UIAI System for Short-Duration Speaker Verification Challenge 2020

TL;DR

This paper describes the UIAI system for short-duration speaker verification, focusing on feature extraction, modeling, and fusion strategies, achieving state-of-the-art results in the 2020 challenge.

Contribution

It introduces a fusion of seven subsystems and a pass-phrase identification model with phone-discriminative features, improving verification performance.

Findings

Fusion of seven subsystems yields low minDCF and EER

Pass-phrase identification model with phone-discriminative features improves results

Achieves 19% relative improvement over baseline

Abstract

In this work, we present the system description of the UIAI entry for the short-duration speaker verification (SdSV) challenge 2020. Our focus is on Task 1 dedicated to text-dependent speaker verification. We investigate different feature extraction and modeling approaches for automatic speaker verification (ASV) and utterance verification (UV). We have also studied different fusion strategies for combining UV and ASV modules. Our primary submission to the challenge is the fusion of seven subsystems which yields a normalized minimum detection cost function (minDCF) of 0.072 and an equal error rate (EER) of 2.14% on the evaluation set. The single system consisting of a pass-phrase identification based model with phone-discriminative bottleneck features gives a normalized minDCF of 0.118 and achieves 19% relative improvement over the state-of-the-art challenge baseline.