Mymtv: a Personalized and Interactive Music Channel

In this document, we present MyMTV, an Interactive TV which adapts to one’s habits, tastes and moods: A music recommendation channel, tailored to the users tastes. Users create a personalized channel of videos by simply selecting a song or an artist they like. The system identifies and log what music video the user is watching. Based on this information, the system builds a user profile to improve the quality of future recommendations. User profiles are stored and managed in order to adapt the personalized stream of content and fit the users tastes, habits and moods. Add-on services are also offered related to the content, such as song lyrics and information about artists. Existing recommendation services are all technically reliant on collaborative filtering techniques. Here we present a Hybrid Recommendation Framework combining userprofiling and collaborate filtering with the multimodal audiovisual content-based approach, thus leveraging the benefits of both worlds in a single system. Both audio features and collaborative filtering are used for recommendation. Users can rate the songs and this information is employed to improve the recommendation quality. Similarity measures between music videos are computed, as well as mood annotation, which helps the user to find the content he would like and that fit to his mood. This demo features all these functionalities.