Pith. sign in

REVIEW 4 major objections 6 minor 60 references

MECASA: Motor Execution Classification using Additive Self-Attention for Hybrid EEG-fNIRS Data

T0 review · 4 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

Pith's one-line read The paper claims that fusing EEG and fNIRS with a convolutional additive self-attention network, MECASA, outperforms standard baselines on rest-versus-task motor-execution classification, reaching 87.34% fused accuracy.

desk verdict The architecture is a reasonable adaptation of CAS-ViT to EEG-fNIRS fusion, but the leaked evaluation protocol makes the accuracy claims untrustworthy until a clean split is shown. read the letter →

arxiv 2501.05525 v1 pith:FANZ4FNX submitted 2025-01-09 cs.HC

classification cs.HC
keywords MotorexecutionEEG-fNIRSfusionAdditiveself-attentionBrain-computerinterfaceCASAmoduleRestversustaskclassificationSMRHybridBCIdataset
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 argues that combining EEG and fNIRS signals improves rest-versus-task classification in a motor-execution brain-computer interface, and that the proposed MECASA architecture does this better than existing baselines. On the SMR Hybrid BCI dataset, MECASA reports 75.07% accuracy for EEG, 86.52% for fNIRS, and 87.34% for the fused data, with fusion consistently beating each single modality across most compared methods. The authors also find that fNIRS alone outperforms EEG alone, and that the best results come from embedding dimensions of 64-128 for EEG and upsampled optical density (OD128) for fNIRS. If the results hold, the architecture offers a practical decoder that exploits complementary temporal and hemodynamic information at linear attention cost.

What carries the argument

The load-bearing component is the Convolutional Additive Self-Attention (CASA) module, which computes similarity between queries and keys as a sum of convolutional context mappings, $\text{Sim}(Q,K)=\Phi(Q)+\Phi(K)$, with $\Phi$ composed of Sigmoid-based channel attention and spatial attention, and forms output as $O=\Gamma(\Phi(Q)+\Phi(K))\cdot V$. Because the similarity uses convolutions instead of the quadratic $QK^\top$ product, the module scales as $O(N)$. Two separate MECASA streams process EEG and fNIRS signals, and a fusion network concatenates their penultimate-layer features before two fully connected layers produce the binary rest-versus-task output.

What would settle it

Retrain and evaluate MECASA under a strict trial-wise or subject-wise split (for example, leave-one-subject-out or holding out whole trials) with all hyperparameters unchanged, and check whether fused accuracy still exceeds the single-modality results by the reported margin.

Watch

Extended reading notes

Core claim

The central claim is that replacing softmax self-attention with a convolutional additive similarity function produces a model, MECASA, that outperforms EEGNet, TSception, Transformer, LMDA-Net, and ConvNeXT on EEG, fNIRS, and their fusion for motor-execution classification. On the SMR Hybrid BCI dataset, the reported accuracies are 75.07±3.89% for EEG, 86.52±1.38% for fNIRS, and 87.34±0.42% for the fused representation, with fusion improving over single-modality accuracy for every compared method except LMDA. The paper further claims that fNIRS yields higher accuracy than EEG alone and that the choice of fNIRS representation matters, with OD128 outperforming OD10 and total hemoglobin.

Load-bearing premise

The evaluation assumes that one-second epochs cut from the same 12-second trial are independent when the dataset is randomly split into training, validation, and test sets; if overlapping epochs from the same trial appear in more than one partition, the reported accuracies are inflated by temporal correlation.

Editorial extensions

If this is right

  • If the results hold, hybrid EEG-fNIRS decoding for motor execution can be improved by fusing the two modalities at the feature level rather than relying on either signal alone.
  • The linear-complexity CASA module makes the architecture a candidate for real-time BCI use, where quadratic self-attention is often too slow.
  • The fNIRS ablation suggests that upsampling optical density to 128 Hz is a more informative input representation than lower-rate optical density or total hemoglobin.
  • The embedding-dimension ablation points to a practical default of 64-128 dimensions for EEG-based variants of this architecture.

Reading between the lines

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

  • The reported accuracy gap likely depends on the evaluation split; if random epoch splitting allows temporally overlapping epochs from the same 12-second trial to land in both training and test sets, all methods' numbers may be inflated by correlation leakage.
  • A strict trial-wise or subject-wise cross-validation experiment, keeping everything else fixed, would settle whether the fusion advantage is a genuine architectural effect or a split artifact.
  • The additive-attention design could transfer to other long physiological time series where quadratic attention is prohibitive, provided the evaluation protocol is leakage-free.
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 / 6 minor

Summary. The paper proposes MECASA, a deep-learning architecture for classifying rest versus task states in a motor-execution paradigm using hybrid EEG-fNIRS data. MECASA adapts the CAS-ViT convolutional additive self-attention module, giving an O(N) attention complexity, and combines separate EEG and fNIRS streams with a late fusion network. The authors evaluate MECASA on the SMR Hybrid BCI dataset against EEGNet, TSception, Transformer, LMDA, and ConvNeXT, and report accuracy with 95% confidence intervals, plus ablations on embedding dimensions and fNIRS data representations. The central claims are that MECASA consistently outperforms all baselines across EEG, fNIRS, and fused modalities, and that fusion consistently improves accuracy over single modalities.

Significance. If the claims were supported, MECASA would be a practically useful decoder: it uses a public dataset, compares against several established baselines, provides ablation experiments, and the CASA module has attractive O(N) complexity for real-time BCI use. The manuscript therefore addresses a relevant problem in multimodal BCI classification. However, the central claims are empirical and currently rest on an evaluation protocol that is internally contradictory and potentially leaky, and on accuracy differences that are not statistically corroborated. The strengths of the paper — a clear architecture description and broad baseline coverage — are undermined by these evaluation issues, so the significance of the claimed consistent superiority cannot be assessed until the evaluation is corrected.

major comments (4)
  1. [§4A–4B] The evaluation protocol is contradictory and does not establish independent test samples. Section 4A states that the 41,228 one-second epochs per modality were split into 70% training, 15% validation, and 15% test sets, while Section 4B states that stratified five-fold cross-validation was employed. These two protocols cannot both be the final evaluation procedure, and no details are given about how they are reconciled. More importantly, the epochs are cut from 12-second trials with a 0.5-second overlap, but the split is described only as a random split of the epoch pool. Adjacent epochs from the same trial and the same subject therefore appear in both training and test sets, sharing half of their signal. This temporal and subject-level overlap allows the models to memorize test content during training, and different architectures can exploit this leakage to different degrees. The relative ranking in Table 1 is thus not a trustworthy basis for the claim that MECASA consistently outperforms the baselines. The authors need to re-run the evaluation with trial-wise or subject-wise splits that keep overlapping epochs in the same partition, and report the actual protocol used.
  2. [Table 1] No significance tests are reported, and the EEG results do not even show non-overlapping confidence intervals. For EEG, MECASA achieves 75.07±3.89 while EEGNet achieves 72.40±0.52; the lower bound of the MECASA interval overlaps the upper bound of the EEGNet interval, so the headline 'consistently outperforms' claim is not supported for EEG at the 95% level, even before considering the data-split problem. Similar overlap exists for other comparisons (for example, in Table 2, fNIRS accuracy for embedding dims 48-56 is 87.57±0.69 versus 86.52±1.38 for 64-128). The paper should report pairwise significance tests or proper confidence intervals for the accuracy differences, and the text should be adjusted to state which comparisons are actually statistically distinguishable.
  3. [Table 1, Abstract, Conclusion] The claim that fusion consistently improves accuracy compared to single modalities is directly contradicted by the reported LMDA results. In Table 1, LMDA has EEG accuracy 69.60±0.80, fNIRS accuracy 74.47±0.54, but fused accuracy 57.47±0.50, which is far lower than either unimodal result. The abstract and conclusion state that 'fusion consistently improving accuracy compared to single-modality approaches,' but the paper's own data show at least one counterexample. The evaluation section hedges with 'generally resulted in improved accuracy,' but the abstract and conclusion overstate the finding. This inconsistency must be resolved either by correcting the claim or by explaining why the LMDA fusion result is excluded from the general pattern.
  4. [§4E, Table 2] The ablation conclusion that embedding dimensions of 64-128 'consistently outperform other configurations' is not supported by the tabulated results. For fNIRS, the 48-56 configuration achieves 87.57±0.69, which is higher in mean accuracy than 64-128's 86.52±1.38, and the confidence intervals overlap. The claim of consistency appears only in the discussion and is not a safe reading of Table 2. The authors should either restrict the claim to the EEG column, perform a statistical comparison, or otherwise temper the conclusion.
minor comments (6)
  1. [§2] There is a duplicated section heading: 'B Classification Algorithms for Motor Execution' appears twice, once without a period and once as 'B. Classification Algorithms for Motor Execution.' One of these should be removed or renumbered.
  2. [§4E/Table 3] There is a stray heading 'E Ablation Study' immediately before Table 3; this appears to be a formatting artifact and should be removed.
  3. [Figure 2 caption] The caption for Figure 2 reads 'Convolution Adaptive Separable Attention,' but the text uniformly refers to the module as 'Convolutional Additive Self-Attention (CASA).' The caption should be made consistent with the text.
  4. [§4C] The classifiers subsection states that the work uses established classifiers 'for EEG-based emotion classification,' but the task in this paper is motor-execution rest-versus-task classification. This looks like a copy-paste from a prior paper and should be corrected.
  5. [§4B] The implementation details do not state how baseline hyperparameters were chosen or whether they were tuned on the validation set. Since the main claim is comparative, the paper should state the hyperparameter selection procedure for every model, or at least note that the same training budget and optimizer settings were used.
  6. [§3C] The notation for the architecture is introduced only informally: the feature-map channel counts C1 and C2, and the meaning of 'Ch×T' for the input, are not precisely defined. A short notation table or explicit values for C1 and C2 would improve reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: MECASA's accuracy claims are empirical comparisons on a public dataset, and the self-cited baselines are not load-bearing.

full rationale

The paper's central claim is an empirical accuracy comparison, not a derivational chain. MECASA is trained and evaluated on epochs from the public SMR Hybrid BCI dataset, and its reported accuracies are direct test-set measurements rather than quantities defined in terms of the inputs or of fitted parameters relabeled as predictions. The ablation studies select embedding dimensions and the OD128 fNIRS representation using the same test partition, which risks optimistic selection bias, but this is not circularity: the reported accuracies are not equal to the selection criterion by construction. The two self-citations (references 58 and 60) provide baseline architectures, ConvNeXT and Transformer, whose numbers in Table 1 are produced by the present comparison; they are not invoked as unverified theorems or as forced architectural premises. The CASA module is attributed to the external CAS-ViT work, not to the authors' own prior claims. The ambiguous evaluation protocol (70/15/15 split in Section 4A versus five-fold cross-validation in Section 4B, with overlapping one-second epochs) is a serious validity threat to the reported margins, but a leaky or inconsistent split would inflate measured accuracy rather than make the claimed result tautological. No equation is reused as both premise and conclusion, and no fitted constant is renamed as a prediction. Therefore no circular step is exhibited, and the appropriate score is 0.

Assumptions & free parameters 5 free parameters · 3 assumptions · 0 invented entities

The paper's central claim is empirical; the main hidden choices are the hyperparameters and preprocessing settings selected via ablation on the test split, plus the implicit independence assumption for overlapping epochs. No new physical or conceptual entities are introduced.

free parameters (5)
  • embedding dimensions = 64-128
    Selected after ablation on test data (Table 2); the best configuration for the EEG stream in the main results.
  • fNIRS data representation = OD128
    Upsampled optical density chosen after ablation on test data (Table 3) yields the highest accuracy.
  • learning rate = 1e-4
    Set by hand, reported in Section 4B.
  • batch size = 16
    Set by hand, reported in Section 4B.
  • fusion FC layer size = 64
    Chosen for the first fully connected layer in the fusion network (Figure 3).
assumptions (3)
  • domain assumption Modified Beer-Lambert Law converts raw fNIRS intensity to optical density or concentration changes
    Invoked in Section 4A preprocessing without validation against ground truth; cited to ref 54.
  • domain assumption The SMR Hybrid dataset is representative of motor execution in healthy adults
    All 15 participants are right-handed males; the paper generalizes BCI implications from this narrow sample.
  • ad hoc to paper Random split of overlapping epochs yields independent samples
    The 70/15/15 split assumes epoch independence despite 0.5s overlap; if false, test accuracy is inflated. This is a load-bearing assumption for all reported numbers.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MECASA: Motor Execution Classification using Additive Self-Attention for Hybrid EEG-fNIRS Data." pith.science (2026). https://pith.science/paper/FANZ4FNX

@misc{pith2026250105525,
  author       = {Pith},
  title        = {Pith review of: MECASA: Motor Execution Classification using Additive Self-Attention for Hybrid EEG-fNIRS Data},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FANZ4FNX}},
  note         = {Machine review of arXiv:2501.05525}
}
read the original abstract

Motor execution, a fundamental aspect of human behavior, has been extensively studied using BCI technologies. EEG and fNIRS have been utilized to provide valuable insights, but their individual limitations have hindered performance. This study investigates the effectiveness of fusing electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) data for classifying rest versus task states in a motor execution paradigm. Using the SMR Hybrid BCI dataset, this work compares unimodal (EEG and fNIRS) classifiers with a multimodal fusion approach. It proposes Motor Execution using Convolutional Additive Self-Attention Mechanisms (MECASA), a novel architecture leveraging convolutional operations and self-attention to capture complex patterns in multimodal data. MECASA, built upon the CAS-ViT architecture, employs a computationally efficient, convolutional-based self-attention module (CASA), a hybrid block design, and a dedicated fusion network to combine features from separate EEG and fNIRS processing streams. Experimental results demonstrate that MECASA consistently outperforms established methods across all modalities (EEG, fNIRS, and fused), with fusion consistently improving accuracy compared to single-modality approaches. fNIRS generally achieved higher accuracy than EEG alone. Ablation studies revealed optimal configurations for MECASA, with embedding dimensions of 64-128 providing the best performance for EEG data and OD128 (upsampled optical density) yielding superior results for fNIRS data. This work highlights the potential of deep learning, specifically MECASA, to enhance EEG-fNIRS fusion for BCI applications.

Figures

Figures reproduced from arXiv: 2501.05525 by the authors.

Figure 1
Figure 1. Illustration of the proposed classification backbone. Two stages downsample the original signal. Linear Linear Linear S S S + C N Q K V Spatial Channel Interaction S Sigmoid S [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Convolution Adaptive Separable Attention Feature 1 Feature 2 Fusion Layer (Concatenate) Fully Connected (64) Fully Connected (2) ReLU [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Architecture of the proposed fusion network used for classification of combined features of EEG and fNIRS data The process begins with feature extraction, where separate, modality-specific networks are used to process data from EEG and fNIRS. The EEG-specific network is designed to capture the temporal and spectral characteristics of electrical brain activity. Concurrently, the fNIRS-specific network ex￾tracts featu… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

60 extracted references · 58 canonical work pages

  1. [1]

    Exploring the role of primary and supplementary motor areas in simple motor tasks with fnirs

    Sabrina Brigadoi, Simone Cutini, Fabio Scarpa, Pietro Scatturin, and Roberto Dell’Acqua. Exploring the role of primary and supplementary motor areas in simple motor tasks with fnirs. Cognitive processing, 13:97–101, 2012

  2. [2]

    Eeg-based neurophysiological in- dices for expert psychomotor performance–a review

    Jazmin M Morrone and Charles R Pedlar. Eeg-based neurophysiological in- dices for expert psychomotor performance–a review. Brain and Cognition , 175:106132, 2024

  3. [3]

    The extraction of motion-onset vep bci features based on deep learning and compressed sensing

    Teng Ma, Hui Li, Hao Y ang, Xulin Lv, Peiyang Li, Tiejun Liu, Dezhong Y ao, and Peng Xu. The extraction of motion-onset vep bci features based on deep learning and compressed sensing. Journal of neuroscience methods , 275: 80–92, 2017

  4. [4]

    Dynamics of the eeg power in the frequency and spatial domains during observation and execution of manual movements

    Silvi Frenkel-Toledo, Shlomo Bentin, Anat Perry, Dario G Liebermann, and Nachum Soroker. Dynamics of the eeg power in the frequency and spatial domains during observation and execution of manual movements. Brain re- search, 1509:43–57, 2013

  5. [5]

    Decoding eeg rhythms during action observation, motor imagery, and execution for standing and sitting.IEEE sensors journal, 20(22):13776–13786, 2020

    Rattanaphon Chaisaen, Phairot Autthasan, Nopparada Mingchinda, Pitsha- porn Leelaarporn, Narin Kunaseth, Suppakorn Tammajarung, Poramate Manoonpong, Subhas Chandra Mukhopadhyay, and Theerawit Wilaiprasit- porn. Decoding eeg rhythms during action observation, motor imagery, and execution for standing and sitting.IEEE sensors journal, 20(22):13776–13786, 2020

  6. [6]

    A deep learning approach for brain computer interaction-motor execution eeg signal classification

    Nesma E Elsayed, Ahmed S Tolba, Magdi Z Rashad, Tamer Belal, and Sha- henda Sarhan. A deep learning approach for brain computer interaction-motor execution eeg signal classification. IEEE Access, 9:101513–101529, 2021

  7. [7]

    On the suitability of near-infrared (nir) systems for next-generation brain–computer interfaces

    Shirley Coyle, Tomás Ward, Charles Markham, and Gary McDarby. On the suitability of near-infrared (nir) systems for next-generation brain–computer interfaces. Physiological measurement, 25(4):815, 2004

  8. [8]

    Saskia Koehler, Johanna Egetemeir, Prisca Stenneken, Stefan P Koch, Paul Pauli, Andreas J Fallgatter, and Martin J Herrmann. The human execu- tion/observation matching system investigated with a complex everyday task: a functional near-infrared spectroscopy (fnirs) study.Neuroscience letters, 508 (2):73–77, 2012

Show all 60 references
  1. [9]

    Real time detection of cognitive load using fnirs: A deep learning approach

    Subashis Karmakar, Supreeti Kamilya, Prasenjit Dey, Parag K Guhathakurta, Mamata Dalui, Tushar Kanti Bera, Suman Halder, Chiranjib Koley, Tandra Pal, and Anupam Basu. Real time detection of cognitive load using fnirs: A deep learning approach. Biomedical Signal Processing and ...

  2. [10]

    Enhanced drowsiness detection using deep learning: an fnirs study

    M Asjid Tanveer, M Jawad Khan, M Jahangir Qureshi, Noman Naseer, and Keum-Shik Hong. Enhanced drowsiness detection using deep learning: an fnirs study. IEEE access, 7:137920–137929, 2019

  3. [11]

    A hybrid bci based on eeg and fnirs signals improves the perfor- mance of decoding motor imagery of both force and speed of hand clenching

    Xuxian Yin, Baolei Xu, Changhao Jiang, Yunfa Fu, Zhidong Wang, Hongyi Li, and Gang Shi. A hybrid bci based on eeg and fnirs signals improves the perfor- mance of decoding motor imagery of both force and speed of hand clenching. Journal of neural engineering, 12(3):036004, 2015

  4. [12]

    Bimodal data fusion of simultaneous measurements of eeg and fnirs during lower limb movements

    Maged S Al-Quraishi, Irraivan Elamvazuthi, Tong Boon Tang, Muhammad Al- Qurishi, Syed Hasan Adil, and Mansoor Ebrahim. Bimodal data fusion of simultaneous measurements of eeg and fnirs during lower limb movements. Brain Sciences, 11(6):713, 2021

  5. [13]

    Identification of lower-limb motor tasks via brain–computer interfaces: a topical overview

    Víctor Asanza, Enrique Peláez, Francis Loayza, Leandro L Lorente-Leyva, and Diego H Peluffo-Ordóñez. Identification of lower-limb motor tasks via brain–computer interfaces: a topical overview. Sensors, 22(5):2028, 2022

  6. [14]

    Analyzing classification per- formance of fnirs-bci for gait rehabilitation using deep neural networks

    Huma Hamid, Noman Naseer, Hammad Nazeer, Muhammad Jawad Khan, Rayyan Azam Khan, and Umar Shahbaz Khan. Analyzing classification per- formance of fnirs-bci for gait rehabilitation using deep neural networks. Sen- sors, 22(5):1932, 2022

  7. [15]

    Decoding multi-class motor imagery and motor execution tasks using rieman- nian geometry algorithms on large eeg datasets.Sensors, 23(11):5051, 2023

    Zaid Shuqfa, Abdelkader Nasreddine Belkacem, and Abderrahmane Lakas. Decoding multi-class motor imagery and motor execution tasks using rieman- nian geometry algorithms on large eeg datasets.Sensors, 23(11):5051, 2023

  8. [16]

    Eeg motor imagery classification with sparse spectrotemporal decomposition and deep learning

    Biao Sun, Xing Zhao, Han Zhang, Ruifeng Bai, and Ting Li. Eeg motor imagery classification with sparse spectrotemporal decomposition and deep learning. IEEE Transactions on Automation Science and Engineering , 18(2):541–551, 2020

  9. [17]

    Transfer learning with data alignment and optimal transport for eeg based mo- tor imagery classification

    Chao Chu, Lei Zhu, Aiai Huang, Ping Xu, Nanjiao Ying, and Jianhai Zhang. Transfer learning with data alignment and optimal transport for eeg based mo- tor imagery classification. Journal of Neural Engineering, 21(1):016015, 2024

  10. [18]

    A diagonal masking self-attention-based multi-scale network for motor imagery classification

    Kaijun Y ang, Jihong Wang, Liantao Y ang, Lifeng Bian, Zi Jiang Luo, and Chen Y ang. A diagonal masking self-attention-based multi-scale network for motor imagery classification. Journal of Neural Engineering, 2024

  11. [19]

    Msfnet: A multi-scale space-time frequency fusion network for motor imagery eeg classification

    Chang Wang, Y ang Wu, Chen Wang, Y aning Ren, Jiefen Shen, Ting Pang, Chee Seng Chan, Wenjie Ren, and Yi Yu. Msfnet: A multi-scale space-time frequency fusion network for motor imagery eeg classification. IEEE Access, 2024

  12. [20]

    Optimal channel selection of multiclass motor imagery classification based on fusion convolutional neural network with attention blocks

    Joharah Khabti, Saad AlAhmadi, and Adel Soudani. Optimal channel selection of multiclass motor imagery classification based on fusion convolutional neural network with attention blocks. Sensors, 24(10):3168, 2024

  13. [21]

    Brain-computer interface using neural network Siddhad et al

    Gan Wang and Moran Cerf. Brain-computer interface using neural network Siddhad et al. | MECASA: Motor Execution Classification using Additive Self-Attention for Hybrid EEG-fNIRS Data arXiv | 7 and temporal-spectral features. Frontiers in Neuroinformatics , 16:952474, 2022

  14. [22]

    Motor imagery classification based on eeg sensing with visual and vibrotactile guidance.Sen- sors, 23(11):5064, 2023

    Luka Batisti ´c, Diego Sušanj, Domagoj Pinˇci´c, and Sandi Ljubic. Motor imagery classification based on eeg sensing with visual and vibrotactile guidance.Sen- sors, 23(11):5064, 2023

  15. [23]

    A novel method for classification of multi-class motor imagery tasks based on feature fusion.Neu- roscience Research, 176:40–48, 2022

    Yimin Hou, Tao Chen, Xiangmin Lun, and Fang Wang. A novel method for classification of multi-class motor imagery tasks based on feature fusion.Neu- roscience Research, 176:40–48, 2022

  16. [24]

    Mar- tins, and Vicente A de Sousa Jr

    Tarciana C de Brito Guerra, Taline Nóbrega, Edgard Morya, Allan de M. Mar- tins, and Vicente A de Sousa Jr. Electroencephalography signal analysis for human activities classification: A solution based on machine learning and mo- tor imagery. Sensors, 23(9):4277, 2023

  17. [25]

    Deep learning for eeg motor imagery classification based on multi-layer cnns feature fusion

    Syed Umar Amin, Mansour Alsulaiman, Ghulam Muhammad, Mo- hamed Amine Mekhtiche, and M Shamim Hossain. Deep learning for eeg motor imagery classification based on multi-layer cnns feature fusion. Future Generation computer systems, 101:542–554, 2019

  18. [26]

    Deep learning for eeg-based motor imagery classification: To- wards enhanced human-machine interaction and assistive robotics

    Nejia Boutarfaia, Samuele Russo, Ahmed Tibermacine, and Imad Eddine Tibermacine. Deep learning for eeg-based motor imagery classification: To- wards enhanced human-machine interaction and assistive robotics. life, 2(3): 4, 2023

  19. [27]

    Eeg clas- sification of motor imagery using a novel deep learning framework

    Mengxi Dai, Dezhi Zheng, Rui Na, Shuai Wang, and Shuailei Zhang. Eeg clas- sification of motor imagery using a novel deep learning framework. Sensors, 19(3):551, 2019

  20. [28]

    Hs-cnn: a cnn with hybrid convolution scale for eeg motor imagery classification.Journal of neural engineering, 17(1):016025, 2020

    Guanghai Dai, Jun Zhou, Jiahui Huang, and Ning Wang. Hs-cnn: a cnn with hybrid convolution scale for eeg motor imagery classification.Journal of neural engineering, 17(1):016025, 2020

  21. [29]

    Adaptive transfer learning for eeg motor imagery classification with deep con- volutional neural network

    Kaishuo Zhang, Neethu Robinson, Seong-Whan Lee, and Cuntai Guan. Adaptive transfer learning for eeg motor imagery classification with deep con- volutional neural network. Neural Networks, 136:1–10, 2021

  22. [30]

    Deep learning for motor imagery eeg-based classification: A review

    Ali Al-Saegh, Shefa A Dawwd, and Jassim M Abdul-Jabbar. Deep learning for motor imagery eeg-based classification: A review. Biomedical Signal Pro- cessing and Control, 63:102172, 2021

  23. [31]

    A cross-space cnn with customized characteristics for motor imagery eeg classification

    Ying Hu, Y an Liu, Siqi Zhang, Ting Zhang, Bin Dai, Bo Peng, Hongbo Y ang, and Y akang Dai. A cross-space cnn with customized characteristics for motor imagery eeg classification. IEEE Transactions on Neural Systems and Reha- bilitation Engineering, 31:1554–1565, 2023

  24. [32]

    Subject-independent deep architecture for eeg-based motor imagery classification

    Shadi Sartipi and Mujdat Cetin. Subject-independent deep architecture for eeg-based motor imagery classification. IEEE Transactions on Neural Sys- tems and Rehabilitation Engineering, 2024

  25. [33]

    Fusion of deep features from 2d-dost of fnirs signals for subject-independent classifi- cation of motor execution tasks

    Pouya Khani, Vahid Solouk, Hashem Kalbkhani, and Farid Ahmadi. Fusion of deep features from 2d-dost of fnirs signals for subject-independent classifi- cation of motor execution tasks. International Journal of Intelligent Systems , 2023(1):3178284, 2023

  26. [34]

    Functional near-infrared spectroscopy for the clas- sification of motor-related brain activity on the sensor-level

    Alexander E Hramov, Vadim Grubov, Artem Badarin, Vladimir A Maksimenko, and Alexander N Pisarchik. Functional near-infrared spectroscopy for the clas- sification of motor-related brain activity on the sensor-level. Sensors, 20(8): 2362, 2020

  27. [35]

    Classification of motor imagery and execution signals with population-level feature sets: implications for probe design in fnirs based bci

    Sinem Burcu Erdo ˆgan, Eran Özsarfati, Burcu Dilek, Kübra Soˆgukkanlı Kadak, Lütfü Hano ˆglu, and Ata Akın. Classification of motor imagery and execution signals with population-level feature sets: implications for probe design in fnirs based bci. Journal of neural engineering...

  28. [36]

    Single-trial classification of fnirs signals in four directions mo- tor imagery tasks measured from prefrontal cortex

    Hong Peng, Jinlong Chao, Sirui Wang, Jie Dang, Fengqi Jiang, Bin Hu, and Dennis Majoe. Single-trial classification of fnirs signals in four directions mo- tor imagery tasks measured from prefrontal cortex. IEEE transactions on nanobioscience, 17(3):181–190, 2018

  29. [37]

    An fnirs-based motor imagery bci for als: A subject-specific data-driven approach

    S M Hosni, S B Borgheai, J Mclinden, and Y Shahriari. An fnirs-based motor imagery bci for als: A subject-specific data-driven approach. IEEE Transac- tions on Neural Systems and Rehabilitation Engineering , 28(12):3063–3073, 2020

  30. [38]

    Exploiting neurovascular coupling: a bayesian sequential monte carlo approach applied to simulated eeg fnirs data

    Pierpaolo Croce, Filippo Zappasodi, Arcangelo Merla, and Antonio Maria Chiarelli. Exploiting neurovascular coupling: a bayesian sequential monte carlo approach applied to simulated eeg fnirs data. Journal of neural engi- neering, 14(4):046029, 2017

  31. [39]

    Motor imagery decoding enhancement based on hybrid eeg-fnirs signals

    Tao Xu, Zhengkang Zhou, Yuliang Y ang, Yu Li, Junhua Li, Anastasios Beze- rianos, and Hongtao Wang. Motor imagery decoding enhancement based on hybrid eeg-fnirs signals. IEEE Access, 11:65277–65288, 2023

  32. [40]

    Fganet: fnirs-guided at- tention network for hybrid eeg-fnirs brain-computer interfaces

    Y oungchul Kwak, Woo-Jin Song, and Seong-Eun Kim. Fganet: fnirs-guided at- tention network for hybrid eeg-fnirs brain-computer interfaces. IEEE Transac- tions on Neural Systems and Rehabilitation Engineering, 30:329–339, 2022

  33. [41]

    Hybrid inte- grated wearable patch for brain eeg-fnirs monitoring.Sensors (Basel, Switzer- land), 24(15):4847, 2024

    Boyu Li, Mingjie Li, Jie Xia, Hao Jin, Shurong Dong, and Jikui Luo. Hybrid inte- grated wearable patch for brain eeg-fnirs monitoring.Sensors (Basel, Switzer- land), 24(15):4847, 2024

  34. [42]

    A generalised at- tention mechanism to enhance the accuracy performance of neural networks

    Pengcheng Jiang, Ferrante Neri, Yu Xue, and Ujjwal Maulik. A generalised at- tention mechanism to enhance the accuracy performance of neural networks. International journal of neural systems, page 2450063, 2024

  35. [43]

    Hybrid attention network for epileptic eeg classification

    Y anna Zhao, Jiatong He, Fenglin Zhu, Tiantian Xiao, Y ongfeng Zhang, Ziwei Wang, Fangzhou Xu, and Yi Niu. Hybrid attention network for epileptic eeg classification. International Journal of Neural Systems, 33(06):2350031, 2023

  36. [44]

    A multi-scale fusion convolutional neural network based on attention mechanism for the vi- sualization analysis of eeg signals decoding

    Donglin Li, Jiacan Xu, Jianhui Wang, Xiaoke Fang, and Ying Ji. A multi-scale fusion convolutional neural network based on attention mechanism for the vi- sualization analysis of eeg signals decoding. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 28(12):26...

  37. [45]

    Wei Tao, Ze Wang, Chi Man Wong, Ziyu Jia, Chang Li, Xun Chen, CL Philip Chen, and Feng Wan. Adfcnn: attention-based dual-scale fusion convolutional neural network for motor imagery brain-computer interface.IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2023

  38. [46]

    Tcja-snn: Temporal-channel joint attention for spiking neural net- works

    Rui-Jie Zhu, Malu Zhang, Qihang Zhao, Haoyu Deng, Yule Duan, and Liang- Jian Deng. Tcja-snn: Temporal-channel joint attention for spiking neural net- works. IEEE Transactions on Neural Networks and Learning Systems, 2024

  39. [47]

    A review on the attention mech- anism of deep learning

    Zhaoyang Niu, Guoqiang Zhong, and Hui Yu. A review on the attention mech- anism of deep learning. Neurocomputing, 452:48–62, 2021

  40. [48]

    Cas-vit: Convolutional additive self-attention vision transformers for efficient mobile applications

    Tianfang Zhang, Lei Li, Y ang Zhou, Wentao Liu, Chen Qian, and Xiangyang Ji. Cas-vit: Convolutional additive self-attention vision transformers for efficient mobile applications. arXiv preprint arXiv:2408.03703, 2024

  41. [49]

    Efficientvit: Memory efficient vision transformer with cascaded group attention

    Xinyu Liu, Houwen Peng, Ningxin Zheng, Yuqing Y ang, Han Hu, and Yixuan Yuan. Efficientvit: Memory efficient vision transformer with cascaded group attention. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 14420–14430, 2023

  42. [50]

    Edgevits: Competing light-weight cnns on mobile devices with vision transformers

    Junting Pan, Adrian Bulat, Fuwen Tan, Xiatian Zhu, Lukasz Dudziak, Hong- sheng Li, Georgios Tzimiropoulos, and Brais Martinez. Edgevits: Competing light-weight cnns on mobile devices with vision transformers. In European Conference on Computer Vision, pages 294–311. Springer, 2022

  43. [51]

    Swiftformer: Efficient addi- tive attention for transformer-based real-time mobile vision applications

    Abdelrahman Shaker, Muhammad Maaz, Hanoona Rasheed, Salman Khan, Ming-Hsuan Y ang, and Fahad Shahbaz Khan. Swiftformer: Efficient addi- tive attention for transformer-based real-time mobile vision applications. arXiv preprint arXiv:2303.15446, 2023

  44. [52]

    Deep sparse rectifier neural networks

    Xavier Glorot, Antoine Bordes, and Y oshua Bengio. Deep sparse rectifier neural networks. In Proceedings of the fourteenth international conference on artificial intelligence and statistics , pages 315–323. JMLR Workshop and Conference Proceedings, 2011

  45. [53]

    Hybrid eeg- fnirs asynchronous brain-computer interface for multiple motor tasks

    Alessio Paolo Buccino, Hasan Onur Keles, and Ahmet Omurtag. Hybrid eeg- fnirs asynchronous brain-computer interface for multiple motor tasks. PloS one, 11(1):e0146610, 2016

  46. [54]

    The modified beer–lambert law revisited

    Laszlo Kocsis, Peter Herman, and Andras Eke. The modified beer–lambert law revisited. Physics in Medicine & Biology, 51(5):N91, 2006

  47. [55]

    EEGNet: a compact convolutional neural network for EEG-based brain–computer interfaces

    Vernon J Lawhern, Amelia J Solon, Nicholas R Waytowich, Stephen M Gor- don, Chou P Hung, and Brent J Lance. EEGNet: a compact convolutional neural network for EEG-based brain–computer interfaces. Journal of Neural Engineering, 15(5):056013, 2018

  48. [56]

    TS- ception: Capturing temporal dynamics and spatial asymmetry from EEG for emotion recognition

    Yi Ding, Neethu Robinson, Su Zhang, Qiuhao Zeng, and Cuntai Guan. TS- ception: Capturing temporal dynamics and spatial asymmetry from EEG for emotion recognition. IEEE Transactions on Affective Computing, 2022

  49. [57]

    A convnet for the 2020s

    Zhuang Liu, Hanzi Mao, Chao-Yuan Wu, Christoph Feichtenhofer, Trevor Dar- rell, and Saining Xie. A convnet for the 2020s. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022

  50. [58]

    Neural networks meet neural activity: Utilizing eeg for mental workload estimation

    Gourav Siddhad, Partha Pratim Roy, and Byung-Gyu Kim. Neural networks meet neural activity: Utilizing eeg for mental workload estimation. arXiv preprint arXiv:2408.13930, 2024

  51. [59]

    Lmda-net: A lightweight multi-dimensional attention network for general eeg-based brain- computer interfaces and interpretability

    Zhengqing Miao, Meirong Zhao, Xin Zhang, and Dong Ming. Lmda-net: A lightweight multi-dimensional attention network for general eeg-based brain- computer interfaces and interpretability. NeuroImage, 276:120209, 2023

  52. [60]

    Efficacy of transformer networks for classification of eeg data

    Gourav Siddhad, Anmol Gupta, Debi Prosad Dogra, and Partha Pratim Roy. Efficacy of transformer networks for classification of eeg data. Biomedical Signal Processing and Control, 87:105488, 2024. 8 | arXiv Siddhad et al. | MECASA: Motor Execution Classification using Additive S...

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.