Pith. sign in

REVIEW 4 major objections 5 minor 34 references

Energy-based features and bi-LSTM neural network for EEG-based music and voice classification

T0 review · 4 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 The paper claims relative EEG channel energies read by a bidirectional LSTM label 400-millisecond trials as music or speech at 98.66% accuracy, distinguish four genres at 61.59%, and detect musical taste at 96.96%.

desk verdict The reported accuracies are inflated by a trial-level random split of overlapping windows; the model memorizes stimulus and subject identity, so the central claims are unsupported. read the letter →

arxiv 2411.13217 v1 pith:Y46JL3OQ submitted 2024-11-20 eess.SP cs.SDeess.AS

Isaac Ariza , Ana M. Barbancho , Lorenzo J. Tardon , Isabel Barbancho This is my paper · ORCID
classification eess.SPcs.SDeess.AS
keywords EEGbi-LSTMenergy-basedfeaturesmusicversusspeechclassificationmusicalgenretastebrain-computerinterfaceauditorystimuli
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 what a person is listening to can be read from brief EEG recordings: pairwise differences in per-channel signal energy, arranged in a matrix and processed by a bidirectional long short-term memory network, are enough to label single 400-millisecond trials as music or speech, to tell among four music genres, and to tell whether the listener likes the song. On their own recordings from six participants, the scheme reaches 98.66% accuracy for the music-versus-speech task and up to 99.97% when the temporal evolution of energy differences is included. Accuracy for four-way genre classification reaches 61.59%, and binary like/dislike classification reaches 96.96%. The authors present these numbers as evidence that relative energy across brain areas carries consistent information about auditory content and preference, which would open low-cost EEG-based applications in music recommendation and brain-computer interaction.

What carries the argument

The central object is the channel-energy-difference matrix E(n): a 61-by-61 matrix whose entry at row i and column j is the difference in dB between the energy of channel i and the energy of channel j during trial n, so each row describes one channel's energy relative to every other channel and the diagonal is zero. A second matrix E'(n) approximates the discrete derivative of E(n) between consecutive trials, capturing how relative channel energies evolve over time. These matrices are fed as sequences into a bidirectional LSTM layer with 20 hidden units, followed by a fully connected layer and softmax, which lets the network use both past and future context in deciding the class of each trial.

What would settle it

Re-run the pipeline while holding out entire subjects, or using only non-overlapping trials from different songs, and compare the accuracy to the reported 98.66%, 61.59%, and 96.96%. If accuracy on held-out subjects or held-out stimuli falls to near chance, the central claim that the energy relations generalize across people and audio would be unsupported.

Watch

Extended reading notes

Core claim

At the paper's core is the claim that a matrix of all pairwise differences in per-channel EEG energy over short trials, read as a sequence by a bi-LSTM, is a sufficient characterization of the brain's reaction to auditory content. For each trial, the energy of each of the 61 channels is computed through the Fourier transform and converted to decibels, and the 61-by-61 matrix E(n) stores every channel-minus-channel difference; the companion matrix E'(n) does the same for the change in these differences between consecutive trials. The bi-LSTM with 20 hidden units then consumes these matrices as time series. The best distinguishing power comes from E'(n) for voice-versus-music (99.97%) and for genre (61.59%), while the plain 400 ms E(n) matrix gives the best taste classification (96.96%). The authors interpret the overall pattern as evidence that energy relations between brain areas change measurably with the kind of sound and with the listener's preference.

Load-bearing premise

The load-bearing premise is that a random 50/50 split of overlapping 400-millisecond trials into training and test sets measures generalization to new listening; because the trials come from only six subjects and a small set of repeated songs and sentences, this premise is what the reported accuracies rest on.

Editorial extensions

If this is right

  • Real-time EEG-based audio tagging becomes plausible: a 400 ms window is short enough to label what a user is hearing while it happens, without waiting for a longer stimulus.
  • Since the energy-difference feature is computed with an FFT and pairwise subtraction, it is cheap enough for consumer EEG hardware and needs no per-channel supervised feature learning.
  • Decoding like-versus-dislike at 96.96% from single-trial responses suggests a neural correlate of musical preference that recommendation systems could exploit without explicit user ratings.
  • The improvement from E'(n) over E(n) in the genre task indicates that the temporal change of relative channel energies carries information that static energy maps miss, guiding where to look for better auditory EEG features.

Reading between the lines

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

  • A stricter evaluation not reported in the paper would hold out entire subjects or entire songs rather than randomly splitting overlapping trials; until that is done, the reported accuracies should be read as upper bounds on how well the features generalize to new listeners and new audio.
  • Because E(n) is antisymmetric with a zero diagonal, the 61-by-61 matrix is highly redundant; a natural testable extension is to select a small subset of channel pairs and measure how accuracy degrades, which would also make the approach more portable.
  • The network was trained for only five epochs with twenty hidden units, so the feature construction is likely doing most of the work; one could test whether a simpler classifier such as a linear model on the same matrices recovers most of the accuracy, which would separate the contribution of the features from that of the recurrent architecture.
  • A further experiment could separate musical taste from song familiarity and genre preference, since each subject heard a different set of songs and the like/dislike label may be entangled with how well the song is known.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The manuscript proposes an EEG-based classification scheme that characterizes each trial by a 61x61 matrix of pairwise log-energy differences between EEG channels (E(n)) and a temporal-difference variant (E'(n)), and feeds these matrices into a bidirectional LSTM for three tasks: binary music/voice classification, 4-class musical genre classification, and binary/multi-class classification of musical taste. The data come from two experiments with 6 subjects, using 20 music fragments and 150 spoken sentences, segmented into 400-ms or 1-s trials with 50% overlap. Reported accuracies reach 98.66% for music/voice (and up to 99.97% for the E'(n) variant), 61.59% for 4-class genre, and 96.96% for binary musical taste. The central claim is that these results demonstrate satisfactory performance of the proposed scheme. The paper also includes a discussion of energy topoplots and comparisons across trial durations and feature variants.

Significance. If the reported accuracies were valid, the paper would be a useful contribution to EEG-based auditory stimulus classification, particularly the energy-difference feature matrix and the bi-LSTM architecture. The dataset description and the feature extraction procedure are reasonably detailed, and the authors report multiple experimental variations. However, the evaluation protocol is invalid for the claims made: random splitting of overlapping trials from a small number of stimuli and subjects allows the network to memorize stimulus and subject identity rather than learn generalizable auditory categories. Because every headline accuracy in the abstract and Section 4 rests on this protocol, the central claim is unsupported. The paper does not provide machine-checked proofs, reproducible code, or baseline comparisons that would partially compensate for the methodological issue.

major comments (4)
  1. [Section 4] The evaluation protocol is invalid for measuring generalization. The manuscript states: 'The dataset used in each test is divided randomly into two subsets, one subset with 50% of the trials to train the network, and the other 50% of the trials are used to test the performance.' Because trials are 400-ms or 1-s windows with 50% overlap (Section 3.1), overlapping trials from the same recording share half of their raw samples. A random split places adjacent, nearly identical trials in both training and test sets, so the test accuracy reflects memorization of the training trials rather than classification of new brain responses. This flaw affects all results in Tables 6, 7, and 8 and the abstract's headline accuracies.
  2. [Section 4] The claim of an 'inter-subject scenario' is contradicted by the actual split. The random trial-level split pools all trials from all 6 subjects and divides them randomly, so every subject contributes to both training and test partitions. The model can therefore learn subject-specific EEG characteristics and use them to classify test trials from the same subjects. A true inter-subject evaluation would require training on some subjects and testing on held-out subjects; with only 6 subjects, such an evaluation would have only a few folds and would likely yield far lower accuracy. As it stands, the test procedure is not inter-subject and the phrase is misleading.
  3. [Section 4.1] There is also stimulus-level leakage. The music/voice task uses only 20 distinct music fragments (5 per genre) and 150 distinct sentences (30 per language), and all trials derived from a single stimulus share the same class label. The random split assigns trials from the same song or sentence to both training and test, so the network can memorize the identity of the specific audio excerpt rather than learning the abstract class 'music' or 'voice.' This is especially severe for the binary music/voice task, where 98.66% accuracy can be trivially achieved by recognizing the 20 songs against the 150 sentences. The same reasoning applies to the musical taste labels, which are tied to individual songs.
  4. [Section 4] No baseline or chance-level comparison is provided. The paper reports accuracies but does not compare against a simple classifier on the same features, a random classifier, or a cross-validation scheme with non-overlapping trials. Given the leakage described above, the reader cannot determine whether the proposed bi-LSTM adds any value beyond stimulus memorization. Additionally, the 10 repeated runs are averaged without reporting standard deviations, confidence intervals, or statistical significance tests, so even the internal variability of the reported numbers is unknown.
minor comments (5)
  1. [Section 3.1] The variable D is used in Eqs. (6) and (7) before it is defined as equal to 1; please define D before its first use.
  2. [Section 3.1] The integration limits in Eq. (2) appear as '—1' and are likely a typographical artifact; they should presumably be -∞ to ∞ or the full frequency range. Please correct the notation.
  3. [Abstract] There is a typo: 'acuiqre' should be 'acquire.'
  4. [Appendix] The ballad and reggaeton tables list 5 songs each, but the four genre tables overall show 5 excerpts per genre; however, the text in Section 2.2.1 says 'Five fragments of different songs from each musical genre are played,' which is consistent. Please verify that the total count of 20 fragments matches the trial counts in Table 4.
  5. [General] The phrase 'inter-subject scenario' in Section 4 should be removed or redefined, since the actual protocol is intra-subject with random trial splitting. If a subject-independent evaluation is intended, a leave-one-subject-out or subject-disjoint split should be used and clearly described.

Circularity Check

1 steps flagged · score 6.0 of 10

Reported accuracies are produced by a random split of 50%-overlapping trials, so test trials are near-duplicates of training trials; the 'satisfactory performance' claim is not an independent prediction.

  1. fitted input called prediction [Section 4 (evaluation protocol); Section 2.3 and Section 3.1 (trial segmentation and feature extraction)]
    "Signal excerpts are divided into trials of 400ms of duration with an overlap of 50%. ... The dataset used in each test is divided randomly into two subsets, one subset with 50% of the trials to train the network, and the other 50% of the trials are used to test the performance."

    Because trials overlap by 50%, a test trial shares half its raw EEG samples with an adjacent training trial. The feature vector E(n) is computed as per-channel energy over the whole trial window, so overlapping train/test trials have near-identical inputs. The random split also places trials from the same song or sentence and the same subject in both partitions, allowing the bi-LSTM to memorize stimulus or subject identity rather than learn the intended auditory category. The abstract's central claim of 'satisfactory performance' (98.66%, 61.59%, 96.96%) is therefore a reconstruction of training information, not an independent prediction; the reported test accuracy reduces, by construction of the split protocol, to a near-memorization check.

full rationale

The paper's feature pipeline (FFT energy, pairwise dB-difference matrices, bi-LSTM) is not itself circular: E(n) is defined from EEG signals, not from the labels, and the network is trained on labels rather than fitted to the reported accuracies. However, the central empirical claim rests entirely on a random 50/50 split of 50%-overlapping trials. Section 4 says the split is by trials, not by stimuli or subjects, while Sections 2.3 and 3.1 define trials with 50% overlap. Adjacent train/test trials therefore share half their raw samples, and same-song/same-subject trials appear in both partitions, so the test accuracy is not evidence of generalization. This is a fitted-input-called-prediction circularity: the model is fit to data whose time-domain content overlaps the test inputs. The self-citation [30] only motivates the bi-LSTM choice and is not load-bearing; no uniqueness theorem or target-derived definition is involved. The evaluation leakage is the sole circular step, but it undermines the paper's central claim, so a score of 6 is appropriate rather than a minor-citation score.

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

The central claim rests on several domain assumptions about EEG stationarity, P300 relevance, and the validity of self-report labels, plus one ad hoc methodological assumption (random split of overlapping trials is valid) that is demonstrably false. No free parameters are fitted to predict a target, but hyperparameters such as trial duration, overlap, hidden units, epochs, and split ratio are chosen without validation. No new entities are introduced.

free parameters (5)
  • Trial duration = 400 ms or 1 s
    Two durations are tested and results vary; the choice is a hyperparameter, not derived from theory.
  • Trial overlap = 50%
    Chosen by the authors; the heavy overlap creates dependent trials and is central to the data-leakage problem.
  • Bi-LSTM hidden units = 20
    Set by the authors; no hyperparameter search or validation is reported.
  • Number of epochs = 5
    Set by the authors; no early stopping or validation is reported.
  • Train/test split ratio = 50/50
    Chosen by the authors; the random split method is the main methodological weakness.
assumptions (5)
  • domain assumption EEG signals are stationary within 400 ms or 1 s trials
    The FFT-based energy measure assumes the signal is stationary within the trial (Section 3.1).
  • domain assumption The P300 component is captured in 400 ms trials
    Section 2.3 states this, but no ERP/P300 analysis is performed, and energy features are used instead.
  • domain assumption Subjects' self-reports of liking are accurate
    Musical taste labels are derived from three-option self-report after each song (Section 2.2.1).
  • ad hoc to paper Random split of overlapping trials is a valid evaluation protocol
    Section 4 assumes this; it is false because overlapping windows and shared stimuli create train/test dependence.
  • domain assumption The 61-channel energy-difference matrix contains class-discriminative information
    The whole method rests on this premise (Section 3.1).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Energy-based features and bi-LSTM neural network for EEG-based music and voice classification." pith.science (2026). https://pith.science/paper/Y46JL3OQ

@misc{pith2026241113217,
  author       = {Pith},
  title        = {Pith review of: Energy-based features and bi-LSTM neural network for EEG-based music and voice classification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Y46JL3OQ}},
  note         = {Machine review of arXiv:2411.13217}
}
read the original abstract

The human brain receives stimuli in multiple ways; among them, audio constitutes an important source of relevant stimuli for the brain regarding communication, amusement, warning, etc. In this context, the aim of this manuscript is to advance in the classification of brain responses to music of diverse genres and to sounds of different nature: speech and music. For this purpose, two different experiments have been designed to acquiere EEG signals from subjects listening to songs of different musical genres and sentences in various languages. With this, a novel scheme is proposed to characterize brain signals for their classification; this scheme is based on the construction of a feature matrix built on relations between energy measured at the different EEG channels and the usage of a bi-LSTM neural network. With the data obtained, evaluations regarding EEG-based classification between speech and music, different musical genres, and whether the subject likes the song listened to or not are carried out. The experiments unveil satisfactory performance to the proposed scheme. The results obtained for binary audio type classification attain 98.66% of success. In multi-class classification between 4 musical genres, the accuracy attained is 61.59%, and results for binary classification of musical taste rise to 96.96%.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 27 canonical work pages

  1. [1]

    Wiley, West Sussex

    Sanei S, Chambers JA (2008) EEG signal processing. Wiley, West Sussex

    work page 2008

  2. [2]

    In: 2018 IEEE international symposium on multimedia (ISM), pp 188-189

    Yu Y, Beuret S, Zeng D, Oyama K (2018) Deep learning of human perception in audio event classification. In: 2018 IEEE international symposium on multimedia (ISM), pp 188-189. https://doi.org/10.1109/ISM.2018.00-11

    work page doi:10.1109/ism.2018.00-11 2018

  3. [3]

    In: 2017 IEEE SENSORS, pp 1-3

    Darmawan FF, Arifianto D, Huda MA, Taruno WP (2017) Human brain auditory activity observation using electrical capacitance volume tomography. In: 2017 IEEE SENSORS, pp 1-3. https://doi.org/10.1109/ICSENS.2017.8234433

    arXiv 2017

  4. [4]

    In: 2020 National conference on communications (NCC), pp 1-5

    Aggarwal S, Sharon R, Murthy HA (2020) P300 based stereo localization of single frequency audio stimulus. In: 2020 National conference on communications (NCC), pp 1-5. https://doi.org/10. 1109/NCC48643.2020.9056052

    work page arXiv 2020

  5. [5]

    In: 2019 3rd international conference on trends in electronics and informatics (ICOEI), pp 1248-1253

    Kumar SD, Subha D (2019) Prediction of depression from EEG signal using long short term memory (LSTM). In: 2019 3rd international conference on trends in electronics and informatics (ICOEI), pp 1248-1253. https://doi.org/10.1109/ICOEI.2019. 8862560

    work page doi:10.1109/icoei.2019 2019

  6. [6]

    In: 2019 2nd international conference on computer applications information security (ICCAIS), pp 1-4

    Alturki FA, AlSharabi K, Aljalal M, Abdurraqeeb AM (2019) A DWT-band power-SVM based architecture for neurological brain disorders diagnosis using EEG signals. In: 2019 2nd international conference on computer applications information security (ICCAIS), pp 1-4. https://doi.org/10.1109/CAIS.2019.8769492

    work page arXiv 2019

  7. [7]

    In: 2013 International winter workshop on brain-computer interface (BCI), pp 46-48

    Kim K, Carlson T, Lee S (2013) Design of a robotic wheelchair with a motor imagery based brain-computer interface. In: 2013 International winter workshop on brain-computer interface (BCI), pp 46-48. https://doi.org/10.1109/IWW-BCI.2013. 6506625

    work page doi:10.1109/iww-bci.2013 2013

  8. [8]

    In: 2020 IEEE international con­ ference on consumer electronics—Asia (ICCE-Asia), pp 1-3

    Moschona DS (2020) An affective service based on multi-modal emotion recognition, using EEG enabled emotion tracking and speech emotion recognition. In: 2020 IEEE international con­ ference on consumer electronics—Asia (ICCE-Asia), pp 1-3. https://doi.org/10.1109/ICCE-Asia49877.2020.9277291

    arXiv 2020

Show all 34 references
  1. [9]

    In: 2017 4th IAPR Asian conference on pattern recognition (ACPR), pp 906-910

    Kee Y, Lee M, Williamson J, Lee S (2017) A hierarchical clas­ sification strategy for robust detection of passive/active mental state using user-voluntary pitch imagery task. In: 2017 4th IAPR Asian conference on pattern recognition (ACPR), pp 906-910. https://doi.org/10.1109/...

    2017 doi

  2. [10]

    IEEE Trans Neural Syst Rehabil Eng 21(2):225-232

    Wang Q, Sourina O (2013) Real-time mental arithmetic task recognition from EEG signals. IEEE Trans Neural Syst Rehabil Eng 21(2):225-232. https://doi.org/10.1109/TNSRE.2012. 2236576

    2013 doi

  3. [11]

    IEEE Rev Biomed Eng 14:204-218

    Hosseini M-P, Hosseini A, Ahi K (2021) A review on machine learning for EEG signal processing in bioengineering. IEEE Rev Biomed Eng 14:204-218. https://doi.org/10.1109/RBME.2020. 2969915

    2021 doi

  4. [12]

    IEEE Trans Neural Syst Rehabil Eng 27(1):31-42

    Zhang P, Wang X, Zhang W, Chen J (2019) Learning spatial- spectral-temporal EEG features with recurrent 3D convolutional neural networks for cross-task mental workload assessment. IEEE Trans Neural Syst Rehabil Eng 27(1):31-42. https://doi.org/10. 1109/TNSRE.2018.2884641

    2019

  5. [13]

    In: 2021 national conference on communications (NCC), pp 1-5

    Patel P, Satija U (2021) Performance analysis of convolutional neural network based EEG epileptic seizure classification in presence of ocular artifacts. In: 2021 national conference on communications (NCC), pp 1-5. https://doi.org/10.1109/ NCC52529.2021.9530053

    2021

  6. [14]

    In: 2021 3rd international conference on electronics representation and algorithm (ICERA), pp 89-94

    Pratiwi M, Wibawa AD, Purnomo MH (2021) EEG-based happy and sad emotions classification using LSTM and bidirectional LSTM. In: 2021 3rd international conference on electronics representation and algorithm (ICERA), pp 89-94. https://doi.org/ 10.1109/ICERA53111.2021.9538698

    2021

  7. [15]

    IEEE Access 8:163269-163279

    Lu W, Wei Y, Yuan J, Deng Y, Song A (2020) Tractor assistant driving control method based on EEG combined with RNN-TL deep learning algorithm. IEEE Access 8:163269-163279. https:// doi.org/10.1109/ACCESS.2020.3021051

    2020

  8. [16]

    Neurocomputing

    Liu Y, Liu W, Obaid M, Abbas I (2015) Exponential stability of Markovian jumping Cohen-Grossberg neural networks with mixed mode-dependent time-delays. Neurocomputing. https:// doi.org/10.1016/j.neucom.2015.11.046

    2015 doi

  9. [17]

    J Frankl Inst 353(2):448-461

    Du B, Liu Y, Atiatallah Abbas I (2016) Existence and asymptotic behavior results of periodic solution for discrete-time neutral-type neural networks. J Frankl Inst 353(2):448-461. https://doi.org/10. 1016/j.jfranklin.2015.11.013

    2016

  10. [18]

    Electronics

    Seo Y-S, Huh J-H (2019) Automatic emotion-based music clas­ sification for supporting intelligent IoT applications. Electronics. https://doi.org/10.3390/electronics8020164

    2019 doi

  11. [19]

    In: 2009 IEEE international conference on acoustics, speech and signal pro­ cessing, pp 489-492

    Lin Y-P, Wang C-H, Wu T-L, Jeng S-K, Chen J-H (2009) EEG­ based emotion recognition in music listening: a comparison of schemes for multiclass support vector machine. In: 2009 IEEE international conference on acoustics, speech and signal pro­ cessing, pp 489-492. IEEE

    2009

  12. [20]

    Neuroscience

    Poikonen H, Alluri V, Brattico E, Lartillot O, Tervaniemi M, Huotilainen M (2015) Event-related brain responses while lis­ tening to entire pieces of music. Neuroscience. https://doi.org/10. 1016/j.neuroscience.2015.10.061

    2015

  13. [21]

    In: 2009 4th international IEEE/EMBS conference on neural engineering, pp 440-443

    Shi S-J, Lu B-L (2009) EEG signal classification during listening to native and foreign languages songs. In: 2009 4th international IEEE/EMBS conference on neural engineering, pp 440-443. https://doi.org/10.1109/NER.2009.5109327

    2009

  14. [22]

    BrainVision, accessed 27 July

    BrainVision (2021) actiCHamp Plus. BrainVision, accessed 27 July. https://brainvision.com/

    2021

  15. [23]

    https://pstnet.com/products/e-prime/

    Psychology Software Tools (2022) E-Prime, accessed 8 March. https://pstnet.com/products/e-prime/

    2022

  16. [24]

    Springer, London

    Miranda ER, Castet J (2014) Guide to brain-computer music interfacing. Springer, London

    2014

  17. [25]

    SIAM J Control Optim 46:496-540

    Oberst U (2007) The fast Fourier transform. SIAM J Control Optim 46:496-540. https://doi.org/10.1137/060658242

    2007 doi

  18. [26]

    Teoría y problemas de analisis matematico en una variable, 3rd edn

    García A, Garcia F, Lopez A, Rodríguez G, de la Villa A (2007) Calculo I. Teoría y problemas de analisis matematico en una variable, 3rd edn. Editorial Clagsa, Madrid

    2007

  19. [27]

    Neural Comput 9(8):1735-1780

    Hochreiter S, Schmidhuber J (1997) Long short-term memory. Neural Comput 9(8):1735-1780

    1997

  20. [28]

    In: 2018 IEEE EMBS international conference on biomedical health informatics (BHI), pp 104-107

    Hasib MM, Nayak T, Huang Y (2018) A hierarchical LSTM model with attention for modeling EEG non-stationarity for human decision prediction. In: 2018 IEEE EMBS international conference on biomedical health informatics (BHI), pp 104-107. https://doi.org/10.1109/BHI.2018.8333380

    2018

  21. [29]

    Neural Comput Appl 34(15):12527-12557

    Houssein EH, Hammad A, Ali AA (2022) Human emotion recognition from EEG-based brain-computer interface using machine learning: a comprehensive review. Neural Comput Appl 34(15):12527-12557

    2022

  22. [30]

    Biomed Signal Process Control 78:103885

    Ariza I, Tardoí n LJ, Barbancho AM, De-Torres I, Barbancho I (2022) Bi-LSTM neural network for EEG-based error detection in musicians’ performance. Biomed Signal Process Control 78:103885. https://doi.org/10.1016/j.bspc.2022.103885

    2022

  23. [31]

    Mach Learn

    Provost F, Kohavi R (1998) Glossary of terms. Mach Learn. https://doi.org/10.1023/A:1017181826899

    1998 doi

  24. [32]

    Mach Learn Technol 2

    Powers D (2008) Evaluation: from precision, recall and F-factor to ROC, informedness, markedness and correlation. Mach Learn Technol 2

    2008

  25. [33]

    In: Proceedings of the 2nd inter­ national symposium on music information retrieval, Indiana, vol

    Tzanetakis G, Essl G, Cook P (2001) Automatic musical genre classification of audio signals. In: Proceedings of the 2nd inter­ national symposium on music information retrieval, Indiana, vol

    2001

  26. [144]

    http://ismir2001.ismir.net/pdf/tzanetakis.pdf 1 3

Pith tools

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