speaker differentiation

The vad function MATLAB code for pyBK implements speaker differentiation on a list of audio files by performing speaker binarization (speech segmentation and clustering in multi-speaker scenarios). The system utilizes a binary key background model (KBM), which is trained on conference data, eliminating the need for external training datasets. This results in a system that is easy to operate and adjust for speaker differentiation tasks. Additionally, the implementation includes useful features for the speaker digitization system pipeline. The code was developed and tested in Python 3.6 using conda, relying on common packages for audio processing, feature extraction, and speech activity detection. Installation steps:1. $ conda create -n pyBK python=3.62. $ source activate pyBK3. $ conda install numpy4. $ conda install -c conda-forge librosa5. $ pip install webrtcvad6. $ git clone h...

This submission includes useful MATLAB functions for speaker recognition using adapted GMM. The implementation details for steps (i)-(iii) can be found in [1]. The fourth function, gmm2sv.m, connects the means (i.e., centers) of the GMM. The cascade means of the adapted GMM are referred to as the GMM supervector (GSV), which is used in the GMM-SVM based speaker recognition system. More information about the GMM-SVM based speaker recognition system can be found in [2]. These codes require the Netlab toolbox. You can access it at: Netlab Toolbox. References: [1] DA Reynolds, TF Quatieri, and RB Dunn, “Speaker Verification Using Adapted Gaussian Mixture Models,” Digital Signal Processing, Vol. 10, pp. 19–41, 2000. [2] Campbell, W. M.; Sturim, D. E.; Reynolds, D. A.; “Support Vector Machines Using GMM Supervectors for Speaker Verification,” Signal Processing Letters.