Download Dispersion Modeling in Waveguide Piano Synthesis Using Tunable Allpass Filters This paper extends a previously proposed method for designing filters simulating the dispersion phenomenon occurring in string instruments. In digital waveguide synthesis, the phenomenon is traditionally modeled by inserting an allpass filter to the string model feedback loop. In this paper, the concept of tunable dispersion filter design, which provides a closed-form formula to design a dispersion filter, is applied to a cascade of first-order allpass filters. Moreover, the method is extended to design a filter cascade including an arbitrary number of first-order filters. In addition, it is shown how the designed dispersion filter can be used in a waveguide piano synthesis model.
Download Timbre-Constrained Recursive Time-Varying Analysis for Musical Note Separation Note separation in music signal processing becomes difficult when there are overlapping partials from co-existing notes produced by either the same or different musical instruments. In order to deal with this problem, it is necessary to involve certain invariant features of musical instrument sounds into the separation processing. For example, the timbre of a note of a musical instrument may be used as one possible invariant feature. In this paper, a timbre estimate is used to represent this feature such that it becomes a constraint when note separation is performed on a mixture signal. To demonstrate the proposed method, a timedependent recursive regularization analysis is employed. Spectral envelopes of different notes are estimated and a modified parameter update strategy is applied to the recursive regularization process. The experiment results show that the flaws due to the overlapping partial problem can be effectively reduced through the proposed approach.
Download Automatic Equalization for Individual Instrument Tracks Using Convolutional Neural Networks We propose a novel approach for the automatic equalization of individual musical instrument tracks. Our method begins by identifying the instrument present within a source recording in order to choose its corresponding ideal spectrum as a target. Next, the spectral difference between the recording and the target is calculated, and accordingly, an equalizer matching model is used to predict settings for a parametric equalizer. To this end, we build upon a differentiable parametric equalizer matching neural network, demonstrating improvements relative to previously established state-of-the-art. Unlike past approaches, we show how our system naturally allows real-world audio data to be leveraged during the training of our matching model, effectively generating suitably produced training targets in an automated manner mirroring conditions at inference time. Consequently, we illustrate how fine-tuning our matching model on such examples considerably improves parametric equalizer matching performance in realworld scenarios, decreasing mean absolute error by 24% relative to methods relying solely on random parameter sampling techniques as a self-supervised learning strategy. We perform listening tests, and demonstrate that our proposed automatic equalization solution subjectively enhances the tonal characteristics for recordings of common instrument types.
Download Fast Approximation of the Lambert W Function for Virtual Analog Modelling When modelling circuits one has often to deal with equations containing both a linear and an exponential part. If only a single exponential term is present or predominant, exact or approximate closed-form solutions can be found in terms of the Lambert W function. In this paper, we propose reformulating such expressions in terms of the Wright Omega function when specific conditions are met that are customary in practical cases of interest. This eliminates the need to compute an exponential term at audio rate. Moreover, we propose simple and real-time suitable approximations of the Omega function. We apply our approach to a static and a dynamic nonlinear system, obtaining digital models that have high accuracy, low computational cost, and are stable in all conditions, making the proposed method suitable for virtual analog modelling of circuits containing semiconductor devices.
Download On sinusoidal parameter estimation This paper contains a review of the issues surrounding sinusoidal parameter estimation which is a vital part of many audio manipulation algorithms. A number of algorithms which use the phase of the Fourier transform for estimation (e.g. [1]) are explored and shown to be identical. Their performance against a classical interpolation estimator [2] and comparison with the Cramer Rao Bound (CRB) is presented. Component detection is also considered and various methods of improving these algorithms are discussed.
Download On Finding Melodic Lines in Audio Recordings The paper presents our approach to the problem of finding melodic line(s) in polyphonic audio recordings. The approach is composed of two different stages, partially rooted in psychoacoustic theories of music perception: the first stage is dedicated to finding regions with strong and stable pitch (melodic fragments), while in the second stage, these fragments are grouped according to their properties (pitch, loudness...) into clusters which represent melodic lines of the piece. Expectation Maximization algorithm is used in both stages to find the dominant pitch in a region, and to train Gaussian Mixture Models that group fragments into melodies. The paper presents the entire process in more detail and provides some initial results.
Download Audio Visualization via Delay Embedding and Subspace Learning We describe a sequence of methods for producing videos from audio signals. Our visualizations capture perceptual features like harmonicity and brightness: they produce stable images from periodic sounds and slowly-evolving images from inharmonic ones; they associate jagged shapes to brighter sounds and rounded shapes to darker ones. We interpret our methods as adaptive FIR filterbanks and show how, for larger values of the complexity parameters, we can perform accurate frequency detection without the Fourier transform. Attached to the paper is a code repository containing the Jupyter notebook used to generate the images and videos cited. We also provide code for a realtime C++ implementation of the simplest visualization method. We discuss the mathematical theory of our methods in the two appendices.
Download WDF Modeling of a Korg MS-50 Based Non-linear Diode Bridge VCF The voltage-controlled low-pass filter of the Korg MS-50 synthesizer is built around a non-linear diode bridge as the cutoff frequency control element, which greatly contributes to the sound of this vintage synthesizer. In this paper, we introduce the overall filter circuitry and give an in-depth analysis of this diode bridge. It is further shown how to turn the small signal equivalence circuit of the bridge into the necessary two-resistor configuration to uncover the underlying Sallen-Key structure. In a second step, recent advances in the field of WDFs are used to turn a simplified version of the circuit into a virtual-analog model. This model is then examined both in the small-signal linear domain as well as in the non-linear region with inputs of different amplitudes and frequencies to characterize the behavior of such diode bridges as cutoff frequency control elements.
Download A Nonlinear Method for Manipulating Warmth and Brightness In musical timbre, two of the most commonly used perceptual dimensions are warmth and brightness. In this study, we develop a model capable of accurately controlling the warmth and brightness of an audio source using a single parameter. To do this, we first identify the most salient audio features associated with the chosen descriptors by applying dimensionality reduction to a dataset of annotated timbral transformations. Here, strong positive correlations are found between the centroid of various spectral representations and the most salient principal components. From this, we build a system designed to manipulate the audio features directly using a combination of linear and nonlinear processing modules. To validate the model, we conduct a series of subjective listening tests, and show that up to 80% of participants are able to allocate the correct term, or synonyms thereof, to a set of processed audio samples. Objectively, we show low Mahalanobis distances between the processed samples and clusters of the same timbral adjective in the low-dimensional timbre space.
Download Sound synthesis by physical modelling using the functional transformation method: Efficient implementations with polyphase-filterbanks The Functional Transformation Method (FTM) is a recently introduced method for sound synthesis by physical modeling. Based on integral transformations, it provides a parallel system description for any linear physical model, usually described by a set of partial differential equations. Such parallel descriptions can be directly implemented by a set of recursive systems in full rate. In this PSfrag replacem paper we present a new and very ef£cient method for this implementation which bene£ts from the spectral decomposition of the system. All recursive systems are working at a subsampled rate and are summed up by the application of a polyphase £lterbank. Performance measurements on a real time implementation show, that a ¤exible and ef£cient realization is achieved. Compared to the direct implementation it is over nine times faster at the cost of nine milliseconds of delay and even faster with more delay.