Speech Anomaly Detection (Normal Human Voice Detection)

This library has been implemented for voice activity detection (VAD) by human voice detection aspect. Therefore, the model is classified voices into two class, no-human (noise) and human voice. The library is based on C/C++17 (Clang compiler is preferred).

Pre-requirements

This library is highly depends on Flashlight library version 0.3. Then take to following step to install it:

1. Install libraries boost, glog, gflag, and json,

2. Install Flashlight from following link (version 0.3):

3. Install GRPC to build it as service.

How to install

Run following commands to build executable file:

git clone https://github.com/mmbejani/Voice-Activity-Detection

cd Voice-Activity-Detection

mkdir build && cd build

cmake .. [OPTION]

make -j 2

./run_[corrosponding built file]

Instead of OPTION use the following option

OPTION	Description	Result File
TRAIN	Compile the files for training phase	run_train
TEST	Compile the corresponding files for test some utterance locally	run_test
INFER	Compile the files for making the model as library (compile all dependency in shared manner). Then the result can be installed (run make install).	sad.so
SERV	Compile the corresponding files for GRPC server	run_server

One can choose one the above options.