In this paper, we build a hybrid personalized music recommendation system that takes into account the user's previous listening history, real-time body parameters such as heart rate and interest of similar users. We build a unique unsupervised hierarchical model that combines all these attributes and computes the best song recommendation to the user.