An Intelligent Personalized Music Recommendation System Using Content-Based Filtering with Convolutional Recurrent Neural Network

N owadays, the emotion of every individual human is detected according to the user's facial expressions. The rapid growth of technologies in music information retrieval, personalized music recommendation systems remain in the beginning stage only. However, the personalized music suggestions are common, but, recommending songs based on emotions remains a significant challenge. Hence, this research proposes a Deep Learning (DL) based Convolutional Recurrent Neural Network with Content-Based Filtering (CRNN-CBF) model to recommend personalized music based on emotion. At first, the data gathered from Facial Emotion Recognition 2013 (FER2013) dataset and the collected images are preprocessed by median filter, histogram equalization, and a wiener filter. Then, features are extracted by using Linear Binary Patterns (LBP) method and then classified using CRNN, categorizing images into seven distinct types of user emotions. Next, spotify Application Programming Interface (API) is employed to deliver playlists and finally, CBF is introduced to recommend personalized music according to user interests. From the results, the proposed CRNN-CBF model achieved outstanding results in accuracy 95.73%, precision 94.86%, and recall 96.78% when compared with existing model such as Emotion-Aware Music Recommendation System by using Deep Neural Network (EA-DNN).