用内容嵌入匹配实现语音匿名，保留语义同时隐藏身份。
Listen to the Features: Voice Anonymization Driven by Content Embedding Matching over Signal Reconstruction
- 通过wav2vec2提取内容嵌入，用向量量化与HiFi-GAN生成匿名语音
- 匿名化后语音内容识别率低至WER 2.53，声纹识别错误率EER 13.39
- 无需情感训练目标，仍部分保留情绪信息，适合隐私保护场景
本文提出一种语音匿名化模型，重点在于保留语音内容而非生成逼真语音。模型使用冻结的预训练wav2vec2编码器提取内容嵌入，通过向量量化与HiFi-GAN声码器解码为匿名语音信号，两者均在LibriTTS上训练，未采用波形重建损失或说话人嵌入映射。训练目标强制匿名语音的嵌入与原始语音嵌入匹配。训练时引入带有梯度反转层的辅助说话人分类分支，以消除说话人特定信息。实验表明，该基于嵌入的简单方法在内容识别上达到极低的词错误率（WER 2.53），匿名化性能（EER 13.39）在VPC榜单中位列第一梯队。值得注意的是，即便无专门的情感训练目标，情绪识别准确率仍达UAR 43.91，且匿名语音清晰可听，无重建损失影响。
原文摘要 · Abstract (English)
The paper presents a voice anonymization model focusing on preserving content rather than producing realistic speech. It relies on content embeddings extracted from a frozen pretrained wav2vec2 encoder. These embeddings are decoded into an anonymized signal using vector quantization and a HiFi-GAN vocoder, both trained on LibriTTS without any waveform reconstruction loss or speaker embedding mapping. The training objective enforces that embeddings of the anonymized signal match those of the original one. While training, an auxiliary speaker classification branch with a gradient reversal layer is used to discard speakerspecific information. Results show that this straightforward embedding-based approach achieves very low WER (2.53) with an anonymization performance (EER 13.39) ranking within first level for VPC. Notably, emotions are partially preserved (UAR 43.91), even without a supporting training objective, while the anonymized voice is audible without reconstruction loss.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。