Pith. sign in

REVIEW 3 major objections 6 minor 54 references

Revisiting SSL for sound event detection: complementary fusion and adaptive post-processing

T0 review · 3 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read On DCASE 2023, one SSL pairing lifts F1 and an adaptive post-processor matches grid search at 1/14 the runtime.

desk verdict A useful SED benchmark with an adaptive post-processing trick; the accuracy edge over cSEBBs is small and statistically unsupported. read the letter →

arxiv 2505.11889 v2 pith:VM7GKXHQ submitted 2025-05-17 eess.AS cs.AIcs.SD

Hanfang Cui , Longfei Song , Li Li , Dongxing Xu , Yanhua Long This is my paper · ORCID
classification eess.AScs.AIcs.SD
keywords soundeventdetectionself-supervisedlearningBEATsHuBERTWavLMSEBBsnPSDS
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

The paper tries to settle two practical questions in sound event detection: which self-supervised models are worth using, and how to post-process their outputs without expensive parameter searches. It reports that BEATs is the best single SSL feature extractor for a CRNN detector, and that fusing BEATs with WavLM embeddings gives a modest but consistent boost to event-level F1, while fusing all five SSL models hurts every metric. Its new post-processor, nSEBBs, replaces grid search with per-class statistics and matches the grid-searched baseline on PSDS1 while cutting runtime from about 130 s to 8 s. If these results hold, a practitioner can pick one strong SSL backbone, consider a single complementary pairing, and skip the parameter search.

What carries the argument

The load-bearing mechanism is nSEBBs, a normalized variant of the sound event bounding boxes post-processor. For each event class it derives two statistics from the frame-level scores: the posterior contrast ratio (PCR), a log-ratio of high-confidence to low-confidence posterior mass, and the average event duration (avgED), read off a high-score mask. A fixed mapping table converts those statistics into the step-filter length and the relative merging threshold, replacing the three-parameter grid search of cSEBBs. The underlying boundary detection still uses cSEBBs' delta-score change points and segment merging. The fusion results rest on a simpler mechanism: frame-level SSL embeddings from different encoders are temporally aligned and combined by element-wise addition or concatenation before the CRNN classifier.

What would settle it

Run the same CRNN+BEATs system with nSEBBs on a different strongly labeled SED benchmark, such as DCASE 2024 Task 4 or a held-out split of DESED, without retuning the PCR/duration-to-parameter mapping; if nSEBBs no longer matches or beats cSEBBs in PSDS1, the adaptive mapping is overfit to the 2023 validation set.

Watch

Extended reading notes

Core claim

On the DCASE 2023 Task 4 validation set, the paper establishes three claims. First, among individual SSL embeddings, CRNN+BEATs is clearly the best, reaching PSDS1 0.497 and event-F1 55.7%, while speech-pretrained models (wav2vec2.0, HuBERT, WavLM) stay near the MT-CRNN baseline. Second, dual-modal fusion helps only in specific pairings: concatenating BEATs with WavLMbase raises event-F1 to 58.6% and inter-F1 to 80.2%, but every dual or full fusion keeps PSDS1 below the single-BEATs baseline, and full fusion drops to PSDS1 0.448. Third, the adaptive nSEBBs post-processor matches or slightly exceeds the grid-searched cSEBBs on PSDS1 for standalone SSL models (e.g., HuBERTbase 0.393 vs 0.390) and for dual fusion (WavLMbase 0.500 vs 0.499), while running in about 8 s instead of 130 s.

Load-bearing premise

The hand-built tables converting each class's confidence and duration statistics into post-processing parameters were tuned on the DCASE 2023 validation set and are assumed to transfer to other recordings and datasets.

Editorial extensions

If this is right

  • CRNN+BEATs is a strong default: no tested SSL addition improves its PSDS1, so for boundary-focused evaluation the single backbone is enough.
  • A practitioner who wants higher event-level F1 should concatenate BEATs with WavLMbase, which raises event-F1 from 55.7% to 58.6% and inter-F1 from 79.0% to 80.2%.
  • nSEBBs removes the cost of grid search: per-class PCR and avgED pick the same quality parameters in about 8 s versus 130 s, so post-processing can run on-device.
  • Dual-threshold versions (nSEBBsD and cSEBBsD) consistently lose PSDS1 to single-threshold versions, so adding an absolute threshold is not a free improvement.
  • Aggregating all five SSL representations is counterproductive, dropping PSDS1 to 0.448 and event-F1 to 53.7%, which argues for selective fusion rather than model soup.

Reading between the lines

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

  • The authors' mapping from PCR and avgED to parameters is presented without an ablation; a natural check is to learn that mapping from the validation set rather than fixing it, and to test on a later DCASE task to see whether the constants transfer.
  • The WavLM complementarity suggests that pretraining-domain diversity (speech vs audio) is the useful axis rather than audio-domain pretraining per se; fusing BEATs with another non-speech model trained on a distinct corpus would test whether the F1 gain is due to WavLM specifically.
  • Because nSEBBs needs only per-class score statistics, the same recipe could be lifted to other temporal detection tasks with variable event durations, such as anomaly or bioacoustic event detection; the paper does not test this.
  • Full-fusion collapse hints that element-wise addition of many SSL streams saturates or misaligns; an attention-weighted or learned combination might avoid the drop, but that is an extrapolation from the paper's evidence.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper evaluates five self-supervised learning (SSL) models (BEATs, HuBERT, wav2vec 2.0, WavLM, Dasheng) as feature extractors for sound event detection (SED) on the DCASE 2023 Task 4 development set, using a CRNN+BEATs baseline. It proposes three SSL fusion strategies: individual SSL embedding integration, dual-modal fusion (BEATs plus one other model), and full aggregation of all SSL embeddings. It further introduces nSEBBs, an adaptive post-processing method that replaces cSEBBs' grid-searched thresholds with a hand-specified mapping from per-class statistics (posterior contrast ratio and average event duration) to the SEBBs parameters. The main reported results are that dual-modal fusion with WavLM yields the highest event-F1 (58.6% vs. 55.7% for CRNN+BEATs, Table 5), and nSEBBs achieves PSDS1 comparable to cSEBBs at roughly 1/14 the runtime (8 s vs. 130 s, §5.4), with the largest PSDS1 gain over the no-post-processing baseline being about 0.04 (e.g., CRNN+HuBERTbase from 0.352 to 0.393, Table 7).

Significance. If the results hold, the paper offers a useful systematic benchmark of SSL representations for SED and a computationally cheaper post-processing alternative that retains most of the benefit of cSEBBs. The per-class analysis in Table 6 and the explicit algorithmic descriptions are strengths. The paper is honest about the limited gains from fusion and the degradation from full aggregation. However, the empirical claims rest on single-point estimates from three-seed runs, and the adaptive mapping is presented without sensitivity or transfer analysis, so the generality of the speed-accuracy trade-off is not yet established. With added uncertainty quantification and mapping validation, this could be a solid contribution to the SED community.

major comments (3)
  1. [§4.3, Tables 3–8] The paper states in §4.3 that 'Each experiment was repeated three times with different random seeds to ensure reproducibility,' yet all tables report a single number per configuration. This is load-bearing for both headline claims: the fusion claim in Table 5 (CRNN+BEATs+WavLMbase concat event-F1 55.7→58.6) and the post-processing claim in Table 7 (CRNN+HuBERTbase PSDS1 0.352→0.393) involve small differences that could easily be within seed noise in SED. The 'nSEBBs is comparable to cSEBBs' assertion is especially fragile, as Table 7 reports differences as small as 0.001–0.016 PSDS1 (e.g., 0.393 vs. 0.390 for HuBERT; 0.504 vs. 0.520 for CRNN+BEATs). Without per-seed values, means, and standard deviations (or significance tests), the central quantitative claims are not established. Please report the three-seed statistics for at least Tables 4, 5, 7, and 8, and discuss whether the observed differences are consistent across seeds.
  2. [§3.3, Eq. (3), Tables 1–2] The adaptive parameter mapping is presented as a principled replacement for cSEBBs' grid search, but the constants in Eq. (3) and Tables 1–2 (θ_base ∈ {2.0, 2.4, 2.8, 3.2}, lstep ∈ {0.384, 0.512, 0.640, 0.800}, θ_abs ∈ {0.12, 0.18, 0.24, 0.30}, and the piecewise slopes 0.1/0.3) are introduced without any sensitivity analysis, ablation, or cross-dataset check. Since the runtime advantage of nSEBBs (8 s vs. 130 s, §5.4) comes precisely from replacing the search with this fixed mapping, the mapping's generality is load-bearing for the paper's central efficiency claim. If these constants were chosen by inspecting the DCASE 2023 validation set, the reported nSEBBs numbers are not independent predictions. Please provide (a) a sensitivity analysis showing how PSDS1 varies with each constant, (b) the distribution of (PCR, avgED) over the validation set to justify the bin boundaries, and (c) ideally, a cross-dataset evaluation (e.g., DCASE 2024 Task 4) to test transfer.
  3. [Abstract, §5.4, Table 7] The claim that nSEBBs 'improving PSDS1 by up to 4%' is ambiguous because the comparison baseline is not stated. In Table 7, the improvements over the no-post-processing baseline are up to 0.041 points (CRNN+HuBERTbase: 0.352→0.393), which is about 4 percentage points, but cSEBBs achieves 0.390 on the same row. For CRNN+BEATs, nSEBBs (0.504) is actually worse than cSEBBs (0.520). The abstract and conclusion should clarify that the improvement is over the raw baseline, not over cSEBBs, and should quantify the nSEBBs-vs-cSEBBs gap (which is often within 0.01–0.02 PSDS1, tying into the missing uncertainty estimates in the first major comment).
minor comments (6)
  1. [§3.3, Algorithm 2 / Eq. (3)] The variable names are inconsistent: Algorithm 2 returns 'PCR' as the first statistic, but Eq. (3) uses 'pcrc'; the text also refers to 'PCR (Percentage of Coverage Rate)' in one place while earlier defining it as 'Posterior Contrast Ratio'. Please unify terminology and define the acronym at first use.
  2. [Table 6] Table 6 lists 'CRNN+BEATs+WavLMbase' without specifying whether the additive or concatenative fusion variant is used, even though Table 5 shows different results for the two fusion modes. Please state the fusion method used for the class-wise results.
  3. [§5.4, Table 7] The 'cSEBBsD' variant is said to 'result from the concurrent application of both thresholds,' but no description is given of how θ_abs and θ_rel interact or how the cSEBBsD thresholds were chosen. A brief algorithmic note or a reference to the same mapping as nSEBBsD would clarify the comparison.
  4. [§3.3, Algorithm 2] The definition of 'nl← Percentile(sc[t], 10)' is unclear: is this the 10th percentile of the score values for that class? If so, for scores that are mostly near zero (as is typical for sigmoid outputs), this quantity can be extremely small, making the PCR unstable. Please state the percentile computation and discuss numerical stability.
  5. [General] The paper contains several grammatical and typographical errors (e.g., 'all these performance difference may due to the differ in their model structures', 'The system employs three principal operational parameters: 1) ... 2) ...', and the inconsistent use of 'nSEBBsD' with a subscript). A careful language edit is needed.
  6. [§1, Contribution list] The second bullet claims 'a novel normalization-based post-processing method,' but the algorithm does not explicitly normalize the scores; it computes mean and variance and applies a mask. Please clarify what 'normalized' refers to (e.g., the use of statistical moments to derive parameters) to avoid overclaiming.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the paper's claims are empirical benchmark results with external baselines; nSEBBs is an algorithmic extension, not a fitted identity.

full rationale

Walking the paper's derivation chain, the main claims are empirical comparisons on the DCASE 2023 Task 4 validation set. The fusion results compare trained CRNN models using externally published SSL embeddings (BEATs, HuBERT, WavLM, etc.), and no equation is defined in terms of the reported PSDS1 or event-F1 values. The nSEBBs post-processor computes per-recording statistics (PCR, avgED) via Algorithm 2 and maps them to boundary-detection parameters through the fixed piecewise mapping in Eq. (3) and Tables 1-2. This mapping is an ansatz with stated constant values, not a parameter fitted to the reported metric values, and the comparison against cSEBBs is against an external prior method (Ebbers et al. 2024), not a self-citation chain. The 'up to 4%' PSDS1 improvement is read directly from Table 7 by comparing nSEBBs to the no-post-processing baseline, so it is not forced by construction. No load-bearing self-citation, no imported uniqueness theorem, and no renaming of a known result as a new one. Remaining concerns such as the absence of seed-level variance and the use of a single validation set for both development and reporting are statistical robustness and generalizability issues, not circularity, and the paper explicitly describes its three-seed repetition in Section 4.3 even though the tables report single numbers. Under the hard rule that circularity requires quoting a specific reduction or fitted parameter renamed as prediction, no such step is present.

Assumptions & free parameters 5 free parameters · 5 assumptions · 2 invented entities

The nSEBBs post-processor depends on a hand-specified piecewise mapping from per-class statistics (PCR, avgED) to thresholds and filter lengths, plus several fixed constants inside Algorithm 2. These are not learned on held-out data, and the paper does not show a sensitivity analysis. The fusion benchmark itself contributes no new free parameters beyond standard training hyperparameters.

free parameters (5)
  • lstep duration mapping = 0.384, 0.512, 0.640, 0.800 s for duration buckets [0,20), [20,40), [40,90), [90,200)
    Hand-chosen in Table 1 to map average event duration to step-filter length; replaces cSEBBs grid search but is itself a fixed design choice.
  • theta_base mapping = 2.0, 2.4, 2.8, 3.2 for PCR buckets [0,10), [10,20), [20,30), [30,50)
    Hand-chosen in Table 1 to modulate the relative merge threshold in Eq. (3).
  • piecewise regression slopes in Eq. (3) = 0.1 and 0.3 per PCR unit
    Scaling factors selected by the authors, not learned or cross-validated.
  • theta_abs mapping for nSEBBsD = 0.12, 0.18, 0.24, 0.30 per duration bucket
    Hand-chosen in Table 2 for the dual-threshold variant; tied to the same duration buckets.
  • Algorithm 2 statistics thresholds = 0.5 sigma mask threshold; 10th percentile noise floor
    Mask and noise floor parameters in Algorithm 2 are chosen without ablation.
assumptions (5)
  • domain assumption DCASE 2023 Task 4 DESED validation set is a representative SED benchmark
    All conclusions are drawn from one dataset; no cross-dataset validation.
  • domain assumption Pre-trained SSL embeddings are used as fixed feature extractors with no fine-tuning
    Sections 3.1-3.2; results depend on the pretraining corpora and on the embedding alignment procedure.
  • ad hoc to paper Local extrema of delta scores correspond to event onsets and offsets
    Algorithm 4 defines boundaries as local maxima and minima of the step-filtered score difference; this heuristic is assumed to hold across event types.
  • ad hoc to paper The mapping from PCR and avgED to thresholds is universal
    Eq. (3) and Tables 1-2 are presented without a transfer or sensitivity study.
  • standard math Percentile and variance computations are well-defined for bounded frame scores
    Algorithm 1 uses per-class mean, variance, and percentile; this is standard and uncontroversial.
invented entities (2)
  • Posterior Contrast Ratio (PCR)
    purpose: Summarizes how concentrated high-confidence predictions are, used to set the merge threshold theta_rel.
    Defined in Algorithm 2 on the model's own posteriors; no external or theoretical justification is provided beyond performance on the validation set.
  • Average Event Duration (avgED)
    purpose: Controls the step-filter length lstep for each sound class.
    Computed from a 0.5-sigma mask of the frame scores; no external validation that this estimates true event duration.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Revisiting SSL for sound event detection: complementary fusion and adaptive post-processing." pith.science (2026). https://pith.science/paper/VM7GKXHQ

@misc{pith2026250511889,
  author       = {Pith},
  title        = {Pith review of: Revisiting SSL for sound event detection: complementary fusion and adaptive post-processing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VM7GKXHQ}},
  note         = {Machine review of arXiv:2505.11889}
}
read the original abstract

Self-supervised learning (SSL) models offer powerful representations for sound event detection (SED), yet their synergistic potential remains underexplored. This study systematically evaluates state-of-the-art SSL models to guide optimal model selection and integration for SED. We propose a framework that combines heterogeneous SSL representations (e.g., BEATs, HuBERT, WavLM) through three fusion strategies: individual SSL embedding integration, dual-modal fusion, and full aggregation. Experiments on the DCASE 2023 Task 4 Challenge reveal that dual-modal fusion (e.g., CRNN+BEATs+WavLM) achieves complementary performance gains, while CRNN+BEATs alone delivers the best results among individual SSL models. We further introduce normalized sound event bounding boxes (nSEBBs), an adaptive post-processing method that dynamically adjusts event boundary predictions, improving PSDS1 by up to 4% for standalone SSL models. These findings highlight the compatibility and complementarity of SSL architectures, providing guidance for task-specific fusion and robust SED system design.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

54 extracted references · 21 canonical work pages

  1. [1]

    write newline

    " write newline " cite write " FUNCTION editor.postfix editor num.names #1 > "( )" "( )" if FUNCTION editor.trans.postfix editor num.names #1 > "( )" "( )" if FUNCTION trans.postfix translator num.names #1 > "( )" "( )" if FUNCTION authors.editors.reflist.apa5 'field := 'dot := field num.names 'numnames := numnames 'format.num.names := format.num.names na...

  2. [2]

    sn-aps.bst

    FUNCTION identify.aps.version "sn-aps.bst" " [2024/07/19 v1.1 APS bibliography style]" * top ENTRY address author booktitle chapter doi edition editor eid howpublished institution journal key keywords month note number organization pages publisher school series title type url volume year eprint archive archivePrefix primaryClass adsurl adsnote version lab...

    2024

  3. [3]

    write newline

    " write newline "" before.all 'output.state := FUNCTION if.digit duplicate "0" = swap duplicate "1" = swap duplicate "2" = swap duplicate "3" = swap duplicate "4" = swap duplicate "5" = swap duplicate "6" = swap duplicate "7" = swap duplicate "8" = swap "9" = or or or or or or or or or FUNCTION n.separate 't := "" #0 'numnames := t empty not t #-1 #1 subs...

  4. [4]

    , Elizalde , B

    bchapter Ashraf , K. , Elizalde , B. , Iandola , F. , Moskewicz , M. , Bernd , J. , Friedland , G. , Keutzer , K. : Audio-based multimedia event detection with dnns sparse sampling . In: Proceedings of the 5th ACM International Conference on Multimedia Retrieval , pp. 611 -- 614 ( 2015 ). 10.1145/2671188.2749396 bchapter

    arXiv 2015

  5. [5]

    , Ferroni , G

    bchapter Bilen , C . , Ferroni , G. , Tuveri , F. , Azcarreta , J. , Krstulovi \'c , S. : A framework for the robust evaluation of sound event detection . In: ICASSP 2020 - IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , pp. 61 -- 65 ( 2020 ). 10.1109/ICASSP40776.2020.9052995 bchapter

    arXiv 2020

  6. [6]

    , Mydlarz , C

    bchapter Bello , J.P. , Mydlarz , C. , Salamon , J. : Sound analysis in smart cities . In: Virtanen , T. , Plumbley , M.D. , Ellis , D. (eds.) Computational Analysis of Sound Scenes and Events , pp. 373 -- 397 . Springer , Cham ( 2018 ) bchapter

    work page 2018

  7. [7]

    , Silva , C

    barticle Bello , J.P. , Silva , C. , Nov , O. , Dubois , R.L. , Arora , A. , Salamon , J. , Mydlarz , C. , Doraiswamy , H. : Sonyc: A system for monitoring, analyzing, and mitigating urban noise pollution . Communications of the ACM 62 , 68 -- 77 ( 2019 ) 10.1145/3224204 barticle

    doi:10.1145/3224204 2019

  8. [8]

    , Zhou , Y

    bchapter Baevski , A. , Zhou , Y. , Mohamed , A. , Auli , M. : wav2vec 2.0: A framework for self-supervised learning of speech representations . In: Advances in Neural Information Processing Systems , vol. 33 , pp. 12449 -- 12460 ( 2020 ) bchapter

    work page 2020

Show all 54 references
  1. [9]

    , Pentl , A

    barticle Clarkson , B. , Pentl , A. , Sawhney , N. : Auditory context awareness via wearable computing . Energy 400 , 20 ( 1998 ) barticle

    1998

  2. [10]

    , Song , Y

    bchapter Cai , P. , Song , Y. , Li , K. , Song , H. , McLoughlin , I. : MAT-SED : A masked audio transformer with masked-reconstruction based pre-training for sound event detection . In: Proceedings of Interspeech 2024 , pp. 557 -- 561 ( 2024 ). 10.21437/Interspeech.2024-714 bchapter

    2024 doi

  3. [11]

    , Wang , C

    barticle Chen , S. , Wang , C. , Chen , Z. , Wu , Y. , Liu , S. , Chen , Z. , Li , J. , Kanda , N. , Yoshioka , T. , Xiao , X. , Wu , J. , Zhou , L. , Ren , S. , Qian , Y. , Qian , Y. , Wu , J. , Zeng , M. , Yu , X. , Wei , F. : WavLM : Large-scale self-supervised pre-training...

    2022

  4. [12]

    , Wu , Y

    botherref Chen , S. , Wu , Y. , Wang , C. , Liu , S. , Tompkins , D. , Chen , Z. , Che , W. , Yu , X. , Wei , F. : BEATs : Audio pre-training with acoustic tokenizers. arXiv preprint arXiv:2212.09058 (2022) botherref

    2022 arXiv

  5. [13]

    https://dcase.community/challenge2023/task-sound-event-detection-with-weak-labels-and-synthetic-soundscapes

    botherref DCASE Community : DCASE Event Detection with Weak Soundscapes . https://dcase.community/challenge2023/task-sound-event-detection-with-weak-labels-and-synthetic-soundscapes. Accessed 12 Oct 2024 (2023) botherref

    2023

  6. [14]

    , Imoto , K

    botherref Dohi , K. , Imoto , K. , Harada , N. , Niizumi , D. , Koizumi , Y. , Nishida , T. , Purohit , H. , Tanabe , R. , Endo , T. , Kawaguchi , Y. : Description and discussion on DCASE 2023 Challenge Task 2 : First-shot unsupervised anomalous sound detection for machine con...

    2023 arXiv

  7. [15]

    , Merentitis , A

    barticle Debes , C. , Merentitis , A. , Sukhanov , S. , Niessen , M. , Frangiadakis , N. , Bauer , A. : Monitoring activities of daily living in smart homes: Understanding human behavior . IEEE Signal Processing Magazine 33 , 81 -- 94 ( 2016 ) 10.1109/MSP.2015.2503881 barticle

    2016

  8. [16]

    , Yan , Z

    bchapter Dinkel , H. , Yan , Z. , Wang , Y. , Zhang , J. , Wang , Y. , Wang , B. : Scaling up masked audio encoder learning for general audio classification . In: Proceedings of Interspeech 2024 , pp. 547 -- 551 ( 2024 ). 10.21437/Interspeech.2024-246 bchapter

    2024 doi

  9. [17]

    , Germain , F.G

    bchapter Ebbers , J. , Germain , F.G. , Wichern , G. , Le Roux , J. : Sound event bounding boxes . In: Proceedings of Interspeech 2024 , pp. 562 -- 566 ( 2024 ). 10.21437/Interspeech.2024-2075 bchapter

    2024 doi

  10. [18]

    , Haeb-Umbach , R

    botherref Ebbers , J. , Haeb-Umbach , R. : Pre-training and self-training for sound event detection in domestic environments. Technical report, DCASE2022 Challenge (2022) botherref

    2022

  11. [19]

    , Chung , Y.-A

    bchapter Gong , Y. , Chung , Y.-A. , Glass , J. : AST : Audio spectrogram transformer . In: Proceedings of Interspeech 2021 , pp. 571 -- 575 ( 2021 ). 10.21437/Interspeech.2021-698 bchapter

    2021 doi

  12. [20]

    , Bolte , B

    barticle Hsu , W.-N. , Bolte , B. , Tsai , Y.-H.H. , Lakhotia , K. , Salakhutdinov , R. , Mohamed , A. : HuBERT : Self-supervised speech representation learning by masked prediction of hidden units . IEEE/ACM Transactions on Audio, Speech, and Language Processing 29 , 3451 -- ...

    2021

  13. [21]

    , Chen , X

    bchapter He , K. , Chen , X. , Xie , S. , Li , Y. , Doll \'a r , P. , Girshick , R. : Masked autoencoders are scalable vision learners . In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pp. 16000 -- 16009 ( 2022 ). 10.1109/CVPR52688.2022....

    2022

  14. [22]

    , Cao , Y

    barticle Kong , Q. , Cao , Y. , Iqbal , T. , Wang , Y. , Wang , W. , Plumbley , M.D. : PANNs : Large-scale pretrained audio neural networks for audio pattern recognition . IEEE/ACM Transactions on Audio, Speech, and Language Processing 28 , 2880 -- 2894 ( 2020 ) 10.1109/TASLP....

    2020

  15. [23]

    , Das , R.K

    barticle Khandelwal , T. , Das , R.K. , Chng , E.S. : Sound event detection: A journey through dcase challenge series . APSIPA Transactions on Signal and Information Processing 13 , 51 ( 2024 ) 10.1561/116.00000051 barticle

    2024 doi

  16. [24]

    , Son , S.W

    botherref Kim , J.W. , Son , S.W. , Song , Y. , Kim , H.K. , Song , I.H. , Lim , J.E. : Semi-supervised learning-based sound event detection using frequency dynamic convolution with large kernel attention for DCASE Challenge 2023 Task 4 . arXiv preprint arXiv:2306.06461 (2023)...

    2023 arXiv

  17. [25]

    , Bottou , L

    barticle LeCun , Y. , Bottou , L. , Bengio , Y. , Haffner , P. : Gradient-based learning applied to document recognition . Proceedings of the IEEE 86 , 2278 -- 2324 ( 1998 ) 10.1109/5.726791 barticle

    1998 doi

  18. [26]

    , Han , B

    botherref Lv , Z. , Han , B. , Chen , Z. , Qian , Y. , Ding , J. , Liu , J. : Unsupervised anomalous detection based on unsupervised pretrained models. Technical report, DCASE2023 Challenge (2023) botherref

    2023

  19. [27]

    , Lewis , P.A.W

    barticle Lawrance , A.J. , Lewis , P.A.W. : An exponential moving-average sequence and point process (ema1) . Journal of Applied Probability 14 , 98 -- 113 ( 1977 ) 10.2307/3213263 barticle

    1977 doi

  20. [28]

    , Long , Y

    barticle Liang , Y. , Long , Y. , Li , Y. , Liang , J. , Wang , Y. : Joint framework with deep feature distillation and adaptive focal loss for weakly supervised audio tagging and acoustic event detection . Digital Signal Processing 123 , 103446 ( 2022 ) 10.1016/j.dsp.2022.103...

    2022

  21. [29]

    , Shao , N

    barticle Li , X. , Shao , N. , Li , X. : Self-supervised audio teacher-student transformer for both clip-level and frame-level tasks . IEEE/ACM Transactions on Audio, Speech, and Language Processing 32 , 1336 -- 1351 ( 2024 ) 10.1109/TASLP.2024.3352248 barticle

    2024

  22. [30]

    , Song , L

    botherref Liu , Q. , Song , L. , Xu , D. , Long , Y. : ICSD : An open-source dataset for infant cry and snoring detection. arXiv preprint arXiv:2408.10561 (2024) botherref

    2024 arXiv

  23. [31]

    , Kim , S.-H

    bchapter Nam , H. , Kim , S.-H. , Ko , B.-Y. , Park , Y.-H. : Frequency dynamic convolution: Frequency-adaptive pattern recognition for sound event detection . In: Proceedings of Interspeech 2022 , pp. 2763 -- 2767 ( 2022 ). 10.21437/Interspeech.2022-10127 bchapter

    2022 doi

  24. [32]

    , Baevski , A

    bchapter Schneider , S. , Baevski , A. , Collobert , R. , Auli , M. : wav2vec: Unsupervised pre-training for speech recognition . In: Proceedings of Interspeech 2019 , pp. 3465 -- 3469 ( 2019 ). 10.21437/Interspeech.2019-1873 bchapter

    2019 doi

  25. [33]

    , Bai , X

    barticle Shi , B. , Bai , X. , Yao , C. : An end-to-end trainable neural network for image-based sequence recognition and its application to scene text recognition . IEEE Transactions on Pattern Analysis and Machine Intelligence 39 , 2298 -- 2304 ( 2016 ) 10.1109/TPAMI.2016.26...

    2016

  26. [34]

    , Dash , D.D

    bchapter Singh , U. , Dash , D.D. , Sharma , M. , Mishra , S. , Malarvizhi , S. , Tiwari , S. , Shankarappa , R.T. : Polyphonic sound event detection and classification using convolutional recurrent neural network with mean teacher . In: 2021 12th International Conference on C...

    2021

  27. [35]

    , Li , X

    bchapter Shao , N. , Li , X. , Li , X. : Fine-tune the pretrained atst model for sound event detection . In: ICASSP 2024 - IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , pp. 911 -- 915 ( 2024 ). 10.1109/ICASSP48485.2024.10446159 bchapter

    2024

  28. [36]

    , Shier , J

    bchapter Turian , J. , Shier , J. , Khan , H.R. , Raj , B. , Schuller , B.W. , Steinmetz , C.J. , Malloy , C. , Tzanetakis , G. , Velarde , G. , McNally , K. , : HEAR : Holistic evaluation of audio representations . In: NeurIPS 2021 Competitions and Demonstrations Track , pp. ...

    2021

  29. [37]

    , Serizel , R

    bchapter Turpault , N. , Serizel , R. , Shah , A.P. , Salamon , J. : Sound event detection in domestic environments with weakly labeled data and soundscape synthesis . In: Workshop on Detection and Classification of Acoustic Scenes and Events (DCASE) , pp. 253 -- 257 ( 2019 )....

    2019 doi

  30. [38]

    , Valpola , H

    bchapter Tarvainen , A. , Valpola , H. : Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results . In: Advances in Neural Information Processing Systems , vol. 30 ( 2017 ) bchapter

    2017

  31. [39]

    , Li , Y

    barticle Oord , A. , Li , Y. , Vinyals , O. : Representation learning with contrastive predictive coding . arXiv preprint arXiv:1807.03748 ( 2018 ) 10.48550/arXiv.1807.03748 barticle

    2018 doi arXiv

  32. [40]

    , Shazeer , N

    bchapter Vaswani , A. , Shazeer , N. , Parmar , N. , Uszkoreit , J. , Jones , L. , Gomez , A.N. , Kaiser , . , Polosukhin , I. : Attention is all you need . In: Advances in Neural Information Processing Systems , vol. 30 , pp. 5998 -- 6008 ( 2017 ) bchapter

    2017

  33. [41]

    , Zheng , H

    bchapter Wang , Y. , Zheng , H. , Sun , Q. , Ma , Y. , Zhu , S. , Zhang , L. , Zhang , W.-Q. : Cross-lingual alzheimer's disease detection based on scale criteria . In: 2024 IEEE 14th International Symposium on Chinese Spoken Language Processing (ISCSLP) , pp. 491 -- 495 ( 202...

    2024

  34. [42]

    , Ding , K

    barticle Zhao , D. , Ding , K. , Qi , X. , Chen , Y. , Feng , H. : Sound event localization and detection based on deep learning . Journal of Systems Engineering and Electronics 35 , 294 -- 301 ( 2024 ) 10.23919/JSEE.2023.000110 barticle

    2024

  35. [43]

    , Song , Y

    bchapter Zheng , X. , Song , Y. , Dai , L.-R. , McLoughlin , I. , Liu , L. : An effective mutual mean teaching based domain adaptation method for sound event detection . In: Proceedings of Interspeech 2021 , pp. 556 -- 560 ( 2021 ). 10.21437/Interspeech.2021-281 bchapter

    2021 doi

  36. [44]

    , Zhang , R

    botherref Zheng , Y. , Zhang , R. , Atito , S. , Yang , S. , Wang , W. , Mei , Y. : ASiT-CRNN : A method for sound event detection with fine-tuning of self-supervised pre-trained asit-based model. Digital Signal Processing, 105055 (2025) 10.1016/j.dsp.2025.105055 botherref

    2025

  37. [45]

    sn-basic.bst

    FUNCTION identify.basic.version "sn-basic.bst" " [2024/07/19 v1.1 bibliography style]" * top ENTRY address archive author booktitle chapter doi edition editor eid eprint howpublished institution journal key keywords month note number organization pages publisher school series ...

    2024

  38. [46]

    write newline

    " write newline "" before.all 'output.state := FUNCTION add.period duplicate empty 'skip "." * add.blank if FUNCTION if.digit duplicate "0" = swap duplicate "1" = swap duplicate "2" = swap duplicate "3" = swap duplicate "4" = swap duplicate "5" = swap duplicate "6" = swap dupl...

  39. [47]

    write newline

    " write newline "" before.all 'output.state := FUNCTION output.doi doi empty skip "doi:" doi * "" * output if FUNCTION format.archive archivePrefix empty "" archivePrefix ":" * if FUNCTION format.primaryClass primaryClass empty "" " [" primaryClass * "] " * if FUNCTION format....

  40. [48]

    write newline

    " write newline "" before.all 'output.state := FUNCTION string.to.integer 't := t text.length 'k := #1 'char.num := t char.num #1 substring 's := s is.num s "." = or char.num k = not and char.num #1 + 'char.num := while char.num #1 - 'char.num := t #1 char.num substring FUNCTI...

  41. [49]

    write newline

    " write newline "" before.all 'output.state := FUNCTION string.to.integer 't := t text.length 'k := #1 'char.num := t char.num #1 substring 's := s is.num s "." = or char.num k = not and char.num #1 + 'char.num := while char.num #1 - 'char.num := t #1 char.num substring FUNCTI...

  42. [50]

    sn-nature.bst

    FUNCTION identify.nature.version "sn-nature.bst" " [2024/07/19 v1.1 bibliography style]" * top ENTRY address archive author booktitle chapter edition editor eprint howpublished institution journal key keywords month note number organization pages publisher school series title ...

    2024

  43. [51]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

  44. [52]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

  45. [53]

    sn-vancouver-num.bst

    FUNCTION identify.vancouver.version "sn-vancouver-num.bst" " [2024/07/19 v1.1 Vancouver bibliography style]" * top ENTRY address assignee author booktitle chapter cartographer day edition editor howpublished institution inventor journal key keywords month note number organizat...

    2024

  46. [54]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.