PERSONALIZED MUSIC RECOMENDATION BASED ON EMOTIONAL STATES USING YOUTUBE

Abstract

Music is a very influential medium
that impacts human emotion in profound ways.
The traditional music players need users to
manually select songs while they lack the
capability to detect emotional states
automatically. This paper presents an AI,
multimodal emotion-based music
recommendation program which uses facial
expression analysis to recommend music
through a chat-bot from YouTube in real-time.
This approach uses a combination of an image-
based Haar cascade classifier for emotion
recognition and a fine-tuned Hugging Face
model to improve accuracy in predictions. The
chat-bot will use keyword-based Natural
Language Processing to simulate a
conversation with the user to improve emotion
detection by analysing text.