Tabla Duos at Distance with Prediction

by Mihir Sarkar
High Quality Audio over Networks Summit (ANET II), The Banff Centre, Banff, Alberta Canada, Apr. 11, 2008.

For playing music such as the highly rhythmic Indian tabla over computer networks, and minimize the effects of network latency, we need intelligent systems that go beyond transmitting compressed audio signals or low-level musical events. In order to cope with the tight synchronization and timing constraints required by such music, I developed a system that draws on machine listening and predictive analysis. Trained for a particular instrument and musical style, my system recognizes musical events like drum strokes and pitch, and sends hierarchical symbolic structures that map to musical features like phrasing and tempo. These symbols shape predicted rhythmic phrases in a musically meaningful way that aims to preserve the performers' intent even as prediction approximations may result in a slightly different experience at both ends of the interaction. This system is being evaluated for distance education, interactive collaboration, and network performance.

More details...