Velvet Noise

Implementation of several types of Velvet Noise as described in http://dafx.de/paper-archive/2019/DAFx2019_paper_53.pdf

An example application implementing the Signal Extrapolation technique described in http://dafx.de/paper-archive/2018/papers/DAFx2018_paper_11.pdf is also provided

Usage

Generate raw velvet noise audio

Example: generate 1s of velvet noise audio as a vector of f32s

let density = 2000.;
let sample_rate = 44100.;
let noise = original_velvet_noise(density, sample_rate);
let samples: Vec<f32> = noise.take(44100).collect();

Generate a velvet convolution kernel

The unique property of velvet noise is that the majority of samples are 0. and can hence be ignored during a convolution. To that end, this crate provides velvet convolution kernels - iterators yielding (usize, f32) where the first element is the index of the non-zero samples and the second element is either 1. or -1.;

let density = 2000;
let sample_rate = 44100;
let kernel = VelvetNoiseKernel(
    OVNImpulseLocations::new(density, sample_rate),
    Choice::classic(),
);
for (index, coefficient) in kernel.take_while(|(index, coefficient)| (*index) < sample_rate / 8) {
    println!("{} {}", index, coefficient);
}