Title: RecMirror: A Multi-Modal Intelligent System for
Multimedia Recommendations
Abstract
Traditional recommendation approaches do not consider the changes of
the users preferences according to context. Recommendations
Can be adapted to the context of the user using a collaborative
filtering approach. In this project, we propose a designing
and developing method to facilitate the interactions between the
user and a recommender system by incorporating different interfaces
and response techniques. In specific, we propose an interactive
mirror to extract the best two-dimensional shape of the human body
skeleton. Then, we can identify who's standing in front of the
mirror using a camera. A user interface is developed to map the
skeleton points tracked by a Kinect camera using a star skeleton
strategy. Then, the user recognition is used into building a
personalized recommendation outputs such as multimedia contents
(e.g., movies and music).