Musical Signal Processing

Curtis Roads, Stephen Travis Pope, Aldo Piccialli, Giovanni De Poli, Curtis Roads, Stephen Travis Pope, Aldo Piccialli, Giovanni De Poli

• Publisher: Routledge
• Year: 2013
• Print ISBN: 9789026514821
• eBook ISBN: 9781134379774
• Topic: Media & Performing Arts
• Subtopic: Music

Compiled by an international array of musical and technical specialists, this book deals with some of the most important topics in modern musical signal processing. Beginning with basic concepts, and leading to advanced applications, it covers such essential areas as sound synthesis (including detailed studies of physical modelling and granular synthesis) ,control signal synthesis, sound transformation (including convolution), analysis/resynthesis (phase vocodor, wavelets, analysis by chaotic functions), object-oriented and artificial intelligence representations, musical interfaces and the integration of signal processing techniques in concert performance.

Part I

Foundations of musical signal processing

Part I

Overview

Curtis Roads

All fields of signal processing have enjoyed success in recent years, as new software techniques ride the surging waves of ever faster computer hardware. Musical signal processing has gained particular momentum, with new techniques of synthesis and transformation being introduced at an increasing pace. These developments make it imperative to understand the foundations in order to put new developments in their proper perspective. This part brings together three important tutorials that serve as introductions to the field of musical signal processing in general.

The development of sound synthesis algorithms proceeded slowly for the first two decades of computer music, due to the primitive state of software and hardware technology. Today the inverse holds: a panoply of analysis and synthesis techniques coexist, and the sheer variety of available methods makes it difficult to grasp as a whole. Chapter 1 by Gianpaolo Borin, Giovanni De Poli, and Augusto Sarti serves as a valuable orientation to this changing domain, putting the main methods in context. Beginning with sampling, their survey touches on fundamental techniques such as additive, granular, and subtractive synthesis, with special attention given to physical modeling—one of the most prominent techniques today, having been incorporated in several commercial synthesizers. As the authors observe, effective use of a synthesis technique depends on the control data that drives it. Therefore they have added a unique section to their chapter devoted to the synthesis of control signals.

First proposed as means of audio data reduction (for which it was not ideally suited), the phase vocoder has evolved over a period of three decades into one of the most important tools of sound transformation in all of musical signal processing. In Chapter 2, Marie-Hélène Serra presents a clear and well-organized explanation of the phase vocoder. The first part of her paper is a review of the theory, while the second part presents the phase vocoder’s primary musical applications: expanding or shrinking the duration of a sound, frequency-domain filtering, and cross-synthesis.

Chapter 3 presents an innovative phase vocoder that divides the processing into two parallel paths: a deterministic model and a stochastic model. As Xavier Serra points out, the deterministic analyzer tracks the frequency trajectories of the most prominent sinusoidal components in the spectrum, while the stochastic analyzer attempts to account for the noise component that is not well tracked by the deterministic part. In past systems this noise component was often left out, which meant that the transformations realized by the phase vocoder were sometimes stained by an artificial sinusoidal quality. In addition to improving the realism of a transformation, separating the noise component lets one alter the deterministic spectrum independently of the stochastic spectrum. This opens up many musical possibilities, but it is a delicate operation that requires skill to realize convincingly.

1

Musical signal synthesis

Gianpaolo Borin, Giovanni De Poli, and Augusto Sarti

The sound produced by acoustic musical instruments is caused by the physical vibration of a resonating structure. This vibration can be described by signals that correspond to the evolution in time of the acoustic pressure generated by the resonator. The fact that sound can be characterized by signals suggests quite naturally that computing equipment could be successfully employed for generating sounds, either for the imitation of acoustic instruments or the creation of new sounds with novel timbral properties.

A wide variety of sound synthesis algorithms are currently available either commercially or in the literature. Each exhibits characteristics that could make it preferable to others, depending on one’s goals and needs. Technological progress has made enormous steps forward in the past few years in terms of delivering low-cost computational power. At the same time, sound synthesis methods have become more and more computationally efficient and the user interface has become “friendlier”. As a consequence, musicians can—without an enormous investment—access a large collection of synthesis techniques and concentrate on exploring their timbral properties.

A sound synthesis algorithm can be thought of as a digital model for a sound itself. Though this observation may seem quite obvious, its meaning for synthe sis is not so straightforward. Indeed, modeling sounds is much more than just generating them, as a digital model can be used for representing and generating a whole class of sounds, depending on the choice of control parameters. The idea of associating a class of sounds to a digital sound model is in complete accordance with the way we tend to classify natural musical instruments according to their sound generation mechanism. For example, strings and woodwinds are normally seen as timbral classes of acoustic instruments characterized by their sound generation mechanism. It should be quite clear that the degree of compactness of a class of sounds is determined, on one hand, by the sensitivity of the digital model to parameter variations and, on the other hand, on the amount of control that is necessary for obtaining a certain desired sound. As an extreme example, we can think of a situation in which a musician is required to generate sounds sample by sample, while the task of the computing equipment is just that of playing the samples. In this case the control signal is represented by the sound itself, therefore the class of sounds that can be produced is unlimited but the instrument is impossible for a musician to control and play. An opposite extremal situation is that in which the synthesis technique is actually the model of an acoustic musical instrument. In this case the class of sounds that can be produced is much more limited (it is characteristic of the mechanism that is being modeled by the algorithm), but the degree of difficulty involved in generating the control parameters is quite modest, as it corresponds to physical parameters that have an intuitive counterpart in the experience of the musician.

An interested conclusion that could be already drawn in the light of what we have stated is that the compactness of the class of sounds associated with a sound synthesis algorithm is somehow in contrast with the “playability” of the algorithm. One should remember that the playability is of crucial importance for the success of a synthesis algorithm as, in order for an algorithm to be suitable for musical purposes, the musician needs an intuitive and easy access to its control parameters during both the sound design process and the performance. Such requirements often represents the reason why a certain synthesis technique is preferred to others.

Some considerations on control parameters are now in order. Varying the control parameters of a sound synthesis algorithm can serve several purposes. The first one is certainly the exploration of a sound space, that is, producing all the different sounds that belong to the class characterized by the algorithm itself. This very traditional way of using control parameters would nowadays be largely insufficient by itself. With the progress in the computational devices that are currently being employed for musical purposes, musicians’ needs have turned more and more toward problems of timbral dynamics. For example, timbral differences between soft (dark) and loud (brilliant) tones are usually obtained through appropriate parameter control. Timbral expression parameters tend to operate at a note-level timescale. As such, they can be suitably treated as signals characterized by a rather slow rate.

Another reason for the importance of time variations in the algorithm parameters is that the musician needs to control musical expression while playing. For example, effects such as staccato, legato, vibrato, etc., are obtained through parameter control. Such parameter variations operate at a phrase-level timescale. For this reason they can be suitably treated as sequences of symbolic events characterized by a very slow rate.

In conclusion, control parameters are signals characterized by their own timescales. Control signals for timbral dynamics are best described as discretetime signals with a slow sampling rate, while controls for musical expression are best described by streams of asynchronous symbolic events. As a consequence, the generation of control signals can once again be seen as a problem of signal synthesis.

In this chapter, the main synthesis techniques in use today will be briefly presented from the point of view of the user. We will try to point out the intrinsic and structural characteristics that determine their musical properties. We will also devote a section to the problem of the synthesis of control signals for both timbral dynamics (signal level) and musical expression (symbol level).

Synthesis of sound signals

Sound synthesis algorithms can be roughly divided into two broad classes: classic direct synthesis algorithms, which include sampling, additive, granular, subtractive, and nonlinear transformation synthesis. The second class includes physical modeling techniques, which contains the whole family of methods that model the acoustics of traditional music instruments.

Sampling

Finding a mathematical model that faithfully imitates a real sound is an extremely difficult task. If an existing reference sound is available, however, it is always possible to reproduce it through recording. Such a method, though simple in its principle, is widely adopted by digital sampling instruments or samplers. Samplers store a large quantity of examples of complete sounds, usually produced by other musical instruments. When we wish to synthesize a sound we just need to directly play one sound of the stored repertoire.

The possibility of modification is rather limited, as it would be for the sound recorded by a tape deck. The most common modification is that of varying the sampling rate (speed) when reproducing the sound, which results in a pitch deviation. However, substantial pitch variations are generally not very satisfactory as a temporal waveform compression or expansion results in unnatural timbral modifications, which is exactly what happens with an varispeed tape recorder. It is thus necessary to limit pitch variations within a range of a few semitones. On the other hand, what makes the sampling m...

Table of contents

• Cover
• Half Title
• Title Page
• Copyright Page
• Contents
• Contributors
• Preface
• In memoriam Aldo Piccialli
• Part I. Foundations of musical signal processing
• Part II. Innovations in musical signal processing
• Part III. Musical signal macrostructures
• Part IV. Composition and musical signal processing
• Name index
• Subject index