Pith. sign in

REVIEW 3 major objections 6 minor 72 references

Training Set Synthesis for Bioacoustic Denoising: A Case Study With Mice

T0 review · 3 major / 6 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read A supervised denoiser for bioacoustic recordings can be trained without clean audio by synthesizing clean targets from automatically tracked frequency ridges, and on mouse ultrasonic vocalizations this improves denoising, ridge tracking…

desk verdict Ridge-based training-set synthesis is a sound, useful bioacoustic denoising trick, but the SI-SDR and ridge-tracking headlines are partly self-referential; trust the modest classification gains. read the letter →

arxiv 2608.10054 v1 pith:LFL7CUGG submitted 2026-08-10 cs.SD eess.AS

Reyhaneh Abbasi , Peter Balazs , Vincent Lostanlen , Clara Hollomey , Dustin J. Penn , Sarah M. Zala , Nicki Holighaus This is my paper · ORCID
classification cs.SDeess.AS
keywords bioacousticdenoisingtrainingsetsynthesisridgetrackingultrasonicvocalizationscomplexratiomaskU-Netridge-guidedlossmouseUSVs
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 aims to solve a chicken-and-egg problem in bioacoustics: supervised denoising needs clean training audio, but clean audio is exactly what field recordings of weakly vocalizing animals do not have. The authors propose to synthesize the clean training signal from the noisy recording itself, using automatically tracked ridges—the time-frequency contours of the fundamental frequency and its first harmonic—as a proxy for the true vocalization. A Gaussian mask around those ridges selects the signal-dominated regions of a reassigned spectrogram, and the masked representation is inverted into audio that serves as the training target. A U-Net is then trained to predict a complex ratio mask from noisy spectrograms, with a ridge-guided loss that up-weights ridge regions and penalizes residual noise elsewhere. In mouse ultrasonic vocalizations this pipeline improves signal-to-distortion ratio, reduces fundamental-ridge tracking error on field recordings from 1.4 ± 3.4 kHz to 0.8 ± 1.5 kHz, and raises syllable classification F1 on denoised-and-retrained data.

What carries the argument

The load-bearing object is the ridge-guided synthesis-from-reassignment pipeline. A ridge is a frequency contour—a curve of local maxima in a time-frequency representation—that here tracks the fundamental frequency and the first harmonic partial of each mouse ultrasonic vocalization. The pipeline produces pseudo-clean training audio by masking the reassigned spectrogram with a truncated Gaussian centered on those ridges and inverting the masked representation through convolution with the Hann-window kernel and fast Griffin-Lim phase retrieval, and it reuses the same ridges to build weight matrices for the loss, emphasizing ridge-adjacent time-frequency bins and penalizing energy elsewhere. The denoiser itself is a U-Net predicting a complex ratio mask—a time-frequency mask applied to the complex spectrogram so that both magnitude and phase are corrected. The argument is carried by the tandem use of ridges as both target generator and loss guide: that is what lets a supervised denoiser be trained from noisy field recordings alone.

What would settle it

Take a synthetic vocalization with a known clean waveform and known ridge and harmonic structure, add real field noise, run the full pipeline—track, mask, synthesize target, train, denoise—and measure whether the denoised output recovers spectral energy that lies outside the 5-bin Gaussian mask but is part of the true synthetic signal; systematic loss of that energy would show the targets are teaching the network to delete genuine signal content.

Watch

Extended reading notes

Core claim

The paper's central claim is that imperfect, automatically tracked ridges can replace clean recordings as supervised training targets for bioacoustic denoising. For each noisy mouse ultrasonic vocalization, the authors track the fundamental-frequency ridge and the first harmonic partial using multitaper reassigned spectrograms, build a truncated Gaussian mask around both ridges (window length L=5 bins), apply this mask to the reassigned spectrogram, and invert the masked representation—via convolution with the analysis window and fast Griffin-Lim phase retrieval—into a synthetic 'clean' signal. That synthetic signal, paired with noise added back at controlled SNRs, trains a U-Net to predict a complex ratio mask; a ridge-guided loss up-weights the ridge neighborhoods and adds a penalty on non-ridge energy. The paper reports that this trained denoiser raises mean SI-SDR of synthetic test signals from -10.17 dB to 16.52 dB at the lowest input SNR, cuts fundamental-ridge frequency deviation on field recordings from 1.4 ± 3.4 kHz to 0.8 ± 1.5 kHz, and, when classifiers are retrained on denoised data, raises macro-F1 from 83% to 89% on in-sample wild-mouse data and from 69% to 72% on out-of-sample noisy wild-mouse data.

Load-bearing premise

The load-bearing premise is that the automatically tracked ridges on noisy recordings are accurate enough that a clean training signal reconstructed from a narrow Gaussian mask around those ridges contains the true vocalization; if a ridge is wrong or a harmonic is missed, the target excludes real signal energy and the denoiser learns to erase it.

Editorial extensions

If this is right

  • Any tonal bioacoustic signal with trackable ridges—marmoset calls, dolphin whistles, bat calls—could be denoised with this recipe, since the ridge tracker is swappable and no clean recordings are required.
  • Retraining downstream classifiers on denoised audio yields consistent gains: macro-F1 rises from 83% to 89% on in-sample wild-mouse USVs, from 69% to 72% on out-of-sample noisy wild-mouse USVs, and from 57% to 61% across a domesticated mouse strain.
  • Ridge tracking itself improves after denoising: fundamental-ridge frequency deviation falls from 1.4 ± 3.4 kHz to 0.8 ± 1.5 kHz, and harmonic-partial recall rises from 58% to 83% relative to the previous signal-processing pipeline.
  • The ridge-guided loss is doing real work: removing it keeps fundamental-ridge precision at 95–96% but raises frequency deviation and lowers harmonic recall, so the weighting is what preserves faint components.
  • Because the synthesis targets are analytic—built from explicit amplitude, instantaneous frequency, and phase parameters—the training data are interpretable and controllable, unlike targets coming from a pretrained denoiser.

Reading between the lines

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

  • Because the synthetic test set is generated by the same ridge-mask-inversion pipeline that creates the training targets, the reported SI-SDR numbers measure self-consistency of the synthesis as much as true recovery; an independent check with real clean recordings of a captive or laboratory animal would separate the two.
  • The narrow 5-bin Gaussian mask and the restriction to fundamental plus first harmonic mean that any genuine signal energy outside those contours is omitted from the targets; species with rich harmonic stacks or strong nonlinear frequency modulations would be a natural stress test for the approach.
  • The paper filters training to the 83.7% of USVs whose ridges are judged correct, so an iterative loop—denoise, retrack, re-synthesize, retrain—is a natural extension that could recover some of the 13.7% incorrect-ridge cases.
  • Preserving multiple ridge trajectories in the output hints at a source-separation use: overlapping calls from two mice, currently excluded as invalid samples, might be disentangled by treating each ridge as a separate synthesis target.
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 proposes a supervised U-Net denoiser for bioacoustic recordings in settings where no clean training data are available. The authors synthesize clean training targets from noisy mouse USV recordings by tracking fundamental and first-harmonic ridges with a multicomponent ridge-tracking method, generating a Gaussian mask around the ridges, and inverting the masked reassigned spectrogram. The same ridge information is used to define a ridge-guided loss that weights the training objective toward vocalization regions. The denoiser predicts a complex ratio mask and is evaluated in three ways: SI-SDR on a synthetic test set, fundamental and harmonic ridge-tracking accuracy on real field recordings with manually refined annotations, and USV classification macro-F1 using a BootSnap classifier retrained on denoised data from wild and domesticated mice. The paper reports large SI-SDR gains (e.g., from −10.17 dB to 16.52 dB at the lowest input SNR), reduced fundamental ridge-tracking frequency deviation (from 1.4 ± 3.4 kHz to 0.8 ± 1.5 kHz), and improved macro-F1 scores (e.g., from 83% to 89% on in-sample wild data and from 57% to 61% on out-of-sample domesticated data).

Significance. If the reported results hold, the paper addresses a genuine practical problem in bioacoustics: training deep denoisers without clean references. The idea of using automatically tracked ridges both for target synthesis and for loss weighting is novel and, because the method is not tied to one ridge tracker, it is potentially transferable to other tonal bioacoustic signals. The authors also provide publicly available code, a GUI, and a manually annotated ridge dataset, which is a concrete contribution. However, the evaluation is substantially entangled with the ridge-tracking pipeline used to build the training targets, and the most independent evidence (classification) is modest and lacks uncertainty quantification. The significance of the central claim therefore depends on whether the authors can add evaluations that break the self-referential loop, especially for SI-SDR and ridge tracking.

major comments (3)
  1. [Section IV-C, 'Scale-Invariant Signal-to-Distortion Ratio (SI–SDR)'] The synthetic test set for SI-SDR is generated using the same ridge-tracking and masked-reassignment inversion procedure (Sections III-C and III-D) that is used to synthesize the training targets. Consequently, a U-Net trained to approximate that target generator can achieve large SI-SDR gains on this test set without recovering the true clean signal; the 'clean' reference is itself an output of the training-target pipeline. The authors acknowledge this when they write that 'SI-SDR results should be interpreted as a controlled, relative comparison under synthetic conditions,' but the abstract and conclusion still cite the SI-SDR numbers as headline evidence of denoising quality. To support the central claim, the authors should either include an independent synthetic or semi-synthetic test set (e.g., clean signals synthesized from independently validated ridges, or real clean recordings with added noise) or explicitly remove SI-SDR from the list of primary evidence and base the main claims on task-based metrics.
  2. [Section IV-C, 'Accuracy of Tracked Ridges'] The manual ridge annotations used as ground truth were 'initially estimated with [16] and subsequently manually refined post hoc,' and [16] is the same multicomponent tracker that is used for training-set synthesis and for evaluating the denoised outputs. Since the network is trained to approximate targets derived from that tracker, improved agreement with these annotations may partly reflect the denoiser conforming to the tracker's own bias rather than to independent ground truth. The manual refinement reduces but does not eliminate this concern. The authors should report the level of agreement between the raw [16] tracks and the refined annotations, or provide an evaluation against independently produced annotations or synthetic ground truth, before the ridge-tracking improvements can be taken at face value. The statement in Section V-D that this evaluation is based on only 193 USVs further limits the strength of the result.
  3. [Section V-C and Section IV-C final paragraph] The classification results are the least entangled evidence in the paper, but they are presented in a way that does not fully support the claim that the denoiser improves classification. As the manuscript itself states, the experiments evaluate the integrated pipeline of denoising plus classifier retraining, and a cross-condition test (e.g., a classifier trained on original audio and evaluated on denoised audio, or vice versa) would be needed to attribute gains to the denoiser alone. Moreover, no confidence intervals, seed variance, or significance tests are reported for the macro-F1 differences in Tables IV and V (89% vs 83%, 72% vs 69%, 61% vs 57%), and the out-of-sample improvements are small. The authors should provide uncertainty estimates and, if they wish to claim a denoising benefit for classification, run the cross-condition experiment they identify as beyond the scope of the present study.
minor comments (6)
  1. [Section VI] The conclusion contains malformed numeric expressions: '1.4 ±0.5 3.5 kHz' and '0.8 ±0.5 1.5 kHz' should be '1.4 ± 3.5 kHz' and '0.8 ± 1.5 kHz', respectively.
  2. [Table III] The table entries '9766 0.8±0.3' and '93830.9±0.2' appear to be missing separators; they should presumably read '97 66' and '93 83' for precision and recall.
  3. [Section IV-C, Biodenoising baseline] The description of the Biodenoising evaluation ('we rewrote recordings at 16 kHz without waveform resampling') is technically unclear and should be explained more precisely, because the sample-rate change affects the mapping of ultrasonic content onto the model's input range.
  4. [Figure 3] The SI-SDR comparison shows error bars but no statistical tests across methods or SNR levels; adding paired significance tests or confidence intervals would strengthen the comparison against baselines.
  5. [Abstract] The abstract says the method enhances ridge tracking 'compared to our previous signal-processing approach,' but the experiments compare against several baselines (DeepSqueak, noisereduce, Biodenoising, and the analytical denoiser); the wording should reflect the full comparison.
  6. [Section III-A] The text uses both 'first harmonic partial' and 'harmonic partial' inconsistently; the authors should define whether the first harmonic is the second frequency component (2f0) or the fundamental itself, and use consistent terminology throughout.

Circularity Check

1 steps flagged · score 4.0 of 10

The headline SI-SDR gain is measured against references synthesized by the very same ridge-mask inversion that defines the training targets, so that metric is partly self-referential; the classification evidence is more independent but modest.

  1. self definitional [Section IV-C, 'Scale-Invariant Signal-to-Distortion Ratio (SI–SDR)' and Sections III-C–III-D]
    "Because original recordings lack clean reference signals, we synthesize a test set to compute improvements in the SI-SDR. To generate this test set, we use ridges from the testing subset of the correct-ridge data described in Table I. We then follow the same procedure as described in Section III-D for the synthesis of the testing set ~s. ... As this test set is generated using a related synthesis procedure, SI-SDR results should be interpreted as a controlled, relative comparison under synthetic conditions."

    The SI-SDR clean reference ~s is produced by the same ridge-tracking plus Gaussian-mask inversion (Sections III-B to III-D) that defines the U-Net's training target ~s. The network is trained to map ~x = ~s + n back to ~s, so a test set whose references come from the same generator measures how well the network has learned to invert its own target-generation procedure, not recovery of the true clean signal s. The conclusion nevertheless presents the SI-SDR improvement (e.g., from -10.17 dB to 16.52 dB) as headline evidence of denoising quality. The paper's caveat that this is a 'controlled, relative comparison' is honest but does not remove the self-referential loop when the number is reported as a main result.

full rationale

The core derivation is not circular in itself: the paper explicitly constructs approximate clean targets by tracking ridges and inverting a masked reassigned spectrogram, then trains a U-Net to predict a complex ratio mask from those targets. That is a legitimate supervised synthesis strategy. The main circularity is confined to the SI-SDR evaluation: the test reference signals are generated by the same ridge-mask inversion pipeline as the training targets, so the reported SI-SDR gains largely reflect the network learning to invert its own target generator. The paper acknowledges this by calling the result a 'controlled, relative comparison,' but the conclusion still headlines the SI-SDR numbers. The ridge-tracking evaluation is a weaker related concern: the manual annotations were initially estimated with [16], the same tracker used for synthesis and also listed as a baseline, although post-hoc manual refinement provides some independence. The classification experiments are the most independent evidence, since they use manual labels and include a genuinely out-of-sample domesticated-mouse dataset, but the gains are modest (macro-F1 57-to-61 and 69-to-72) and no error bars or seed variance are reported. The self-citations to [16] and [64] are not load-bearing circularity: they are concrete published algorithms used as components, not invoked as proof of correctness. Overall, there is one clearly self-referential evaluation loop plus a partially shared-tracker benchmark, so a score of 4 is appropriate rather than 0 or 6.

Assumptions & free parameters 11 free parameters · 5 assumptions · 0 invented entities

The method rests on the assumption that ridges tracked on noisy recordings are good enough to synthesize clean targets; if the tracker fails or the 5-bin Gaussian mask excludes signal energy, the U-Net learns to erase real vocalization content. Many hand-set constants (L, B, C, alpha, gamma, alpha_noise, taper count, SNR probabilities) shape the training distribution and are not independently motivated. No new entities are introduced.

free parameters (11)
  • L (mask window length) = 5 frequency bins
    Manual choice; controls how much spectral energy around ridges is treated as signal in synthesis (Section III-C, IV-B1).
  • sigma = L/4 = 1.25 bins
    Set by L; Gaussian mask spread in Eq. (1).
  • B, C (mask compression) = B=15, C=0.1
    Manual constants in Eq. (4); bound dynamic range of complex ratio mask targets.
  • alpha, gamma (ridge loss weights) = alpha=0.3, gamma=10
    Manual constants in Eq. (5); floor and scaling for ridge-region weight.
  • alpha_noise (noise penalty weight) = 0.8
    Set 'based on pilot tasks' (Section III-F); a tuned hyperparameter.
  • b (ridge weight band) = not specified
    Introduced in Eq. (5) as +/- b bins but no numeric value is given in the text; missing implementation detail.
  • SNR bin probabilities = 0.30, 0.25, 0.20, 0.15, 0.10
    Design choice biasing training toward low SNR (Section IV-B2).
  • USV segment extension = +/-3 ms
    Empirically selected after testing +/-0.5, +/-1.5, +/-3, +/-5 ms (Section IV-A).
  • tfridge penalty = 2e-8
    Hand-set constant for ridge path smoothness (Section IV-B1).
  • Number of tapers = 6
    Chosen to balance noise suppression and spectral concentration (Section IV-B1).
  • Harmonic search band = +/-3 kHz around 2*f0
    Hand-set to account for estimation errors (Section IV-B1).
assumptions (5)
  • domain assumption Additive noise model x = s + n with speech-like contamination
    Section III states x=s+n; ignores reverberation, overlapping conspecifics, and non-stationary noise correlations.
  • domain assumption Mouse USVs are locally sinusoidal and well represented by fundamental plus first harmonic partial ridges
    Section III-A/IV-B1; the synthesis uses only these two components. Higher harmonics and broadband components are excluded by scope.
  • domain assumption Ridges tracked by [16] are moderately accurate and 'correct'/'incorrect' manual categorization is reliable
    Section III and IV-A; training targets and loss weighting inherit tracker and annotation errors.
  • domain assumption Manual ridge annotations in the 193-USV evaluation set are objective ground truth
    Section IV-C; these are post hoc refinements of [16]'s estimates, so they may share systematic bias with the tracker.
  • domain assumption Inversion of the masked reassigned spectrogram [54] yields an approximate clean signal that preserves vocalization content
    Section III-D; phase retrieval and narrow Gaussian masks can distort or discard real signal energy.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Training Set Synthesis for Bioacoustic Denoising: A Case Study With Mice." pith.science (2026). https://pith.science/paper/LFL7CUGG

@misc{pith2026260810054,
  author       = {Pith},
  title        = {Pith review of: Training Set Synthesis for Bioacoustic Denoising: A Case Study With Mice},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LFL7CUGG}},
  note         = {Machine review of arXiv:2608.10054}
}
read the original abstract

Bioacoustic recordings are often degraded by ambient noise, which complicates the analysis of weak or noise-overlapped vocalizations. Convolutional neural networks, particularly U-Net architectures, have shown a strong denoising performance in speech and music processing. However, their direct application to bioacoustic signals is limited by the scarcity of clean training data. To address this issue, we propose a training set synthesis approach and develop a supervised denoising model that predicts a complex ratio mask in the time-frequency domain. The model leverages ridges, or frequency contours, that represent the fundamental frequency together with one or more harmonic partial components of vocalizations. These ridges are used both for the synthesis of training sets and to design a loss function that assigns higher weights to the ridge regions (ridge-guided loss function). This weighting step helps the network better preserve vocalization details during denoising. As a case study, we evaluate our approach using ultrasonic vocalizations (USVs) recordings of house mice, which are widely studied in behavioral biology and neuroscience. In actual field recordings, the proposed method enhances fundamental and harmonic partial ridge tracking compared to our previous signal-processing approach. In addition, a classifier trained on denoised data improves USV classification on out-of-sample, noisy recordings from wild and domesticated mice compared to classifiers trained on noisy recordings. Our proposed method also substantially improves the scale-invariant signal-to-distortion ratio on synthetic testing data across a wide range of input signal-to-noise ratios. Although we focus on USVs, the proposed approach should be broadly applicable to other bioacoustic signals with trackable ridges, and thus enables ridgebased training set synthesis and denoising.

Figures

Figures reproduced from arXiv: 2608.10054 by the authors.

Figure 1
Figure 1. The noisy inputs x˜ are generated by adding noise n to the synthetic data s˜ (Section IV-B2), i.e., x˜ = ˜s + n. A denoiser network is then trained with the noisy input x˜ and its corresponding clean target s˜. The network predicts complex ratio masks (cRMs) [55] (Section III-F) from noisy spectrograms, which are subsequently applied to obtain the denoised signals. A ridge-guided loss is employed to emphasize vocali… view at source ↗
Figure 1
Figure 1. Flowchart of the proposed training set synthesis pipeline. The input noisy signal [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Flowchart of the proposed denoiser training pipeline. The input noisy signal [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figures from the paper (3 more)
Figure 3
Figure 3. Figure 3: SI-SDR (scale-invariant signal-to-distortion ratio) before [PITH_FULL_IMAGE:figures/full_fig_p010_3.png]
Figure 4
Figure 4. Figure 4: Spectrograms of two example USVs (top) with corre [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]
Figure 5
Figure 5. Figure 5: Spectrograms of the original (top) and the denoised (bottom) audio from four datasets: (a) training (wild mouse, correct [PITH_FULL_IMAGE:figures/full_fig_p012_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

72 extracted references · 65 canonical work pages

  1. [16]

    Robust multicomponent tracking of ultrasonic vocalizations,

    R. Abbasi, N. Holighaus, P. Balazs, V . Lostanlen, D. J. Penn, and S. M. Zala, “Robust multicomponent tracking of ultrasonic vocalizations,” ICASSP, vol. 126, no. 5, pp. 200–205, 2025

    work page 2025

  2. [1]

    P. R. Marler and H. Slabbekoorn,Nature’s music: the science of birdsong. Elsevier, 2004

    work page 2004

  3. [2]

    Schnitzler and E

    H.-U. Schnitzler and E. K. Kalko, “Echolocation by insect-eating bats: we define four distinct functional groups of bats and find differences in signal structure that correlate with the typical echolocation tasks faced by each group,”Bioscience, vol. 51, no. 7, pp. 557–569, 2001

    work page 2001

  4. [3]

    Types and functions of ultrasonic vocalizations in laboratory rats and mice,

    C. V . Portfors, “Types and functions of ultrasonic vocalizations in laboratory rats and mice,”Journal of the American Association for Laboratory Animal Science, vol. 46, no. 1, pp. 28–34, 2007

    work page 2007

  5. [4]

    J. W. Bradbury, S. L. Vehrencampet al.,Principles of animal communi- cation. Sinauer Associates Sunderland, MA, 1998, vol. 132

    work page 1998

  6. [5]

    Suppression of acoustic noise in speech using spectral subtrac- tion,

    S. Boll, “Suppression of acoustic noise in speech using spectral subtrac- tion,”IEEE Transactions on acoustics, speech, and signal processing, vol. 27, no. 2, pp. 113–120, 2003. 2https://www.clip.science

    work page 2003

  7. [6]

    Bioacoustic signal denoising: a review,

    J. Xie, J. G. Colonna, and J. Zhang, “Bioacoustic signal denoising: a review,”Artificial Intelligence Review, vol. 54, pp. 3575–3597, 2021

    work page 2021

  8. [7]

    Transfer learning for denoising the echolocation clicks of finless porpoise (neophocaena phocaenoides sunameri) using deep convolutional autoencoders,

    W. Yang, W. Chang, Z. Song, Y . Zhang, and X. Wang, “Transfer learning for denoising the echolocation clicks of finless porpoise (neophocaena phocaenoides sunameri) using deep convolutional autoencoders,”The Journal of the Acoustical Society of America, vol. 150, no. 2, pp. 1243– 1250, 2021

    work page 2021

Show all 72 references
  1. [8]

    Robust north atlantic right whale detection using deep learning models for denoising,

    W. Vickers, B. Milner, D. Risch, and R. Lee, “Robust north atlantic right whale detection using deep learning models for denoising,”The Journal of the Acoustical Society of America, vol. 149, no. 6, pp. 3797–3812, 2021

    2021

  2. [9]

    Synthetic training set generation using text-to-audio models for environmental sound classification,

    F. Ronchini, L. Comanducci, and F. Antonacci, “Synthetic training set generation using text-to-audio models for environmental sound classification,” inProceedings of the Detection and Classification of Acoustic Scenes and Events (DCASE) Workshop, Tokyo, Japan, 2024

    2024

  3. [10]

    Simulated soundscapes and transfer learning boost the performance of acoustic classifiers under data scarcity,

    M. J. Weldy, D. B. Lesmeister, T. Denton, A. Duarte, B. J. Vernasco, A. Gasc, J. C. Rowe, M. J. Adams, and M. G. Betts, “Simulated soundscapes and transfer learning boost the performance of acoustic classifiers under data scarcity,”Methods in Ecology and Evolution, 2025

    2025

  4. [11]

    Age-group estimation in free-ranging african elephants based on acoustic cues of low-frequency rumbles,

    A. S. Stoeger, M. Zeppelzauer, and A. Baotic, “Age-group estimation in free-ranging african elephants based on acoustic cues of low-frequency rumbles,”Bioacoustics, vol. 23, no. 3, pp. 231–246, 2014

    2014

  5. [12]

    Primed to vocalize: Wild-derived male house mice increase vocalization rate and diversity after a previous encounter with a female,

    S. M. Zala, D. Nicolakis, M. A. Marconi, A. Noll, T. Ruf, P. Balazs, and D. J. Penn, “Primed to vocalize: Wild-derived male house mice increase vocalization rate and diversity after a previous encounter with a female,” PloS one, vol. 15, no. 12, p. e0242959, 2020

    2020

  6. [13]

    Ultrasonic courtship vocalizations of male house mice contain distinct individual signatures,

    M. A. Marconi, D. Nicolakis, R. Abbasi, D. J. Penn, and S. M. Zala, “Ultrasonic courtship vocalizations of male house mice contain distinct individual signatures,”Animal Behaviour, vol. 169, pp. 169–197, 2020

    2020

  7. [14]

    Individual vocal recognition across taxa: a review of the literature and a look into the future,

    N. V . Carlson, E. M. Kelly, and I. Couzin, “Individual vocal recognition across taxa: a review of the literature and a look into the future,” Philosophical Transactions of the Royal Society B, vol. 375, no. 1802, p. 20190479, 2020

    2020

  8. [15]

    V ocal individual discrimination in japanese monkeys,

    M. Ceugniet and A. Izumi, “V ocal individual discrimination in japanese monkeys,”Primates, vol. 45, pp. 119–128, 2004

    2004

  9. [17]

    Courtship vocalizations of wild house mice show highly dynamic changes and correlate with male copulatory success,

    T. Klaus, B. Wernisch, S. M. Zala, and D. J. Penn, “Courtship vocalizations of wild house mice show highly dynamic changes and correlate with male copulatory success,”Animal Behaviour, vol. 220, p. 123024, 2025

    2025

  10. [18]

    Marmaudio: A large annotated dataset of vocalizations by common marmosets,

    C. Lamothe, M. Obliger, P. Best, R. Trapeau, S. Ravel, T. Arti `eres, R. Marxer, and P. Belin, “Marmaudio: A large annotated dataset of vocalizations by common marmosets,” 2025. [Online]. Available: https://zenodo.org/records/17191682

    2025

  11. [19]

    F0 estimation for bioacoustics: A benchmark/training dataset of non-human vocalisations with annotated frequency contours,

    P. Best, M. Araya-Salas, A. G. Ekstr ¨om, B. Freitas, F. H. Jensen, A. Kershenbaum, A. R. Lameira, K. D. S. Lehmann, P. Linhart, R. C. Liu, M. Madhavan, A. Markham, M. A. Roch, H. Root- Gutteridge, M. ˇS´alek, G. Smith-Vidaurre, A. Strandburg-Peshkin, M. R. Warren, M. Wijers, ...

    2025 doi

  12. [20]

    Male mice ultrasonic vocalizations enhance female sexual approach and hypothalamic kisspeptin neuron activity,

    A. Asaba, T. Osakada, K. Touhara, M. Kato, K. Mogi, and T. Kikusui, “Male mice ultrasonic vocalizations enhance female sexual approach and hypothalamic kisspeptin neuron activity,”Hormones and behavior, vol. 94, pp. 53–60, 2017

    2017

  13. [21]

    Ultrasonic vocalizations in house mice depend upon genetic relatedness of mating partners and correlate with subsequent reproductive success,

    D. Nicolakis, M. A. Marconi, S. M. Zala, and D. J. Penn, “Ultrasonic vocalizations in house mice depend upon genetic relatedness of mating partners and correlate with subsequent reproductive success,”Frontiers in zoology, vol. 17, pp. 1–19, 2020

    2020

  14. [22]

    Male mice song syntax depends on social contexts and influences female preferences,

    J. Chabout, A. Sarkar, D. B. Dunson, and E. D. Jarvis, “Male mice song syntax depends on social contexts and influences female preferences,” Frontiers in behavioral neuroscience, vol. 9, p. 76, 2015

    2015

  15. [23]

    Speech enhancement based on a priori signal to noise estimation,

    P. Scalartet al., “Speech enhancement based on a priori signal to noise estimation,” in1996 IEEE international conference on acoustics, speech, and signal processing conference proceedings, vol. 2. IEEE, 1996, pp. 629–632

    1996

  16. [24]

    Birdsong denoising using wavelets,

    N. Priyadarshani, S. Marsland, I. Castro, and A. Punchihewa, “Birdsong denoising using wavelets,”PloS one, vol. 11, no. 1, p. e0146790, 2016

    2016

  17. [25]

    Acoustic detection of elephant presence in noisy environments,

    M. Zeppelzauer, A. S. St ¨oger, and C. Breiteneder, “Acoustic detection of elephant presence in noisy environments,” inProceedings of the 2nd ACM international workshop on Multimedia analysis for ecological data, 2013, pp. 3–8

    2013

  18. [26]

    Noisereduce: Domain general noise reduction for time series signals,

    T. Sainburg and A. Zorea, “Noisereduce: Domain general noise reduction for time series signals,”arXiv preprint arXiv:2412.17851, 2024. 15

    2024 arXiv

  19. [27]

    Mu- pet—mouse ultrasonic profile extraction: a signal processing tool for rapid and unsupervised analysis of ultrasonic vocalizations,

    M. Van Segbroeck, A. T. Knoll, P. Levitt, and S. Narayanan, “Mu- pet—mouse ultrasonic profile extraction: a signal processing tool for rapid and unsupervised analysis of ultrasonic vocalizations,”Neuron, vol. 94, no. 3, pp. 465–485, 2017

    2017

  20. [28]

    Usvseg: A robust method for segmentation of ultrasonic vocalizations in rodents,

    R. O. Tachibana, K. Kanno, S. Okabe, K. I. Kobayasi, and K. Okanoya, “Usvseg: A robust method for segmentation of ultrasonic vocalizations in rodents,”PloS one, vol. 15, no. 2, p. e0228907, 2020

    2020

  21. [29]

    Improved wavelet denoising via empirical wiener filtering,

    S. P. Ghael, A. M. Sayeed, and R. G. Baraniuk, “Improved wavelet denoising via empirical wiener filtering,” inWavelet Applications in Signal and Image Processing V, vol. 3169. SPIE, 1997, pp. 389–399

    1997

  22. [30]

    Multitaper time-frequency reassignment for nonstationary spectrum estimation and chirp enhancement,

    J. Xiao and P. Flandrin, “Multitaper time-frequency reassignment for nonstationary spectrum estimation and chirp enhancement,”IEEE Transactions on Signal Processing, vol. 55, no. 6, pp. 2851–2860, 2007

    2007

  23. [31]

    Improving the readability of time-frequency and time-scale representations by the method of re-assigment,

    F. Auger and P. Flandrin, “Improving the readability of time-frequency and time-scale representations by the method of re-assigment,”IEEE Trans. Signal Process., vol. 43, no. 5, pp. 1068–1089, May 1995

    1995

  24. [32]

    Bioacoustic processing and analyses of mouse vocalizations: current methods and future directions,

    R. Abbasi, D. Nicolakis, M. A. Marconi, T. Klaus, B. Wernisch, M. Peng, P. Balazs, D. J. Penn, and S. M. Zala, “Bioacoustic processing and analyses of mouse vocalizations: current methods and future directions,” 2025, manuscript in preparation; to be submitted

    2025

  25. [33]

    Low-latency single channel speech enhancement using u-net convolutional neural networks,

    A. E. Bulut and K. Koishida, “Low-latency single channel speech enhancement using u-net convolutional neural networks,” inICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2020, pp. 6214–6218

    2020

  26. [34]

    Deep learning for denoising,

    S. Yu, J. Ma, and W. Wang, “Deep learning for denoising,”Geophysics, vol. 84, no. 6, pp. V333–V350, 2019

    2019

  27. [35]

    Automatic acoustic detection of birds through deep learning: the first bird audio detection challenge,

    D. Stowell, M. D. Wood, H. Pamuła, Y . Stylianou, and H. Glotin, “Automatic acoustic detection of birds through deep learning: the first bird audio detection challenge,”Methods in Ecology and Evolution, vol. 10, no. 3, pp. 368–380, 2019

    2019

  28. [36]

    Hybridmouse: a hybrid convolutional-recurrent neural network-based model for identification of mouse ultrasonic vocalizations,

    Y . Goussha, K. Bar, S. Netser, L. Cohen, Y . Hel-Or, and S. Wagner, “Hybridmouse: a hybrid convolutional-recurrent neural network-based model for identification of mouse ultrasonic vocalizations,”Frontiers in Behavioral Neuroscience, vol. 15, p. 810590, 2022

    2022

  29. [37]

    Automated bat call classification using deep convolutional neural networks,

    E. Schwab, S. Pogrebnoj, M. Freund, F. Flossmann, S. V ogl, and K.-H. Frommolt, “Automated bat call classification using deep convolutional neural networks,”Bioacoustics, vol. 32, no. 1, pp. 1–16, 2023

    2023

  30. [38]

    Birdvoxdetect: Large-scale detection and classification of flight calls for bird migration monitoring,

    V . Lostanlen, A. Cramer, J. Salamon, A. Farnsworth, B. M. Van Doren, S. Kelling, and J. P. Bello, “Birdvoxdetect: Large-scale detection and classification of flight calls for bird migration monitoring,”IEEE/ACM Transactions on Audio, Speech, and Language Processing, 2024

    2024

  31. [39]

    Automatic bioacoustics noise reduction method based on a deep feature loss network,

    C. Zhang, K. He, X. Gao, and Y . Guo, “Automatic bioacoustics noise reduction method based on a deep feature loss network,”Ecological Informatics, vol. 80, p. 102517, 2024

    2024

  32. [40]

    A deep autoencoder approach to bird call enhancement,

    R. Sinha and P. Rajan, “A deep autoencoder approach to bird call enhancement,” in2018 IEEE 13th International Conference on Industrial and Information Systems (ICIIS). IEEE, 2018, pp. 22–26

    2018

  33. [41]

    Real time speech enhancement in the waveform domain,

    A. Defossez, G. Synnaeve, and Y . Adi, “Real time speech enhancement in the waveform domain,”Interspeech, 2020

    2020

  34. [42]

    Singing voice separation with deep u-net convolutional networks,

    A. Jansson, E. Humphrey, N. Montecchio, R. Bittner, A. Kumar, and T. Weyde, “Singing voice separation with deep u-net convolutional networks,” 2017

    2017

  35. [43]

    Speech derever- beration using fully convolutional networks,

    O. Ernst, S. E. Chazan, S. Gannot, and J. Goldberger, “Speech derever- beration using fully convolutional networks,” in2018 26th European Signal Processing Conference (EUSIPCO). IEEE, 2018, pp. 390–394

    2018

  36. [44]

    U-net: Convolutional networks for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks for biomedical image segmentation,” inMedical image computing and computer-assisted intervention–MICCAI 2015: 18th international conference, Munich, Germany, October 5-9, 2015, proceedings, part III

    2015

  37. [45]

    Springer, 2015, pp. 234–241

    2015

  38. [46]

    Computational bioacoustics with deep learning: a review and roadmap,

    D. Stowell, “Computational bioacoustics with deep learning: a review and roadmap,”PeerJ, vol. 10, p. e13152, 2022

    2022

  39. [47]

    Biodenoising: animal vocalization denoising without access to clean data,

    M. Miron, S. Keen, J.-Y . Liu, B. Hoffman, M. Hagiwara, O. Pietquin, F. Effenberger, and M. Cusimano, “Biodenoising: animal vocalization denoising without access to clean data,” inICASSP 2025-2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP...

    2025

  40. [48]

    Synthetic data enables context-aware bioacoustic sound event detection,

    B. Hoffman, D. Robinson, M. Miron, V . Baglione, D. Canestrari, D. Elias, E. Trapote, and O. Pietquin, “Synthetic data enables context-aware bioacoustic sound event detection,”arXiv preprint arXiv:2503.00296, 2025

    2025 arXiv

  41. [49]

    Automatic recognition of harmonic bird sounds using a frequency track extraction algorithm,

    J. R. Heller and J. D. Pinezich, “Automatic recognition of harmonic bird sounds using a frequency track extraction algorithm,”The Journal of the Acoustical Society of America, vol. 124, no. 3, pp. 1830–1837, 2008

    2008

  42. [50]

    Using deep learning to track time × frequency whistle contours of toothed whales without human-annotated training data,

    P. Li, X. Liu, H. Klinck, P. Gruden, and M. A. Roch, “Using deep learning to track time × frequency whistle contours of toothed whales without human-annotated training data,”The Journal of the Acoustical Society of America, vol. 154, no. 1, pp. 502–517, 2023

    2023

  43. [51]

    Bioacoustic fundamental frequency estimation: a cross-species dataset and deep learning baseline,

    P. Best, M. Araya-Salas, A. G. Ekstr ¨om, B. Freitas, F. H. Jensen, A. Kershenbaum, A. R. Lameira, K. D. Lehmann, P. Linhart, R. C. Liu et al., “Bioacoustic fundamental frequency estimation: a cross-species dataset and deep learning baseline,”Bioacoustics, pp. 1–28, 2025

    2025

  44. [52]

    Automated extraction of dolphin whistles—a sequential monte carlo probability hypothesis density approach,

    P. Gruden and P. R. White, “Automated extraction of dolphin whistles—a sequential monte carlo probability hypothesis density approach,”The Journal of the Acoustical Society of America, vol. 148, no. 5, pp. 3014– 3026, 2020

    2020

  45. [53]

    Maturation of social-vocal communication in prairie vole (microtus ochrogaster) pups,

    M. R. Warren, D. Campbell, A. M. Borie, C. L. Ford IV , A. M. Dharani, L. J. Young, and R. C. Liu, “Maturation of social-vocal communication in prairie vole (microtus ochrogaster) pups,”Frontiers in Behavioral Neuroscience, vol. 15, p. 814200, 2022

    2022

  46. [54]

    Biocppnet: automatic bioacoustic source separation with deep neural networks,

    P. C. Bermant, “Biocppnet: automatic bioacoustic source separation with deep neural networks,”Scientific Reports, vol. 11, no. 1, p. 23502, 2021

    2021

  47. [55]

    Rajbamshi, P

    S. Rajbamshi, P. Balazs, and N. Holighaus,Adhoc method to invert the reassigned time-frequency representation. Proceedings of the 23rd International Congress on Acoustics, Aachen, 2019

    2019

  48. [56]

    Complex ratio masking for monaural speech separation,

    D. S. Williamson, Y . Wang, and D. Wang, “Complex ratio masking for monaural speech separation,”IEEE/ACM transactions on audio, speech, and language processing, vol. 24, no. 3, pp. 483–492, 2015

    2015

  49. [57]

    Short term spectral analysis, synthesis, and modification by discrete fourier transform,

    J. Allen, “Short term spectral analysis, synthesis, and modification by discrete fourier transform,”IEEE transactions on acoustics, speech, and signal processing, vol. 25, no. 3, pp. 235–238, 1977

    1977

  50. [58]

    Calculation of a constant q spectral transform,

    J. C. Brown, “Calculation of a constant q spectral transform,”The Journal of the Acoustical Society of America, vol. 89, no. 1, pp. 425–434, 1991

    1991

  51. [59]

    Uncertainty principles for time-frequency representations,

    K. Gr¨ochenig, “Uncertainty principles for time-frequency representations,” Advances in Gabor analysis, pp. 11–30, 2003

    2003

  52. [60]

    Reassignment and synchrosqueezing for general time–frequency filter banks, subsampling and processing,

    N. Holighaus, Z. Pr ˚uˇsa, and P. L. Søndergaard, “Reassignment and synchrosqueezing for general time–frequency filter banks, subsampling and processing,”Signal Processing, vol. 125, pp. 1–8, 2016

    2016

  53. [61]

    A fast griffin-lim algorithm,

    N. Perraudin, P. Balazs, and P. L. Søndergaard, “A fast griffin-lim algorithm,” in2013 IEEE workshop on applications of signal processing to audio and acoustics. IEEE, 2013, pp. 1–4

    2013

  54. [62]

    Automatic mouse ultrasound detector (A-MUD): A new tool for processing rodent vocalizations,

    S. M. Zala, D. Reitschmidt, A. Noll, P. Balazs, and D. Penn, “Automatic mouse ultrasound detector (A-MUD): A new tool for processing rodent vocalizations,”PLOS ONE, 2017

    2017

  55. [63]

    Deepsqueak: a deep learning-based system for detection and analysis of ultrasonic vocalizations,

    K. R. Coffey, R. E. Marx, and J. F. Neumaier, “Deepsqueak: a deep learning-based system for detection and analysis of ultrasonic vocalizations,”Neuropsychopharmacology, vol. 44, no. 5, pp. 859–868, 2019

    2019

  56. [64]

    Female presence and estrous state influence mouse ultrasonic courtship vocalizations,

    J. L. Hanson and L. M. Hurley, “Female presence and estrous state influence mouse ultrasonic courtship vocalizations,”PloS one, vol. 7, no. 7, p. e40782, 2012

    2012

  57. [65]

    Capturing the songs of mice with an improved detection and classification method for ultrasonic vocalizations (bootsnap),

    R. Abbasi, P. Balazs, M. A. Marconi, D. Nicolakis, S. M. Zala, and D. J. Penn, “Capturing the songs of mice with an improved detection and classification method for ultrasonic vocalizations (bootsnap),”PLoS Computational Biology, vol. 18, no. 5, p. e1010049, 2022

    2022

  58. [66]

    librosa: Audio and music signal analysis in python

    B. McFee, C. Raffel, D. Liang, D. P. Ellis, M. McVicar, E. Battenberg, and O. Nieto, “librosa: Audio and music signal analysis in python.”SciPy, vol. 2015, pp. 18–24, 2015

    2015

  59. [67]

    The large time-frequency analysis toolbox 2.0,

    Z. Pr ˚uˇsa, P. L. Søndergaard, N. Holighaus, C. Wiesmeyr, and P. Balazs, “The large time-frequency analysis toolbox 2.0,” inSound, Music, and Motion: 10th International Symposium, CMMR 2013, Marseille, France, October 15-18, 2013. Revised Selected Papers 10. Springer, 2014, p...

    2013

  60. [68]

    The linear time frequency analysis toolbox,

    P. L. Søndergaard, B. Torr´esani, and P. Balazs, “The linear time frequency analysis toolbox,”International Journal of Wavelets, Multiresolution and Information Processing, vol. 10, no. 04, p. 1250032, 2012

    2012

  61. [69]

    A framework for invertible, real-time constant-Q transforms,

    N. Holighaus, M. D ¨orfler, G. A. Velasco, and T. Grill, “A framework for invertible, real-time constant-Q transforms,”IEEE Trans. Audio Speech Lang. Process., vol. 21, no. 4, pp. 775 –785, 2013

    2013

  62. [70]

    Sdr–half-baked or well done?

    J. Le Roux, S. Wisdom, H. Erdogan, and J. R. Hershey, “Sdr–half-baked or well done?” inICASSP 2019-2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2019, pp. 626–630

    2019

  63. [71]

    Melody extraction from polyphonic music signals: Approaches, applications, and challenges,

    J. Salamon, E. G ´omez, D. P. Ellis, and G. Richard, “Melody extraction from polyphonic music signals: Approaches, applications, and challenges,” IEEE Signal Processing Magazine, vol. 31, no. 2, pp. 118–134, 2014

    2014

  64. [72]

    Improving bird classification with unsupervised sound separation,

    T. Denton, S. Wisdom, and J. R. Hershey, “Improving bird classification with unsupervised sound separation,” inICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2022, pp. 636–640

    2022

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.