Pith. sign in

REVIEW 4 major objections 5 minor 27 references

Beyond Speaker Identity: Text Guided Target Speech Extraction

T0 review · 4 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

Pith's one-line read StyleTSE demonstrates that a natural-language description of speaking style is sufficient to extract a target voice from a two-talker mixture, outperforming the previous text-guided baseline by about 5 dB SI-SDRi.

desk verdict Plausible and useful text-guided TSE system with a strong new dataset, but the paper never shows same-speaker mixtures, so the 'beyond speaker identity' claim remains a gap. read the letter →

arxiv 2501.09169 v1 pith:SKFIYZ7F submitted 2025-01-15 eess.AS cs.SD

classification eess.AScs.SD
keywords targetspeechextractiontext-guidedseparationspeakingstyleTextrolMixnaturallanguagecluegatedfusionSepFormerSI-SDRi
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Target speech extraction normally needs a speaker-specific clue such as an enrollment audio snippet, and fails when no such clue has been collected. This paper argues that a natural-language description of how a person speaks—for example 'shocked pitch and slow speaking rate'—can serve as the clue instead, provided the described style separates the target from the interferer. To make that test possible, the authors build TextrolMix, 120,000 two-talker mixtures (157 hours) with six annotated style attributes per utterance, and StyleTSE, a SepFormer-based extraction model whose gated bi-modality clue network reads audio and text. On TextrolMix, text-only guidance reaches 16.41 dB SI-SDRi with dynamic mixing, and audio-plus-text reaches 16.84 dB, against 10.92 dB for the replicated LLM-TSE baseline. The paper claims this extends target speech extraction beyond speaker identity to how the speech is delivered.

What carries the argument

The central object is the bi-modality clue network working together with the TextrolMix construction rule. The clue network encodes the text with a frozen BERT model, encodes the reference audio with a trainable encoder whose output is attention-pooled across time frames, and combines the two vectors with a sigmoid-gated weighted sum; the fused vector is then expanded and added to the chunked mixture representation at every IntraTransformer block of the SepFormer backbone. The dataset rule pairs each target utterance with an interference utterance that differs in at least one of six style attributes (speaker identity, emotion, pitch, gender, accent, tempo) and requires any reference audio to share an attribute with the target but not with the non-targets, which guarantees that the text clue has a real separator to point at. The gated fusion and attention pooling are the components that the ablations identify as responsible for the gains.

What would settle it

Build mixtures where the target and interference carry identical style-attribute annotations (same emotion, pitch, gender, accent, and tempo, differing only in speaker identity) and test StyleTSE; if SI-SDRi does not fall sharply from the reported 16 dB range, the model is exploiting dataset conventions or hidden identity cues rather than the text description.

Watch

Extended reading notes

Core claim

StyleTSE establishes that target speech extraction can be driven by a style description rather than a speaker identity: the model conditions a SepFormer separation backbone on a 768-dimensional BERT [CLS] embedding of the text clue and an attention-pooled embedding of the audio clue, then fuses the two with a learned gate and injects the result into the chunked mixture representation inside each IntraTransformer block. With text-only clues on TextrolMix the model reaches 16.41 dB SI-SDRi; with a style-matched audio clue added it reaches 16.84 dB, while the replicated LLM-TSE baseline reaches 10.92 dB. The authors attribute the gains to the dataset construction rule (target and interference differ in at least one of six style attributes), to attention pooling over the reference audio, and to two-stage training with dynamic mixing. The paper frames the result as showing that text guidance works not only for who is speaking but for how they are speaking.

Load-bearing premise

The load-bearing premise is that every TextrolMix mixture has a text-discoverable separator: the target and interference differ in at least one of the six annotated style attributes, and any reference audio shares an attribute only with the target.

Editorial extensions

If this is right

  • Text-only clues suffice for target speech extraction on TextrolMix, reaching 16.41 dB SI-SDRi with dynamic mixing and staying above 15 dB across long, mid, and short text descriptions.
  • Adding a reference audio clue that shares a style attribute with the target improves extraction to 16.84 dB SI-SDRi, so the same model covers classical enrollment-based TSE and text-only TSE.
  • Because the model is trained with missing modalities, audio-only inputs perform at roughly the speaker-identity level (around 15.7 dB SI-SDRi in the ablation), meaning a single system can serve both settings.
  • Replacing gated fusion with averaging or concatenation costs about 1 dB, and removing attention pooling in favor of average pooling costs about 2 dB, identifying the mechanism behind the improvement.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the result generalizes beyond TextrolMix, the practical interface for extraction could shift from 'provide a recording of the speaker' to 'say which style you want,' which matters for hearables, meeting assistants, and archival audio where enrollment samples do not exist.
  • The dataset construction guarantees a text-discoverable separator, so the reported 16 dB figures likely represent an upper bound for style-guided extraction; real-world performance depends on whether the target and interference actually differ in a describable attribute.
  • The same gated-fusion pattern could be applied to other query types, such as typed descriptions in general sound-separation systems, where text and audio tags currently compete rather than cooperate.
  • A direct next test would be zero-shot evaluation on naturally recorded (non-TTS) two-talker conversations with human-written style descriptions, which would reveal how much of the result relies on TextrolSpeech's synthesized voices and attribute labels.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes StyleTSE, a target speech extraction model that conditions a SepFormer separation network on natural-language descriptions of speaking style, optionally combined with an audio reference clue through a gated fusion network. The authors also introduce TextrolMix, a two-talker mixture dataset built from TextrolSpeech, in which each mixture has a target utterance, an interference utterance that differs in at least one style attribute, and text/audio clues. The reported results show text-only SI-SDRi of 16.41 dB and audio+text SI-SDRi of 16.84 dB on the TextrolMix test set, outperforming an LLM-TSE replication and a fine-tuned AudioSep baseline. The paper claims that this demonstrates target speech extraction based on how speech sounds, not only on who is speaking.

Significance. If the results are robust, the paper makes a useful contribution: it introduces a public text-guided TSE dataset, proposes a practical bi-modality clue architecture, and reports strong quantitative gains. The two-stage training and dynamic mixing strategy is also a reasonable engineering contribution. The public dataset and generation scripts are a positive feature that should help reproducibility. However, the central claim that StyleTSE separates by speaking style rather than speaker identity is not yet empirically secured: the paper lacks same-speaker evaluation, mismatched-text controls, and architecture-matched baselines, and it reports no variance across runs.

major comments (4)
  1. [Section II, Table I] Section II states that TextrolMix mixtures can include utterances from the same speaker if they exhibit distinguishable style attributes, but Table I reports only aggregate results and attribute-class results, with no breakdown by same-speaker versus different-speaker mixtures. Since TextrolSpeech is a TTS corpus and style attributes such as gender, accent, pitch, and emotion may be strongly correlated with particular TTS speakers, the text-only 16.41 dB SI-SDRi result could be driven by speaker-identity leakage rather than by understanding the text description of speaking style. The claim that the model extracts speech based on 'how they are speaking' requires a same-speaker condition; please report SI-SDRi and PESQ separately for same-speaker and different-speaker test mixtures.
  2. [Section IV-B, Table I] The experiments include no control condition with mismatched text, random text, or no text clue. To establish that the text semantics are actually used for extraction, the authors should compare the text-clue condition against a mismatched-text condition (e.g., a text description of the non-target speaker) and a random-text condition under the same mixture and backbone. If mismatched text does not substantially degrade performance, the model is likely relying on speaker-identity cues rather than on the meaning of the natural-language description. This control is load-bearing for the paper's central 'beyond speaker identity' claim.
  3. [Section IV-B, Table I] The baseline comparison is confounded by architecture: the LLM-TSE replication uses a Conv-TasNet-based SpeakerBeam architecture, while StyleTSE uses SepFormer. The reported gain of roughly 4-6 dB SI-SDRi could reflect the stronger separation backbone rather than the effectiveness of text-guided style conditioning. Please add an architecture-matched ablation, such as the same SepFormer-based model trained with audio-only enrollment, or a text-conditioned model with the same backbone as LLM-TSE, to isolate the contribution of the text clue. In addition, no error bars or repeated-run statistics are reported; small differences in Table I and Table II (for example 15.35 vs 15.10 dB for long vs mid text clues) cannot be interpreted without variance information.
  4. [Section III-C] The dynamic mixing description appears inconsistent with Section II. Section II says that each target utterance is paired with an interference utterance that differs in at least one style attribute, but Section III-C says 'DM pairs utterances such that the interference shares the same style attributes as the predefined ones, preserving clue effectiveness.' Please clarify whether the dynamic-mixing interference shares or differs from the target's attributes, and quantify how many attributes typically differ. This is important because the construction rule determines whether the text clue is discriminative at training time and whether the average of 2.45 alternative interferences per target preserves the intended clue semantics.
minor comments (5)
  1. [Figure 1] Figure 1 contains several typographical artifacts in labels, including 'Es0mate', 'A8en0on', and 'So#max'; these should be corrected.
  2. [References] Reference [1] is rendered as 'V oiceFilter' with an erroneous space, and reference [21] contains a stray 'Y ' in the author list; these should be fixed.
  3. [Table II] The row label 'Gated w/o attpool' should be written consistently as 'Gated w/o AttPool' or 'Gated w/o attention pooling' to match the text in Section IV-C.
  4. [Section IV-A] The paper says training signals are truncated to 3 seconds, but TextrolMix utterances are 3-15 seconds; please clarify how truncation interacts with the dataset construction and whether the test set is truncated in the same way.
  5. [Section IV-B] The statement that audio-only performance 'was similar to the speaker identity class (±0.05 dB SI-SDRi), thus not reported separately' is surprising given the large attribute-conditioned differences in the lower half of Table I; a brief discussion of why the speaker-identity class is the relevant comparison would help the reader.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's claims are empirical, tested on a held-out split, and do not reduce to fitted inputs or self-citations.

full rationale

The paper makes no derivation claim that reduces to its inputs. The central result is an empirical comparison (Table I) of StyleTSE against AudioSep and LLM-TSE on a held-out test split of TextrolMix. SI-SDRi is computed from estimated and reference waveforms; the text clue is a natural-language description encoded by frozen BERT, not a parameter fitted to the test labels. Training uses SI-SDR loss, dynamic mixing, and a two-stage procedure, and no reported number is recomputed from a fitted parameter or from the dataset construction rule. The TextrolMix pairing rule — interference differs in at least one style attribute, and the reference audio shares an attribute with the target but not the non-targets — guarantees that a text-discoverable separator exists, which affects task difficulty and generalization, but it is not circular: it does not make the model's output equal to its input, and evaluation still measures waveform reconstruction error against ground truth. The comparisons against a replicated LLM-TSE and fine-tuned AudioSep are external baselines. No load-bearing self-citation is present: SepFormer, BERT, LibriMix, and TextrolSpeech are external prior works, and the frozen BERT encoder supplies independent language representations. Concerns about speaker-identity leakage or limited speaker diversity are validity and generality issues, not circularity. Therefore no specific circular step can be quoted, and the honest finding is no significant circularity.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The central claim rests on a set of hand-chosen dataset construction rules and training hyperparameters, plus domain assumptions about TextrolSpeech labels and BERT text embeddings. None of these are formal axioms, but they are load-bearing: if the style-attribute construction or labels are wrong, the empirical conclusions do not transfer.

free parameters (5)
  • Target/interference attribute-difference rule = at least one differing attribute
    Every mixture is constructed so target and interference differ in at least one of six style attributes, making text clues informative by construction; this is a hand-selected dataset rule, not an empirical finding.
  • Text description length templates = long, mid, short
    Three template lengths were chosen after preliminary experiments showed poor generalization of long clues; this ad hoc choice shapes the text distribution.
  • Clue-modality training ratio = text-audio:text-only:audio-only = 2:2:1
    Second-stage training mixes clue modalities in this hand-set ratio; the paper does not justify it with an ablation.
  • Dynamic mixing interference selection = avg 2.45 alternatives per target
    Interference candidates are selected to share style attributes with the target while preserving clue effectiveness; this selection criterion is a design choice.
  • Mixture SNR and loudness settings = 0 dB mean, 4 dB std; -25 to -33 LUFS
    Adopted from LibriMix, these values control task difficulty and are not varied in experiments.
assumptions (5)
  • domain assumption TextrolSpeech style attribute labels are accurate reflections of perceived speaking style.
    Used to construct TextrolMix text descriptions and to judge whether a clue is valid; no human listening validation is reported.
  • domain assumption Frozen BERT [CLS] embeddings encode sufficient semantic information about speaking style for separation.
    The text clue is reduced to the [CLS] token and never fine-tuned; if this embedding is too coarse, text-guided extraction would fail.
  • standard math SepFormer dual-path transformer generalizes to clue-conditioned masking when clue embeddings are added.
    The paper adapts SepFormer [13] rather than deriving a new separation model; this is inherited from prior literature.
  • domain assumption Mixtures synthesized via the LibriMix procedure are representative of two-talker speech mixtures.
    The dataset uses utterance mixing with loudness normalization; results are only on such synthetic mixes.
  • domain assumption SI-SDR and PESQ adequately capture extraction quality.
    These are the only objective metrics; no listening test or downstream task is reported.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Beyond Speaker Identity: Text Guided Target Speech Extraction." pith.science (2026). https://pith.science/paper/SKFIYZ7F

@misc{pith2026250109169,
  author       = {Pith},
  title        = {Pith review of: Beyond Speaker Identity: Text Guided Target Speech Extraction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SKFIYZ7F}},
  note         = {Machine review of arXiv:2501.09169}
}
read the original abstract

Target Speech Extraction (TSE) traditionally relies on explicit clues about the speaker's identity like enrollment audio, face images, or videos, which may not always be available. In this paper, we propose a text-guided TSE model StyleTSE that uses natural language descriptions of speaking style in addition to the audio clue to extract the desired speech from a given mixture. Our model integrates a speech separation network adapted from SepFormer with a bi-modality clue network that flexibly processes both audio and text clues. To train and evaluate our model, we introduce a new dataset TextrolMix with speech mixtures and natural language descriptions. Experimental results demonstrate that our method effectively separates speech based not only on who is speaking, but also on how they are speaking, enhancing TSE in scenarios where traditional audio clues are absent. Demos are at: https://mingyue66.github.io/TextrolMix/demo/

Figures

Figures reproduced from arXiv: 2501.09169 by the authors.

Figure 1
Figure 1. Proposed text-guided target speech extraction model StyleTSE features a separation network and a bi-modality clue network. The attention [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

27 extracted references · 19 canonical work pages

  1. [1]

    V oiceFilter: Targeted V oice Separation by Speaker-Conditioned Spectrogram Masking,

    Q. Wang, H. Muckenhirn, K. Wilson, P. Sridhar, Z. Wu, J. R. Hershey, R. A. Saurous, R. J. Weiss, Y . Jia, and I. L. Moreno, “V oiceFilter: Targeted V oice Separation by Speaker-Conditioned Spectrogram Masking,” in INTERSPEECH 2019

  2. [2]

    Speakerbeam: Speaker aware neural network for target speaker extraction in speech mixtures,

    K. ˇZmol´ıkov´a, M. Delcroix, K. Kinoshita, T. Ochiai, T. Nakatani, L. Burget, and J. ˇCernock`y, “Speakerbeam: Speaker aware neural network for target speaker extraction in speech mixtures,” IEEE Journal of Selected Topics in Signal Processing , vol. 13, no. 4, pp. 800–814, 2019

  3. [3]

    Spex+: A complete time domain speaker extraction network,

    M. Ge, C. Xu, L. Wang, E. S. Chng, J. Dang, and H. Li, “Spex+: A complete time domain speaker extraction network,”INTERSPEECH, 2020

  4. [4]

    Looking to listen at the cocktail party: A speaker-independent audio-visual model for speech separation,

    A. Ephrat, I. Mosseri, O. Lang, T. Dekel, K. Wilson, A. Hassidim, W. T. Freeman, and M. Rubinstein, “Looking to listen at the cocktail party: A speaker-independent audio-visual model for speech separation,” Association for Computing Machinery , 2018

  5. [5]

    Multimodal SpeakerBeam: Single channel target speech extraction with audio-visual speaker clues

    T. Ochiai, M. Delcroix, K. Kinoshita, A. Ogawa, and T. Nakatani, “Multimodal SpeakerBeam: Single channel target speech extraction with audio-visual speaker clues.” in INTERSPEECH, 2019

  6. [6]

    Text-driven separation of arbitrary sounds,

    K. Kilgour, B. Gfeller, Q. Huang, A. Jansen, S. Wisdom, and M. Tagliasacchi, “Text-driven separation of arbitrary sounds,” IN- TERSPEECH, 2022

  7. [7]

    CLIPSep: Learning text-queried sound separation with noisy unlabeled videos,

    H.-W. Dong, N. Takahashi, Y . Mitsufuji, J. McAuley, and T. Berg- Kirkpatrick, “CLIPSep: Learning text-queried sound separation with noisy unlabeled videos,” ICLR, 2023

  8. [8]

    Separate anything you describe,

    X. Liu, Q. Kong, Y . Zhao, H. Liu, Y . Yuan, Y . Liu, R. Xia, Y . Wang, M. D. Plumbley, and W. Wang, “Separate anything you describe,” arXiv preprint arXiv:2308.05037 , 2023

Show all 27 references
  1. [9]

    Target sound extraction with variable cross-modality clues,

    C. Li, Y . Qian, Z. Chen, D. Wang, T. Yoshioka, S. Liu, Y . Qian, and M. Zeng, “Target sound extraction with variable cross-modality clues,” in ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2023, pp. 1–5

  2. [10]

    CLAPSep: Leveraging contrastive pre-trained models for multi- modal query-conditioned target sound extraction,

    H. Ma, Z. Peng, M. Shao, J. Liu, X. Li, and X. Wu, “CLAPSep: Leveraging contrastive pre-trained models for multi- modal query-conditioned target sound extraction,” arXiv preprint arXiv:2402.17455, 2024

  3. [11]

    Typing to listen at the cocktail party: Text-guided target speaker extraction,

    X. Hao, J. Wu, J. Yu, C. Xu, and K. C. Tan, “Typing to listen at the cocktail party: Text-guided target speaker extraction,” arXiv preprint arXiv:2310.07284, 2023

  4. [12]

    Target speech diariza- tion with multimodal prompts,

    Y . Jiang, R. Tao, Z. Chen, Y . Qian, and H. Li, “Target speech diariza- tion with multimodal prompts,” arXiv preprint arXiv:2406.07198 , 2024

  5. [13]

    Attention is all you need in speech separation,

    C. Subakan, M. Ravanelli, S. Cornell, M. Bronzi, and J. Zhong, “Attention is all you need in speech separation,” in ICASSP 2021- 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2021, pp. 21–25

  6. [14]

    SpeechBrain: A general-purpose speech toolkit,

    M. Ravanelli, T. Parcollet, P. Plantinga, A. Rouhe, S. Cornell, L. Lugosch, C. Subakan, N. Dawalatabad, A. Heba, J. Zhong, J.-C. Chou, S.-L. Yeh, S.-W. Fu, C.-F. Liao, E. Rastorgueva, F. Grondin, W. Aris, H. Na, Y . Gao, R. D. Mori, and Y . Bengio, “SpeechBrain: A general-purp...

  7. [15]

    Textrolspeech: A text style control speech corpus with codec language text-to-speech models,

    S. Ji, J. Zuo, M. Fang, Z. Jiang, F. Chen, X. Duan, B. Huai, and Z. Zhao, “Textrolspeech: A text style control speech corpus with codec language text-to-speech models,” in ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE,...

  8. [16]

    Optimization of speaker extraction neural network with magnitude and temporal spectrum approximation loss,

    C. Xu, W. Rao, E. S. Chng, and H. Li, “Optimization of speaker extraction neural network with magnitude and temporal spectrum approximation loss,” in ICASSP 2019-2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2019, pp. 6990–6994

  9. [17]

    LibriMix: An open-source dataset for generalizable speech separation,

    J. Cosentino, M. Pariente, S. Cornell, A. Deleforge, and E. Vin- cent, “LibriMix: An open-source dataset for generalizable speech separation,” arXiv preprint arXiv:2005.11262 , 2020

  10. [18]

    Dual-path rnn: efficient long sequence modeling for time-domain single-channel speech sepa- ration,

    Y . Luo, Z. Chen, and T. Yoshioka, “Dual-path rnn: efficient long sequence modeling for time-domain single-channel speech sepa- ration,” in ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2020, pp. 46–50

  11. [19]

    Deep neural networks for acoustic modeling in speech recognition: The shared views of four research groups,

    G. Hinton, L. Deng, D. Yu, G. E. Dahl, A.-r. Mohamed, N. Jaitly, A. Senior, V . Vanhoucke, P. Nguyen, T. N. Sainath et al. , “Deep neural networks for acoustic modeling in speech recognition: The shared views of four research groups,” IEEE Signal processing magazine, vol. 29, ...

  12. [20]

    X-vectors: Robust dnn embeddings for speaker recognition,

    D. Snyder, D. Garcia-Romero, G. Sell, D. Povey, and S. Khudanpur, “X-vectors: Robust dnn embeddings for speaker recognition,” in 2018 IEEE international conference on acoustics, speech and signal processing (ICASSP). IEEE, 2018, pp. 5329–5333

  13. [21]

    Bert: Pre-training of deep bidirectional transformers for language understanding,

    J. Devlin, “Bert: Pre-training of deep bidirectional transformers for language understanding,” arXiv preprint arXiv:1810.04805 , 2018

  14. [22]

    Semi- supervised time domain target speaker extraction with attention,

    Z. Wang, R. Giri, S. Venkataramani, U. Isik, J.-M. Valin, P. Smaragdis, M. Goodwin, and A. Krishnaswamy, “Semi- supervised time domain target speaker extraction with attention,” ICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , 2021

  15. [23]

    Performance measure- ment in blind audio source separation,

    E. Vincent, R. Gribonval, and C. F ´evotte, “Performance measure- ment in blind audio source separation,” IEEE Transactions on Audio, Speech, and Language Processing , vol. 14, no. 4, pp. 1462–1469, 2006

  16. [24]

    Wavesplit: End-to-end speech sep- aration by speaker clustering,

    N. Zeghidour and D. Grangier, “Wavesplit: End-to-end speech sep- aration by speaker clustering,” IEEE/ACM Transactions on Audio, Speech, and Language Processing , vol. 29, pp. 2840–2849, 2021

  17. [25]

    Perceptual evaluation of speech quality (PESQ)-a new method for speech quality assessment of telephone networks and codecs,

    A. W. Rix, J. G. Beerends, M. P. Hollier, and A. P. Hekstra, “Perceptual evaluation of speech quality (PESQ)-a new method for speech quality assessment of telephone networks and codecs,” in 2001 IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP)...

  18. [26]

    Conv-tasnet: Surpassing ideal time– frequency magnitude masking for speech separation,

    Y . Luo and N. Mesgarani, “Conv-tasnet: Surpassing ideal time– frequency magnitude masking for speech separation,” IEEE/ACM Transactions on Audio, Speech, and Language Processing , vol. 27, no. 8, pp. 1256–1266, 2019

  19. [27]

    Self-supervised disentangled representation learning for robust target speech extraction,

    Z. Mu, X. Yang, S. Sun, and Q. Yang, “Self-supervised disentangled representation learning for robust target speech extraction,” in Pro- ceedings of the AAAI Conference on Artificial Intelligence , vol. 38, no. 17, 2024, pp. 18 815–18 823

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.