NoiseModels.jl

Generative noise models

julia> # press ] for pkg mode
pkg> add NoiseModels

Currently the only model implemented is a GaussianNoiseModel. This model generates multichannel colored Gaussian noise given a noise sample or a covariance tensor.

Given training_data as a $T \times N$ matrix of $T$ time samples and $N$ channels, we train a GaussianNoiseModel as follows:

using NoiseModels

# fit a GaussianNoiseModel to training_data
model = fit(GaussianNoiseModel, training_data)

To generate noise samples with the same statistics as the training data:

# generate 10000 time samples of N channel random noise
generated = rand(model, 10000)