UIAI System for Short-Duration Speaker Verification Challenge 2020

Md Sahidullah (1) , Achintya Kumar Sarkar (2) , Ville Vestman (3) , Xuechen Liu (1) , Romain Serizel (1) , Tomi Kinnunen (3) , Zheng-Hua Tan (4) , Emmanuel Vincent (1)

1 MULTISPEECH - Speech Modeling for Facilitating Oral-Based Communication
2 IIIT Sri City - Indian Institute of Information Technology, Sri City
3 University of Eastern Finland
4 AAU - Aalborg University [Denmark]

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.

Mots clés: Text-dependent speaker verification; Utterance verification; System fusion; Bottleneck feature; SdSV challenge 2020; Gaussian mixture model GMM; X-vector; Time-delay neural network; Speech recognition

Domaines: Acoustique [physics.class-ph]; Traitement du signal et de l'image [eess.SP]; Intelligence artificielle [cs.AI]; Apprentissage [cs.LG]

Md Sahidullah, Achintya Kumar Sarkar, Ville Vestman, Xuechen Liu, Romain Serizel, et al.. UIAI System for Short-Duration Speaker Verification Challenge 2020. IEEE Spoken Language Technology Workshop 2021, IEEE, Jan 2021, Shenzhen, China. ⟨hal-02907037v1⟩