Our Learned Lessons from Cross-Lingual Speaker Verification: The CRMI-DKU System Description for the Short-Duration Speaker Verification Challenge 2021
In this paper, we present our CRMI-DKU system description for the Short-duration Speaker Veri fi cation Challenge (SdSVC) 2021. We introduce the whole pipeline of our cross-lingual speaker veri fi cation system, including data preprocessing, training strategy, utterance-level speaker embedding extractor, domain-adaptation, and score calibration. We also propose methods to learn language-invariant features and perform domain adaptation to reduce the cross-lingual mismatch. In addition, we explore a semi-supervised method to utilize the unlabeled training data. The fi nal submitted score level fusion sys-tem achieves 0.0476 minDCF and 0.98% EER on the evaluation set.
Paper
Full text
Our Learned Lessons from Cross-Lingual Speaker Verification: The CRMI-DKU System Description for the Short-Duration Speaker Verification Challenge 2021
Semantic Scholar · Computer Science · 2021
Abstract
In this paper, we present our CRMI-DKU system description for the Short-duration Speaker Veri fi cation Challenge (SdSVC) 2021. We introduce the whole pipeline of our cross-lingual speaker veri fi cation system, including data preprocessing, training strategy, utterance-level speaker embedding extractor, domain-adaptation, and score calibration. We also propose methods to learn language-invariant features and perform domain adaptation to reduce the cross-lingual mismatch. In addition, we explore a semi-supervised method to utilize the unlabeled training data. The fi nal submitted score level fusion sys-tem achieves 0.0476 minDCF and 0.98% EER on the evaluation set.