Emotion-Based Music Recommender System for Tinnitus Patients (EMOTIN)

Publication Date


Document Type

Contribution to a Book

Publication Title

Studies in Computational Intelligence





First Page


Last Page



The chapter presents a personalized recommender system (RS) for tinnitus therapy, called EMOTIN. It is based on the automatic indexing of music by emotions. A framework and an algorithm for emotion-based music recommendations for tinnitus patients are proposed. The contribution is expected to be novel, as nobody yet proposed a music RS specifically to target negative emotions associated with tinnitus. The chapter is organized as follows. First, the medical problem of tinnitus, a common but non-curable hearing disorder, is presented. An overview of existing therapies, based on tinnitus retraining and music interventions, is provided. The background information includes work done previously in this research, including RECTIN (Recommender for Tinnitus) system and its prototype GUI-based version—eTRT (electronic Tinnitus Retraining Therapy). Next, the methods used to enhance an existing prototype by adding the music recommendation functionality are presented. The high-level overview of the content-based approach for music recommendation is described. A formal definition of the music recommendation is given and a recommender algorithm based on neighborhood approach and similarity metrics is described in detail. The methods to build a user profile are proposed, including user input data and a dimensional model of emotions. For item content analysis, an approach based on music emotion recognition and audio notching is outlined. The preliminary results on music item content analysis include datasets preparation, audio feature extraction, and testing regression models for music emotion recognition. Discussion includes future directions for this research, including full implementation of the system and evaluating the system’s recommendations with tinnitus patients.


Computer Science