• Home
Name Date Size #Lines LOC

..--

doc/06-Sep-2024-1914

examples/06-Sep-2024-6131

include/06-Sep-2024-11533

m4/06-Sep-2024-322284

src/06-Sep-2024-15,28914,249

training/06-Sep-2024-239170

AUTHORSD06-Sep-202437 21

Android.bpD06-Sep-2024666 3128

COPYINGD06-Sep-20241.5 KiB3226

METADATAD06-Sep-2024391 2019

MODULE_LICENSE_BSDD06-Sep-20240

Makefile.amD06-Sep-20243.6 KiB13183

READMED06-Sep-2024596 2215

TRAINING-READMED06-Sep-2024311 126

autogen.shD06-Sep-2024227 115

configure.acD06-Sep-20243.7 KiB130101

rnnoise-uninstalled.pc.inD06-Sep-2024320 1411

rnnoise.pc.inD06-Sep-2024293 1512

update_versionD06-Sep-20242.3 KiB6630

README

1RNNoise is a noise suppression library based on a recurrent neural network.
2
3To compile, just type:
4% ./autogen.sh
5% ./configure
6% make
7
8Optionally:
9% make install
10
11While it is meant to be used as a library, a simple command-line tool is
12provided as an example. It operates on RAW 16-bit (machine endian) mono
13PCM files sampled at 48 kHz. It can be used as:
14
15./examples/rnnoise_demo <noisy speech> <output denoised>
16
17The output is also a 16-bit raw PCM file.
18
19The latest version of the source is available from
20https://gitlab.xiph.org/xiph/rnnoise .  The github repository
21is a convenience copy.
22

TRAINING-README

1(1) cd src ; ./compile.sh
2
3(2) ./denoise_training signal.raw noise.raw count > training.f32
4
5    (note the matrix size and replace 500000 87 below)
6
7(3) cd training ; ./bin2hdf5.py ../src/training.f32 500000 87 training.h5
8
9(4) ./rnn_train.py
10
11(5) ./dump_rnn.py weights.hdf5 ../src/rnn_data.c ../src/rnn_data.h
12