Emotion Based Song Suggestion System for Tamil Language

Music has become an integral part of our daily lives. An individual’s mood can often be determined by looking at their face. The aim of this paper is to suggest a playlist by detecting the current emotional state of an individual. A subset of the FER2013 (Facial Emotional Recognition) dataset is used in this study for facial expression detection along with the dataset from music streaming platform. Different machine learning models like Decision Tree, Adaptive Boosting, Extreme Gradient Boosting, Gaussian Naive Bayes, Ensemble Learning models are used to train and test our later dataset. The MobileNet architecture was trained and tested on the former mentioned dataset to detect facial emotions. The proposed models are evaluated using the confusion matrix from which the accuracy, precision, and recall scores are extracted. Ensemble Learning achieved an accuracy of 89.6% while the MobileNet model obtained a training accuracy of 75% and validation accuracy of 62%.