BSAPI: SESpeakersEChannelsFAI Class Reference

Factor analysis that models speaker variability, remaining speaker variability and channel variability. More...

#include <bsapi.h>

Inheritance diagram for SESpeakersEChannelsFAI:

Detailed Description

Factor analysis that models speaker variability, remaining speaker variability and channel variability.

The GMM model (supervector = concatenation of all GMM means) can be factorized to two supervector $M=s+c$ . $s$ describes speaker variability and $c$ describes channel variability. $c=u x$ where $u$ is low rank rectangular matrix (eigenchannels) and $x$ are channel factors. And $s = m + v y + d z$ . $m$ is speaker- and channel-independent supervector, $v$ is rectangular low rank matrix (eigenspeakers) $y$ are speaker factors, $d$ is diagonal matrix and $z$ is a vector with remaining speaker variability.

The documentation for this class was generated from the following file:

bsapi.h