Pith. sign in

REVIEW 2 major objections 5 minor 56 references

Cross-individual Recognition of Emotions by a Dynamic Entropy based on Pattern Learning with EEG features

T0 review · 2 major / 5 minor · reviewed 2026-08-27 · deepseek-v4-flash

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

Pith's one-line read Deep learning reads emotions across people from EEG

desk verdict Plausible 7-10 point cross-subject accuracy gain on two public EEG datasets, but the evaluation protocol leaks test information at two points; worth reviewing, not worth believing as stated. read the letter →

arxiv 2009.12525 v2 pith:ONC6MZTL submitted 2020-09-26 cs.LG eess.SP

Xiaolong Zhong , Zhong Yin This is my paper · ORCID
classification cs.LGeess.SP
keywords cross-subjectemotionrecognitionEEGdifferentialentropyconvolutionalneuralnetworksqueeze-excitationleave-one-subject-outDEAPMAHNOB-HCI
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 seeks to establish that a person's emotional state can be read from the EEG of people the classifier has never met, without per-user calibration. The proposed pipeline, DEPL, converts EEG into differential entropy features per frequency band, smooths those features over time, maps them onto a 2-D layout of electrode positions, and trains a small convolutional network with squeeze-excitation channel recalibration to learn cross-subject patterns. Under leave-one-subject-out testing on DEAP and MAHNOB-HCI, the paper reports participant-averaged accuracies of 66.23% for valence and 68.50% for arousal on DEAP, and 70.25% for valence and 73.27% for arousal on MAHNOB-HCI, significantly above classical shallow classifiers with p<0.001. The claim is that modeling interdependencies between cortical locations is what enables the cross-individual transfer, and that compact networks are sufficient. If true, this would move affective brain-computer interfaces toward systems that adapt to a new user's emotions from the first session.

What carries the argument

The load-bearing object is the DEPL framework, a four-stage pipeline. It first computes differential entropy, a closed-form entropy measure for band-passed EEG treated as Gaussian, separately in the theta, alpha, beta, and gamma bands, and subtracts a baseline estimate. It then smooths the per-second entropy values by averaging over lagged time steps, an operation that acts as a low-pass filter on the feature trajectory and is what makes the features dynamic. The smoothed features are embedded into a 9 by 9 grid that preserves electrode topology, and a two-convolution-layer network with max pooling, squeeze-excitation blocks, Swish activations, dropout, batch normalization, and L2 regularization is trained to classify low versus high valence and arousal. The squeeze-excitation block is the channel-recalibration mechanism that models interdependencies between cortical locations.

What would settle it

Re-run the leave-one-subject-out protocol twice: once with z-score means and variances computed only from the training folds, and once with them computed from all data including the held-out subject. If accuracies are nearly identical, the normalization is clean; if the second run scores higher, the reported cross-individual accuracy gap is partly an artifact of test-subject statistics leaking into feature normalization.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central claim is that DEPL outperforms both classical shallow learning machines and modern deep models on cross-individual emotion recognition while using the fewest trainable parameters. The framework's advantage is attributed to three design components: temporal smoothing of differential entropy features to remove short-term fluctuations, 2-D mapping that preserves adjacency among electrodes, and the squeeze-excitation block that reweights the importance of each channel. The gamma band is shown to be the most informative frequency band on both databases, and the improvements over shallow baselines are statistically significant. The authors present this as evidence that a dynamic entropy-based representation of EEG, rather than raw signals or hand-crafted static features, is a suitable substrate for individual-independent affective computing.

Load-bearing premise

The leave-one-subject-out comparison assumes the z-score normalization is fitted only on training subjects' data; the paper does not state this explicitly, and if test-subject statistics enter the normalization, the reported accuracy gaps are optimistically biased.

Editorial extensions

If this is right

  • A pretrained DEPL model could label a new user's valence and arousal from EEG alone, without any labeled data from that person, at accuracy around 66 to 73 percent.
  • Gamma-band activity is the carrier of cross-individual affective information in this representation; designs that discard high-frequency EEG risk losing the most transferable signal.
  • Channel-dependency modeling via squeeze-excitation adds a measurable gain over plain CNNs and smoothed features, since the DEPL variant with both components scores highest.
  • Compact two-convolution-block networks are sufficient; deeper modern architectures overfit on the limited EEG epochs, so model size is not the bottleneck.
  • The reported comparison establishes a cross-individual baseline on DEAP and MAHNOB-HCI that later EEG emotion recognition work can be measured against.

Reading between the lines

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

  • Because the temporal smoothing averages over past seconds, the reported accuracy is offline; a causal, online version would need to trade smoothing length against response latency, likely lowering accuracy.
  • The authors note they did not fuse bands; a learnable weighting across theta, alpha, beta, and gamma could exceed the single-band gamma result, and this is directly testable with the same code.
  • A stricter test of cross-individual transfer would train on one database and test on another, since both use 32-channel 128 Hz EEG; the paper only evaluates within-database leave-one-subject-out.
  • If the z-score normalization is recomputed from training folds only, the reported gaps can be checked for optimistic bias; this is a reproducibility extension rather than a new method.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Request a human review

A listed scientist reviews the paper for a fee and the review publishes here regardless of verdict. See the reviewers or get listed.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 5 minor

Summary. The paper proposes a deep learning framework, Dynamic Entropy-based Pattern Learning (DEPL), for cross-individual EEG-based emotion recognition. Differential entropy (DE) features are computed for four frequency bands, smoothed over time, mapped to 2D electrode layouts, and fed into a convolutional neural network with squeeze-and-excitation blocks. The method is evaluated with leave-one-subject-out (LOSO) cross-validation on the DEAP and MAHNOB-HCI datasets, reporting participant-averaged accuracies of 66.23% and 68.50% for valence and arousal on DEAP and 70.25% and 73.27% on MAHNOB-HCI, with paired t-tests showing significant improvement over shallow learning baselines. The paper also compares several deep architectures and reports that the gamma band is optimal.

Significance. If the evaluation protocol is clean, the paper's contribution is meaningful: it demonstrates a relatively compact CNN architecture with SE blocks that outperforms classical shallow learners on cross-subject EEG emotion classification on two public benchmarks, and it provides an ablation showing the contribution of feature smoothing and attention. The use of two standard datasets, paired statistical tests, and comparisons with multiple baselines are strengths, and the low parameter count of the proposed architecture is a practical advantage. However, as reported, the central claim is currently not fully supported because the preprocessing normalization and the model/band selection procedure are not shown to be free of test-subject information.

major comments (2)
  1. [§V.A] Section V.A: The description of the z-scoring is underspecified. The text states only that 'the leave-one-subject-out paradigm with z-scored features' was used; it does not state whether the z-score mean and standard deviation are estimated from the training subjects only in each fold, or from the entire database including the held-out test subject. If the latter, the test features contain distributional information from the test subject, and the reported 'cross-individual' accuracies are optimistically biased. Please specify the normalization procedure precisely and, if necessary, re-run the experiments with a training-only z-score estimate.
  2. [§V.B] Section V.B: The optimal frequency band (gamma) and the final network architecture (DEPL versus CNN-1/2/3) are selected by comparing participant-averaged LOSO accuracies on the same test folds that are later used to produce the headline numbers in Table V and the paired t-test results in Fig. 7. The text states that 'the optimal frequency band had been examined' with the accuracies in Fig. 5(a) and that 'the current network structure possessed the highest participant average performance.' Because the test subjects' labels and predictions have already been used in the selection, the reported 66.23%/68.50%/70.25%/73.27% accuracies are the best of several configurations rather than the performance of a pre-specified model, and the p<0.001 comparisons in Fig. 7 are not confirmatory. Please use a nested evaluation scheme (e.g., an inner validation split within the training subjects) for band and architecture selection, and report the corresponding unbiased test accuracies.
minor comments (5)
  1. [§III.B] Equation (2) appears corrupted in typesetting; please fix the integral and the derivation leading to Eq. (3).
  2. [§III.B] The claim that the decomposed sub-band EEG signals are Gaussian is asserted but no Kolmogorov–Smirnov test results are reported; please provide the supporting evidence or explicitly state that Eq. (3) is used as an approximation.
  3. [§IV.E vs Table VI] Section IV.E describes the first convolutional layer as using 3×3 kernels, while Table VI lists Conv(100, 5×5) for DEPL; please reconcile this discrepancy.
  4. [Table VIII] Table VIII cites Alhagry et al. as reference [42], but reference [42] in the list is Nair and Hinton (ReLU); this citation appears to be misnumbered.
  5. [§V.C] The hyper-parameter tuning for the shallow learning baselines is described only as 'optimized in detail'; please specify the tuning procedure (e.g., internal cross-validation within the training subjects) to ensure that no test-subject information was used in the baseline comparisons.

Circularity Check

1 steps flagged · score 3.0 of 10

No derivation-level circularity, but the gamma band and DEPL configuration are selected on the same leave-one-subject-out test folds used for the headline accuracies and p-values, so the reported numbers are selection-optimized rather than independent predictions.

  1. fitted input called prediction [Section V.B, 'Comparison of Different CNN Structures'; reported in Table V and Fig. 7]
    "Before implementing the DEPL, the optimal frequency band had been examined in Fig. 5(a). For the DEAP database, the highest and the lowest accuracies were achieved by the gamma and theta band features, respectively. ... Therefore, we implemented the DEPL model with the use of gamma features."

    The band choice is made by ranking the same leave-one-subject-out participant-averaged accuracies (Fig. 5(a)) that later generate the headline numbers in Table V and the paired t-tests in Fig. 7. Because the decision rule is 'choose the band with the highest observed accuracy,' the claim that gamma is the best band is true on these data by construction of the selection rule, and the subsequent p<0.001 improvements over other bands are not independent confirmations. The reported DEPL accuracy is the maximum of four band-specific configurations evaluated on the same test participants, so it is a selected optimum rather than an unbiased prediction.

full rationale

Most of the paper is an empirical engineering study, not a derivation chain. The differential-entropy formula, moving-average smoothing, 2-D electrode mapping, CNN, and squeeze-excitation block are standard components combined into a new architecture; none of the claims reduces to the definition of another by construction. The external DEAP and MAHNOB-HCI benchmarks provide independent data, and there is no load-bearing self-citation or imported uniqueness theorem. The main caveat is evaluation-circular: Section V.B chooses the gamma band and the DEPL configuration by ranking the very leave-one-subject-out accuracies that later appear as Table V and Fig. 7, so the headline numbers and p<0.001 comparisons are selection-optimized rather than confirmatory. A related reporting gap is that Section V.A states only that 'z-scored features' were used without saying whether the z-score statistics are computed from training folds only; if whole-database statistics were used, the cross-individual claim would be weakened. These are evaluation-logic problems rather than derivation-level circularity, so the score is modest.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

DEPL is an empirical pipeline built from known components. The main free parameters are the smoothing delay d, the post-hoc gamma band choice, and standard CNN hyperparameters. The DE feature formula rests on an unshown Gaussianity assumption, and the LOSO protocol depends on a leakage-free z-scoring procedure.

free parameters (5)
  • smoothing delay d
    Eq. (5) defines the dynamic DE feature as an average over the previous d time steps, but d is never specified nor justified; it is a free parameter of the central method.
  • gamma band selection = gamma
    Section V.B selects the gamma band after comparing theta, alpha, beta, and gamma on the same LOSO test folds; the reported accuracies are conditional on this post hoc choice.
  • L2 penalty strength = 0.6
    Reported in Section V.A; no sensitivity analysis is provided.
  • dropout probability = 0.6
    Reported in Section V.A; no sensitivity analysis is provided.
  • learning rate = 1.0e-05
    Reported in Section V.A; no sensitivity analysis is provided.
assumptions (4)
  • domain assumption Band-pass filtered EEG epochs are Gaussian distributed, making Eq. (3) the correct differential entropy.
    Invoked in Section III.B through an unreported Kolmogorov-Smirnov test; this assumption is load-bearing for the entire DEPL input representation.
  • domain assumption The z-scoring used in leave-one-subject-out is computed on training folds only and does not use test-subject statistics.
    Section V.A mentions 'z-scored features' without specifying the normalization procedure; if test-subject statistics are used, the reported cross-subject accuracies are optimistically biased.
  • domain assumption The 9x9 2D electrode map with zero padding preserves the cortical adjacency structure needed for 2D convolutions.
    Section IV.B constructs the 2D maps; distorted adjacency would undermine the between-electrode dependency learning that the paper claims.
  • domain assumption Leave-one-subject-out accuracy on these datasets measures cross-individual emotion recognition ability.
    Section V.A uses LOSO; the paper does not control for the fact that 60 test epochs per video share the same label, which can inflate apparent accuracy.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Cross-individual Recognition of Emotions by a Dynamic Entropy based on Pattern Learning with EEG features." pith.science (2026). https://pith.science/paper/ONC6MZTL

@misc{pith2026200912525,
  author       = {Pith},
  title        = {Pith review of: Cross-individual Recognition of Emotions by a Dynamic Entropy based on Pattern Learning with EEG features},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ONC6MZTL}},
  note         = {Machine review of arXiv:2009.12525}
}
read the original abstract

Use of the electroencephalogram (EEG) and machine learning approaches to recognize emotions can facilitate affective human computer interactions. However, the type of EEG data constitutes an obstacle for cross-individual EEG feature modelling and classification. To address this issue, we propose a deep-learning framework denoted as a dynamic entropy-based pattern learning (DEPL) to abstract informative indicators pertaining to the neurophysiological features among multiple individuals. DEPL enhanced the capability of representations generated by a deep convolutional neural network by modelling the interdependencies between the cortical locations of dynamical entropy based features. The effectiveness of the DEPL has been validated with two public databases, commonly referred to as the DEAP and MAHNOB-HCI multimodal tagging databases. Specifically, the leave one subject out training and testing paradigm has been applied. Numerous experiments on EEG emotion recognition demonstrate that the proposed DEPL is superior to those traditional machine learning (ML) methods, and could learn between electrode dependencies w.r.t. different emotions, which is meaningful for developing the effective human-computer interaction systems by adapting to human emotions in the real world applications.

Figures

Figures reproduced from arXiv: 2009.12525 by the authors.

Figure 1
Figure 1. An illustration of mapping electroencephalographic (EEG) signals over time to differential entropy (DE) feature vectors. B. Computation of Differential Entropy Applying complexity measures with entropy-based pattern learning has been certified as one of the leading technical means for EEG-based emotion recognition [31]. Entropy measures can be exploited to quantify the nonlinearity, uncertainty, and non-stationarity… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

56 extracted references · 56 canonical work pages

  1. [1]

    Intelligent sensing systems for measuring wellness indices of the daily activities for the elderly,

    N. K. Suryadevara, M. T. Quazi, S. C. Mukhopadhyay, “Intelligent sensing systems for measuring wellness indices of the daily activities for the elderly,” 2012 Eighth Int. Conf. Intell. Environ., IEEE, 2012, pp. 347–350

    work page 2012

  2. [2]

    A novel real-time emotion detection system from audio streams based on bayesian quadratic discriminate classifier for adas ,

    F. Al Machot, A. H. Mosa, K. Dabbour, A. Fasih, C. Schwarzlmü ller, M. Ali and K. Kyamakya, “A novel real-time emotion detection system from audio streams based on bayesian quadratic discriminate classifier for adas ,” in Proc. Joint INDS'11 & ISTET'11. IEEE, 2011, pp. 1–5

    work page 2011

  3. [3]

    Psychologist in a pocket: towards depression screening on mobile phones,

    J. Á. Bitsch, R. Ramos, T.Ix, P. G. Ferrer -Cheng and K. Wehrle, “Psychologist in a pocket: towards depression screening on mobile phones,” pHealth, vol. 211, pp.153–159, 2015

    work page 2015

  4. [4]

    Processing of emotion words by patients with autism spectrum disorders: Evidence from reaction times and EEG,

    A. Lartseva, T. Dijkstra, C. C. Kan and J. K . Buitelaar, “Processing of emotion words by patients with autism spectrum disorders: Evidence from reaction times and EEG,” J. Autism Dev. Disord., vol. 44, no. 11, pp. 2882–2894, 2014

    work page 2014

  5. [5]

    Automatic measurement of affect in dimensional and contin uous spaces: Why, what, and how?

    H. Gunes, M. Pantic, “Automatic measurement of affect in dimensional and contin uous spaces: Why, what, and how?” in Proc. 7th Int. Conf. Methods Techn. Behav. Res., 2010, pp. 1–5

    work page 2010

  6. [6]

    A review on automatic facial expression recognition systems assisted by multimodal sensor data,

    N. Samadiani, G. Huang, B. Cai, W. Luo, C. H. Chi, Y. Xiang and J. He, “A review on automatic facial expression recognition systems assisted by multimodal sensor data,” Sensors, vol. 19, no. 8, 1863, 2019

    work page 2019

  7. [7]

    3D CNN -based speech emotion recognition using K -means clustering and spectrograms,

    N. Hajarolasvadi, H. Demirel, “ 3D CNN -based speech emotion recognition using K -means clustering and spectrograms,” Entropy, vol. 21, no. 5, 479, 2019

    work page 2019

  8. [8]

    Multimodal emotion recognition in response to videos,

    M. Soleymani, M. Pantic, T. Pun, “Multimodal emotion recognition in response to videos,” IEEE Trans. Affective Comput., vol. 3, no. 2, pp. 211–223, 2011

    work page 2011

Show all 56 references
  1. [9]

    Recognition of emotions using multimodal physiological signals and an ensemble deep learning model,

    Z. Yin, M. Zhao, Y. Wang, J. Yang and J. Zhang, “Recognition of emotions using multimodal physiological signals and an ensemble deep learning model,” Comput. Methods Progr. Biomed., vol. 140, pp. 93–110, 2017

    2017

  2. [10]

    MoSa: A modeling and sentiment analysis system for mobile application big data,

    Y. Zhang, W. Ren, T. Zhu and E. Faith, “MoSa: A modeling and sentiment analysis system for mobile application big data,” Symmetry, vol. 11, no. 1, 115, 2019

    2019

  3. [11]

    EEG-based automatic emotion recognition: Feature extraction, selection and classification methods ,

    P. Ackermann, C. Kohlschein, J. A. Bitsch, K. Wehrle and S. Jeschke. “EEG-based automatic emotion recognition: Feature extraction, selection and classification methods ,” 2016 IEEE 18th Int. Conf. E-health Netw., Appl. Serv. (Healthcom), IEEE, 2016, pp. 1–6. 24

    2016

  4. [12]

    A review on the computational methods for emotional state estimation from the human EEG,

    M. K. Kim, M. Kim, E. Oh and S. P. Kim, “A review on the computational methods for emotional state estimation from the human EEG,” Comput. Math. Methods Med., 2013, 2013

    2013

  5. [13]

    Revealing feelings: facets of emotional expressivity in self -reports, peer ratings, and behavior,

    J. J. Gross, O. P. John, “Revealing feelings: facets of emotional expressivity in self -reports, peer ratings, and behavior,” J. Personal. Soc. Psychol., 1997, vol. 72, no. 2, 435

    1997

  6. [14]

    Toward machine emotional intelligence: Analysis of affective physiological state,

    R. W. Picard, E. Vyzas, J. Healey, “Toward machine emotional intelligence: Analysis of affective physiological state,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 23, no. 10, pp. 1175–1191, 2001

    2001

  7. [15]

    Subject-independent emotion recognition based on physiological signals: a three-stage decision method,

    J. Chen, B. Hu, Y. Wang, P. Moore, Y. Dai, L. Feng and Z. Ding, “Subject-independent emotion recognition based on physiological signals: a three-stage decision method,” BMC Med. Inform. Decis. Mak., vol. 17, no. 3, 167, 2017

    2017

  8. [16]

    Emotional state classification from EEG data using machine learning approach,

    X. W. Wang, D. Nie, B. L. Lu, “Emotional state classification from EEG data using machine learning approach,” Neurocomputing, vol, 129, pp. 94–106, 2014

    2014

  9. [17]

    Filtering of spontaneous and low intensity emotions in educational contexts ,

    S. Salmeron-Majadas, M. Arevalillo-Herrá ez, O. C. Santos, M. Saneiro, R. Cabestrero, P. Quiró s and J. G. Boticario, “Filtering of spontaneous and low intensity emotions in educational contexts ,” Int. Conf. Artif. Intell. Education, Springer, Cham, 429–438, 2015

    2015

  10. [18]

    Deep learning with convolutional neural networks for EEG decoding and visualization,

    R. T. Schirrmeister, J. T. Springenberg, L. D. J. Fiederer, M. Glasstetter, K. Eggensperger, M. Tangermann and T. Ball, “Deep learning with convolutional neural networks for EEG decoding and visualization,” Hum. Brain Mapp., vol. 38, no. 11, 5391–5420, 2017

    2017

  11. [19]

    Differential entropy feature for EEG-based vigilance estimation,

    L. Shi, Y. Y. Jiao, B. L. Lu, “Differential entropy feature for EEG-based vigilance estimation,” 2013 35th Ann. Int. Conf. IEEE Eng. Med. Biol. Soc. (EMBC), IEEE, 2013, pp. 6627–6630

    2013

  12. [20]

    Squeeze-and-excitation networks,

    J. Hu, L. Shen, G. Sun, “Squeeze-and-excitation networks,” Proceedings of the IEEE Conf. Comput. Vis. Pattern Recognition, 7132–7141, 2018

    2018

  13. [21]

    Deap: A database for emotion analysis; using physiological signals,

    S. Koelstra, C. Muhl, M. Soleymani, J. S. Lee, A. Yazdani, T. Ebrahimi and I. Patras , “Deap: A database for emotion analysis; using physiological signals,” IEEE Trans. Affective Comput., vol. 3, no. 1, pp. 18–31, 2011

    2011

  14. [22]

    A multimodal database for affect rec ognition and implicit tagging ,

    M. Soleymani, J. Lichtenauer, T. Pun T, “A multimodal database for affect rec ognition and implicit tagging ,” IEEE Trans. Affective Comput., vol. 3, no. 1, pp. 42–55, 2011

    2011

  15. [23]

    Affective space is bipolar,

    J. A. Russell, “Affective space is bipolar,” J. Personal. Soc. Psychol., vol. 37, no. 3, 345, 1979

    1979

  16. [24]

    Improving BCI -based emotion recogni tion by combining EEG feature selection and kernel classifiers,

    J. Atkinson, D. Campos, “Improving BCI -based emotion recogni tion by combining EEG feature selection and kernel classifiers,” Expert Syst. Appl., vol. 47, pp. 35–41, 2016. 25

    2016

  17. [25]

    EEG -based emotion estimation using Bayesian weighted -log-posterior function and perceptron convergence algorithm,

    H. J. Yoon, S. Y. Chung, “EEG -based emotion estimation using Bayesian weighted -log-posterior function and perceptron convergence algorithm,” Comput. Biol. Med., vol. 43, no. 12, pp. 2230–2237, 2013

    2013

  18. [26]

    Emotion classification in arousal valence model using MAHNOB-HCI database,

    M. B. H. Wiem, Z. Lachiri, “Emotion classification in arousal valence model using MAHNOB-HCI database,” Int. J. Adv. Comput. Sci. Appl., vol. 8, no. 3, 2017

    2017

  19. [27]

    Emotion recognition based on EEG using LSTM recurrent neural network,

    S. Alhagry, A. A. Fahmy, R. A. El -Khoribi, “Emotion recognition based on EEG using LSTM recurrent neural network,” Emotion, vol. 8, no. 10, pp. 355–358, 2017

    2017

  20. [28]

    EEG-based emotion recognition using 3D convolutional neural networks,

    E. S. Salama, R. A. El-Khoribi, M. E. Shoman and M. A. W. Shalaby, “EEG-based emotion recognition using 3D convolutional neural networks,” Int. J. Adv. Comput. Sci. Appl., vol. 9, no. 8, pp. 329–337, 2018

    2018

  21. [29]

    Exploring EEG features in cross -subject emotion recognition,

    X. Li, D. Song, P. Zhang, Y. Zhang, Y. Hou and B. Hu, “Exploring EEG features in cross -subject emotion recognition,” Front. Neurosci., vol. 12, 162, 2018

    2018

  22. [30]

    Subject independent emotion recognition from EEG using VMD and deep learning,

    P. Pandey, K. R. Seeja, “Subject independent emotion recognition from EEG using VMD and deep learning,” J. King Saud Univ. Comput. Inform. Sci., 2019

    2019

  23. [31]

    EEG-based emotion recognition using frequency domain features and support vector machines,

    X. W. Wang, D. Nie, B. L. Lu, “EEG-based emotion recognition using frequency domain features and support vector machines,” Int. Conf. Neural Inform. Proc., Springer, Berlin, Heidelberg, 2011, pp. 734-743

    2011

  24. [32]

    Nonlinear characterization and complexity analysis of cardiotocographic exam inations using entropy measures ,

    J. A. L. Marques, P. C. Cortez, J. P. V. Madeiro, V. H. C. de Albuquerque , S. J. Fong and F. S. Schlindwein , “Nonlinear characterization and complexity analysis of cardiotocographic exam inations using entropy measures ,” J. Supercomput., vol. 76, no. 2, pp. 1305–1320, 2020

    2020

  25. [33]

    Analysis of absence seizure generation using EEG spatial-temporal regularity measures,

    N. Mammone, D. Labate, A. Lay-Ekuakille, F. C. Morabito, “Analysis of absence seizure generation using EEG spatial-temporal regularity measures,” Int. J. Neural Syst., vol. 22, no. 06, 1250024, 2012

    2012

  26. [34]

    Entropy index in quantitative EEG measurement for diagnosis accuracy,

    A. Lay-Ekuakille, P. Vergallo, G. Griffo, F. Conversano, S. Casciaro, S. Urooj and A. Trabacca, “Entropy index in quantitative EEG measurement for diagnosis accuracy,” IEEE Trans. Instrum. Meas., vol. 63, no. 6, 1440–1450, 2013

    2013

  27. [35]

    Theta oscillations in the hippocampus,

    G. Buzsá ki, “Theta oscillations in the hippocampus,” Neuron, vol. 33, no. 3, 325–340, 2002

    2002

  28. [36]

    Extraversion–introversion and 8 –13 Hz waves in frontal cortical regions ,

    Y. Tran, A. Craig, P. McIsaac, “Extraversion–introversion and 8 –13 Hz waves in frontal cortical regions ,” Personal. Individ. Differences, vol. 30, no. 2, 205–215, 2001

    2001

  29. [37]

    Devices and methods for maintaining an alert state of consciousness through brain wave monitoring , by R. B. Levin, (2000, Dec 26), U.S. Patent No. 6,167,298, 2000-12-26. 26

    2000

  30. [38]

    Gamma-band responses in the brain: a short review of psychophysiological correlates and functional significance,

    C. Başar-Eroglu, D. Strü ber, M. Schü rmann, M. Stadler and E. Başar , “Gamma-band responses in the brain: a short review of psychophysiological correlates and functional significance,” Int. J. Psychophysiol., vol. 24, no. (1–2), pp. 101-112, 1996

    1996

  31. [39]

    Continuous convolutional neural network with 3D input for EEG -based emotion recognition,

    Y. Yang, Q. Wu, F. Fu and X. Chen , “Continuous convolutional neural network with 3D input for EEG -based emotion recognition,” Int. Conf. Neural Inform. Proc., Springer, Cham, 2018, pp. 433-443

    2018

  32. [40]

    Emotional state classification from EEG data using machine learning approach,

    X. W. Wang, D. Nie, B. L. Lu, “Emotional state classification from EEG data using machine learning approach,” Neurocomputing, vol. 129, pp. 94-106, 2014

    2014

  33. [41]

    Gradient-based learning applied to document recognition,

    Y. LeCun, L. Bottou, Y. Bengio and P. Haffner, “Gradient-based learning applied to document recognition,” Proc. IEEE, vol. 86, no. 11, 2278–2324, 1998

    1998

  34. [42]

    Rectified linear units improve restricted boltzmann machines ,

    V. Nair, G. E. Hinton, “Rectified linear units improve restricted boltzmann machines ,” in Proc. 27th Int. Conf. Mach. Learn. (ICML-10), 2010, pp. 807–814

    2010

  35. [43]

    Swish: a self-gated activation function,

    P. Ramachandran, B. Zoph, Q. V. Le, “Swish: a self-gated activation function,” arXiv preprint arXiv:1710.05941, 7, 2017

    2017 arXiv

  36. [44]

    Deep Learning for EEG motor imagery classification based on multi-layer CNNs feature fusion,

    S. U. Amin, M. Alsulaiman, G. Muhammad, M. A. Mekhtiche and M. S. Hossain, “Deep Learning for EEG motor imagery classification based on multi-layer CNNs feature fusion,” Fut. Gener. Computer Syst., vol. 101, pp. 542–554, 2019

    2019

  37. [45]

    Imagenet classification with deep convolutional neural networks,

    A. Krizhevsky, I. Sutskever, G. E. Hinton, “Imagenet classification with deep convolutional neural networks,” Adv. Neural Inform. Proc. Syst., 2012, pp. 1097–1105

    2012

  38. [46]

    Going deeper with convolutions,

    C. Szegedy, W. Liu, Y. Jia, P., S. Reed, D. Anguelov and A. Rabinovich , “Going deeper with convolutions,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., 2015, 1–9

    2015

  39. [47]

    Continuous convolutional neural network with 3D input for EEG -based emotion recognition,

    Y. Yang, Q. Wu, Y Fu and X. Chen , “Continuous convolutional neural network with 3D input for EEG -based emotion recognition,” Int. Conf. Neural Inform. Proc., Springer, Cham, 2018, pp. 433–443

    2018

  40. [48]

    Emotion recognition from multi-channel EEG through parallel convolutional recurrent neural network,

    Y. Yang, Q. Wu, M. Qiu, Y. Wang and X. Chen, “Emotion recognition from multi-channel EEG through parallel convolutional recurrent neural network,” 2018 Int. Jt. Conf. Neural Netw. (IJCNN), IEEE, 2018, 1–7

    2018

  41. [49]

    Multi-method fusion of cross-subject emotion recognition based on high-dimensional EEG features,

    F. Yang, X. Zhao, W. Jiang, P. Gao and G. Liu, “Multi-method fusion of cross-subject emotion recognition based on high-dimensional EEG features,” Front. Comput. Neurosci., vol. 13, 2019. 27

    2019

  42. [50]

    Cross-corpus EEG-based emotion recognition ,

    S. Rayatdoost, M. Soleymani, “Cross-corpus EEG-based emotion recognition ,” 2018 IEEE 28th Int . Workshop Mach. Learn. Signal Process. (MLSP), IEEE, 2018, pp. 1–6

    2018

  43. [51]

    Emotion recognition from EEG signals using hierarchical Bayesian network with privileged information,

    Z. Gao, S. Wang, “Emotion recognition from EEG signals using hierarchical Bayesian network with privileged information,” Proceedings of the 5th ACM Int. Conf. Multimedia Retr. 2015, pp. 579–582

    2015

  44. [52]

    SVM-based feature selection methods for emotion recognition from multimodal data,

    C. Torres-Valencia, M. Álvarez-Ló pez, A. Orozco-Gutié rrez, “SVM-based feature selection methods for emotion recognition from multimodal data,” J. Multimodal User Interf., vol. 11, no. 1, 9–23, 2017

    2017

  45. [53]

    EEG-based emotion recognition utilizing wavelet coefficients,

    A. Momennezhad, “EEG-based emotion recognition utilizing wavelet coefficients,” Multimedia Tools Appl., vol. 77, no. 20, 27089–27106, 2018

    2018

  46. [54]

    Approximate entropy as a measure of system complexity,

    S. M. Pincus, “Approximate entropy as a measure of system complexity,” Proc. Natl. Academy Sci., vol. 88, no. 6, pp. 2297–2301, 1991

    1991

  47. [55]

    Deep convolutional neural network for the automated detection and diagnosis of seizure using EEG signals,

    U. R. Acharya, S L. Oh, Y. Hagiwara, J. H. Tan and H. Adeli , “Deep convolutional neural network for the automated detection and diagnosis of seizure using EEG signals,” Comput. Biol. Med., vol. 100, pp. 270–278, 2018

    2018

  48. [56]

    Multiscale entropy analysis of biological signals,

    M. Costa, A. L. Goldberger, C. K. Peng, “Multiscale entropy analysis of biological signals,” Phys. Rev. E, vol. 71, no. 2, 021906, 2005. Author Biographies Xiaolong ZHONG is with the School of Optical -Electrical and Computer Engineering , University of Shanghai for Science an...

    2005

Pith tools

Reviewed August 27, 2026 · model on record in the stance chip above.