Spectral Modeling of Musical Sounds

ReSpect is a free software library performing the realtime synthesis of spectral sounds. The origin of this library is an additive synthesis module, implemented using software oscillators and presented in Paris at the JIM'99 conference. On the top of this efficient synthesis module were added high-level functionalities such as a psychoacoustic model and a spatialization system, all working in a spectral representation. Re-Spect accepts several spectral sound models as input. Thus, this library can be used for research in the spectral modeling field, as well as for teaching purposes, since the sources are available. Moreover, its modular approach allows to enable or disable a specific module, which is extremely useful for pedagogy. Its final goal is to serve as a software tool to be included in bigger projects. The ReSpect library is used for example in the Dolabip project, a multi-field project for developing early-learning games for electro-acoustic music at school. It is also used in other computer music software tools of the SCRIME, more oriented towards the composition of electro-acoustic music.

2004

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

1996

The purpose of this thesis is to develop a sound model that can be used as a creative tool by professional musicians. Post-production editing suites are used for compiling and arranging music tracks, and for creating soundtracks and voice-overs for the radio, television and film industries. A sound model would bring a new dimension of flexibility to these systems, allowing the user to stretch and mould sounds as they please.

Proceedings of the …, 1997

Spectral based analysis/synthesis techniques offer powerful tools for the processing of sounds. They include many different approaches, each one with its own advantages and set of compromises. In the design of a general purpose synthesizer more than one technique is needed while maintaining a unified control strategy and a single synthesis engine. In this article we present several complementary spectral representations, techniques to obtain them or generate sound from them, and the design of a general purpose software synthesizer based on these techniques.

Technologies and Applications, 2007

This chapter reviews audio signal processing techniques related to sound generation via additive synthesis. Particular focus will be put on sinusoidal modeling. Each processing stage involved in obtaining a sinusoidal representation for audio signals is described. Then, synthesis techniques that allow reconstructing an audio signal based on a given parametric representation are presented. Finally, some audio applications where sinusoidal modeling is employed are briefly discussed.

Spectral models provide general representations of sound in which many audio effects can be performed in a very natural and musically expressive way. Based on additive synthesis, these models control many sinusoidal oscillators via a huge number of model parameters which are only remotely related to musical parameters as perceived by a listener. The Structured Additive Synthesis (SAS) sound model has the flexibility of additive synthesis while addressing this problem. It consists of a complete abstraction of sounds according to only four parameters: amplitude, frequency, color, and warping. Since there is a close correspondence between the SAS model parameters and perception, the control of the audio effects gets simplified. Many effects thus become accessible not only to engineers, but also to musicians and composers. But some effects are impossible to achieve in the SAS model. In fact structuring the sound representation imposes limitations not only on the sounds that can be represented, but also on the effects that can be performed on these sounds. We demonstrate these relations between models and effects for a variety of models from temporal to SAS, going through well-known spectral models.

Sound and Music Computing, 2022

This article describes the use of representations based on data extracted from the audio signal (sonogram, chromagram, audio descriptor) in musicology. A first part gives the historical context of these representations and discusses their transfer from the field of exact sciences to musicology. In the second part, we propose three new types of representation particularly effective in musical analysis: visualizations based on waveform and amplitude, a multilayered sonogram and chromagram, and an interface to assist the interpretation of selfsimilarity matrices for the analysis of musical forms and microstructures.

PREPRINTS-AUDIO …, 1999

The Sound Description Interchange Format (SDIF) is a recently established standard for the interchange of a variety of sound descriptions including spectral, time-domain, and higher-level models. SDIF consists of a specified data format framework and an extensible set of standard sound descriptions and their official representations. We begin with the history and goals of SDIF, followed by the specification of the data format and the standard sound descriptions. Then we describe some current applications of SDIF and conclude with a look at the future of SDIF.

Organised Sound, 1997

andor the frequency of the partials can, for example, synthesis process since it combines sound processing and be manipulated through spectral analysis. Timealgorithmic synthesis within the same formalism. Its aim is frequency analysis allows the separation of the time to make sound simulators by synthesis methods based on and frequency characteristics associated with the signal models or physical models, the parameters of which sound and is of great interest (Kronland-Martinet are directly extracted from the analysis of natural sounds. 1988). However, this approach conflicts with a very In this article the successive steps for making such systems important mathematical principle which states that are described. These are numerical synthesis and sound one cannot arbitrarily modify a time-frequency repgeneration methods, analysis of natural sounds, particularly resentation of a signal. This constraint is due to the time-frequency and time-scale (wavelet) representations, existence of the so-called 'reproducing kernel' which extraction of pertinent parameters, and the determination of the correspondence between these parameters and those takes into account the redundancy of such represencorresponding to the synthesis models. Additive synthesis, tations (Kronland-Martinet et al. 1987). It correnonlinear synthesis, and waveguide synthesis are discussed.

This paper describes a novel analysis-synthesis method based on estimation and comparison between spectral methods that are specifically designed for modeling of random signals. These methods are an alternative to the discrete Fourier transform, providing high-resolution, smooth and easily interpretable spectral estimation over short signal frames. It is generally acknowledged that musical signals that exhibit both periodicity and variations can be modeled as sinusoidal components with various extents of modulation (generally characterized as band-limited frequency components), and remaining noise or wide-band frequency components that describe unvoiced and other noisy parts of the signal spectrum. Using single frame Fourier analysis makes is difficult to tell if spectral energy at a particular frequency is due to noise or sinusoidal components. Existing methods for such decomposition usually consider properties of Fourier phases, which are noisy. Our method decomposes a signal into sinusoidal, modulated sinusoidal and noise components based on a comparison between two spectral representations, namely autoregressive (AR) and minimum variance distortionless response (MVDR), both calculated from linear prediction coefficients (LPC) in a time varying manner. Using different optimal properties of each model, we develop estimators for frequencies and amplitudes (spectral envelope) of sinusoidal components (spectral lines) in noise and derive a "noisality" index that assigns different weights to contributions of sinusoidal and noise components at every frequency. Examples of synthetic and real sounds are presented in the paper.