A Simplified Granular Processing Interface Based on Perceptual Research

April 16th, 2006

This is my dissertation that was completed to fulfill the requirements of my PhD at Northwestern University. The findings informed my later plug-in design work for Hipno, a set of plugins published by Cycling74. The chapter 2 contain a primer on granular methods that many people have told me they found useful.  Chapter 3 reviews granular research to date and many of the earlier software that implemented these techniques.  Even if you don’t dive into the actual experimental data, these may of interest.

ABSTRACT – Granular processing is a computer music technique that manipulates “grains” of sound to produce a variety of effects. Grains are produced by multiplying short segments of digital audio, typically lasting between 10 and 50 milliseconds, with an amplitude envelope of equivalent length. Software designed to produce granular processing effects often requires the user to manage multiple parameters that lack a clear connection to the audio output. A better understanding of how listeners perceive the processing output should yield insights into how the user interface could be simplified.

A series of three experiments was designed to investigate how listeners perceive differences between granular processing examples. Stimuli were produced using specific program settings to process two distinct sound sources. In each experiment, at least twenty participants were asked to rate the similarity of each possible pair of stimuli including identity pairs that existed among these stimuli. The author then used multidimensional scaling (MDS) to develop a graphical representation of the perceptual organization exhibited by participants.

Differences between stimuli included the processed sound source and settings for the grain duration and grain period parameters. Visual analysis of the MDS solution showed that participants clearly distinguished between the two sound sources. Processing descriptors based primarily on the review of literature were tested for correlation to the MDS dimensions. This analysis revealed the significance of three processing features: (1) a base-2 logarithmic scaling for differences in grain duration, (2) the minimum and maximum boundaries for randomized grain durations, and (3) the mean value and total deviation for randomized grain periods.

Between-subject variables relating to experience with electroacoustic music were also examined. However, the results of this secondary inquiry were deemed inconclusive overall based on the relationship between participants’ responses to pre-experiment questions and a priori operational definitions.

The findings were used to inform the design of a new graphical user interface (GUI) for granular processing. The resulting GUI helped to verify this study’s conclusions by successfully demonstrating their practical application to software development. The GUI features unique controls for managing randomization and a feedback display for monitoring differences between the control input and audio output.

Leave a Reply