FoCal Multi-class
Tools for evaluation, calibration and fusion of, and decision-making with, multi-class statistical pattern recognition scores.
This is a free, open-source MATLAB toolkit, with general multi-class statistical pattern recognition applicability, but specialized for application in language recognition. It is a set of tools to work with recognition scores, in order to evaluate, calibrate and fuse them and also to finally make hard recognition decisions.
The fusion/calibration tools include discriminative logistic regression as well as generative Gaussian backends, with PPCA, Factor-Analysis and HLDA covariance regularization.
The evaluation tools include Cdet and various flavours of Cllr.
This toolkit is different from, but similar in design principles to the original FoCal Toolkit that was used by several NIST SRE'06 participants to fuse and calibrate their scores.
Version 1.0, with MATLAB code, readme with installation instructions and (pdf) user manual is available here: FoCal_MultiClass_V1.zip (237 KB).
This code has been tested on versions 5 and 7 of MATLAB.
Version 0.1 MATLAB code is available here: FoCal.zip. Install all files on your MATLAB path and start by running ../examples/nary_focal_example.m.
This code has been tested on versions 5 and 7 of MATLAB.
For documentation, see the example above, as well as the help comments in each .m-file.