mlsanityData to 14 Oct 2026

AegisVox: Factorized Privacy and Clinical Preservation for Long-Form Speech

arXiv 2610.14643 ↗First posted 10 Oct 2026cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

Clinical speech reveals identity through both spoken identifiers and voice, yet its pauses, delivery, and speaker roles can also carry useful clinical evidence. Protecting that speech therefore requires more than preserving its words. We present AegisVox, a factorized framework that assigns consistent surrogate voices and edits identifier-bearing turns using document-consistent replacements. Across 390 public simulated or non-patient conversations totaling 63.6 hours, a matched comparison of two speech renderers separates replacement realization from retained identifiers, speaker linkage, and clinical preservation. Text-to-speech rendering realizes 535/601 planned replacements versus 453/601 for the codec-based speech editor. However, similar aggregate clinical preservation conceals more model-labeled contradictions, and conversation-wide speaker scores conceal greater edited-speech linkage under a score-direction sensitivity analysis. Fresh transcription of original audio provides a reference for interpreting the clinical measurement chain; an independent judge tests sensitivity to model-based assessment. Better replacement realization can therefore coexist with worse clinical or speaker-linkage measurements. AegisVox supports use-specific evaluation of those tradeoffs, not a guarantee of identifier removal or clinical safety.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 19 of its 38 references that are in the map. First posted on arXiv 10 Oct 2026.

Builds on

19 of 38 references.

References · 19

  1. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,626
  2. 2022Robust Speech Recognition via Large-Scale Weak Supervisioncited by 4,057
  3. 2023G-Eval: NLG Evaluation using GPT-4 with Better Human Alignmentcited by 2,276
  4. 2022Large-scale Contrastive Language-Audio Pretraining with Feature Fusion and Keyword-to-Caption Augmentationcited by 719
  5. 2024F5-TTS: A Fairytaler that Fakes Fluent and Faithful Speech with Flow Matchingcited by 361
  6. 2024VoiceCraft: Zero-Shot Speech Editing and Text-to-Speech in the Wildcited by 120
  7. 2023ACI-BENCH: a Novel Ambient Clinical Intelligence Dataset for Benchmarking Automatic Visit Note Generationcited by 118
  8. 2025Canary-1B-v2 & Parakeet-TDT-0.6B-v3: Efficient and High-Performance Models for Multilingual ASR and ASTcited by 60
  9. 2020Introducing the VoicePrivacy Initiativecited by 49
  10. 2024Zero-shot Voice Conversion with Diffusion Transformerscited by 43
  11. 2021EditSpeech: A Text Based Speech Editing System Using Partial Inference and Bidirectional Fusioncited by 32
  12. 2022PriMock57: A Dataset Of Primary Care Mock Consultationscited by 31
7 more
  1. 2019Audio De-identification: A New Entity Recognition Taskcited by 16
  2. 2022CampNet: Context-Aware Mask Prediction for End-to-End Text-Based Speech Editingcited by 16
  3. 2024Sortformer: A Novel Approach for Permutation-Resolved Speaker Supervision in Speech-to-Text Systemscited by 15
  4. 2023Generating medically-accurate summaries of patient-provider dialogue: A multi-stage approach using large language modelscited by 7
  5. 2025RedactOR: An LLM-Powered Framework for Automatic Clinical Data De-Identificationcited by 5
  6. 2025Content Anonymization for Privacy in Long-form Audiocited by 3
  7. 2026Anonymization, Not Elimination: Utility-Preserved Speech Anonymizationcited by 1

Built on by

Not cited yet.