Pith. sign in

REVIEW 3 major objections 5 minor 57 references

Raw Audio Classification with Cosine Convolutional Neural Network (CosCovNN)

T0 review · 3 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read Replacing CNN kernels with two-parameter cosine filters is claimed to make raw-audio classifiers both smaller and more accurate, with a memory-augmented version reaching state-of-the-art results.

desk verdict A simple two-parameter cosine filter for every conv layer is a plausible efficiency idea, but best-of-five selection on the evaluation datasets and a self-contradicted SOTA claim make the reported gains unverified. read the letter →

arxiv 2412.00312 v1 pith:ZPX2B43W submitted 2024-11-30 cs.SD cs.AIeess.AS

Kazi Nazmul Haque , Rajib Rana , Tasnim Jarin , Bjorn W. Schuller Jr This is my paper · ORCID
classification cs.SDcs.AIeess.AS
keywords rawaudioclassificationcosineconvolutionalneuralnetworkparametricfilterdesignlearnablefrontendvectorquantizationmemory-augmentednetworksspeechemotionrecognitionacousticscene
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 sets out to show that a convolutional network for raw audio does not need thousands of learned filter weights: each filter can be generated by a cosine function with just two learnable parameters, an amplitude and a frequency. The resulting model, CosCovNN, is reported to be more accurate than an equivalently structured CNN on all five audio classification datasets tested, while using 77.66% fewer parameters. The paper further claims that adding a vector-quantisation layer and a learned memory module, producing VQCCM, lifts accuracy to state-of-the-art levels on those same datasets. If these results hold, they imply that building periodicity into the filter, rather than letting the network discover it from scratch, makes raw-waveform audio classifiers both smaller and more reliable.

What carries the argument

The load-bearing object is the cosine filter $g[n]=\theta_1\cos(\theta_2 n)$, a length-$L$ kernel generated by evaluating a periodic function with two trainable scalars: amplitude $\theta_1$ and angular frequency $\theta_2$. A standard 1D convolution would learn all $L$ tap values; CosCovNN instead fills the taps from this formula, so a layer's parameter cost is 2 per filter rather than $L$ per filter. The bounded periodic values are preserved through the network with tanh activations, and the classifier replaces a large fully connected layer with $Z$ cosine-convolutional layers followed by global average pooling. In VQCCM, the supporting mechanisms are a vector-quantisation codebook, which replaces each feature vector with its nearest codebook entry and copies gradients around the non-differentiable argmin, and a memory module whose learned vector is read, gated, multiplied into each layer's features, and written back additively.

What would settle it

Re-run the five-dataset comparisons with the same architectures but report the mean or median over five independent seeds instead of the maximum; if CosCovNN no longer beats the equivalent CNN on most tasks, or VQCCM no longer exceeds LEAF on typical runs, the central efficiency-accuracy claim fails. A complementary check is to move the S09 architecture search to a held-out validation split and see whether the chosen filter and pooling sizes still transfer to the other four datasets.

Watch

Extended reading notes

Core claim

The central claim is that the standard learned kernel of a 1D CNN can be replaced, for audio, by a cosine filter $g[n]=\theta_1\cos(\theta_2 n)$ without sacrificing accuracy—on the contrary, CosCovNN is reported to beat the equivalent CNN on all five tasks. Because every filter of length $L$ is specified by only two scalars, the convolutional parameter count falls from $L$ per filter to 2 per filter, which the paper computes as a 77.66% reduction for its five-layer architecture. The augmented model VQCCM inserts a vector-quantisation codebook after the first cosine layer and a memory reader-writer at every layer; the paper reports that VQCCM surpasses the LEAF learnable-frontend baseline on all five datasets, ties it on acoustic scenes at 99.1% with a smaller standard deviation, and achieves state-of-the-art accuracy in several categories.

Load-bearing premise

The protocol records the maximum accuracy over five training runs and uses those maxima to choose filter and pooling sizes on the S09 dataset and to compare models, so if the best of five runs is a lucky outlier rather than the typical outcome, the reported margins over the CNN baseline and over LEAF could shrink.

Editorial extensions

If this is right

  • If the cosine parametrisation is as effective as reported, any raw-waveform CNN can cut its convolutional parameter count by roughly the ratio of filter length to two, with no accuracy loss.
  • Because the architecture search was run once on the S09 spoken-digit dataset and then transferred to speech emotion, speaker identification, acoustic scenes, and music, the paper's recipe is claimed to be dataset-agnostic within audio.
  • The ablation study suggests that the memory module helps consistently, while vector quantisation alone does not, but the two together give the largest gains—so they appear to solve complementary problems.
  • A 77.66% parameter reduction makes CosCovNN and VQCCM natural candidates for memory-constrained and on-device audio classification pipelines.

Reading between the lines

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

  • A sharper comparison than the paper's would pit CosCovNN against a CNN with matched total parameter count rather than the same architecture; that would isolate whether the cosine form itself, rather than smaller model size, drives the accuracy gain.
  • Because a cosine filter is a continuous function, it can be evaluated at any kernel length for the same two parameters, which suggests a route to very long filters for low-frequency structure without extra training cost.
  • The two-parameter form is effectively a learned Fourier feature; initialising $\theta_2$ from DCT frequencies or letting each channel carry its own frequency could make the layer behave like a trainable DCT basis for a wider range of signals.
  • The memory size was tuned on one-second utterances, but acoustic scenes are ten-second clips, so a direct test is whether a fixed memory dimension still suffices when input length grows tenfold.
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 / 5 minor

Summary. The manuscript introduces CosCovNN, a convolutional architecture for raw-waveform audio classification in which each filter is a learnable cosine function g[n] = θ₁ cos(θ₂ n), requiring only two parameters per filter instead of L. It also presents VQCCM, an extension that adds vector quantization and a memory layer. The authors report that CosCovNN outperforms an equivalent CNN with 77.66% fewer parameters on five audio datasets (Speech Command, Spoken Digit, Speech Emotion, Acoustic Scenes, Musical Instrument, Speaker Id), and that VQCCM achieves state-of-the-art accuracy on all five datasets. The supporting material includes an architecture search on the S09 dataset, comparisons to TD-fbanks, SincNet, LEAF, and a CNN baseline, and an ablation study of the memory and VQ components.

Significance. If the core claims hold, the cosine-filter parameterization is an elegant and practically attractive idea: it drastically reduces the number of trainable parameters in convolutional layers while preserving or improving accuracy on raw-waveform audio. The paper's central comparison is internally coherent in its parameter-count arithmetic (the 77.66% reduction follows from the filter sizes and the two-parameter form), and the use of a straightforward Fourier/DCT-inspired filter is a clear, easily communicated contribution. The authors also provide a concrete architecture and an ablation that isolates the effects of memory and VQ components. However, the experimental protocol and result reporting as presented do not currently support the paper's headline claims: the selection-on-the-evaluation-set protocol introduces selection bias, and the SOTA claim is contradicted by the paper's own Table 4 on Acoustic Scenes. The contribution is worth pursuing, but the empirical case needs to be rebuilt with proper validation methodology and corrected claims.

major comments (3)
  1. [§5.1.1, §5.4.1, Table 4] The model-selection protocol uses the maximum accuracy over five training runs on S09 and IEMOCAP, and those very datasets then appear in the final evaluation in Table 4. Section 5.1.1 states that the S09 architecture search 'is conducted five times and the maximum accuracy is recorded for comparison,' and Section 5.4.1 says that VQCCM memory and VQ sizes were chosen using 'maximum accuracy based on five runs' on S09 and IEMOCAP. The reported mean±std for the chosen configuration does not correct for the multiplicity of runs and architectures tested, so the reported numbers are optimistically biased. Because several margins in Table 4 are small (Spoken Digit: VQCCM 97.1 vs LEAF 96.7; Acoustic Scenes: 99.1 vs 99.1), selection bias could plausibly account for the claimed advantages. Please report the full distribution of results (e.g., mean and standard deviation over all runs used for selection), perform model selection on a held-out validation split that is disjoint from the test evaluation, or provide an explicit correction for selection.
  2. [§5.3.2, Table 4, Abstract] The abstract's statement that VQCCM 'achieves state-of-the-art (SOTA) performance across five different datasets' is contradicted by Table 4 itself: on Acoustic Scenes, TD-fbanks reaches 99.5 ± 0.4, which is higher than VQCCM's 99.1 ± 0.3, and VQCCM merely ties LEAF at 99.1. The sentence in §5.3.2 that 'VQCCM has outperformed LEAF for all tasks' is also false for Acoustic Scenes, where the means are equal (99.1 vs 99.1). The claims should be revised to specify the datasets on which VQCCM actually achieves a new best result, and the text in §5.3.2 should be made consistent with the numbers in Table 4.
  3. [§5.1.1, §5.2.2, Table 4] The architecture (filter sizes 100, 50, 12, 6, 3 and pooling sizes 10, 8, 4, 4) was selected on the S09 dataset for CosCovNN using the max-of-five protocol, and the same architecture is then used for the CNN baseline without any tuning of the CNN's filter sizes or pooling. The parameter-count reduction is a simple arithmetic fact, but the accuracy comparison between CosCovNN and CNN is confounded: the architecture search may have favored the cosine filters, and the CNN may not be at its best configuration. Please either run an equivalent architecture search for the CNN under the same protocol or compare both models on a fixed, pre-specified architecture and report the full distribution of results over runs.
minor comments (5)
  1. [Section 2] The text from 'The field of audio classification has been fundamentally transformed...' through '...reduces the model's complexity by approximately 77' appears twice verbatim (once on pages 3–4 and once on pages 5–6). This duplication should be removed.
  2. [Eq. (7)] Equation (7) writes 'Fi = F'i = Ek', which is confusing; it should define F'i = Ek with k = argmin_j ||Fi − Ej||2, and then state that the gradient is copied from F'i to Fi during backpropagation.
  3. [References] Reference [41] is cited for both CGCNN (Noé et al.) and CLDNN (Sainath et al.), which are different works, and the citation for TD-filterbanks in Table 1 is shown as an empty pair of brackets. The reference list and in-text citations should be corrected.
  4. [General presentation] There are several typos and formatting issues: 'fare' should be 'fair' in §5.2.1; 'consisted of' should be 'consists of' in §4.1; 'comprises of' should be 'comprises' in §4.5; and the claim in §5.2.2 that cosine filters 'are less impacted by noise' is not supported by any experiment in the paper.
  5. [Reproducibility] The paper does not report whether code is available, and training hyperparameters such as learning rate, number of epochs, batch size, and optimizer are not specified, which makes the results difficult to reproduce or compare against.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the cosine-filter derivation is an algebraic reparameterization and all headline accuracy claims are empirical comparisons on external datasets.

full rationale

The paper's only formal derivation is the cosine filter itself: starting from y[n]=A cos(2π n/λ), the authors set A=θ1 and 2π/λ=θ2 to obtain g[n,θ1,θ2]=θ1 cos(θ2 n). This is a renaming of constants, not a result that presupposes the paper's accuracy conclusions. The 77.66% parameter reduction is computed directly from the stated filter sizes and two-parameter-per-filter design, and the accuracy comparisons in Table 4 are empirical outcomes on standard public datasets, not quantities that appear as inputs to the model. No load-bearing self-citation, uniqueness theorem, or ansatz-importing citation was found; references to DCT/Fourier theory provide motivation rather than a forced derivation. The main legitimate concern is statistical protocol, not circularity: the architecture search (Section 5.1.1) and VQ/memory tuning (Section 5.4.1) record maximum accuracy over five runs on S09 and IEMOCAP, and those datasets also appear in the final comparison table, so the reported numbers may carry selection bias; the abstract's state-of-the-art claim is also internally inconsistent with the Acoustic Scenes row (TD-fbanks 99.5 vs VQCCM 99.1). These concerns affect evidential strength but do not make any reported prediction equal, by construction, to a fitted input. The derivation chain is therefore self-contained and no circular step is exhibited.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The central efficiency claim rests on the two-parameter cosine filter and on an architecture whose hyperparameters were selected on one dataset. The most consequential untested assumptions are the transferability of the S09-tuned architecture and the use of maximum-of-five accuracy as a performance measure. No new physical entities are introduced.

free parameters (4)
  • Cosine filter parameters θ1 (amplitude) and θ2 (angular frequency) = Learned during training; final values not reported
    Every CosCovNN filter is a cosine with two learnable parameters; this is the source of the parameter-count reduction and is fitted to the data.
  • Architecture filter sizes (100, 50, 12, 6, 3) and pooling sizes (10, 8, 4, 4) = Selected via search on S09 dataset
    Chosen by maximizing accuracy on S09 and then transferred to the other four datasets with only layer/filter-count changes.
  • Memory size and VQ codebook size = Memory 100/500, embeddings 256/512 across experiments
    Tuned on S09 and IEMOCAP (Section 5.4); final values for each dataset are not fully specified.
  • VQ commitment loss weight β and training hyperparameters (learning rate, epochs, batch size) = Not reported
    Required to reproduce the loss in Eq. (10) and training; omitted from the text.
assumptions (4)
  • domain assumption Audio signals can be effectively represented by cosine filters at every convolutional layer
    Section 3.2.1 defines the filter as θ1 cos(θ2 n); the whole method assumes this two-parameter periodic basis captures the informative structure of raw waveforms.
  • domain assumption An architecture tuned on the S09 dataset generalizes to other audio datasets with only layer/filter-count changes
    Section 5.1.1 states the search is done once on S09 and then applied elsewhere; cross-dataset results depend on this transfer.
  • ad hoc to paper Maximum accuracy over five runs is a valid model-selection and reporting criterion
    Section 5.1.1 says 'every experiment is conducted five times and the maximum accuracy is recorded'; this inflates apparent performance and is not a standard summary.
  • standard math Standard convolution and back-propagation definitions hold for the cosine filter
    Eq. (1) and Eq. (6) rely on the standard 1D convolution operation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Raw Audio Classification with Cosine Convolutional Neural Network (CosCovNN)." pith.science (2026). https://pith.science/paper/ZPX2B43W

@misc{pith2026241200312,
  author       = {Pith},
  title        = {Pith review of: Raw Audio Classification with Cosine Convolutional Neural Network (CosCovNN)},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZPX2B43W}},
  note         = {Machine review of arXiv:2412.00312}
}
abstract

This study explores the field of audio classification from raw waveform using Convolutional Neural Networks (CNNs), a method that eliminates the need for extracting specialised features in the pre-processing step. Unlike recent trends in literature, which often focuses on designing frontends or filters for only the initial layers of CNNs, our research introduces the Cosine Convolutional Neural Network (CosCovNN) replacing the traditional CNN filters with Cosine filters. The CosCovNN surpasses the accuracy of the equivalent CNN architectures with approximately $77\%$ less parameters. Our research further progresses with the development of an augmented CosCovNN named Vector Quantised Cosine Convolutional Neural Network with Memory (VQCCM), incorporating a memory and vector quantisation layer VQCCM achieves state-of-the-art (SOTA) performance across five different datasets in comparison with existing literature. Our findings show that cosine filters can greatly improve the efficiency and accuracy of CNNs in raw audio classification.

Figures

Figures reproduced from arXiv: 2412.00312 by the authors.

Figure 1
Figure 1. The present figure depicts an exposition of the intricate architecture of the [PITH_FULL_IMAGE:figures/full_fig_p009_1.png] view at source ↗
Figure 2
Figure 2. The figure illustrates a comparative analysis between the convolutional layer [PITH_FULL_IMAGE:figures/full_fig_p011_2.png] view at source ↗
Figure 3
Figure 3. The diagram presents the structure of the VQCCM model. In this architecture, [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: The architectural details of the VQ layer are shown in this figure. Here, the [PITH_FULL_IMAGE:figures/full_fig_p014_4.png]
Figure 5
Figure 5. Figure 5: The figure provides a detailed view of the Memory Reader and Writer compo [PITH_FULL_IMAGE:figures/full_fig_p016_5.png]
Figure 6
Figure 6. Figure 6: The figure presents a comparative visualization of the trained filter outputs from [PITH_FULL_IMAGE:figures/full_fig_p021_6.png]
Figure 7
Figure 7. Figure 7: This figure shows the impact of the Memory and VQ layer of the VQCCM model [PITH_FULL_IMAGE:figures/full_fig_p023_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

57 extracted references · 46 canonical work pages

  1. [1]

    A survey on the new generation of deep learning in image processing,

    L. Jiao and J. Zhao, “A survey on the new generation of deep learning in image processing,” IEEE Access, vol. 7, pp. 172231–172263, 2019

    work page 2019

  2. [2]

    Comparison and analysis of SampleCNN architectures for audio classification,

    T. Kim, J. Lee, and J. Nam, “Comparison and analysis of SampleCNN architectures for audio classification,” IEEE J. Sel. Topics Signal Pro- cess., vol. 13, no. 2, pp. 285–297, 2019

    work page 2019

  3. [3]

    Convolutional Recurrent Neural Networks for Urban Sound Classification Using Raw Waveforms,

    J. Sang, S. Park, and J. Lee, “Convolutional Recurrent Neural Networks for Urban Sound Classification Using Raw Waveforms,” in 2018 26th European Signal Processing Conference (EUSIPCO) , 2018, pp. 2444– 2448, doi: 10.23919/EUSIPCO.2018.8553247

    work page arXiv 2018

  4. [4]

    Rawnet: Advanced end-to-end deep neural network using raw waveforms for text- independent speaker verification,

    J.-W. Jung, H.-S. Heo, J.-H. Kim, H.-J. Shim, and H.-J. Yu, “Rawnet: Advanced end-to-end deep neural network using raw waveforms for text- independent speaker verification,” in Interspeech, 2019, pp. 1268–1272

    work page 2019

  5. [5]

    Learning the Front-End Speech Feature with Raw Waveform for End-to-End Speaker Recogni- tion,

    N. Liang, W. Xu, C. Luo, and W. Kang, “Learning the Front-End Speech Feature with Raw Waveform for End-to-End Speaker Recogni- tion,” in Proceedings of the 2020 6th International Conference on Com- puting and Artificial Intelligence , Tianjin, China, 2020, pp. 317–322, doi: 10.1145/3404555.3404571

    work page arXiv 2020

  6. [6]

    Learning environmental sounds with end- to-end convolutional neural network,

    Y. Tokozume and T. Harada, “Learning environmental sounds with end- to-end convolutional neural network,” in 2017 IEEE International Con- ference on Acoustics, Speech and Signal Processing (ICASSP), 2017, pp. 2721–2725, doi: 10.1109/ICASSP.2017.7952651

    arXiv 2017

  7. [7]

    Acoustic modeling with deep neural networks using raw time signal for L VCSR,

    Z. T¨ uske, P. Golik, R. Schl¨ ut, and H. Ney, “Acoustic modeling with deep neural networks using raw time signal for L VCSR,” in Fifteenth Annual Conference of the International Speech Communication Associ- ation, 2014

    work page 2014

  8. [8]

    Very deep convolutional neural networks for raw waveforms,

    W. Dai, C. Dai, S. Qu, J. Li, and S. Das, “Very deep convolutional neural networks for raw waveforms,” in2017 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , 2017, pp. 421–425, doi: 10.1109/ICASSP.2017.7952190

    arXiv 2017

Show all 57 references
  1. [9]

    Speaker recognition from raw waveform with sincnet,

    M. Ravanelli and Y. Bengio, “Speaker recognition from raw waveform with sincnet,” in 2018 IEEE Spoken Language Technology Workshop (SLT), 2018, pp. 1021–1028. 24

    2018

  2. [10]

    LEAF: A Learnable Frontend for Audio Classification,

    N. Zeghidour, O. Teboul, F. de Chaumont Quitry, and M. Tagliasacchi, “LEAF: A Learnable Frontend for Audio Classification,” in Interna- tional Conference on Learning Representations , 2021. [Online]. Avail- able: https://openreview.net/forum?id=jM76BCb6F9m

    2021

  3. [11]

    Discrete Cosine Transform,

    N. Ahmed, T. Natarajan, and K. R. Rao, “Discrete Cosine Transform,” IEEE Transactions on Computers, vol. C-23, no. 1, pp. 90–93, 1974, doi: 10.1109/T-C.1974.223784

    1974

  4. [12]

    An Audio Scene Classification Frame- work with Embedded Filters and a DCT-based Temporal Module,

    H. Chen, P. Zhang, and Y. Yan, “An Audio Scene Classification Frame- work with Embedded Filters and a DCT-based Temporal Module,” in ICASSP 2019 - 2019 IEEE International Conference on Acous- tics, Speech and Signal Processing (ICASSP) , 2019, pp. 835–839, doi: 10.1109/ICASSP....

    2019

  5. [13]

    End-to-end speech enhancement based on dis- crete cosine transform,

    C. Geng and L. Wang, “End-to-end speech enhancement based on dis- crete cosine transform,” in 2020 IEEE International Conference on Arti- ficial Intelligence and Computer Applications (ICAICA) , 2020, pp. 379– 383

    2020

  6. [14]

    J. O. Smith, Mathematics of the Discrete Fourier Transform (DFT): With Audio Applications . Julius Smith, 2008

    2008

  7. [15]

    Gaussian Mixture Modeling of Short- Time Fourier Transform Features for Audio Fingerprinting,

    A. Ramalingam and S. Krishnan, “Gaussian Mixture Modeling of Short- Time Fourier Transform Features for Audio Fingerprinting,” IEEE Transactions on Information Forensics and Security , vol. 1, no. 4, pp. 457–463, 2006, doi: 10.1109/TIFS.2006.885036

    2006

  8. [16]

    Time- Frequency Filter Bank: A Simple Approach for Audio and Music Sepa- ration,

    N. Yang, M. Usman, X. He, M. A. Jan, and L. Zhang, “Time- Frequency Filter Bank: A Simple Approach for Audio and Music Sepa- ration,” IEEE Access, vol. 5, pp. 27114–27125, 2017, doi: 10.1109/AC- CESS.2017.2761741

    2017

  9. [17]

    Convolutional Neural Network Accelerator with Vector Quantization,

    H. Lee, Y.-H. Wu, Y.-S. Lin, and S.-Y. Chien, “Convolutional Neural Network Accelerator with Vector Quantization,” in 2019 IEEE Interna- tional Symposium on Circuits and Systems (ISCAS) , 2019, pp. 1–5, doi: 10.1109/ISCAS.2019.8702105

    2019

  10. [18]

    Neural Turing Machines,

    A. Graves, G. Wayne, and I. Danihelka, “Neural Turing Machines,” arXiv preprint arXiv:1410.5401 , 2014. 25

    2014 arXiv

  11. [19]

    Meta-Learning with Memory-Augmented Neural Networks,

    A. Santoro, S. Bartunov, M. Botvinick, D. Wierstra, and T. Lillicrap, “Meta-Learning with Memory-Augmented Neural Networks,” in Pro- ceedings of The 33rd International Conference on Machine Learning , New York, NY, USA, 20–22 Jun. 2016, pp. 1842–1850. PMLR

    2016

  12. [20]

    Quan- tized Memory-Augmented Neural Networks,

    S. Park, S. Kim, S. Lee, H. Bae, and S. Yoon, “Quan- tized Memory-Augmented Neural Networks,” Proceedings of the AAAI Conference on Artificial Intelligence , vol. 32, no. 1, Apr. 2018, doi: 10.1609/aaai.v32i1.11704. [Online]. Available: https://ojs.aaai.org/index.php/AAAI/arti...

    2018 doi

  13. [21]

    Content-based audio classification and retrieval by support vector machines,

    G. Guo and S. Z. Li, “Content-based audio classification and retrieval by support vector machines,” IEEE Transactions on Neural Networks , vol. 14, no. 1, pp. 209–215, 2003

    2003

  14. [22]

    Mixed type audio classification with support vector machine,

    L. Chen, S. Gunduz, and M. T. Ozsu, “Mixed type audio classification with support vector machine,” in 2006 IEEE International Conference on Multimedia and Expo , 2006, pp. 781–784

    2006

  15. [23]

    Content analysis for audio classi- fication and segmentation,

    L. Lu, H.-J. Zhang, and H. Jiang, “Content analysis for audio classi- fication and segmentation,” IEEE Transactions on Speech and Audio Processing, vol. 10, no. 7, pp. 504–516, 2002

    2002

  16. [24]

    Content based audio classification: A neural network approach,

    V. Mitra and C.-J. Wang, “Content based audio classification: A neural network approach,” Soft Computing , vol. 12, pp. 639–646, 2008

    2008

  17. [25]

    Audio environment classifi- cation for hearing aids using artificial neural networks with windowed input,

    C. Freeman, R. Dony, and S. M. Areibi, “Audio environment classifi- cation for hearing aids using artificial neural networks with windowed input,” in 2007 IEEE Symposium on Computational Intelligence in Im- age and Signal Processing , 2007, pp. 183–188

    2007

  18. [26]

    Applying neural network on the content-based audio classification,

    X. Shao, C. Xu, and M. S. Kankanhalli, “Applying neural network on the content-based audio classification,” in Fourth International Conference on Information, Communications and Signal Processing, 2003 and the Fourth Pacific Rim Conference on Multimedia. Proceedings of the 200...

    2003

  19. [27]

    A comprehen- sive review of synthetic data generation in smart farming by us- ing variational autoencoder and generative adversarial network,

    Y. Akkem, S. K. Biswas, and A. Varanasi, “A comprehen- sive review of synthetic data generation in smart farming by us- ing variational autoencoder and generative adversarial network,” 26 Engineering Applications of Artificial Intelligence , vol. 131, p. 107881, 2024, doi: 10....

    2024

  20. [28]

    Smart farm- ing using artificial intelligence: A review,

    Y. Akkem, S. K. Biswas, and A. Varanasi, “Smart farm- ing using artificial intelligence: A review,” Engineering Ap- plications of Artificial Intelligence , vol. 120, p. 105899, 2023, doi: 10.1016/j.engappai.2023.105899. [Online]. Available: https://www.sciencedirect.com/scienc...

    2023

  21. [29]

    Sound classification using convolutional neural network and tensor deep stacking network,

    A. Khamparia, D. Gupta, N. G. Nguyen, A. Khanna, B. Pandey, and P. Tiwari, “Sound classification using convolutional neural network and tensor deep stacking network,”IEEE Access, vol. 7, pp. 7717–7727, 2019

    2019

  22. [30]

    CNN architectures for large-scale audio classification,

    S. Hershey, S. Chaudhuri, D. P. W. Ellis, J. F. Gemmeke, A. Jansen, R. C. Moore, M. Plakal, D. Platt, R. A. Saurous, B. Seybold, et al. , “CNN architectures for large-scale audio classification,” in 2017 IEEE International Conference on Acoustics, Speech and Signal Processing ...

    2017

  23. [31]

    Music genre classification using duplicated convolutional layers in neural networks,

    H. Yang and W.-Q. Zhang, “Music genre classification using duplicated convolutional layers in neural networks,” in Interspeech, 2019, pp. 3382– 3386

    2019

  24. [32]

    Convolutional recurrent neural networks for the classification of cetacean bioacoustic patterns,

    D. N. Makropoulos, A. Tsiami, A. Prospathopoulos, D. Kassis, A. Frantzis, E. Skarsoulis, G. Piperakis, and P. Maragos, “Convolutional recurrent neural networks for the classification of cetacean bioacoustic patterns,” in ICASSP 2023 - 2023 IEEE International Conference on Acou...

    2023

  25. [33]

    Recurrent neural network transducer for audio-visual speech recognition,

    T. Makino, H. Liao, Y. Assael, B. Shillingford, B. Garcia, O. Braga, and O. Siohan, “Recurrent neural network transducer for audio-visual speech recognition,” in 2019 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU) , 2019, pp. 905–912

    2019

  26. [34]

    Attention based convolutional recurrent neural network for environmental sound classi- fication,

    Z. Zhang, S. Xu, S. Zhang, T. Qiao, and S. Cao, “Attention based convolutional recurrent neural network for environmental sound classi- fication,” Neurocomputing, vol. 453, pp. 896–903, 2021

    2021

  27. [35]

    A Survey of Audio Classification using Deep Learning,

    K. Zaman, M. Sah, C. Direkoglu, and M. Unoki, “A Survey of Audio Classification using Deep Learning,” IEEE Access, 2023. 27

    2023

  28. [36]

    P. C. Vakkantula, Speech Mode Classification using the Fusion of CNNs and LSTM Networks , West Virginia University, 2020

    2020

  29. [37]

    Multi-channel spectrograms for speech processing applications using deep learning methods,

    T. Arias-Vergara, P. Klumpp, J. C. Vasquez-Correa, E. N¨ oth, J. R. Orozco-Arroyave, and M. Schuster, “Multi-channel spectrograms for speech processing applications using deep learning methods,” Pattern Analysis and Applications , vol. 24, pp. 423–431, 2021

    2021

  30. [38]

    Deep neural networks for small footprint text-dependent speaker verification,

    E. Variani, X. Lei, E. McDermott, I. Lopez Moreno, and J. Gonzalez- Dominguez, “Deep neural networks for small footprint text-dependent speaker verification,” in 2014 IEEE International Conference on Acous- tics, Speech and Signal Processing (ICASSP) , 2014, pp. 4052–4056

    2014

  31. [39]

    End-to-end text- dependent speaker verification,

    G. Heigold, I. Moreno, S. Bengio, and N. Shazeer, “End-to-end text- dependent speaker verification,” in 2016 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , 2016, pp. 5115– 5119

    2016

  32. [40]

    On Learning Interpretable CNNs with Parametric Modulated Kernel-Based Filters,

    E. Loweimi, P. Bell, and S. Renals, “On Learning Interpretable CNNs with Parametric Modulated Kernel-Based Filters,” in Interspeech, 2019, pp. 3480–3484

    2019

  33. [41]

    Cgcnn: Complex Gabor con- volutional neural network on raw speech,

    P.-G. No´ e, T. Parcollet, and M. Morchid, “Cgcnn: Complex Gabor con- volutional neural network on raw speech,” in ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2020, pp. 7724–7728

    2020

  34. [42]

    Learning filterbanks from raw speech for phone recognition,

    N. Zeghidour, N. Usunier, I. Kokkinos, T. Schaiz, G. Synnaeve, and E. Dupoux, “Learning filterbanks from raw speech for phone recognition,” in 2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2018, pp. 5509–5513

    2018

  35. [43]

    Audio set: An ontology and human-labeled dataset for audio events,

    J. F. Gemmeke, D. P. W. Ellis, D. Freedman, A. Jansen, W. Lawrence, R. C. Moore, M. Plakal, and M. Ritter, “Audio set: An ontology and human-labeled dataset for audio events,” in 2017 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2017, pp. 776–780

    2017

  36. [44]

    Efficientleaf: A faster learnable audio frontend of questionable use,

    J. Schl¨ uter and G. Gutenbrunner, “Efficientleaf: A faster learnable audio frontend of questionable use,” in 2022 30th European Signal Processing Conference (EUSIPCO), 2022, pp. 205–208. 28

    2022

  37. [45]

    Analysis of CNN-based speech recognition system using raw speech as input,

    D. Palaz and R. Collobert, “Analysis of CNN-based speech recognition system using raw speech as input,” Idia, Tech. Rep., 2015

    2015

  38. [46]

    Speech acoustic modeling from raw multichannel waveforms,

    Y. Hoshen, R. J. Weiss, and K. W. Wilson, “Speech acoustic modeling from raw multichannel waveforms,” in2015 IEEE International Confer- ence on Acoustics, Speech and Signal Processing (ICASSP) , 2015, pp. 4624–4628

    2015

  39. [47]

    Speaker location and microphone spacing in- variant acoustic modeling from raw multichannel waveforms,

    T. N. Sainath, R. J. Weiss, K. W. Wilson, A. Narayanan, M. Bac- chiani, and A. Senior, “Speaker location and microphone spacing in- variant acoustic modeling from raw multichannel waveforms,” in 2015 IEEE Workshop on Automatic Speech Recognition and Understanding (ASRU), 2015,...

    2015

  40. [48]

    Audio compression using a modified discrete cosine transform with tem- poral auditory masking,

    A. O. Salau, I. Oluwafemi, K. F. Faleye, and S. Jain, “Audio compression using a modified discrete cosine transform with tem- poral auditory masking,” in 2019 International Conference on Sig- nal Processing and Communication (ICSC) , 2019, pp. 135–142, doi: 10.1109/ICSC45622.2...

    2019

  41. [49]

    Efficient object localization using convolutional networks,

    J. Tompson, R. Goroshin, A. Jain, Y. LeCun, and C. Bregler, “Efficient object localization using convolutional networks,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 2015, pp. 648–656

    2015

  42. [50]

    Neural discrete representa- tion learning,

    A. Van Den Oord, O. Vinyals, and others, “Neural discrete representa- tion learning,” Advances in Neural Information Processing Systems, vol. 30, 2017

    2017

  43. [51]

    1D convolutional neural networks and applications: A survey,

    S. Kiranyaz, O. Avci, O. Abdeljaber, T. Ince, M. Gabbouj, and D. J. Inman, “1D convolutional neural networks and applications: A survey,” Mechanical Systems and Signal Processing , vol. 151, p. 107398, 2021

    2021

  44. [52]

    Network in network,

    M. Lin, Q. Chen, and S. Yan, “Network in network,” arXiv preprint arXiv:1312.4400, 2013

    2013 arXiv

  45. [53]

    Speech Commands: A Dataset for Limited-Vocabulary Speech Recognition,

    P. Warden, “Speech Commands: A Dataset for Limited-Vocabulary Speech Recognition,” CoRR, vol. abs/1804.03209, 2018. [Online]. Avail- able: http://arxiv.org/abs/1804.03209. 29

    2018 arXiv

  46. [54]

    IEMOCAP: Interactive emotional dyadic motion capture database,

    C. Busso, M. Bulut, C.-C. Lee, A. Kazemzadeh, E. Mower, S. Kim, J. N. Chang, S. Lee, and S. S. Narayanan, “IEMOCAP: Interactive emotional dyadic motion capture database,” Language Resources and Evaluation , vol. 42, no. 4, pp. 335–359, 2008

    2008

  47. [55]

    Voxceleb: Large- scale speaker verification in the wild,

    A. Nagrani, J. S. Chung, W. Xie, and A. Zisserman, “Voxceleb: Large- scale speaker verification in the wild,” Computer Speech and Language , vol. 60, p. 101027, 2020, doi: 10.1016/j.csl.2019.101027

    2020

  48. [56]

    A multi-device dataset for urban acoustic scene classification,

    A. Mesaros, T. Heittola, and T. Virtanen, “A multi-device dataset for urban acoustic scene classification,” in Workshop on Detection and Classification of Acoustic Scenes and Events , 2018. [Online]. Available: https://api.semanticscholar.org/CorpusID:50785886

    2018

  49. [57]

    Neural audio synthesis of musical notes with wavenet au- toencoders,

    J. Engel, C. Resnick, A. Roberts, S. Dieleman, M. Norouzi, D. Eck, and K. Simonyan, “Neural audio synthesis of musical notes with wavenet au- toencoders,” in International Conference on Machine Learning, PMLR, 2017, pp. 1068–1077. 30 Table 1: Comparison of VQCCM with the liter...

    2017

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.