Pith. sign in

REVIEW 42 references

Pretraining Multi-Speaker Identification for Neural Speaker Diarization

T0 review · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Pretraining a neural diarization encoder on multi-speaker identification, using short fully overlapped mixtures of up to two speakers, outperforms conventional diarization pretraining on simulated conversations and removes the need to…

desk verdict Useful pretraining alternative that likely matches DIA pretraining; the claimed superiority needs error bars to be convincing. read the letter →

arxiv 2505.24545 v1 pith:5U32EANB submitted 2025-05-30 eess.AS cs.SD

Shota Horiguchi , Atsushi Ando , Marc Delcroix , Naohiro Tawara This is my paper
classification eess.AScs.SD
keywords speakerdiarizationend-to-endneuralidentificationpretrainingmulti-speakerrecursiveattentivepoolingsimulateddatalocalself-supervisedspeechrepresentations
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

This paper claims that a neural speaker-diarization system can be pretrained on a multi-speaker identification task instead of on simulated conversational recordings. The proposed pretraining mixes zero, one, or two single-speaker utterances into short fully overlapped audio clips, trains the encoder with recursive attentive pooling to identify every speaker in the mix, and then reuses that encoder as the diarization encoder after attaching a small backend and finetuning on real conversations. In experiments across six meeting and conversation corpora, this identification-based pretraining achieves lower diarization error rates than conventional diarization pretraining on simulated or compounded real data, and using mixtures with a variable number of speakers helps further. The paper reports that a ReDimNet-B2-based model with about 5.4 million parameters matches or beats a WavLM-BLSTM baseline with about 96.5 million parameters, so the approach also offers a parameter-efficient route toward local diarization without large simulated corpora.

What carries the argument

The load-bearing object is the recursive attentive pooling mechanism of [14], which turns a frame-wise encoder into a multi-speaker identifier: it iteratively computes attention weights to pool separate speaker embeddings from one mixture and decides when to stop, so the encoder learns to separate speakers internally. Pretraining is done on 3-second fully overlapped mixtures of zero, one, or two speakers built by reusing utterances and noise already in the minibatch, which creates the multi-speaker training signal without any simulated conversation data. At diarization time the pooling and linear layers are removed, the pretrained encoder is kept, and a small backend consisting of a single LSTM and linear layer is attached; the whole network is then finetuned with powerset cross-entropy on 10-second chunks of real conversations.

What would settle it

Take a real meeting corpus with segments in which three or more speakers talk at once, finetune the proposed SID-pretrained encoder on it, and compare the diarization error rate with a DIA-pretrained baseline; if the SID-pretrained model fails to reach at least the baseline performance, the claimed transfer from fully overlapped two-speaker pretraining to real overlap conditions is contradicted.

Watch

Extended reading notes

Core claim

The central discovery is that the encoder of an end-to-end diarization system can be pretrained to solve multi-speaker identification, recognizing all speakers in a mixture containing zero, one, or two fully overlapped voices, and that this transfers to diarization better than pretraining the same encoder on the diarization task itself. Using speech from a large speaker-recognition corpus, the paper builds 3-second training samples by summing two utterances and reusing noise segments, so no conversational simulation is needed. After pretraining, the pooling and linear layers are discarded, a single LSTM plus linear layer is attached as the diarization backend, and the whole model is finetuned on each target domain. On six meeting and conversation corpora, the multi-speaker identification pretraining gives the lowest macro-average diarization error rate among all compared pretraining strategies, and a second stage of diarization pretraining on a compound of real datasets improves it further. The paper reports that the resulting ReDimNet-B2-based model, with roughly 5.4 million parameters, matches or beats a WavLM-BLSTM baseline that uses about 96.5 million parameters.

Load-bearing premise

The load-bearing premise is that training on 3-second, fully overlapped zero-to-two-speaker mixtures teaches an encoder to separate speakers in a way that transfers to real conversations, where overlaps are partial and more than two speakers can appear; this transfer is only verified empirically through finetuning and not derived or otherwise justified.

Editorial extensions

If this is right

  • Pretraining for neural diarization no longer needs hundreds of gigabytes of simulated conversations; a standard speaker-identification corpus plus on-the-fly mixing within minibatches is sufficient.
  • Encoders designed for speaker-embedding extraction can be reused directly as diarization encoders, so a diarization system can be built with around five million parameters instead of tens of millions.
  • Adding a second stage of conventional diarization pretraining on real compound data still helps when the first stage was identification-based, closing part of the gap between single-speaker and multi-speaker identification pretraining.
  • A 10-second local window makes the simulation protocol less relevant, because dialogue-act patterns are too long to appear within a short window; simulation quality thus becomes less critical for pretraining.
  • A lightweight local diarization model pretrained without simulated data can serve as the building block in hybrid clustering-based diarization, which already handles unlimited speakers through clustering.

Reading between the lines

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

  • One extension the paper does not test is pretraining with partial overlaps or more than two simultaneous speakers; the same recursive-pooling mechanism could in principle handle those, and denser overlap in pretraining might transfer even better to real conversations with heavy overlap.
  • The complementary gains from a second diarization-pretraining stage suggest identification and diarization objectives teach different skills; a single joint objective that combines both might reach the same performance in one pass without the two-stage schedule.
  • The parameter efficiency relative to a frozen self-supervised encoder hints that task-specific identification pretraining could be a cheaper alternative to large self-supervised features for on-device diarization, where model size and storage matter.
  • Because the pretraining only needs single-speaker utterances and noise, the recipe could be ported to new domains or languages by collecting a small amount of single-speaker audio, without simulating domain-specific conversations.
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.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the SID-pretraining advantage is an empirical result against external benchmarks, and the self-citation to recursive pooling is not load-bearing.

full rationale

The claimed derivation is empirical rather than analytic: a pretrained multi-speaker identification encoder is transferred to local diarization by removing the pooling/linear head, adding a diarization backend, and fine-tuning on each target real dataset. The pretraining loss (speaker identification plus counting loss) and the fine-tuning loss (powerset cross-entropy diarization loss) are different objectives, so the final DER is not forced by definition. The central comparison against DIA pretraining is measured on six external meeting corpora with oracle clustering, and no test-set parameter is fitted. The only notable self-citation is [14], which supplies the recursive attentive pooling mechanism used to extract multiple speaker embeddings. That citation is not load-bearing for the paper's central claim: the mechanism is re-validated in this paper's Table 2 on standard VoxCeleb verification sets, and its value for diarization is established by the downstream DER results rather than by importing [14]'s conclusions. Consequently, no step in the derivation chain reduces to its own inputs. The small macro-DER differences and absence of significance testing flagged in the skeptical review concern statistical robustness, not circularity.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

The central claim rests on a small set of domain assumptions about overlap counts, the transferability of identification representations, and the suitability of VoxCeleb. No free parameters are fitted to the evaluation data, and no new physical or architectural entities are postulated.

assumptions (3)
  • domain assumption It is often assumed in diarization studies that the maximum number of speakers speaking simultaneously is two.
    Quoted from §3.3. The pretraining uses audio containing 0 to 2 speakers, and the powerset loss is limited to at most two overlapping speakers. This limits the model's exposure to more than two overlapping speakers, which may occur in some real recordings.
  • domain assumption A model pretrained to identify speakers from fully overlapped 3-second mixtures learns representations useful for diarization of real conversations.
    Core transfer assumption. It is validated only empirically by finetuning results in Section 5; no theoretical argument is given for why the identification objective should transfer to frame-wise diarization.
  • domain assumption VoxCeleb 1&2 is a sufficient source of speaker diversity for pretraining.
    Used as the only pretraining corpus. The paper does not compare with other large speaker recognition corpora, so the result may depend on the specific properties of VoxCeleb.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Pretraining Multi-Speaker Identification for Neural Speaker Diarization." pith.science (2026). https://pith.science/paper/5U32EANB

@misc{pith2026250524545,
  author       = {Pith},
  title        = {Pith review of: Pretraining Multi-Speaker Identification for Neural Speaker Diarization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5U32EANB}},
  note         = {Machine review of arXiv:2505.24545}
}
read the original abstract

End-to-end speaker diarization enables accurate overlap-aware diarization by jointly estimating multiple speakers' speech activities in parallel. This approach is data-hungry, requiring a large amount of labeled conversational data, which cannot be fully obtained from real datasets alone. To address this issue, large-scale simulated data is often used for pretraining, but it requires enormous storage and I/O capacity, and simulating data that closely resembles real conversations remains challenging. In this paper, we propose pretraining a model to identify multiple speakers from an input fully overlapped mixture as an alternative to pretraining a diarization model. This method eliminates the need to prepare a large-scale simulated dataset while leveraging large-scale speaker recognition datasets for training. Through comprehensive experiments, we demonstrate that the proposed method enables a highly accurate yet lightweight local diarization model without simulated conversational data.

Figures

Figures reproduced from arXiv: 2505.24545 by the authors.

Figure 1
Figure 1. Comparison of pretraining strategies. SID: speaker identification, DIA: speaker diarization. Early models use a position-wise feed-forward network with sigmoid activation for the diarization backend, i.e., gdiar : R D → (0, 1)S , limiting EEND to at most S speakers [5]. To break this limitation, some methods introduced block-wise pro￾cessing followed by clustering to integrate the block-wise re￾sults [6, 7]. Cluster… view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

42 extracted references · 35 canonical work pages

  1. [1]

    Introduction Speaker diarization, which estimates who is speaking when, plays an essential role in multi-speaker applications such as speech separation [1] and speech recognition [2,3]. Speaker di- arization methods fall into three directions: clustering speaker embeddings from short segments [4], end-to-end neural net- works identifying speaker-wise spee...

    work page

  2. [2]

    Pretraining Multi-Speaker Identification for Neural Speaker Diarization

    Review of conventional methods 2.1. End-to-end neural diarization EEND is initially proposed as a single-modeled clustering-free diarization method [5]. It generates frame-wise posteriors of speech activity for each of S speakers [pt]T t=1 ∈ (0, 1)S×T from input frame-wise acoustic features[xt]T t=1 ∈ RD×T using a neural network. The earliest model consis...

    work page Pith review arXiv 2025

  3. [3]

    Pretraining strategies of EEND models 3.1. Baseline 1: Diarization pretraining Most EEND methods rely on pretraining using large-scale con- versational datasets with permutation-free diarization loss [5– 12], referred to here as DIA pretraining (Fig. 1(a)). Pretraining and finetuning share the same model architecture: an encoder (f (DIA) enc in (1)) using...

    work page

  4. [4]

    Dataset Table 1 lists the datasets used in our experiments, all monaural with a 16 kHz sampling rate and 16 bit depth

    Experimental setup 4.1. Dataset Table 1 lists the datasets used in our experiments, all monaural with a 16 kHz sampling rate and 16 bit depth. The V oxCeleb 1&2 dataset was used for SID pretraining and mixture gener- ation. We used two simulation protocols. The first follows original EEND training: concatenating utterances interleaved by silence to genera...

    work page

  5. [5]

    Results 5.1. Preliminary results of speaker verification We first report speaker verification results under three condi- tions: whether two single-speaker recordings are of the same speaker ( s vs. s ), whether a two-speaker recording has the speaker in a single-speaker recording (s vs. s), and whether two two-speaker recordings have the same speaker ( m ...

    work page arXiv

  6. [6]

    The method is storage-friendly, simulation-agnostic, and outperformed diarization-based pre- training, with further gains from additional DIA pretraining

    Conclusion This paper demonstrated the effectiveness of multi-speaker SID pretraining for EEND. The method is storage-friendly, simulation-agnostic, and outperformed diarization-based pre- training, with further gains from additional DIA pretraining. Future work will include the method to perform local diariza- tion and speaker embedding extraction in a s...

    work page

  7. [7]

    Front-end processing for the CHiME-5 dinner party scenario,

    C. Boeddeker, J. Heitkaemper, J. Schmalenstoeer, L. Drude, J. Heymann, and R. Haeb-Umbach, “Front-end processing for the CHiME-5 dinner party scenario,” in Proc. CHiME-5, 2018, pp. 35–40

    work page 2018

  8. [8]

    BUT/JHU system description for CHiME-8 NOTSOFAR-1 challenge,

    A. Polok, D. Klement, J. Han, ˇS. Sedl ´aˇcek, B. Yusuf, M. Ma- ciejewski, M. S. Wiesner, and L. Burget, “BUT/JHU system description for CHiME-8 NOTSOFAR-1 challenge,” in Proc. CHiME, 2024, pp. 18–22

    work page 2024

Show all 42 references
  1. [9]

    Di- CoW: Diarization-conditioned whisper for target speaker auto- matic speech recognition,

    A. Polok, D. Klement, M. Kocour, J. Han, F. Landini, B. Yusuf, M. Wiesner, S. Khudanpur, J. ˇCernock`y, and L. Burget, “Di- CoW: Diarization-conditioned whisper for target speaker auto- matic speech recognition,” arXiv:2501.00114, 2024

    2024 arXiv

  2. [10]

    Bayesian HMM clustering of x-vector sequences (VBx) in speaker diarization: Theory, implementation and analysis on standard tasks,

    F. Landini, J. Profant, M. Diez, and L. Burget, “Bayesian HMM clustering of x-vector sequences (VBx) in speaker diarization: Theory, implementation and analysis on standard tasks,” Com- puter Speech & Language, vol. 71, p. 101254, 2022

    2022

  3. [11]

    End-to-end neural speaker diarization with permutation-free objectives,

    Y . Fujita, N. Kanda, S. Horiguchi, K. Nagamatsu, and S. Watan- abe, “End-to-end neural speaker diarization with permutation-free objectives,” in Proc. Interspeech, 2019, pp. 4300–4304

    2019

  4. [12]

    Integrating end-to- end neural and clustering-based diarization: Getting the best of both worlds,

    K. Kinoshita, M. Delcroix, and N. Tawara, “Integrating end-to- end neural and clustering-based diarization: Getting the best of both worlds,” in Proc. ICASSP, 2021, pp. 7198–7202

    2021

  5. [13]

    Towards neural diarization for unlimited num- bers of speakers using global and local attractors,

    S. Horiguchi, S. Watanabe, P. Garc ´ıa, Y . Xue, Y . Takashima, and Y . Kawaguchi, “Towards neural diarization for unlimited num- bers of speakers using global and local attractors,” inProc. ASRU, 2021, pp. 98–105

    2021

  6. [14]

    pyannote.audio 2.1 speaker diarization pipeline: prin- ciple, benchmark, and recipe,

    H. Bredin, “pyannote.audio 2.1 speaker diarization pipeline: prin- ciple, benchmark, and recipe,” in Proc. Interspeech, 2023, pp. 1983–1987

    2023

  7. [15]

    Powerset multi-class cross entropy loss for neural speaker diarization,

    A. Plaquet and H. Bredin, “Powerset multi-class cross entropy loss for neural speaker diarization,” in Proc. Interspeech, 2023, pp. 3222–3226

    2023

  8. [16]

    End-to-end diarization for variable number of speakers with local-global networks and discriminative speaker embeddings,

    S. Maiti, H. Erdogan, K. Wilson, S. Wisdom, S. Watanabe, and J. R. Hershey, “End-to-end diarization for variable number of speakers with local-global networks and discriminative speaker embeddings,” in Proc. ICASSP, 2021, pp. 7183–7187

    2021

  9. [17]

    Improving the nat- uralness of simulated conversations for end-to-end neural diariza- tion,

    N. Yamashita, S. Horiguchi, and T. Homma, “Improving the nat- uralness of simulated conversations for end-to-end neural diariza- tion,” in Proc. Odyssey, 2022, pp. 133–140

    2022

  10. [18]

    From simu- lated mixtures to simulated conversations as training data for end- to-end neural diarization,

    F. Landini, A. Lozano-Diez, M. Diez, and L. Burget, “From simu- lated mixtures to simulated conversations as training data for end- to-end neural diarization,” in Proc. Interspeech, 2022, pp. 5095– 5099

    2022

  11. [19]

    Leveraging self-supervised learning for speaker diarization,

    J. Han, F. Landini, J. Rohdin, A. Silnova, M. Diez, and L. Burget, “Leveraging self-supervised learning for speaker diarization,” in Proc. ICASSP, 2025

    2025

  12. [20]

    Recursive attentive pooling for ex- tracting speaker embeddings from multi-speaker recordings,

    S. Horiguchi, A. Ando, T. Moriya, T. Ashihara, H. Sato, N. Tawara, and M. Delcroix, “Recursive attentive pooling for ex- tracting speaker embeddings from multi-speaker recordings,” in Proc. SLT, 2024, pp. 1219–1226

    2024

  13. [21]

    Frame-wise and overlap-robust speaker em- beddings for meeting diarization,

    T. Cord-Landwehr, C. Boeddeker, C. Zoril ˘a, R. Doddipatla, and R. Haeb-Umbach, “Frame-wise and overlap-robust speaker em- beddings for meeting diarization,” in Proc. ICASSP, 2023

    2023

  14. [22]

    Leverag- ing speaker embeddings in end-to-end neural diarization for two- speaker scenarios,

    J. I. Alvarez-Trejos, B. Labrador, and A. Lozano-Diez, “Leverag- ing speaker embeddings in end-to-end neural diarization for two- speaker scenarios,” in Proc. Odyssey, 2024, pp. 107–114

    2024

  15. [23]

    ECAPA- TDNN: Emphasized channel attention, propagation and aggrega- tion in TDNN based speaker verification,

    B. Desplanques, J. Thienpondt, and K. Demuynck, “ECAPA- TDNN: Emphasized channel attention, propagation and aggrega- tion in TDNN based speaker verification,” in Proc. Interspeech, 2020, pp. 3830–3834

    2020

  16. [24]

    Reshape dimensions network for speaker recognition,

    I. Yakovlev, R. Makarov, A. Balykin, P. Malov, A. Okhotnikov, and N. Torgashov, “Reshape dimensions network for speaker recognition,” in Proc. Interspeech, 2024, pp. 3235–3239

    2024

  17. [25]

    Advances in inte- gration of end-to-end neural and clustering-based diarization for real conversational speech,

    K. Kinoshita, M. Delcroix, and N. Tawara, “Advances in inte- gration of end-to-end neural and clustering-based diarization for real conversational speech,” inProc. Interspeech, 2021, pp. 3565– 3569

    2021

  18. [26]

    BUT system for the Second DIHARD Speech Diarization Challenge,

    F. Landini, S. Wang, M. Diez, L. Burget, P. Mat ˇejka, K. ˇZmol´ıkov´a, L. Mo ˇsner, A. Silnova, O. Plchot, O. Novotn `y, H. Zeinali, and J. Rohdin, “BUT system for the Second DIHARD Speech Diarization Challenge,” inProc. ICASSP, 2020, pp. 6529– 6533

    2020

  19. [27]

    Overlap-aware diarization: Resegmentation using neural end-to-end overlapped speech detection,

    L. Bullock, H. Bredin, and L. P. Garcia-Perera, “Overlap-aware diarization: Resegmentation using neural end-to-end overlapped speech detection,” in Proc. ICASSP, 2020, pp. 7114–7118

    2020

  20. [28]

    End-to-end speaker diarization as post-processing,

    S. Horiguchi, P. Garcia, Y . Fujita, S. Watanabe, and K. Naga- matsu, “End-to-end speaker diarization as post-processing,” in Proc. ICASSP, 2021, pp. 7188–7192

    2021

  21. [29]

    V oxCeleb: Large-scale speaker verification in the wild,

    A. Nagrani, J. S. Chung, W. Xie, and A. Zisserman, “V oxCeleb: Large-scale speaker verification in the wild,”Computer Speech & Language, vol. 60, p. 101027, 2020

    2020

  22. [30]

    AISHELL-4: An open source dataset for speech enhancement, separation, recognition and speaker diarization in conference scenario,

    Y . Fu, L. Cheng, S. Lv, Y . Jv, Y . Kong, Z. Chen, Y . Hu, L. Xie, J. Wu, H. Bu, X. Xu, J. Du, and J. Chen, “AISHELL-4: An open source dataset for speech enhancement, separation, recognition and speaker diarization in conference scenario,” in Proc. Inter- speech, 2021, pp. 3665–3669

    2021

  23. [31]

    M2MeT: The ICASSP 2022 multi-channel multi-party meeting transcription challenge,

    F. Yu, S. Zhang, Y . Fu, L. Xie, S. Zheng, Z. Du, W. Huang, P. Guo, Z. Yan, B. Ma, X. Xu, and H. Bu, “M2MeT: The ICASSP 2022 multi-channel multi-party meeting transcription challenge,” in Proc. ICASSP, 2022, pp. 6167–6171

    2022

  24. [32]

    Unleashing the killer corpus: experiences in creating the multi-everything AMI Meeting Corpus,

    J. Carletta, “Unleashing the killer corpus: experiences in creating the multi-everything AMI Meeting Corpus,”Language Resources and Evaluation, vol. 41, no. 2, pp. 181–190, 2007

    2007

  25. [33]

    Open source MagicData-RAMC: A rich annotated Mandarin conversational (RAMC) speech dataset,

    Z. Yang, Y . Chen, L. Luo, R. Yang, L. Ye, G. Cheng, J. Xu, Y . Jin, Q. Zhang, P. Zhang, L. Xie, and Y . Yan, “Open source MagicData-RAMC: A rich annotated Mandarin conversational (RAMC) speech dataset,” in Proc. Interspeech, 2022, pp. 1736– 1740

    2022

  26. [34]

    MSDWild: Multi- modal speaker diarization dataset in the wild,

    T. Liu, S. Fan, X. Xiang, H. Song, S. Lin, J. Sun, T. Han, S. Chen, B. Yao, S. Liu, Y . Wu, Y . Qian, and K. Yu, “MSDWild: Multi- modal speaker diarization dataset in the wild,” in Proc. Inter- speech, 2022, pp. 1476–1480

    2022

  27. [35]

    Spot the conversation: Speaker diarisation in the wild,

    J. S. Chung, J. Huh, A. Nagrani, T. Afouras, and A. Zisserman, “Spot the conversation: Speaker diarisation in the wild,” in Proc. Interspeech, 2020, pp. 299–303

    2020

  28. [36]

    Mamba-based segmentation model for speaker diariza- tion,

    A. Plaquet, N. Tawara, M. Delcroix, S. Horiguchi, A. Ando, and S. Araki, “Mamba-based segmentation model for speaker diariza- tion,” in Proc. ICASSP, 2025

    2025

  29. [37]

    FunASR: A fundamental end-to- end speech recognition toolkit,

    Z. Gao, Z. Li, J. Wang, H. Luo, X. Shi, M. Chen, Y . Li, L. Zuo, Z. Du, Z. Xiao, and S. Zhang, “FunASR: A fundamental end-to- end speech recognition toolkit,” in Proc. Interspeech, 2023, pp. 1593–1597

    2023

  30. [38]

    Adam: A method for stochastic opti- mization,

    D. P. Kingma and J. Ba, “Adam: A method for stochastic opti- mization,” in Proc. ICLR, 2015

    2015

  31. [39]

    Speaker recognition from raw wave- form with SincNet,

    M. Ravanelli and Y . Bengio, “Speaker recognition from raw wave- form with SincNet,” in Proc. SLT, 2018, pp. 1021–1028

    2018

  32. [40]

    WavLM: Large-scale self- supervised pre-training for full stack speech processing,

    S. Chen, C. Wang, Z. Chen, Y . Wu, S. Liu, Z. Chen, J. Li, N. Kanda, T. Yoshioka, X. Xiaoet al., “WavLM: Large-scale self- supervised pre-training for full stack speech processing,” IEEE Journal of Selected Topics in Signal Processing , vol. 16, no. 6, pp. 1505–1518, 2022

    2022

  33. [41]

    pyan- note.audio speaker diarization pipeline at V oxSRC 2023,

    S. Baroudi, H. Bredin, A. Plaquet, and T. Pellegrini, “pyan- note.audio speaker diarization pipeline at V oxSRC 2023,” The V oxCeleb Speaker Recognition Challenge, 2023

    2023

  34. [42]

    NTT speaker diarization system for CHiME-7: Multi-domain, multi- microphone end-to-end and vector clustering diarization,

    N. Tawara, M. Delcroix, A. Ando, and A. Ogawa, “NTT speaker diarization system for CHiME-7: Multi-domain, multi- microphone end-to-end and vector clustering diarization,” inProc. ICASSP, 2024, pp. 11 281–11 285

    2024

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.