Language-depedent I-Vectors for LRE15

TL;DR

This paper introduces an alternative method for propagating uncertainty in i-vector based spoken language recognition, addressing limitations of the standard Gaussian backend especially for short utterances.

Contribution

It presents a novel approach to incorporate i-vector uncertainty into the backend, improving language recognition accuracy over existing methods.

Findings

Enhanced recognition performance for short utterances

Effective uncertainty propagation in i-vector systems

Improved robustness over standard Gaussian backend

Abstract

A standard recipe for spoken language recognition is to apply a Gaussian back-end to i-vectors. This ignores the uncertainty in the i-vector extraction, which could be important especially for short utterances. A recent paper by Cumani, Plchot and Fer proposes a solution to propagate that uncertainty into the backend. We propose an alternative method of propagating the uncertainty.