep learning algorithms to classify human feelings. Convolution Neural Networks (CNN) are the groundbreaking technology to process images in an efficient manner. This paper explores deep learning models on FER-2013 dataset for emotion prediction to analyze facial expression and classify them into different emotions such as Angry, Disgust, Fearful, Happy, Sad, Surprise and Neutral. Convolutional Neural Network (CNN) and ResNet-50 are the two architectures used for this research work. Both the models were trained at 35 epochs. CNN gives 97.83% accuracy and ResNet-50 gives 97.74% accuracy on the FER-2013 dataset. Thus, the result shows that both models are doing well in learning from training data and even have nearly the same performance. The findings of the study hence emphasizes that CNN models have high potential for performing efficient and accurate emotion recognition, but they still require more improvement in terms of generalization on unseen data.