Time-series analysis of microtiming deviations in the Rosanna shuffle

Abstract

Musical rhythms played by humans are not perfectly precise compared to a metronome. The deviations from a perfect temporal grid are called microtiming deviations. They are unavoidable in music performed by humans and range from a few to tens of milliseconds. According to previous studies, these deviations are have a complex structure and they are long-range correlated to a certain degree.

In this thesis, different methods in time series analysis are applied to one of the most famous drum grooves of all time, the Rosanna shuffle played by the legendary Jeff Porcaro. The long range correlations of the interbeat intervals and amplitudes are studied with detrended fluctuation analysis, a widely used method in non-stationary time series analysis. The results suggest that the previously found long-range correlations in Jeff Porcaro’s playing are also present here, both in timing and dynamics.

Additionally, the amplitude and interval variations in the two bar phrase constructing the Rosanna shuffle are analysed, and specific structural patterns are found. The Rosanna shuffle is usually written in a triplet-feel shuffle, but the intervals differ from the musical notation, and the swing ratio of the song is found to be lower than expected.