Variational Bayes Factor Analysis for i-Vector Extraction

TL;DR

This paper derives equations for a Variational Bayes approach to i-vector extraction, aiming to improve robustness and adaptability in speaker recognition systems by reducing overfitting and enabling domain adaptation.

Contribution

It introduces a Variational Bayes framework for i-vector extraction, extending existing methods with a probabilistic approach for better performance and flexibility.

Findings

Provides derivation of Variational Bayes equations for i-vector extraction

Enables extraction of longer i-vectors to reduce overfitting

Facilitates adaptation of i-vector extractors with limited data

Abstract

In this document we are going to derive the equations needed to implement a Variational Bayes i-vector extractor. This can be used to extract longer i-vectors reducing the risk of overfittig or to adapt an i-vector extractor from a database to another with scarce development data. This work is based on Patrick Kenny's joint factor analysis and Christopher Bishop's variational principal components.