An Environmental Context-Responsive Music Recommendation System with Scenic Images
Atsushi Kirimoto, Shiori Sasaki, Yasushi Kiyoki · 2008
This paper presents an implementation method of an environmental context-responsive music recommendation system. First, a user's environmental context is identified by a scenic image input by users in this system. Second, the correlation between environmental context and impression (feeling data) are calculated. Third, the highly-correlated music data to the impression are selected and recommended to the user. The main features of this system are (1) to set the relations between real scenery, environmental context and colors as color-environment database by numerical values, (2) to identify the environmental context of users and extract the impression of an environmental context from scenic images acquired by camera devices in real time, (3) to set a space which represents the relations between the environmental context and music through the attributes of impressions, and (4) to calculate the correlation between the impression of the environmental context and music data on the space. To examine the feasibility of the system, we have performed several experiments by setting images of sky as data, identifying the high-correlated whether to the target data, extracting the impression of the whether, and selecting the highly-correlated music data to the impression of the whether. Keyword Information Recommendation, Multimedia, Personalization, Semantic Associative Search