Combining timbric and rhythmic features for semantic music tagging

In this thesis we propose a novel approach to semantic music tagging. The project uses a modified Hidden Markov Model to semantically link two acoustic features. We make the assumption that acoustically similar songs have similar tags. We model our known collection as a graph where the states represent the songs and the model's probabilities are related\nto the timbric and rhythmic similarity. Tags are inferred from songs in acoustically meaningful paths, all starting from the query song