The fork improved on the original version to support Python 3, fixed a few bugs, and made it importable from other Python scripts

logmmse

A python implementation of the LogMMSE speech enhancement/noise reduction alogrithm

Installation

pip install logmmse

Basic Example

from logmmse import logmmse_from_file
out = logmmse_from_file('sample.wav')
print(out)

API Reference

logmmse(data, sampling_rate, output_file=None, initial_noise=6, window_size=0, noise_threshold=0.15)

data (Mandatory)

A 1d or 2d (number of frames, channels) Numpy array representing the audio signal

Supported format: 32-bit floating-point, 32-bit PCM, 16-bit PCM, and 8-bit PCM, see https://docs.scipy.org/doc/scipy/reference/generated/scipy.io.wavfile.read.html

sampling_rate (Mandatory)

Sampling rate of the data (e.g. 16k/44.1 kHz)

output_filename

filename of the wave output. If none, no file would be generated

the output format would be the same as input

return

A 1d or 2d (number of frames, channels) Numpy array representing the enhanced signal

logmmse_from_file(data, output_filename=None, initial_noise=6, window_size=0, noise_threshold=0.15)

input_filename (Mandatory)

Filename of the wave input

supported format: 32-bit floating-point, 32-bit PCM, 16-bit PCM, and 8-bit PCM, see https://docs.scipy.org/doc/scipy/reference/generated/scipy.io.wavfile.read.html

output_filename

See above

return

A 1d or 2d (number of frames, channels) Numpy array representing the enhanced signal

Contribute

After opening an issue or pull request, please also send me an email for notification

Reference

This file has been ported from logmmse.m file found at the following link. https://raw.githubusercontent.com/braindead/Noise-reduction/master/logmmse.m