mlsanityData to 19 Oct 2026

Enroll-on-Wakeup: A First Comparative Study of Target Speech Extraction for Seamless Interaction in Real Noisy Human-Machine Dialogue Scenarios

arXiv 2602.15519 ↗First posted 17 Feb 2026eess.AS cs.SDcited by 0

On the map: Speech and audio language models → Audio language models and spoken dialogue

Abstract (arXiv)

Target speech extraction (TSE) typically relies on pre-recorded high-quality enrollment speech, which disrupts user experience and limits feasibility in spontaneous interaction. In this paper, we propose Enroll-on-Wakeup (EoW), a novel framework where the wake-word segment, captured naturally during human-machine interaction, is automatically utilized as the enrollment reference. This eliminates the need for pre-collected speech to enable a seamless experience. We perform the first systematic study of EoW-TSE, evaluating advanced discriminative and generative models under real diverse acoustic conditions. Given the short and noisy nature of wake-word segments, we investigate enrollment augmentation using LLM-based TTS. Results show that while current TSE models face performance degradation in EoW-TSE, TTS-based assistance significantly enhances the listening experience, though gaps remain in speech recognition accuracy.

Key facts. No citing papers in the map yet (data through 19 Oct 2026). Placed in Audio language models and spoken dialogue, one of 11 directions in Speech and audio language models. Builds on 31 of its 36 references that are in the map. First posted on arXiv 17 Feb 2026.

Builds on

31 of 36 references.

References · 31

  1. 2020ECAPA-TDNN: Emphasized Channel Attention, Propagation and Aggregation in TDNN Based Speaker Verificationcited by 980
  2. 2018SDR - half-baked or well done?cited by 626
  3. 2019Why gradient clipping accelerates training: A theoretical justification for adaptivitycited by 360
  4. 2021DNSMOS P.835: A Non-Intrusive Perceptual Objective Speech Quality Metric to Evaluate Noise Suppressorscited by 292
  5. 2015Deep clustering: Discriminative embeddings for segmentation and separationcited by 281
  6. 2019WHAM!: Extending Speech Separation to Noisy Environmentscited by 246
  7. 2024XTTS: a Massively Multilingual Zero-Shot Text-to-Speech Modelcited by 239
  8. 2020LibriMix: An Open-Source Dataset for Generalizable Speech Separationcited by 219
  9. 2025CosyVoice 3: Towards In-the-wild Speech Generation via Scaling-up and Post-trainingcited by 215
  10. 2022Wespeaker: A Research and Production oriented Speaker Embedding Learning Toolkitcited by 179
  11. 2025IndexTTS2: A Breakthrough in Emotionally Expressive and Duration-Controlled Auto-Regressive Zero-Shot Text-to-Speechcited by 142
  12. 2018VoiceFilter: Targeted Voice Separation by Speaker-Conditioned Spectrogram Maskingcited by 122
19 more
  1. 2022TF-GridNet: Integrating Full- and Sub-Band Modeling for Speech Separationcited by 121
  2. 2020Dual-Path Transformer Network: Direct Context-Aware Modeling for End-to-End Monaural Speech Separationcited by 105
  3. 2019WHAMR!: Noisy and Reverberant Single-Channel Speech Separationcited by 83
  4. 2020SpEx+: A Complete Time Domain Speaker Extraction Networkcited by 55
  5. 2020SpEx: Multi-Scale Time Domain Speaker Extraction Networkcited by 52
  6. 2023MeetEval: A Toolkit for Computation of Word Error Rates for Meeting Transcription Systemscited by 49
  7. 2025Fun-ASR Technical Reportcited by 28
  8. 2023AV-SepFormer: Cross-Attention SepFormer for Audio-Visual Target Speaker Extractioncited by 20
  9. 2024TF-Locoformer: Transformer with Local Modeling by Convolution for Speech Separation and Enhancementcited by 20
  10. 2022L-SpEx: Localized Target Speaker Extractioncited by 12
  11. 2025SoloSpeech: Enhancing Intelligibility and Quality in Target Speech Extraction through a Cascaded Generative Pipelinecited by 12
  12. 2024Target Speaker Extraction by Directly Exploiting Contextual Information in the Time-Frequency Domaincited by 11
  13. 2024USEF-TSE: Universal Speaker Embedding Free Target Speaker Extractioncited by 10
  14. 2023MC-SpEx: Towards Effective Speaker Extraction with Multi-Scale Interfusion and Conditional Speaker Modulationcited by 7
  15. 2023Scenario-Aware Audio-Visual TF-GridNet for Target Speech Extractioncited by 7
  16. 2025SEF-PNet: Speaker Encoder-Free Personalized Speech Enhancement with Local and Global Contexts Aggregationcited by 5
  17. 2024MoMuSE: Momentum Multi-modal Target Speaker Extraction for Real-time Scenarios with Impaired Visual Cuescited by 3
  18. 2025Listen to Extract: Onset-Prompted Target Speaker Extractioncited by 2
  19. 2025ELEGANCE: Efficient LLM Guidance for Audio-Visual Target Speech Extractioncited by 2

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025AU-Harness: An Open-Source Toolkit for Holistic Evaluation of Audio LLMs
  2. 2025End-to-End Target Speaker Speech Recognition Using Context-Aware Attention Mechanisms for Challenging Enrollment Scenario
  3. 2025Enhancing Intelligibility for Generative Target Speech Extraction via Joint Optimization with Target Speaker ASR
  4. 2026Endpoint Anticipation for Low-Latency Spoken Dialogue
  5. 2026Unified Target-Speaker ASR with Text and Enrollment Speech Cues
  6. 2026Not All Speech Is Intent: Adaptive Self-Correcting Inference Layer for Post-ASR False Wake-Up
  7. 2024SLAM-Omni: Timbre-Controllable Voice Interaction System with Single-Stage Training
  8. 2024Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks