Recurrent Neural Network Voice activity detection (RNN VAD) example project, extract from WebRTC.

This repository just contains the scripts to test RNN VAD library from WebRTC source code.

[1]https://webrtc.org/ [2]https://jmvalin.ca/demo/rnnoise/

1. Get Abseil-cpp source code

git clone https://github.com/abseil/abseil-cpp.git

2. Build with cmake

mkdir bin
cmake ..

you need wav file which is 16khz sample rate for test RNNVAD

webrtc_rnnvad.exe -i test.wav -f features.file -o vadprob.file

for linux env
in common_audio/resampler/sinc_resampler.cc
convolve_SSE2 is disabled for convenience