The Thesis describes an approach to generating continuous control data for shaping parameter motion of a synthesizer via a gestalt theory of structural time and of melodic expectation. The set of motions arises from analysis of implicative contents acquired from a musical performer’s alteration of a linear potentiometer. The results prove musically relevant, allowing a performer to generate a set of motions from a single slider, each motion operating on different a structural timescale, and allowing a sense of non-linear motion with overall direction to appear in the resultant sound quality.

The dissertation delineates several notions and conceptual tools of fundamental importance in our computer music practice, and proposes concrete applications of these concepts.

The first chapter is centered around the dialectic of matter and form, from the perspective of aesthetic philosophy, and afterward within the musical discourse. In particular, we discuss the dualism empiricism vs. formalism, a prevalent binary opposition in computer music aesthetics. We generalize this dualism as the opposition signs vs. concepts. While acknowledging the pertinence of this opposition, both from the perspective of the aesthetics and the composition methods, we show that, in practice, the situation is more nuanced, and most computer music pieces exist somewhere on an axis drawn between pure empiricism and strict formalism. The following section discusses three compositional strategies to organize the sound material in a musical form: the form as a result of a process, sound transformations and morphological variations as structuring processes, and articulation of space as structuring element.

The second chapter is centered around the notions of performance, bodily presence and human agency in computer music. We show that the increasing use of electronics and digital technologies in musical practices has dislocated our notion of liveness and magnified the possible interactions with the sound material. We discuss the consequences of the absence of the body or of bodily gestures in computer music performances, and we establish a network of relationships between timbre, attribution of causality, inference of intention, and emergence of a virtual presence, that we consider are at the source of fruitful and creative ambiguities in computer music.

The third chapter presents our answers to these questions, in the form of six electronic music studies within a performance for dance, video and live electronics. These studies are based on a combination of direct interventions, improvisation and algorithmic processes. Several tools and techniques used in these pieces are presented, notably a modal synthesis model with morphing capabilities, a fractal-based algorithm to generate material at different time scales, a virtual analog extended model of the Xpander filter with morphing capabilities, and two reverberations able to generate abstract but cohesive spaces.

Compared with more traditionally oriented mediums of artistic creation, art that is time-based, technologically specific, site specific, and intended to be “permanent” poses unique challenges when considering its potential longevity. Institutions engaged in holding digital cultural objects must address the same problems of technological obsolescence that have been central within the development of digital technologies generally, as well as similar questions of physical preservation that arise in the conservation of any cultural object. Such works require rigorous and specific ongoing maintenance while also contributing significant economic stresses to the collecting capacities of the institutions that hold them. John Luther Adams’ The Wind Garden and The Place Where You Go To Listen, with which I continue to be closely involved, face very real and complex problems for their continued existence. Both of these works rely on a computing architecture which, given enough time, will cease to run on any computer without significant translation and redesign. While both works have institutional affiliations, the former as part of the Stuart Collection at the University of California, San Diego, and the latter under the auspices of the University of Alaska’s Museum of the North, each demands a continuous allocation of knowledge, resources, and attention that outstrips either institutions’ abilities to simply “hold” them as they would more traditional media. Similar questions of conservation and longevity are currently being asked by the Nam June Paik Art Center regarding the future of Paik's Something Pacific, and other related works by Paik. In discussing the specifics of my own co-creation of these works, I address the wide array of issues surrounding the genesis and preservation of this type of work, including problems of the very definition of a work’s “objecthood”, the economic, technological, and cultural forces that threaten our ability to preserve this work, our cultural conceptions of “permanence”, and the various philosophical binds encountered by other related artworks.

In this paper, I first develop real-time applications including virtual instruments and plate reverb using the Kirchhoff plate model with loss and tensioning terms. Finite-difference time-domain (FDTD) methods are employed, and they are implemented on central processing units (CPUs) and optimized by loop unrolling or advanced vector extensions (AVX), enabling these programs to be executed in real time at fast speeds. Then I present a numerical method for solving the biharmonic equation using finite difference methods, which can be used for solving part of the von-Kármán plate equation. With the simply supported boundary condition, the linear system could be regarded as a composition of two Poisson's equations, and these Poisson's equations are solved by the Thomas algorithm for a series of tridiagonal systems after transpositions and linear transformations for vectors in the systems and all non-empty blocks of the Laplacian matrix. We also point out that the matrix decomposition used for these linear transformations has a closed-form formula, which is easy to be pre-computed and also space-saving. Furthermore, since this solver is computed block by block and does not need sparse matrix operations, this method is good for single instruction multiple data (SIMD) parallelization using AVX intrinsics on CPUs, which makes it possible to execute at fast speeds for real-time music applications. We also show that this solver for the simply supported boundary condition can also be easily adapted for other boundary conditions using Woodbury matrix identity with a little extra complexity. Numerical experiments show that the C++ implementation of this method is faster than decomposition-based solvers (like LU or Cholesky decomposition) of some well-known C++ libraries at the scale of applications in the field of musical acoustics. Finally, applications such as virtual plate instruments and plate reverb using FDTD schemes are developed as puredata (Pd) externals which can serve as objects in puredata, a real-time graphical programming environment for audio and graphics. In these applications, multiple inputs (excitation or audio signal) and outputs whose positions are free to change in real time are allowed, and physical parameters can be dynamically manipulated in real time, allowing users the ability to produce both realistic sound and new sounds not possible to generate in the real world. Additionally, these Pd externals can also be used as modules to build Pd patches, which provides more possibilities for experimental artists.

The cocktail party effect describes the human ability to detect a specific sound of interest in a noisy environment. Scientists have evaluated different influential parameters affecting auditory stream segregation. Previous studies have focused on individual cues while the interaction between them appears to be poorly studied. This thesis investigates the interactions between the temporal and spatial

cues, and the spectral and spatial cues in a cocktail party scenario. Two streams were presented at the same time from ±7.5◦ or ±45◦ or ±90◦ . One stream (the target) was to be attended, and the other stream (the masker) was to be ignored. The stimuli were frequency-modulated narrowband noise. The modulation-rates were randomly chosen between 0.7 and 3.3 Hz (prosodic fluctuations in natural speech) so that the interaction of modulation-rate difference and stream segregation could be examined. Transient gaps were embedded in both streams and the task was to report only the gaps in the target. If the subject could detect gaps in the target (not masker), the stream segregation was accurate. Performance was evaluated as a function of source separation using signal detection theory. The overall performance was significantly poorer at ±7.5◦ than at ±45◦ and ±90◦ . The task was evaluated as a function of the difference between the carrier frequency of target and masker that was chosen randomly. The results showed that the effect of the frequency difference on performance was significant at all the angles and it was significantly stronger at ±7.5◦ than at ±45◦ . he task was analyzed as a function of the modulation-rate difference between the streams and showed it did not play the role of a segregation cue at any angle. The reaction time at ±7.5◦ was significantly slower than that at ±45◦ and ±90◦ . In summary, consistent with previous work on stream segregation, spatial cues influenced performance. When they were not strong enough cues, making a decision took a longer time. Frequency was a stronger cue for stream segregation when spatial cues were less clear. Regardless of the location in space, no interaction between the rate and space was found.

Coupled Oscillator Systems (COS) are presented as a method for sound synthesis and musical instrumentation. Seven systems ranging in size from 3 to 9 oscillators are described with different arrangements all containing a topology of velocity based coupling terms. First, each system is described mathematically. Then, each system is analyzed with recurrence plots and recurrence quantification analysis. The analysis characterizes the non-linear dynamics of the output signals and shows why certain ways of structuring COS may be preferable to others. The goal of the COS designs presented is to increase the non-stationarity of the signal, while also increasing the coherence of the phase transistors over an (quasi-) attractor landscape. Varieties of dynamical phenomena are located and quantified, such as drifts, phase transitions and higher order periodicities. The durations of residence in attractors also reflected higher order temporal proportions. Of key interest, as found by analysis, was the presence of chaotic itinerancy (CI). Orderly CI in neural dynamics corresponds to production of intentional sequences of action and thought. Here cases of synthesis dynamics are located that reflect the quantitative aspects of orderly itinerancy and as a result the production of abstract musical sequences with perceived intentional organization. A theoretical chapter considers novel concepts, such as, extended and embedded intentionality in conjunction with auditory cues to discuss the implications of this tool as an instrument for musical improvisation. Finally, if the phenomena produced by a COS sound musical and are consistent with the phenomenon of CI observed in the brain, can we consider the brain and its dynamics to be a kind of musical phenomenon itself?

A thesis about how to construct digital audio synthesis unit generators using dynamical models and what they sound like.

The goal of this research is to explore the use of genetic algorithms to evolve waveforms. Genetic algorithms are introduced with a simple application that evolves novel tuning systems. Research involving the application of genetic algorithms to musical situations is reviewed. A framework for applying genetic algorithms is described in terms of virtual biology and virtual ecology. Virtual biology applies the central concepts of genetic algorithms (genetic representation, reproduction, fitness, mutation) to waveforms. Methods for implementing virtual biology are described in detail for time-domain waveforms and are proposed for spectral domain waveforms. Virtual ecologies replace the simple fitness function of conventional genetic algorithms. These ecologies can be designed to produce formal musical structure in an algorithm's output. Several algorithms that employ the framework are described in detail. The output of a simple version of the algorithm is systematically evaluated by running several fixed sets of sounds and environments with active evolutionary parameters. Works produced by more complex algorithms are evaluated subjectively. The success of this project is not in this algorithm's ability to make a population of sounds sound more like their environment, but rather in its ability to create novel sounds that are intimately tied to the process of their creation.

An analysis and synthesis technique is described for the estimation and modeling of a particular class of nonlinear acoustical systems with energy-preserving nonlinearities. It is situated in the field of nonlinear sound synthesis techniques. The rich mathematical structure behind the technique is explored, with particular attention to the effects of scaling to arbitrarily high-dimensional systems. The technique is then applied to simulations and recordings of nonlinear vibratory behavior. After estimating model parameters from the timeseries, the models are validated using various assessment techniques, including resynthesis. Applications to computer music are discussed.

Cross-synthesis is the process of blending two or more audio signals to produce a hybrid signal with timbral characteristics of the originals. There are a number of methods to perform cross-synthesis of digital audio including but not limited to vocoding, phase vocoding and convolution. Convolution is a common approach to perform cross-synthesis on arbitrary sound files but the process suffers from inflexibility. Having no control over the process itself, a musician interested in convolutional cross-synthesis is left to modify the original sounds to influence the outcome.

There are a number of drawbacks with the process of convolutional cross-synthesis that impede its musical usefulness. One such drawback is the lack of control over which sound the outcome more closely resembles. Another is a perceived absence of high frequency energy in the outcome. This thesis will introduce novel extensions to the process of discrete convolution for the purposes of offline cross-synthesis that attempt to remedy these concerns.

Discrete convolution and the proposed extensions will be analyzed for their effect on acoustic features. This black-box analysis consists of subjecting a novel, heterogeneous dataset of sounds dubbed RFS1k to convolutional cross-synthesis and examining the average effects on a set of features.