Sentiment based music play system

Abstract

Most of us hear music to experience emotions. Music can soothe your bad mood. Music systems available at present times allow you to play selected songs and recommends songs in genres based on your tastes or tastes of users similar to others. Since such music systems are not designed with the emotions evoked in mind, music listeners cannot completely rely on such systems and do not enjoy listening to songs on the station or website. In this paper, we demonstrate a sentiment based music system. Our raspberry pi based system, in conjunction with a speaker and a microphone plays songs on the basis of the mood in the room. The captured background voices are converted to text and their sentiment is determined using machine learning based classification problem. We use a naive Bayesian classifier for this classification. Songs with similar sentiments are determined using the tempo of the song in BPM (Bits per Minute).