Asymmetric and trial-dependent modeling: the contribution of LIA to SdSV Challenge Task 2
This addresses incremental improvements for speaker verification systems in real-life applications.
The paper tackled mismatches between enrollment/test data and between evaluation trial subsets in speaker verification, showing the proposed approaches were relevant and efficient on the SdSV Challenge Task 2 evaluation.
The SdSv challenge Task 2 provided an opportunity to assess efficiency and robustness of modern text-independent speaker verification systems. But it also made it possible to test new approaches, capable of taking into account the main issues of this challenge (duration, language, ...). This paper describes the contributions of our laboratory to the speaker recognition field. These contributions highlight two other challenges in addition to short-duration and language: the mismatch between enrollment and test data and the one between subsets of the evaluation trial dataset. The proposed approaches experimentally show their relevance and efficiency on the SdSv evaluation, and could be of interest in many real-life applications.