Pith. sign in

REVIEW 3 major objections 5 minor 83 references

SigWavNet: Learning Multiresolution Signal Wavelet Network for Speech Emotion Recognition

T0 review · 3 major / 5 minor · reviewed 2026-08-09 · deepseek-v4-flash

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

Pith's one-line read SigWavNet proposes a fully learnable multiresolution wavelet network for speech emotion recognition, reporting accuracy and F1 gains over prior methods on IEMOCAP and EMO-DB.

desk verdict The architecture is interesting and the ablation is useful, but the speaker-independent claim is not supported by the described stratified random split on 10-speaker datasets. read the letter →

arxiv 2502.00310 v1 pith:ZPEIKTOX submitted 2025-02-01 cs.SD cs.AIcs.CLeess.AS

Alaa Nfissi , Wassim Bouachir , Nizar Bouguila , Brian Mishara This is my paper · ORCID
classification cs.SDcs.AIcs.CLeess.AS
keywords speechemotionrecognitionfastdiscretewavelettransformlearnablefiltersasymmetrichardthresholdingdilatedCNNBi-GRUattentionmechanismIEMOCAP
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 tries to establish that a speech emotion recognition system can be built end-to-end from raw waveforms by replacing the fixed wavelet transform with a learnable fast discrete wavelet transform, and that this design beats previously published methods on two standard acted-speech datasets. The authors report speaker-independent accuracy of 84.8% and F1 of 85.1% on IEMOCAP, and 90.1% accuracy and 90.3% F1 on EMO-DB, without fixed-length segmentation or separate denoising. If true, the result suggests that wavelet-inspired filter banks can be trained as part of a neural network rather than hand-designed, and that such a front end helps emotion recognition by localizing cues in both time and frequency.

What carries the argument

The load-bearing mechanism is the learnable fast discrete wavelet transform (FDWT) layer: a cascade of convolutional filter pairs, initialized as Daubechies-10 wavelet coefficients and refined by training, that decompose raw speech into approximation and detail coefficients at several levels. The conjugate quadrature filter identity $g[n]=(-1)^n h[-n]$ derives the high-pass kernel from the low-pass kernel, halving parameters while preserving orthogonality. A learnable asymmetric hard-thresholding (LAHT) activation applies two sigmoid-based gates with separate positive and negative biases to mimic wavelet denoising inside the network. Downstream, 1D dilated convolutions with spatial attention and a bidirectional GRU with temporal attention extract spatial and sequential patterns from each band, and a channel-weighting layer reweights the bands before global average pooling and log-softmax classification.

What would settle it

Run SigWavNet with leave-one-speaker-out cross-validation on IEMOCAP and EMO-DB, training on nine speakers and testing on the held-out tenth, and compare accuracy and F1 with the reported 84.8%/85.1% and 90.1%/90.3%; a noticeably lower score would indicate that the original numbers rely on speaker leakage.

Watch

Extended reading notes

Core claim

The central claim is that making the discrete wavelet transform itself trainable removes the main obstacle to using wavelets in deep learning for speech emotion recognition. SigWavNet treats the low-pass and high-pass filters of each decomposition level as convolutional kernels initialized with Daubechies-10 coefficients, constrains the high-pass kernel to be the alternating flip of the low-pass kernel through the conjugate quadrature filter property, and applies a learnable asymmetric hard-thresholding activation to the detail coefficients at each level. The resulting multiresolution coefficients are fed to a dilated CNN with spatial attention and a Bi-GRU with temporal attention, then pooled and classified with log-softmax. The authors argue that this learns a data-driven filter bank resembling a Mel-scale emphasis on low frequencies while preserving high-frequency details, and they report that the full model outperforms all compared methods on both datasets.

Load-bearing premise

The load-bearing premise is that stratified random utterance-level splits give a speaker-independent evaluation, even though each dataset has only ten speakers, so the same speaker's voice almost certainly appears in both training and test folds.

Editorial extensions

If this is right

  • If the reported results hold under proper speaker separation, raw-waveform SER no longer needs fixed-length segmentation, hand-designed features, or a separate denoising stage.
  • The learnable FDWT front end offers a compact inductive bias: an $L$-level cascade uses only $2L$ filters, far fewer than typical 1D CNNs, which could make the architecture attractive for on-device emotion recognition.
  • The same wavelet-to-CNN design could be applied to other 1D signal classification tasks where multiscale denoising and time-frequency localization matter.
  • The reported margins over prior methods on IEMOCAP and EMO-DB would make SigWavNet the default baseline to beat in future SER comparisons, assuming the evaluation protocol is confirmed.

Reading between the lines

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

  • The evaluation claim should be tested with leave-one-speaker-out splits. IEMOCAP and EMO-DB contain only ten speakers each, so stratified random utterance splits almost certainly mix the same speaker into training and test, and the reported 'speaker-independent' numbers may partly reflect speaker-identity cues; a speaker-disjoint evaluation would show how much of the margin survives.
  • A natural extension is a noise-robustness test: add babble, music, or channel noise to the raw waveforms and measure whether the learnable thresholding maintains its advantage over fixed hard thresholding.
  • The learnable FDWT idea is not limited to emotions; it could be transferred to other variable-length physiological and acoustic signals, such as ECG, EEG, or environmental audio, where learnable multiscale denoising is likely to behave similarly.
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 paper introduces SigWavNet, an end-to-end multiresolution deep learning architecture for speech emotion recognition (SER). The model operates on raw waveforms, using a learnable fast discrete wavelet transform (FDWT) with learnable asymmetric hard thresholding (LAHT), followed by a 1D dilated CNN with spatial attention and a Bi-GRU with temporal attention, then channel weighting and global average pooling before classification. The authors report state-of-the-art speaker-independent accuracy of 84.8% on IEMOCAP and 90.1% on EMO-DB, with F1-scores of 85.1% and 90.3%, respectively. The paper includes a per-class performance analysis, a comparison with prior methods, and an ablation study examining the contributions of the learnable wavelet kernels, LAHT, and the attention components. Source code is provided via a GitHub repository.

Significance. The architectural ideas in this paper are interesting and potentially useful: a learnable wavelet filter bank that adapts during training, a differentiable asymmetric hard-thresholding activation, and the combination of multiresolution decomposition with attention-based sequence modeling are well-motivated. The ablation study is systematic and informative, showing the incremental contributions of LAHT, kernel learning strategies, and the Bi-GRU/temporal-attention module. The open-source release of the code is a strength. However, the central empirical claim—superiority over state-of-the-art methods in speaker-independent SER—is not supported by the evaluation protocol described in Section IV-B. Because the data splits are stratified random utterance-level splits rather than speaker-disjoint splits, and both datasets have only 10 speakers, the reported accuracies likely benefit from speaker identity leakage. If the experiments were redone with a proper speaker-independent protocol, the method could be a valuable contribution, but as presented the headline result is unsubstantiated.

major comments (3)
  1. [Section IV-B] The evaluation protocol is not speaker-independent despite being labeled as such. The text states that data are partitioned using "stratified random sampling"[55] with strata based on emotion classes, and no step ensures that utterances from the same speaker are confined to either training or test sets. IEMOCAP and EMO-DB each contain only 10 speakers; with a 90/10 stratified random split, the same speaker's utterances will almost certainly appear in both training and test folds. Under these conditions, the model can exploit speaker-specific vocal characteristics to predict emotion, inflating accuracy relative to a true speaker-independent evaluation. Consequently, the claimed speaker-independent accuracies of 84.8% (IEMOCAP) and 90.1% (EMO-DB) in Tables II and IV do not establish generalization to unseen speakers. The authors should re-run the experiments with speaker-disjoint splits, for example leave-one-speaker-out or grouped k-fold by speaker, and report results under that protocol.
  2. [Tables II and IV] The comparison with state-of-the-art methods is not controlled across protocols. The manuscript does not state whether each cited method in Tables II and IV used a speaker-independent, speaker-dependent, or random split, nor does it specify the exact folds or cross-validation procedures for those methods. On small datasets like IEMOCAP and EMO-DB, protocol differences can substantially change reported performance. For example, if some compared methods used a more difficult speaker-disjoint protocol while SigWavNet used a random split, the relative ranking in these tables would be misleading. The authors should compare all methods under an identical speaker-disjoint protocol, or at minimum transparently report the evaluation protocol of each compared work and discuss the compatibility of the numbers.
  3. [Section IV-B and Tables I-IV] The reported gains lack measures of variability. The text says each test is repeated 10 times with different random seeds, but Tables I-IV report only a single accuracy and F1-score with no standard deviation, confidence interval, or per-fold results. With only 10 speakers and random splits, the variance across seeds could be large, and the difference between 84.8% and the next-best IEMOCAP result of 80.8% may not be statistically significant. Reporting mean and standard deviation across the 10 repetitions, or per-fold results under a speaker-disjoint split, is needed to support the claim of outperforming prior methods.
minor comments (5)
  1. [Section III-C] The phrase "dilated 1D dilated convolutional layers" is redundant; it should be "dilated 1D convolutional layers".
  2. [Equation (16)] The condition "α · β < 0" appears inconsistent with the statement that to emulate linear activation the biases can be set to zero "while allowing α and β to be real numbers"; if α and β are both positive, their product is positive, violating the condition. Please clarify the intended range of α and β.
  3. [Section IV-A.1] The sentence "totaling 10039 utterances each" is unclear; IEMOCAP has 10,039 utterances in total, not per speaker or per session. Please rephrase.
  4. [Figure 7] The caption "Daubechies Wavelets kernel size" does not explain what is plotted (e.g., the scaling and wavelet function, or the filter coefficients). Clarify the figure and its relation to the initialization of Conv_h and Conv_g.
  5. [References] Reference [2] is a bare URL for a CDC page; it should be formatted consistently with the other references (author, title, year, and access date).

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: SigWavNet is trained and evaluated on external benchmarks; the sole self-citation is motivational, not load-bearing.

full rationale

This is an empirical deep-learning paper. The reported accuracies (84.8% on IEMOCAP, 90.1% on EMO-DB, Tables II and IV) come from training a concrete network on public corpora and testing on a held-out 10% partition (Section IV-B). No reported number is mathematically defined as the value of a fitted parameter or as an equation whose input already contains the target result. The learnable FDWT filters (Eqs. 12-14) are initialized from Daubechies coefficients and updated by supervised loss; LAHT (Eq. 16) is a new parameterized activation proposed in this paper, so no ansatz is smuggled in via citation. The CQF relation (Eq. 14) is standard filter-bank mathematics, not a uniqueness claim from the authors. The only self-citation is reference [1], used in the introduction to motivate emergency-call and mental-health applications; it does not support any architectural, comparative, or predictive claim. The main caveat — Section IV-B describes a stratified random split rather than a speaker-disjoint split, which weakens the 'speaker-independent' label on these 10-speaker datasets — is an evaluation-protocol or validity concern, not circularity: the numbers would be inflated or non-comparable, but they are not forced to equal their own inputs by construction. The ablation study selects among the paper's own configurations on the same benchmark and is not a derivation from the definitions of the components. No circular step is identified.

Assumptions & free parameters 7 free parameters · 4 assumptions · 0 invented entities

The tuned hyperparameters and architecture choices are listed as free parameters because their values are selected by grid search or by observation and mostly not reported. The domain assumptions about emotional cues and wavelet sparsity are standard for this kind of model. The questionable assumption that a random stratified split is speaker-independent is the most consequential entry, since it undermines the central claim. No new physical entities are introduced.

free parameters (7)
  • Focal loss class-balance weight alpha_t = not reported
    Appears in eq. (19); chosen via ASHA grid search (Section IV-B), value not given in the paper.
  • Focal loss focusing exponent gamma = not reported
    Appears in eq. (19); tuned by grid search, not reported.
  • L2 regularization strength lambda = not reported
    Appears in eq. (20); tuned by grid search, not reported.
  • FDWT decomposition levels L = 8
    Section IV-D selects an 8-level LFDWTB 'based on the performance observed in our experiments'.
  • Wavelet kernel size = 20 (Daubechies-10)
    Section IV-D: 'an 8-LFDWTB with a kernel size of 20, emulating Daubechies-10 wavelets'.
  • Number of dilated CNN layers = 3
    Section IV-D: 'the subsequent dilated CNN module consists of 3 convolutional layers'.
  • Number of Bi-GRU layers = 6
    Section IV-D: 'the third module, which employs a Bi-GRU, incorporates 6 layers'.
assumptions (4)
  • domain assumption Emotional information in speech spans long durations and is not confined to fixed segments
    Stated in Section III-A as motivation for processing full utterances without segmentation, citing physiological/psychological research [4].
  • domain assumption A random stratified split by emotion class is a valid speaker-independent evaluation
    Assumed in Section IV-B, which claims SI experiments while describing only stratified random sampling; this is likely false for 10-speaker datasets.
  • standard math Wavelet decomposition of natural signals is sparse, and noise produces small-magnitude coefficients
    The denoising rationale in Section III-B2, a standard wavelet denoising assumption from [47].
  • standard math Filter pairs satisfying the conjugate quadrature filter relation form an orthonormal wavelet basis with perfect reconstruction
    Invoked in Section III-C1 eq. (14) to derive Conv g from Conv h; standard wavelet theory result [48].

how reviews work

0 comments
Cite this review

Pith. "Pith review of SigWavNet: Learning Multiresolution Signal Wavelet Network for Speech Emotion Recognition." pith.science (2026). https://pith.science/paper/ZPEIKTOX

@misc{pith2026250200310,
  author       = {Pith},
  title        = {Pith review of: SigWavNet: Learning Multiresolution Signal Wavelet Network for Speech Emotion Recognition},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZPEIKTOX}},
  note         = {Machine review of arXiv:2502.00310}
}
read the original abstract

In the field of human-computer interaction and psychological assessment, speech emotion recognition (SER) plays an important role in deciphering emotional states from speech signals. Despite advancements, challenges persist due to system complexity, feature distinctiveness issues, and noise interference. This paper introduces a new end-to-end (E2E) deep learning multi-resolution framework for SER, addressing these limitations by extracting meaningful representations directly from raw waveform speech signals. By leveraging the properties of the fast discrete wavelet transform (FDWT), including the cascade algorithm, conjugate quadrature filter, and coefficient denoising, our approach introduces a learnable model for both wavelet bases and denoising through deep learning techniques. The framework incorporates an activation function for learnable asymmetric hard thresholding of wavelet coefficients. Our approach exploits the capabilities of wavelets for effective localization in both time and frequency domains. We then combine one-dimensional dilated convolutional neural networks (1D dilated CNN) with a spatial attention layer and bidirectional gated recurrent units (Bi-GRU) with a temporal attention layer to efficiently capture the nuanced spatial and temporal characteristics of emotional features. By handling variable-length speech without segmentation and eliminating the need for pre or post-processing, the proposed model outperformed state-of-the-art methods on IEMOCAP and EMO-DB datasets. The source code of this paper is shared on the Github repository: https://github.com/alaaNfissi/SigWavNet-Learning-Multiresolution-Signal-Wavelet-Network-for-Speech-Emotion-Recognition.

Figures

Figures reproduced from arXiv: 2502.00310 by the authors.

Figure 1
Figure 1. Traditional block diagram of wavelet filter analysis [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. L-LFDWTB SigWavNet General Architecture 1) Learnable FDWT: In our implementation, we introduce a series of convolutional layers, Convh and Convg, that func￾tion as learnable scaling (low-pass) and wavelet (high-pass) filters at each decomposition level. These layers facilitate the automatic extraction of meaningful and sparse representations from the raw speech waveform, tailored to the nuances of SER. To enhance th… view at source ↗
Figure 3
Figure 3. SigWavNet Three Levels Learnable Fast Discrete Wavelet Transform Block (3-LFDWTB) [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Learnable Asymmetric Hard Threshold (LAHT) Function [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Architecture of a 1D Dilated CNN Enhanced with Spatial Attention, Followed by a Bi-GRU and Temporal Attention Mechanism. This diagram [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: 1D Global Average Pooling The superiority of Log Softmax lies in its enhanced numerical stability and computational efficiency. This function effectively addresses numerical instabilities that can arise from exponenti￾ating large or small input values, which is a commo…
Figure 7
Figure 7. Figure 7: Daubechies Wavelets kernel size The initial weights of our model are set randomly, as it does not rely on any pre-trained models. We evaluate our model’s performance using speaker-independent (SI) experiments on two publicly available datasets, IEMOCAP and EMO-DB. In t…
Figure 8
Figure 8. Figure 8: SigWavNet confusion matrix on IEMOCAP [PITH_FULL_IMAGE:figures/full_fig_p011_8.png]
Figure 9
Figure 9. Figure 9: presents a confusion matrix that details SigWavNet’s ability to identify various emotions in the EMO-DB dataset. SigWavNet shows good performance in detecting ’Anger’, with a high correct recognition rate of 92.31%. The model has minimal errors in this category, miside…
Figure 10
Figure 10. Figure 10: Ablation study configurations and performance [PITH_FULL_IMAGE:figures/full_fig_p014_10.png]
Figure 11
Figure 11. Figure 11: Low- and High-pass kernels of the first 3 levels after training [PITH_FULL_IMAGE:figures/full_fig_p014_11.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

83 extracted references · 69 canonical work pages

  1. [55]

    A study of stratified random sampling,

    H. Aoyama, “A study of stratified random sampling,” Ann. Inst. Stat. Math, vol. 6, no. 1, pp. 1–36, 1954. 18

    work page 1954

  2. [1]

    Unlocking the emotional states of high-risk suicide callers through speech analysis,

    A. Nfissi, W. Bouachir, N. Bouguila, and B. Mishara, “Unlocking the emotional states of high-risk suicide callers through speech analysis,” in 2024 IEEE 18th International Conference on Semantic Computing (ICSC). IEEE Computer Society, 2024, pp. 33–40

    2024

  3. [2]

    Suicide prevention (2024),

    “Suicide prevention (2024),” centers for Disease Control and Preven- tion. Available at: https://www.cdc.gov/suicide/ (Accessed: February 10, 2024)

    2024

  4. [3]

    Speech emotion recognition and sentiment analysis based thera- pist bot,

    Y . Bhangdia, R. Bhansali, N. Chaudhari, D. Chandnani, and M. L. Dhore, “Speech emotion recognition and sentiment analysis based thera- pist bot,” in 2021 Third International Conference on Inventive Research in Computing Applications (ICIRCA) , 2021, pp. 96–101

    2021

  5. [4]

    Development of speech emotion recognition system based on discrete wavelet transform (dwt) and voice segmentation,

    H. Nugroho and R. N. N. Fuadiyah, “Development of speech emotion recognition system based on discrete wavelet transform (dwt) and voice segmentation,” International Journal on Electrical Engineering and Informatics, vol. 14, no. 3, pp. 593–607, 2022

    work page 2022

  6. [5]

    Survey on speech emotion recognition: Features, classification schemes, and databases,

    M. El Ayadi, M. S. Kamel, and F. Karray, “Survey on speech emotion recognition: Features, classification schemes, and databases,” Pattern recognition, vol. 44, no. 3, pp. 572–587, 2011

    work page 2011

  7. [6]

    Missing-class-robust domain adaptation by unilateral alignment,

    Q. Wang, G. Michau, and O. Fink, “Missing-class-robust domain adaptation by unilateral alignment,” IEEE Transactions on Industrial Electronics, vol. 68, no. 1, pp. 663–671, 2020. 17

    work page 2020

  8. [7]

    A deep learning driven method for fault classification and degradation assessment in mechanical equipment,

    Z. Li, Y . Wang, and K. Wang, “A deep learning driven method for fault classification and degradation assessment in mechanical equipment,” Computers in industry , vol. 104, pp. 1–10, 2019

    work page 2019

Show all 83 references
  1. [8]

    Potential, challenges and future directions for deep learning in prog- nostics and health management applications,

    O. Fink, Q. Wang, M. Svensen, P. Dersin, W.-J. Lee, and M. Ducoffe, “Potential, challenges and future directions for deep learning in prog- nostics and health management applications,” Engineering Applications of Artificial Intelligence , vol. 92, p. 103678, 2020

    2020

  2. [9]

    Unsupervised transfer learning for anomaly detection: Application to complementary operating condition transfer,

    G. Michau and O. Fink, “Unsupervised transfer learning for anomaly detection: Application to complementary operating condition transfer,” Knowledge-Based Systems, vol. 216, p. 106816, 2021

    2021

  3. [10]

    A survey of speech emotion recognition in natural environment,

    M. S. Fahad, A. Ranjan, J. Yadav, and A. Deepak, “A survey of speech emotion recognition in natural environment,” Digital signal processing , vol. 110, p. 102951, 2021

    2021

  4. [11]

    Research on speech emotion recognition based on the fractional fourier transform,

    L. Huang and X. Shen, “Research on speech emotion recognition based on the fractional fourier transform,” Electronics, vol. 11, no. 20, p. 3393, 2022

    2022

  5. [12]

    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

  6. [13]

    A deep convolutional neural network with new training methods for bearing fault diagnosis under noisy environment and different working load,

    W. Zhang, C. Li, G. Peng, Y . Chen, and Z. Zhang, “A deep convolutional neural network with new training methods for bearing fault diagnosis under noisy environment and different working load,” Mechanical sys- tems and signal processing , vol. 100, pp. 439–453, 2018

    2018

  7. [14]

    Convolutional neural networks analyzed via convolutional sparse coding,

    V . Papyan, Y . Romano, and M. Elad, “Convolutional neural networks analyzed via convolutional sparse coding,” The Journal of Machine Learning Research, vol. 18, no. 1, pp. 2887–2938, 2017

    2017

  8. [15]

    Fourier neural networks: A comparative study,

    M. Uteuliyeva, A. Zhumekenov, R. Takhanov, Z. Assylbekov, A. J. Castro, and O. Kabdolov, “Fourier neural networks: A comparative study,” Intelligent Data Analysis , vol. 24, no. 5, pp. 1107–1120, 2020

    2020

  9. [16]

    Waveletkernelnet: An interpretable deep neural network for industrial intelligent diagnosis,

    T. Li, Z. Zhao, C. Sun, L. Cheng, X. Chen, R. Yan, and R. X. Gao, “Waveletkernelnet: An interpretable deep neural network for industrial intelligent diagnosis,” IEEE Transactions on Systems, Man, and Cyber- netics: Systems, vol. 52, no. 4, pp. 2302–2312, 2021

    2021

  10. [17]

    Multi-level wavelet convolu- tional neural networks,

    P. Liu, H. Zhang, W. Lian, and W. Zuo, “Multi-level wavelet convolu- tional neural networks,” IEEE Access, vol. 7, pp. 74 973–74 985, 2019

    2019

  11. [18]

    An end-to-end multi-level wavelet convolu- tional neural networks for heart diseases diagnosis,

    M. Khalil, A. Adib et al., “An end-to-end multi-level wavelet convolu- tional neural networks for heart diseases diagnosis,” Neurocomputing, vol. 417, pp. 187–201, 2020

    2020

  12. [19]

    Speech emotion recognition using local and global features,

    Y . Gao, B. Li, N. Wang, and T. Zhu, “Speech emotion recognition using local and global features,” in Brain Informatics: International Confer- ence, BI 2017, Beijing, China, November 16-18, 2017, Proceedings . Springer, 2017, pp. 3–13

    2017

  13. [20]

    Open-source media interpretation by large feature-space extraction,

    F. Eyben, F. Weninger, M. W ¨ollmer, and B. Shuller, “Open-source media interpretation by large feature-space extraction,” TU Munchen, MMK , 2016

    2016

  14. [21]

    Speech emotion recognition using deep convolutional neural network and discriminant temporal pyramid matching,

    S. Zhang, S. Zhang, T. Huang, and W. Gao, “Speech emotion recognition using deep convolutional neural network and discriminant temporal pyramid matching,” IEEE Transactions on Multimedia , vol. 20, no. 6, pp. 1576–1590, 2017

    2017

  15. [22]

    Speech emotion recognition using deep 1d & 2d cnn lstm networks,

    J. Zhao, X. Mao, and L. Chen, “Speech emotion recognition using deep 1d & 2d cnn lstm networks,” Biomedical signal processing and control , vol. 47, pp. 312–323, 2019

    2019

  16. [23]

    Hidden markov model-based speech emotion recognition,

    B. Schuller, G. Rigoll, and M. Lang, “Hidden markov model-based speech emotion recognition,” in 2003 IEEE International Confer- ence on Acoustics, Speech, and Signal Processing, 2003. Proceed- ings.(ICASSP’03)., vol. 2. Ieee, 2003, pp. II–1

    2003

  17. [24]

    Deep learning is robust to massive label noise,

    D. Rolnick, A. Veit, S. Belongie, and N. Shavit, “Deep learning is robust to massive label noise,” arXiv preprint arXiv:1705.10694 , 2017

    2017 arXiv

  18. [25]

    A cnn-assisted enhanced audio signal pro- cessing for speech emotion recognition,

    Mustaqeem and S. Kwon, “A cnn-assisted enhanced audio signal pro- cessing for speech emotion recognition,” Sensors, vol. 20, no. 1, p. 183, 2019

    2019

  19. [26]

    Recursive interferometric representation,

    S. Mallat, “Recursive interferometric representation,” in Proc. of EU- SICO conference, Danemark , vol. 3, 2010

    2010

  20. [27]

    Group invariant scattering,

    ——, “Group invariant scattering,” Communications on Pure and Ap- plied Mathematics, vol. 65, no. 10, pp. 1331–1398, 2012

    2012

  21. [28]

    Deep scattering spectrum,

    J. And ´en and S. Mallat, “Deep scattering spectrum,” IEEE Transactions on Signal Processing , vol. 62, no. 16, pp. 4114–4128, 2014

    2014

  22. [29]

    Joint time–frequency scattering,

    J. And ´en, V . Lostanlen, and S. Mallat, “Joint time–frequency scattering,” IEEE Transactions on Signal Processing, vol. 67, no. 14, pp. 3704–3718, 2019

    2019

  23. [30]

    Hybrid network for end-to- end text-independent speaker identification,

    W. Ghezaiel, L. Brun, and O. L ´ezoray, “Hybrid network for end-to- end text-independent speaker identification,” in 2020 25th International conference on pattern recognition (ICPR). IEEE, 2021, pp. 2352–2359

    2020

  24. [31]

    Discriminative wavelet packet filter bank selection for pattern recognition,

    J. Silva and S. S. Narayanan, “Discriminative wavelet packet filter bank selection for pattern recognition,” IEEE Transactions on Signal Processing, vol. 57, no. 5, pp. 1796–1810, 2009

    2009

  25. [32]

    Discrete wavelet transforms,

    K. D. Rao, M. Swamy, K. D. Rao, and M. Swamy, “Discrete wavelet transforms,” Digital Signal Processing: Theory and Practice , pp. 619– 691, 2018

    2018

  26. [33]

    The wavelet transform, time-frequency localization and signal analysis,

    I. Daubechies, “The wavelet transform, time-frequency localization and signal analysis,” IEEE transactions on information theory, vol. 36, no. 5, pp. 961–1005, 1990

    1990

  27. [34]

    Time-frequency feature and ams- gmm mask for acoustic emotion classification,

    L. Z ˜ao, D. Cavalcante, and R. Coelho, “Time-frequency feature and ams- gmm mask for acoustic emotion classification,” IEEE signal processing letters, vol. 21, no. 5, pp. 620–624, 2014

    2014

  28. [35]

    Particle swarm optimization based feature enhancement and feature selection for improved emotion recognition in speech and glottal signals,

    H. Muthusamy, K. Polat, and S. Yaacob, “Particle swarm optimization based feature enhancement and feature selection for improved emotion recognition in speech and glottal signals,” PloS one , vol. 10, no. 3, p. e0120344, 2015

    2015

  29. [36]

    Effectiveness of tuned q-factor wavelet trans- form in emotion recognition among left-brain damaged stroke patients,

    B. S. Zheng, W. Khairunizam, S. Murugappan Murugappan, Z. Razlan, I. Zunaidi, and C. Yean, “Effectiveness of tuned q-factor wavelet trans- form in emotion recognition among left-brain damaged stroke patients,” Int J Simul Syst Sci Technol , vol. 19, no. 3, p. 2, 2018

    2018

  30. [37]

    Emotion recognition in human-computer interaction,

    R. Cowie, E. Douglas-Cowie, N. Tsapatsoulis, G. V otsis, S. Kollias, W. Fellenz, and J. G. Taylor, “Emotion recognition in human-computer interaction,” IEEE Signal processing magazine, vol. 18, no. 1, pp. 32–80, 2001

    2001

  31. [38]

    Speech emotion recognition: Comparison of speech segmentation approaches,

    M. Mansoorizadeh and N. M. Charkari, “Speech emotion recognition: Comparison of speech segmentation approaches,” Proceedings of IKT, Mashad, Iran, 2007

    2007

  32. [39]

    Wavelet theory and applica- tions: a literature study,

    M. Steinbuch and M. Van de Molengraft, “Wavelet theory and applica- tions: a literature study,” Eindhoven: Eindhoven University Technology Department of Mechanical Engineering Control System Group , 2005

    2005

  33. [40]

    Wavelet based feature combination for recognition of emotions,

    H. K. Palo and M. N. Mohanty, “Wavelet based feature combination for recognition of emotions,” Ain shams engineering journal , vol. 9, no. 4, pp. 1799–1806, 2018

    2018

  34. [41]

    Egyptian arabic speech emotion recognition using prosodic, spectral and wavelet features,

    L. Abdel-Hamid, “Egyptian arabic speech emotion recognition using prosodic, spectral and wavelet features,” Speech Communication , vol. 122, pp. 19–30, 2020

    2020

  35. [42]

    Wavelet packet analysis for speaker-independent emotion recognition,

    K. Wang, G. Su, L. Liu, and S. Wang, “Wavelet packet analysis for speaker-independent emotion recognition,” Neurocomputing, vol. 398, pp. 257–264, 2020

    2020

  36. [43]

    Emotion recognition in speech using mfcc and wavelet features,

    K. K. Kishore and P. K. Satish, “Emotion recognition in speech using mfcc and wavelet features,” in 2013 3rd IEEE International Advance Computing Conference (IACC) . IEEE, 2013, pp. 842–847

    2013

  37. [44]

    Feature fusion methods research based on deep belief networks for speech emotion recognition under noise condition,

    Y . Huang, K. Tian, A. Wu, and G. Zhang, “Feature fusion methods research based on deep belief networks for speech emotion recognition under noise condition,” Journal of ambient intelligence and humanized computing, vol. 10, pp. 1787–1798, 2019

    2019

  38. [45]

    Emotion recognition from speech under environmental noise conditions using wavelet de- composition,

    J. C. V ´asquez-Correa, N. Garc ´ıa, J. R. Orozco-Arroyave, J. D. Arias- Londo˜no, J. F. Vargas-Bonilla, and E. N ¨oth, “Emotion recognition from speech under environmental noise conditions using wavelet de- composition,” in 2015 International Carnahan Conference on Security T...

    2015

  39. [46]

    Heisenberg’s uncertainty princi- ple,

    P. Busch, T. Heinonen, and P. Lahti, “Heisenberg’s uncertainty princi- ple,” Physics reports, vol. 452, no. 6, pp. 155–176, 2007

    2007

  40. [47]

    Mallat, A Wavelet Tour of Signal Processing: The Sparse Way

    S. Mallat, A Wavelet Tour of Signal Processing: The Sparse Way . Elsevier Science, 2008. [Online]. Available: https://books.google.ca/ books?id=5qzeLJljuLoC

    2008

  41. [48]

    Perfect channel splitting by use of interpola- tion/decimation/tree decomposition techniques,

    A. Croisier, “Perfect channel splitting by use of interpola- tion/decimation/tree decomposition techniques,” in Proc. Int. Symp. on Info., Circuits and Systems,(Patras, Greece) , 1976

    1976

  42. [49]

    Network in network,

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

    2013 arXiv

  43. [50]

    A systematic literature review of speech emotion recognition approaches,

    Y . B. Singh and S. Goel, “A systematic literature review of speech emotion recognition approaches,” Neurocomputing, 2022

    2022

  44. [51]

    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

  45. [52]

    Speech emotion recognition with acoustic and lexical features,

    Q. Jin, C. Li, S. Chen, and H. Wu, “Speech emotion recognition with acoustic and lexical features,” in 2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , 2015, pp. 4749– 4753

    2015

  46. [53]

    Deep learning for robust feature generation in audiovisual emotion recognition,

    Y . Kim, H. Lee, and E. M. Provost, “Deep learning for robust feature generation in audiovisual emotion recognition,” in 2013 IEEE interna- tional conference on acoustics, speech and signal processing . IEEE, 2013, pp. 3687–3691

    2013

  47. [54]

    A database of german emotional speech

    F. Burkhardt, A. Paeschke, M. Rolfes, W. F. Sendlmeier, B. Weiss et al., “A database of german emotional speech.” in Interspeech, vol. 5, 2005, pp. 1517–1520

    2005

  48. [56]

    Massively parallel hyperparameter tuning,

    L. Li, K. Jamieson, A. Rostamizadeh, E. Gonina, M. Hardt, B. Recht, and A. Talwalkar, “Massively parallel hyperparameter tuning,” 2018

    2018

  49. [57]

    A system for massively parallel hyperpa- rameter tuning,

    L. Li, K. Jamieson, A. Rostamizadeh, E. Gonina, J. Ben-Tzur, M. Hardt, B. Recht, and A. Talwalkar, “A system for massively parallel hyperpa- rameter tuning,” Proceedings of Machine Learning and Systems , vol. 2, pp. 230–246, 2020

    2020

  50. [58]

    Focal loss for dense object detection,

    T.-Y . Lin, P. Goyal, R. Girshick, K. He, and P. Doll ´ar, “Focal loss for dense object detection,” in Proceedings of the IEEE international conference on computer vision , 2017, pp. 2980–2988

    2017

  51. [59]

    Adam: A method for stochastic optimization,

    D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,”

  52. [60]

    Light- sernet: A lightweight fully convolutional neural network for speech emo- tion recognition,

    A. Aftab, A. Morsali, S. Ghaemmaghami, and B. Champagne, “Light- sernet: A lightweight fully convolutional neural network for speech emo- tion recognition,” in ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2022, pp. ...

    2022

  53. [61]

    Speech emotion recognition with deep convolutional neural networks,

    D. Issa, M. F. Demirci, and A. Yazici, “Speech emotion recognition with deep convolutional neural networks,” Biomedical Signal Processing and Control, vol. 59, p. 101894, 2020

    2020

  54. [62]

    3-d convolutional recurrent neural networks with attention model for speech emotion recognition,

    M. Chen, X. He, J. Yang, and H. Zhang, “3-d convolutional recurrent neural networks with attention model for speech emotion recognition,” IEEE Signal Processing Letters , vol. 25, no. 10, pp. 1440–1444, 2018

    2018

  55. [63]

    Speech emotion recognition from 3d log-mel spectrograms with deep learning network,

    H. Meng, T. Yan, F. Yuan, and H. Wei, “Speech emotion recognition from 3d log-mel spectrograms with deep learning network,” IEEE access, vol. 7, pp. 125 868–125 881, 2019

    2019

  56. [64]

    Improved end-to-end speech emotion recognition using self attention mechanism and multitask learning

    Y . Li, T. Zhao, and T. Kawahara, “Improved end-to-end speech emotion recognition using self attention mechanism and multitask learning.” in Interspeech, 2019, pp. 2803–2807

    2019

  57. [65]

    Deep scattering network for speech emotion recognition,

    P. Singh, G. Saha, and M. Sahidullah, “Deep scattering network for speech emotion recognition,” in 2021 29th European Signal Processing Conference (EUSIPCO). IEEE, 2021, pp. 131–135

    2021

  58. [66]

    Speech emotion recognition using wavelet packet reconstruction with attention-based deep recurrent neutral networks,

    H. Meng, T. Yan, H. Wei, and X. Ji, “Speech emotion recognition using wavelet packet reconstruction with attention-based deep recurrent neutral networks,” Bulletin of the Polish Academy of Sciences. Technical Sciences, vol. 69, no. 1, 2021

    2021

  59. [67]

    Mfcc and machine learning based speech emotion recog- nition over tess and iemocap datasets,

    M. Z. Iqbal, “Mfcc and machine learning based speech emotion recog- nition over tess and iemocap datasets,” 2020

    2020

  60. [68]

    A cross-corpus experiment in speech emotion recognition

    C. Parlak, B. Diri, and F. G ¨urgen, “A cross-corpus experiment in speech emotion recognition.” in SLAM@ INTERSPEECH, 2014, pp. 58–61

    2014

  61. [69]

    Emotion recognition using speech data with convolutional neural network,

    M. H. Pham, F. M. Noori, and J. Torresen, “Emotion recognition using speech data with convolutional neural network,” in 2021 IEEE 2nd International Conference on Signal, Control and Communication (SCC). IEEE, 2021, pp. 182–187

    2021

  62. [70]

    Improved speech emotion recognition with mel frequency magnitude coefficient,

    J. Ancilin and A. Milton, “Improved speech emotion recognition with mel frequency magnitude coefficient,” Applied Acoustics , vol. 179, p. 108046, 2021

    2021

  63. [71]

    Speech emotion recognition using deep convolutional neural networks improved by the fast continuous wavelet transform,

    B. E. Van Zwol, M. A. Langezaal, L. Arts, A. Gatt, and E. L. Van Den Broek, “Speech emotion recognition using deep convolutional neural networks improved by the fast continuous wavelet transform,” in Workshop Proceedings of the 19th International Conference on Intelligent Envi...

    2023

  64. [72]

    Automated accurate speech emotion recognition system using twine shuffle pattern and iterative neighborhood component analysis techniques,

    T. Tuncer, S. Dogan, and U. R. Acharya, “Automated accurate speech emotion recognition system using twine shuffle pattern and iterative neighborhood component analysis techniques,” Knowledge-Based Sys- tems, vol. 211, p. 106547, 2021

    2021

  65. [73]

    The amalgamation of wavelet packet information gain entropy tuned source and system parameters for improved speech emotion recognition,

    H. K. Palo, S. Subudhiray, and N. Das, “The amalgamation of wavelet packet information gain entropy tuned source and system parameters for improved speech emotion recognition,” Speech Communication, vol. 149, pp. 11–28, 2023

    2023

  66. [74]

    Speech emotion recognition system based on wavelet transform and multi-task learning,

    Y . Liu and Z. Kexin, “Speech emotion recognition system based on wavelet transform and multi-task learning,” in 2022 7th International Conference on Intelligent Informatics and Biomedical Science (ICI- IBMS), vol. 7. IEEE, 2022, pp. 141–149

    2022

  67. [75]

    The effects of normalisation methods on speech emotion recognition,

    T. J. Sefara, “The effects of normalisation methods on speech emotion recognition,” in 2019 International multidisciplinary information tech- nology and engineering conference (IMITEC) . IEEE, 2019, pp. 1–8

    2019

  68. [76]

    Empirical evaluation of rectified activations in convolutional network,

    B. Xu, N. Wang, T. Chen, and M. Li, “Empirical evaluation of rectified activations in convolutional network,” arXiv preprint arXiv:1505.00853, 2015

    2015 arXiv

  69. [77]

    Rectifier nonlinearities improve neural network acoustic models,

    A. L. Maas, A. Y . Hannun, A. Y . Ng et al. , “Rectifier nonlinearities improve neural network acoustic models,” in Proc. icml, vol. 30, no. 1. Citeseer, 2013, p. 3

    2013

  70. [78]

    Learn to pay attention,

    S. Jetley, N. A. Lord, N. Lee, and P. H. Torr, “Learn to pay attention,” arXiv preprint arXiv:1804.02391 , 2018

    2018 arXiv

  71. [79]

    Hadamard products and multivariate statistical analysis,

    G. P. Styan, “Hadamard products and multivariate statistical analysis,” Linear algebra and its applications , vol. 6, pp. 217–240, 1973

    1973

  72. [80]

    Empirical evaluation of gated recurrent neural networks on sequence modeling,

    J. Chung, C. Gulcehre, K. Cho, and Y . Bengio, “Empirical evaluation of gated recurrent neural networks on sequence modeling,” arXiv preprint arXiv:1412.3555, 2014. Alaa Nfissi received an engineering degree in com- puter science and data science from the Private High School o...

    2014 arXiv

  73. [83]

    He is the author or co-author of over 500 publications in several prestigious journals and conferences

    He is currently the holder of a Concordia University Research Chair Tier 1 in Applied Artificial Intelligence. He is the author or co-author of over 500 publications in several prestigious journals and conferences. He is a regular reviewer for many international journals and s...

  74. [2014]

    Available: https://arxiv.org/abs/1412.6980

    [Online]. Available: https://arxiv.org/abs/1412.6980

    arXiv

  75. [2020]

    Currently, he is pursuing a Ph.D

    He then earned his MSc degree in advanced engineering, internet of things, and data processing, from the Polytechnic School of Tunisia in 2021. Currently, he is pursuing a Ph.D. at Concordia University, Montreal, Quebec, Canada. His research in artificial intelligence includes...

    2021

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.