This is a free library for Speaker Recognition (Verification), which is completely implemented in C/C++. In addition, Speaker Gender Recognition is also implemented by GMM-UBM.
Simultaneously, you can extract speech features by FreeSR such as mfcc, fbank etc. More detailed information can be referred to python_speech_features. Most of implementations are similar between them, only in different language.
| Windows (win32) | Linux (X64) | Android | |
|---|---|---|---|
| >=VS2013 | gcc>=4.9.4 | -- | |
| I-Vector | * | * | * |
| GMM-UBM | * | * | * |
mkdir build; cd build; cmake ..
The input is preferably raw data without normalization from wav (bitspersample 16bits and sample rate 16K). More detailed information about usage of FreeSR can be referred to speaker_recognizer.h.
For Windows, it is compiled by VS2013/Release/32bit. The debug version can be found here.
For Android, the so library can be found here.
When use VS2013, maybe you need to change the line wrap format of header files in include to Windows from UNIX,
The sample english.wav obtained from python_speech_features. The english.pcm is raw data from english.wav