Measuring groove: A Computational Analysis of Timing and Dynamics in Drum Recordings

TitleMeasuring groove: A Computational Analysis of Timing and Dynamics in Drum Recordings
Publication TypeMaster Thesis
Year of Publication2017
AuthorsTroes, T.
AbstractThis thesis presents the implementation of a generative groove system, improving on currently available humanisers for electronic music production. This is achieved by complementing perceptual observations about rhythm and groove with the computational analysis of musical recordings of real-life drummers. First, an extensive literature review on rhythm perception, groove and groove generation is carried out and a working definition for groove is given. Providing additional downbeat annotations for the ENST-Drums dataset, we measure the micro-timing deviations from an idealised metrical grid as well as the amplitude fluctuations of three different drummers. These results are interpreted through various graphical representations and validated by a statistical analysis, making use of various machine learning techniques. A better understanding of each drummer’s behaviour on a sixteenth-stepand beat-level is gained; furthermore, it is found that the genres of “disco” and “rock” can be characterised by their micro-timing deviations. We then design an algorithm for the generation of genre-specific micro-timing deviation patterns, which then trigger events in a digital drum machine. The resulting loops can be correctly classified, thus providing us with a proof of concept that we can better understand the generation of drum grooves and build more natural, multi-dimensional computer-based systems.
KeywordsDrums, dynamics, Electronic Music Production, Groove, Micro-timing, music perception