Explore the challenges and solutions in text-dependent speaker verification through this comprehensive talk on the Brno University of Technology (BUT) system for the Short-duration Speaker Verification (SdSV) Challenge 2020. Delve into the complexities of verifying both speaker identity and spoken phrases in short audio recordings. Learn about the innovative approach combining x-vector and i-vector systems with MFCCs and bottleneck features, as well as the implementation of a phrase-dependent PLDA backend for scoring. Discover how these techniques, along with a simple phrase recognizer, contributed to the system's top performance in the challenge. Gain insights from Dr. Alicia Lozano-Diez, an expert in speaker recognition and diarization, as she shares her research findings and experiences in the field of speech technology.
Text-Dependent Speaker Verification: BUT System for the SdSV Challenge 2020
Center for Language & Speech Processing(CLSP), JHU via YouTube
Overview
Syllabus
Text-Dependent Speaker Verification: BUT system for the SdSV Challenge 2020 - Alicia Lozano-Diez
Taught by
Center for Language & Speech Processing(CLSP), JHU