Facial expression analysis is largely driven by the capabilities of machine learning algorithms: A Comprehensive Study

Facial expressions have consistently revealed emotions to observers. Facial expression Recognition (FER) has made significant progress in the computing field due to advancements in image processing and machine learning. Machine learning algorithms have led to numerous innovations in human-computer interaction and other areas. Facial expression is non-verbal communication that includes thoughts and emotions without using any words. We mainly express ourselves by expressing 7 facial emotions like: happy, sad, fear etc. This paper seeks to elucidate the fundamental principles of Facial Emotion Recognition (FER) and to compare current research in the field of Machine learning approaches such as transfer learning, CNN, RNN and traditional machine learning approaches including SVM and k-NN. The result for each algorithm on facial expression classification reveal that the LSTM-CNN technique achieved the highest accuracy at 99.43%. In contrast, the DNN classification method achieved an average accuracy of 89%. Conversely, the use of the DCNN technique resulted in a lower accuracy of 49.8%.