Fine-tuning an SSL content encoder on Japanese, especially when the encoder is pre-trained multilingually, makes anonymized Japanese and Mandarin speech much more intelligible while keeping speaker privacy at usable levels.
This study was carried out using the TSUB- AME4.0 supercomputer at Institute of Science Tokyo
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
eess.AS 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Mitigating Language Mismatch in SSL-Based Speaker Anonymization
Fine-tuning an SSL content encoder on Japanese, especially when the encoder is pre-trained multilingually, makes anonymized Japanese and Mandarin speech much more intelligible while keeping speaker privacy at usable levels.