Existing deterministic+stochastic spectral models assume that the sounds are with low noise levels. The stochastic part of the sound is generally estimated by subtraction of the deterministic part: It is assumed to be the residual. Inevitable errors in the estimation of the parameters of the deterministic part result in errors – often worse – in the estimation of the stochastic part. We propose a new method that avoids these errors. Our method analyzes the stochastic part without any prior knowledge of the deterministic part. It relies on the study of the distribution of the amplitude values in successive short-time spectra. Computations of the statistical moments or the maximum likelihood lead to an estimation of the noise power density. Experimentations on synthetic or natural sounds show that this method is promising.

Download

This paper presents a new fast sound synthesis method using polynomials. This is an additive method, where polynomials are used to approximate sine functions. Traditional additive synthesis requires each sample to be generated for each partial oscillator. Then all these partial samples are summed up to obtain the resulting sound sample, thus making the synthesis time proportional to the product of the number of oscillators and the sampling rate. By using polynomial approximations, we instead sum up only the oscillator coefficients and then generate directly the sound sample from these new coefficients. Most of computation time is consumed by a data structure that manages the update of the generator coefficients as a priority queue. Practical implementations show that Polynomial Additive Sound Synthesis (PASS) is particularly efficient for low-frequency signals.

Download

In this paper we propose a complete computational system for Auditory Scene Analysis. This time-frequency system localizes, separates, and spatializes an arbitrary number of audio sources given only binaural signals. The localization is based on recent research frameworks, where interaural level and time differences are combined to derive a confident direction of arrival (azimuth) at each frequency bin. Here, the power-weighted histogram constructed in the azimuth space is modeled as a Gaussian Mixture Model, whose parameter structure is revealed through a weighted Expectation Maximization. Afterwards, a bank of Gaussian spatial filters is configured automatically to extract the sources with significant energy accordingly to a posterior probability. In this frequency-domain framework, we also inverse a geometrical and physical head model to derive an algorithm that simulates a source as originating from any azimuth angle.

Download

In this paper, we introduce two original evaluation methods in the context of sinusoidal modeling. The first one assesses the quality of the extraction of sinusoidal components from short-time signals, whereas the second one focuses on the quality of the tracking of these sinusoidal components over time. Each proposed method intends to use a unique cost function that globally reflects the performance of the tested algorithm in a realistic framework. Clearly defined evaluation protocols are then proposed with several test cases to evaluate most of the desired properties of extractors or trackers of sinusoidal components. This paper is a first proposal to be used as a starting point in a sinusoidal analysis / synthesis contest to be held at DAFx’07.

Download