Content-based Representations, Indexing and Retrieval of Music

The proposed research involves development of methods for content-based indexing of music databases using a combination of signal processing and knowledge-based methods, design of statistical algorithms for enabling queries using sung or hummed melodies, and design of robust search techniques for retrieving the queried information especially in the presence of uncertainty. The research approach is user-centric, and can be subdivided into three major topics:

• Representation and Indexing: Algorithms for organizing musical information utilize representations that exploit theories and knowledge about human musical intuition, and music perception and cognition.
• Query Formulation and Interaction Modality: The user interacts with the music database through humming, a natural activity. Algorithms mapping the user's query to music symbols are designed to cater to all levels of musical abilities. Expert knowledge of the desired piece of music is not required.
• Search and Retrieval: Algorithms for retrieval i.e., matching user query against the database, will tolerate and be robust to uncertainties and errors. Such errors occur in the user's query and can also be introduced in the audio signal-to-symbol transduction.

NSF Report
Poster
More Info
Laboratory