Paper Archive

Air absorption effects lead to significant attenuation in high frequencies over long distances and this is critical to model in wide-band virtual acoustic simulations. Air absorption is commonly modelled using filter banks applied to an impulse response or to individual impulse events (rays or image sources) arriving at a receiver. Such filter banks require non-trivial fitting to air absorption attenuation curves, as a function of time or distance, in the case of IIR approximations, or may suffer from overlap-add artefacts in the case of FIR approximations. In this study, a filter method is presented which avoids the aforementioned issues. The proposed approach relies on a time-varying diffusion kernel that is found in an approximate Green’s function solution to Stokes’ equation in free space. This kernel acts as a low-pass filter that is parametrised by physical constants, and can be applied to an impulse response using time-varying convolution. Numerical examples are presented demonstrating the utility of this approach for adding air absorption effects to room impulse responses simulated using geometrical acoustics or wave-based methods.

Download

Physical modelling sound synthesis is a powerful method for constructing virtual instruments aiming to mimic the sound of realworld counterparts, while allowing for the possibility of engaging with these instruments in ways which may be impossible in person. Such a case is explored in this paper: particularly the simulation of a friction drum inspired instrument. It is an instrument played by causing the membrane of a drum head to vibrate via friction. This involves rubbing the membrane via a stick or a cord attached to its center, with the induced vibrations being transferred to the air inside a sound box. This paper describes the development of a real-time audio application which models such an instrument as a bowed membrane connected to an acoustic tube. This is done by means of a numerical simulation using finite-difference time-domain (FDTD) methods in which the excitation, whose position is free to change in real-time, is modelled by a highly non-linear elasto-plastic friction model. Additionally, the virtual instrument allows for dynamically modifying physical parameters of the model, thereby allowing the user to generate new and interesting sounds that go beyond a realworld friction drum.

Download

The article performs a generic comparison of integrator- and differentiator based continuous-time systems as well as their discretetime models, aiming to answer the reoccurring question in the music DSP community of whether there are any benefits in using differentiators instead of conventionally employed integrators. It is found that both kinds of models are practically equivalent, but there are certain reservations about differentiator based models.

Download

This paper presents a discrete-time model of the spherical harmonics expansion describing a sound field. The so-called radial functions are realized as digital filters, which characterize the spatial impulse responses of the individual harmonic orders. The filter coefficients are derived from the analytical expressions of the timedomain radial functions, which have a finite extent in time. Due to the varying degrees of discontinuities occurring at their edges, a time-domain sampling of the radial functions gives rise to aliasing. In order to reduce the aliasing distortion, the discontinuities are replaced with the higher-order anti-derivatives of a band-limited step function. The improved spectral accuracy is demonstrated by numerical evaluation. The proposed discrete-time sound field model is applicable in broadband applications such as spatial sound reproduction and active noise control.

Download

The distributed nature of coupling in helical springs presents specific challenges in obtaining efficient computational structures for accurate spring reverb simulation. For direct simulation approaches, such as finite-difference methods, this is typically manifested in significant numerical dispersion within the hearing range. Building on a recent study of a simpler spring model, this paper presents an alternative discretisation approach that employs higher-order spatial approximations and applies centred stencils at the boundaries to address the underlying linear-system eigenvalue problem. Temporal discretisation is then applied to the resultant uncoupled mode system, rendering an efficient and flexible modal reverb structure. Through dispersion analysis it is shown that numerical dispersion errors can be kept extremely small across the hearing range for a relatively low number of system nodes. Analysis of an impulse response simulated using model parameters calculated from a measured spring geometry confirms that the model captures an enhanced set of spring characteristics.

Download

A method is proposed to encode the acoustic scattering of objects for virtual acoustic applications through a multiple-input and multiple-output framework. The scattering is encoded as a matrix in the spherical harmonic domain, and can be re-used and manipulated (rotated, scaled and translated) to synthesize various sound scenes. The proposed method is applied and validated using Boundary Element Method simulations which shows accurate results between references and synthesis. The method is compatible with existing frameworks such as Ambisonics and image source methods.

Download

A perceptual study revealing a novel connection between modal properties of feedback delay networks (FDNs) and colorless reverberation is presented. The coloration of the reverberation tail is quantified by the modal excitation distribution derived from the modal decomposition of the FDN. A homogeneously decaying allpass FDN is designed to be colorless such that the corresponding narrow modal excitation distribution leads to a high perceived modal density. Synthetic modal excitation distributions are generated to match modal excitations of FDNs. Three listening tests were conducted to demonstrate the correlation between the modal excitation distribution and the perceived degree of coloration. A fourth test shows a significant reduction of coloration by the colorless FDN compared to other FDN designs. The novel connection of modal excitation, allpass FDNs, and perceived coloration presents a beneficial design criterion for colorless artificial reverberation.

Download

Decomposing a sound-field into its individual components and respective parameters can represent a convenient first-step towards offering the user an intuitive means of controlling spatial audio effects and sound-field modification tools. The majority of such tools available today, however, are instead limited to linear combinations of signals or employ a basic single-source parametric model. Therefore, the purpose of this paper is to present a parametric framework, which seeks to overcome these limitations by first dividing the sound-field into its multi-source and ambient components based on estimated spatial parameters. It is then demonstrated that by manipulating the spatial parameters prior to reproducing the scene, a number of sound-field modification and spatial audio effects may be realised; including: directional warping, listener translation, sound source tracking, spatial editing workflows and spatial side-chaining. Many of the effects described have also been implemented as real-time audio plug-ins, in order to demonstrate how a user may interact with such tools in practice.

Download

We release synth1B1, a multi-modal audio corpus consisting of 1 billion 4-second synthesized sounds, paired with the synthesis parameters used to generate them. The dataset is 100x larger than any audio dataset in the literature. We also introduce torchsynth, an open source modular synthesizer that generates the synth1B1 samples on-the-fly at 16200x faster than real-time (714MHz) on a single GPU. Finally, we release two new audio datasets: FM synth timbre and subtractive synth pitch. Using these datasets, we demonstrate new rank-based evaluation criteria for existing audio representations. Finally, we propose a novel approach to synthesizer hyperparameter optimization.

Download

This paper proposes a novel way of doing audio synthesis at the waveform level using Transformer architectures. We propose a deep neural network for generating waveforms, similar to wavenet . This is fully probabilistic, auto-regressive, and causal, i.e. each sample generated depends on only the previously observed samples. Our approach outperforms a widely used wavenet architecture by up to 9% on a similar dataset for predicting the next step. Using the attention mechanism, we enable the architecture to learn which audio samples are important for the prediction of the future sample. We show how causal transformer generative models can be used for raw waveform synthesis. We also show that this performance can be improved by another 2% by conditioning samples over a wider context. The flexibility of the current model to synthesize audio from latent representations suggests a large number of potential applications. The novel approach of using generative transformer architectures for raw audio synthesis is, however, still far away from generating any meaningful music similar to wavenet, without using latent codes/meta-data to aid the generation process.

Download