UIAI System for Short-Duration Speaker Verification Challenge 2020
This work addresses speaker verification for short-duration speech in a specific challenge, presenting incremental improvements through system fusion and modeling.
The paper tackled text-dependent speaker verification in a short-duration challenge by investigating feature extraction, modeling, and fusion strategies, resulting in a fused system achieving a normalized minDCF of 0.072 and EER of 2.14%, with a single system showing a 19% relative improvement over the baseline.
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.