Graph-based audio looping and granulation

Gerard Roma, Pierre Alexandre Tremblay, Owen Green
CeReNem, University of Huddersfield

In this paper we describe similarity graphs computed from the analysis of audio as a guide for audio playback, with the aim of extending the content of fixed recordings. We explain the creation of the graph from time-frequency analysis, as well as methods for onset detection and cluster analysis derived from the same distance measure. Several playback algorithms can be devised based on conditional pruning of the graph using these measures. We describe examples for looping, granulation, and automatic montage.

This page contains audio examples for the proposed approach.

Implementation

Source code can be donloaded from github
Precompiled binaries can be downloaded here:

Audio examples

Example 1: Looping

Source file Query Result

Example 2: Granulation (music box)

Source file Result

Example 3: Granulation (drums)

Source file Result

Example 4: Granulation (deterministic graph loops)

Source file Result

Example 5: Montage (applause)

Source file Result

Example 6: Montage (helicopter)

Source file Result