jybaek / rnnoise

Recurrent neural network for audio noise reduction

rnnoise

RNNoise is a noise suppression library based on a recurrent neural network

Usage

To compile, just type:

% ./autogen.sh
% ./configure
% make

Optionally:

% make install

denoise training

Compile first as follows:

% cd src
% ./compile.sh

train denoise. If you want to learn various noise, you only need to modify a very small part of the source code. But I will not mention it here separately.

% ./denoise_training <speech> <noise> <output denoised>
...
matrix size: N x M // This value is immediately reused as the next argument.

make denoise_data9.h5

Change the previously created binary to the HDF5 binary data format.

% cd training
% ./bin2hdf5.py output.f32 N M denoise_data9.h5

rnn train

% cd training
% ./rnn_train.py

This will save the model under the name newweights9i.hdf5.

save the hyperparameter

Save the hyperparameter from the newweights9i.hdf5 file to the C file.

% cd training
% ./dump_rnn.py newweights9i.hdf5 ../src/rnn_data.c ../src/rnn_data.h

recompile demo

Recompile the demo based on the newly written rnn_data.c.

% make clean && make

demo

While it is meant to be used as a library, a simple command-line tool is provided as an example. It operates on RAW 16-bit (machine endian) mono PCM files sampled at 48 kHz. It can be used as:

% ./examples/rnnoise_demo input.pcm output.pcm

The output is also a 16-bit raw PCM file.