Song Playlist Generator System Based on Facial Expression and Song Mood

Because of the hectic pace that people have nowadays, life is incredibly hectic. People are increasingly inclined to listen to music while performing their daily duties which help them relax after a stressful day. As a result, songs become important part of daily lifestyle. Due to the huge demand several music players have entered to the market and try to attempt to deliver the best possible music recommendation for the customer. This paper proposes a Deep Learning based approach for the playlist generation based on human current mood with the help of user’s past history of song selection. In this approach we are trying to generate playlist from the emotion of the user to add touch of current situation of user mood and user personal choices of the songs for providing more personalized experience. After introduction of the Convolutional Neural Network object detection, Image classification, Emotion detection tasks reaches great height. In the proposed method, we use convolution neural network (CNN) for emotion detection task and artificial neural network (ANN) for the song classification task. Experiment result says that our suggested model achieve 84% accuracy on FER-13 dataset which contain around 14k facial images. For song classification task we have used different song-features which is extracted from Spotify music player. We have achieved 82% accuracy in song classification task. Currently this system is only with Spotify music player. Motivation of this approach is to provide better song recommended playlist based on user current mood.