I am a part of the CompMusic project, which aims to advance in the automatic description of music by emphasizing cultural specificity. It carries research within the field of music information processing with a domain knowledge approach. Within CompMusic, I focus of rhythm related problems in the music cultures under study. My work so far has been on Bayesian models for meter inference and pattern discovery, mainly in Indian art music. The research goals include a complete description of rhythm through these structures and patterns, to define content based culture specific rhythm similarity measures. 

My current work is on developing novel Bayesian models for tracking metrical structures from audio music recordings and has been applied mainly on Indian art music. Rhythm in Indian art music (Carnatic and Hindustani music) is organized around the framework of tala, which consists of cyclical metrical structures. Automatic estimation of different components of the tala, such as the cycle length, tempo, beats and the downbeat (sama) are some of the tasks I have been focusing on. 

Percussion in some music cultures including Indian art music are based on onomatopoeic oral mnemonic syllables. These syllables are vocalized and extensively used in music training, and hence provide a language for percussion and are ideal for representation of percussion patterns. My current work has also been to explore the use of such syllables for representation and discovery of percussion patterns from audio music recordings. With a significant analogy between these percussion patterns and speech, I have been exploring the speech recognition framework for the task. 

A significant effort in CompMusic has been to build accurately labeled and well curated research corpora and datasets, for use with various Music Information Research tasks. Some of the datasets created within the context of CompMusic are here: