The Impact of Language on Automatic Speaker Verification Systems
The project explores how language changes the performance of a speaker verification system and aims to develop a system for non-English language.
Project GitHub: Not Ready.
The project is conducted by my 2nd year student, Kuznetsov Stepan, at HSE (Coursework) in 24/25 academic year.
Automatic Speaker Verification (ASV) systems are mostly trained on datasets with English language, meaning that their performance on other languages might be poor due to out-of-domain input. The student’s goal is to explore the impact of language on ASV systems and come up with the ways to improve the quality of the models.