Description: This package contains a set of speaker recognition
research utilities, including Gaussian mixture models, VQ codebook
designing program and MLP network. They can also be used as general
classifiers.
The utilities are divided into the following categories:
Feature extraction and dimensional reduction
cepstrum -- extract features from speech sigals (LPCC, MFCC, etc.).
search -- select effective features (SFS, SBS method).
randline -- randomize the a sequence, auxiliary utility.
bin2asc -- binary to ASCII, auxiliary utility.
MLP network
mlptrain -- MLP network training program.
mlptest -- MLP network test program.
VQ codebook training and test programs
lbglvq -- VQ codebook training program.
nearest -- VQ codebook test program.
Gaussian mixture model (GMM)
gmmtrain -- GMM training program.
gmmtest -- GMM test program.
Note: this is a research tool not a true speaker recognition system.