Auto Regression Model for Network Music Perfromance Audio Packet Loss Concealment

(ARMNMPAPLC:)

2024 spiral.ok.ubc.ca An auto regression model for generating audio samples given a training signal. Takes audio file as input and outputs a generated audio file

Requires:

numpy sklearn

Usage:

Wave file format is a mono 44100Hz 16bit

python arPlc.py wavFile.wav

In the execute def

from starting point in seonds, how many samples in the training

train = song.getFrames(0, 88200)

the order (lag) of the ar model. larger lag captures temporal changes but increases training time

ar = AR(22500)

same as above but for the test to generate samples after

test = song.getFrames(1, 1024)

how many samples to generate, monte carlo sim depth (the MC sim blows out samples, to is disabled in code-leave as one)

res = ar.predict(np.array(test), 44100, 1)