A global pandemic of rhythmic restlessness has struck regular concert-goers the world over. This illness typically manifests itself in a desire to conduct recordings of the Berlin Philharmonic, or some other renowned orchestra, from one's living room sofa. Many who fall foul to this disease are left with the bitter aftertaste of unrequited love however, as the lack of interactivity in most of today's music playback devices precludes any response from the music.
This report details the development of an interactive, vision-based system that addresses this problem, providing an application through which amateur conductors can improve their conducting technique. The focus is on determining the conductor's [potentially varying] tempo and playing back the music in time with it. Achieving this requires baton tracking, beat detection and beat prediction.
A suitable tracking algorithm for this application must be able to track the baton in real time against a cluttered environment. We will explore the use of the condensation algorithm as a potential baton tracker. To evaluate its accuracy, we will compare its performance to that of a shape recognition-based tracking algorithm that performs well in uncluttered environments. We will also investigate methods for detecting beats after they have occurred and predicting the time at which the next beat will occur.