Performance Enhancement of P300 Detection by Multiscale-CNN
The DOI matches its Crossref record; 55 of the 56 references it cites were found in a registry or an index.
Abstract
The P300-based spelling system is one of the most popular brain–computer interface applications. Its success largely depends on performance, including the information transmission rate (ITR) and detection rate (i.e., accuracy). To achieve good performance, we proposed a multiscale convolutional neural network (MS-CNN) model that consists of seven layers. First, an upfront data set was used to train the MS-CNN, aiming to obtain a subject-unspecific model (universal model) for P300 detection. Second, this universal model was adapted by a portion of data derived from a subject to update the model to obtain a subject-specific model by incorporating a transfer learning technique. We applied the proposed model in the brain–computer interface (BCI) Controlled Robot Contest at the 2019 World Robot Conference, and our performance was the best among the teams in the contest. In the contest, ten healthy young subjects were randomly assigned by the contest committee to assess the model. Our model achieved the best P300 detection performance (higher accuracy with less repetition time). The ITR for the subject-unspecific case was 13.49 bits/min, while the ITR for the subject-specific case was 12.13 bits/min when the repetitions were fewer than six. It is believed that our method may pave a promising path for taking a further step toward efficient implementation of the P300-based spelling system.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“the learned spatial attention maps generally accords with those of previous studies [15], [23], [24].”
From ST-CapsNet: Linking Spatial and Temporal Attention With Capsule Network for P300 Detection Improvement (2023) · cites this paper as [24]
In the citing paper
“An important domain knowledge regarding ERP signals is that the SNR can be enhanced through trial averaging. [26], [27].”
From A Cascade xDAWN EEGNet Structure for Unified Visual-Evoked Related Potential Detection (2024) · cites this paper as [27]
In the citing paper
“which is consistent with the results in [23] and [24].”
From ST-CapsNet: Linking Spatial and Temporal Attention With Capsule Network for P300 Detection Improvement (2023) · cites this paper as [24]
In the citing paper
“However, some valuable information that would help in classification will be lost during the forward propagation, because it employed the max pooling operation to reduce feature maps which only retains the most active features and discards the rest.”
From ST-CapsNet: Linking Spatial and Temporal Attention With Capsule Network for P300 Detection Improvement (2023) · cites this paper as [24]
Citations
Cites
U-Net: Convolutional Networks for Biomedical Image Segmentation
Ronneberger, O., Fischer, P., & Brox, T. (2015) · Lecture notes in computer science · doi:10.1007/978-3-319-24574-4_28 · Cited by 99,404
Gradient-based learning applied to document recognition
Lecun, Y., Bottou, L., Bengio, Y., & Haffner, P. (1998) · Proceedings of the IEEE · doi:10.1109/5.726791 · Cited by 58,805
Pan, S. J., & Yang, Q. (2010) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2009.191 · Cited by 23,387 · cited 2×
“Using subject-specific data sets, we can transfer the individual information to the universal MS-CNN model by transfer learning techniques [49].”
BackgroundMethodBrain–computer interfaces for communication and control
Wolpaw, J. R., Birbaumer, N., McFarland, D. J., Pfurtscheller, G., & Vaughan, T. M. (2002) · Clinical Neurophysiology · doi:10.1016/s1388-2457(02)00057-3 · Cited by 7,927 · cited 1×
“B RAIN–computer interfaces (BCIs) provide an alternative method for controlling external devices or communicating between human beings and the outside world by translating brain activities into commands or information [1].”
MethodCNN Features Off-the-Shelf: An Astounding Baseline for Recognition
Razavian, A. S., Azizpour, H., Sullivan, J., & Carlsson, S. (2014) · 2014 IEEE Conference on Computer Vision and Pattern Recognition Workshops · doi:10.1109/cvprw.2014.131 · Cited by 4,451
Res2Net: A New Multi-Scale Backbone Architecture
Gao, S.-H., Cheng, M.-M., Zhao, K., Zhang, X.-Y., Yang, M.-H., & Torr, P. (2021) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2019.2938758 · Cited by 3,385 · cited 1×
“In addition, there has been an increasing trend in the use of multiscale convolution in recent years [39]–[43].”
BackgroundFarwell, L. A., & Donchin, E. (1988) · Electroencephalography and Clinical Neurophysiology · doi:10.1016/0013-4694(88)90149-6 · Cited by 3,228 · cited 1×
“Farwell and Donchin [13] proposed using correlation coefficients between EEG time series and P300 templates as features and achieved a detection accuracy of 80%.”
MethodLearning Hierarchical Features for Scene Labeling
Farabet, C., Couprie, C., Najman, L., & LeCun, Y. (2013) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2012.231 · Cited by 2,620
Cognitive and biological determinants of P300: an integrative review
Polich, J., & Kok, A. (1995) · Biological Psychology · doi:10.1016/0301-0511(95)05130-9 · Cited by 1,613 · cited 1×
“Moreover, the P300 peak is influenced by users’ age [11] and the surrounding environment [12].”
BackgroundDeep learning in bioinformatics
Min, S., Lee, B., & Yoon, S. (2016) · Briefings in Bioinformatics · doi:10.1093/bib/bbw068 · Cited by 1,465 · cited 1×
“Deep learning is successful in diverse applications, such as image retrieval, speech recognition, and biomedical signal classification [24]–[28].”
Background
Research works printed but not linked: 2
- wang, 2012, Asynchronous TV remote control system based on event-related potential brain-computer interface, Control Theory Appl, 29, 1507
- 2008, BCI Competition III Final Results
Cited by
Wang, Z., Chen, C., Li, J., Wan, F., Sun, Y., & Wang, H. (2023) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2023.3237319 · Cited by 75
“We can find that the average spatial filter has higher values in the parietal and occipital regions, which is consistent with the results in [23] and [24].”
BackgroundResultSelf-citationMOCNN: A Multiscale Deep Convolutional Neural Network for ERP-Based Brain-Computer Interfaces
Jin, J., Xu, R., Daly, I. P., Zhao, X., Wang, X., & Cichocki, A. S. (2024) · IEEE Transactions on Cybernetics · doi:10.1109/tcyb.2024.3390805 · Cited by 73
“Finally, multiscaling is usually achieved by convolution of different kernel sizes to the same input, which can lead to information redundancy [20], [36], [37], [38].”
BackgroundMethodFu, B., Chu, W., Gu, C., & Liu, Y. (2024) · IEEE Journal of Biomedical and Health Informatics · doi:10.1109/jbhi.2024.3419043 · Cited by 41
Gao, W., Huang, W., Li, M., Gu, Z., Pan, J., Yu, T., Yu, Z. L., & Li, Y. (2023) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2023.3259991 · Cited by 32
“For instance, a classifier based on an xDAWN filter [14], a CNN [15], [16], [17], and a reinforcement learning model [18] were previously trained on data acquired from a pool of subjects and then adapted with subject-specific labeled data via incremental training or model fine-tuning.”
BackgroundMethodPan, J., Cai, H., Huang, H., He, Y., & Li, Y. (2023) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2023.3267367 · Cited by 35
“Their information transmission rate (ITR) is low, however, because it takes many repetitions to recognize one character [22], and it is very difficult for patients with DoC to complete such long-term BCI experiments due to their own ease of fatigue and lack of concentration, making it difficult to…”
BackgroundMethodKleih, S. C., & Botrel, L. (2024) · Frontiers in Human Neuroscience · doi:10.3389/fnhum.2024.1400336 · Cited by 30
Entropy-Based Thunderstorm Imaging System With Real-Time Prediction and Early Warning
Xu, Y., Xing, H., Su, X., & Ji, X. (2022) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2022.3164167 · Cited by 22
“In order to reduce invalid data features input to the BiL-STM [29], [30], we decide to add the CNN [31] before the BiLSTM to deeply explore 3DAEF spatial features.”
MethodA Cascade xDAWN EEGNet Structure for Unified Visual-Evoked Related Potential Detection
Wang, H., Wang, Z., Sun, Y., Yuan, Z., Xu, T., & Li, J. (2024) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2024.3415474 · Cited by 20
“An important domain knowledge regarding ERP signals is that the SNR can be enhanced through trial averaging [26], [27].”
BackgroundSelf-citationJi, Z., Li, S., Zhang, H., Chen, C., Xu, Q., Li, J., & Wang, H. (2025) · Brain-Apparatus Communication A Journal of Bacomics · doi:10.1080/27706710.2025.2489396 · Cited by 24
Self-citationArpaïa, P., Callegaro, L., Cultrera, A., Espósito, A., & Ortolano, M. (2021) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2021.3127650 · Cited by 16
“Among possible paradigms, reactive BCIs are the most performing ones [6], [7].”
BackgroundQian, L., Ge, X., Zhao, F., Wang, S., Yuan, J., Pan, Y., Shi, H., Xu, J., & Sun, Y. (2023) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2023.3242771 · Cited by 15
“target stimulus and distractors [16], it can be inferred that the fixation-related P300 could provide discriminative information in terms of FRP classification.”
BackgroundSelf-citationTajmirriahi, M., Amini, Z., Rabbani, H., & Kafieh, R. (2022) · IEEE Sensors Journal · doi:10.1109/jsen.2022.3159475 · Cited by 15
Multifeature Fusion-Based Thunderstorm Prediction System With Switchable Patterns
Xu, Y., Xing, H., Ji, X., Zhao, D., Su, X., & Witold Pedrycz, C. (2023) · IEEE Sensors Journal · doi:10.1109/jsen.2023.3291397 · Cited by 12
“7) Fixed Models: The established eight CNN-BiLSTM models (M1–M8) [10], [20], [21], [22], [23] are used to train E z data in Fig.”
MethodTou, S. L. J., Warschausky, S. A., Karlsson, P., & Huggins, J. E. (2026) · Frontiers in Human Neuroscience · doi:10.3389/fnhum.2026.1720969 · Cited by 12
“Similarly, neural network-based approaches can integrate information across multiple electrode locations to learn latent representations and have demonstrated notable gains in classification accuracy (Wang et al., 2021, 2020; Ji et al., 2025).”
Peng, Y. K., Zhang, Y., Kong, W., Nie, F., Lu, B., & Cichocki, A. S. (2022) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2022.3165741 · Cited by 23
“Therefore, EEG provides a new path for objective emotion recognition and some other brain–computer interface applications [5], which have been drawing a lot of attention from academic and industrial communities in past decades.”
BackgroundBhandari, V., Londhe, N. D., & Kshirsagar, G. B. (2023) · IEEE Transactions on Cognitive and Developmental Systems · doi:10.1109/tcds.2023.3288201 · Cited by 11
“Wang et al. [10] averaged the original target data, and data produced from averaging were included with an imbalanced data set in oversampling-based techniques for P300-based speller systems.”
BackgroundMethodInfluentialIIMCNet: Intra- and Inter-Modality Correlation Network for Hybrid EEG-fNIRS Brain-Computer Interface
Yuan, X., Zhang, Y., & Rolfe, P. J. (2025) · IEEE Journal of Biomedical and Health Informatics · doi:10.1109/jbhi.2025.3594203 · Cited by 8
“4) Industrial applications, such as controlling robotic arms [10], wheelchairs [11], and spelling devices [12].”
Borra, D., & Magosso, E. (2025) · Computer Methods and Programs in Biomedicine · doi:10.1016/j.cmpb.2025.109008 · Cited by 5
Park, S., Ha, J., & Kim, L. (2024) · Frontiers in Physiology · doi:10.3389/fphys.2024.1325784 · Cited by 6
“Additionally, the performance of detecting ERP components (i.e., P3) has been enhanced by improving the feature extraction and classi fi cation algorithms using deep learning methods (Wang H. et al., 2019; Wang H. et al., 2021; Wang et al., 2023a).”
MethodMao, J., Wei, W., Li, X., Qiu, S., & He, H. (2025) · IEEE Sensors Journal · doi:10.1109/jsen.2025.3607814 · Cited by 2
“In contrast, MS-CNN [24] utilized multiscale temporal kernels to capture diverse features on a minimal cohort (ten subjects; 6 × 6 matrix, 175-ms flash).”
Cao, Y., Yang, H., Xue, Y., Wang, F., Li, T., Zhao, L., & Fu, Y. (2026) · Cognitive Neurodynamics · doi:10.1007/s11571-025-10380-5 · Cited by 2
Application of CNN Classifier for Somatosensory ERP-Based Brain-Computer Interface
Wang, J., Lu, X., Fei, N., Li, X., & Hu, Y. S. (2024) · IEEE International Conference on Computational Intelligence and Virtual Environments for Measurement Systems and Applications (CIVEMSA) · doi:10.1109/civemsa58715.2024.10586614 · Cited by 2
“The multi-scale kernels inside the MBConv were inspired by [14, 19], while the original MobileNet/EfficientNet [15-18] only used single-size kernels.”
MethodEffect of Dilated Convolution on Performance and Parameters of Devanagari Script-based P300 Speller
Bhandari, V., Londhe, N. D., & Kshirsagar, G. B. (2022) · IEEE India Conference · doi:10.1109/indicon56171.2022.10040171 · Cited by 2
“After that different variants of CNN such as CNN with batch normalization [12], 3D input CNN [27], multiscale CNN [28], etc. have been employed by researchers to boost P300 classifier performance.”
MethodAn improved multi-scale convolution and transformer network for EEG-based motor imagery decoding
Zhu, L., Wang, Y., Huang, A., Tan, X., & Zhang, J. (2025) · International Journal of Machine Learning and Cybernetics · doi:10.1007/s13042-025-02556-6 · Cited by 4
“Although some models have utilized multi-scale convo-lutions for EEG signal feature extraction [9, 26–31], they generally use only a limited set of convolutional kernels to process the EEG signals.”
Vakili, A., Jahanshiri, S., & Salimi-Badr, A. (2026) · PLoS ONE · doi:10.1371/journal.pone.0358241 · Cited by 2
“Section IV discusses the implications of our findings, and Section V concludes the paper.”
Methods for Explaining CNN-Based BCI: A Survey of Recent Applications
Ivanovs, M., Banga, B., Abolins, V., & Nesenbergs, K. (2022) · IEEE International Scientific Conference on Informatics (Informatics) · doi:10.1109/informatics57926.2022.10083473 · Cited by 3
“[52] EEG Weight visualisation of the first hidden layer”
BackgroundWang, F., Wan, Y., Li, Z., Qi, F., & Li, J. (2023) · Frontiers in Neuroscience · doi:10.3389/fnins.2023.1167125 · Cited by 2
“The S/N of the signal can be enhanced effectively while the spatial information of redundancy is further removed (Wang et al., 2021).”
BackgroundSelf-citationWang, J., Wang, F., Yan, H., & Zheng, Z. (2024) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn60899.2024.10650125 · Cited by 3
“Wang et al. [7], introduced the Multiscale-CNN, which utilizes multiple temporal kernels in its temporal convolution layer to extract discriminative time features.”
Cheng, H., Peng, X., Du, Q., & Zhang, H. (2024) · International Conference on Bioinformatics and Biomedical Technology (ICBBT) · doi:10.1145/3674658.3674690 · Cited by 1
“Among these, P300, a positive wave, typically appears around 300 millisec-onds after stimulus presentation and is associated with cognitive processes such as attention, memory, and decision-making[15].”
(2024) · Frontiers research topics · doi:10.3389/978-2-8325-4365-8
Wang, F., Yang, X., & Gao, W. (2025) · Lecture notes in computer science · doi:10.1007/978-981-96-9805-9_4
Zhao, X., Xu, R., Zhang, Y., Lau, A. T., Xu, R., Wang, X., Cichocki, A. S., & Jin, J. (2025) · Journal of Neuroscience Methods · doi:10.1016/j.jneumeth.2025.110546
Bhandari, V., Londhe, N. D., & Kshirsagar, G. B. (2026) · IEEE Transactions on Human-Machine Systems · doi:10.1109/thms.2025.3639248
“In this context, various subject-independent approaches have been developed by prior researchers to address subject variability [13], [14], [15], [16], [17], [18], [19], [20] effectively.”
ATCRN: Attention-guided Temporal Convolutional Remix Network for P300 speller
Shi, B., Liu, M., & Wang, Y. (2026) · Journal of Neuroscience Methods · doi:10.1016/j.jneumeth.2026.110727
Kim, M., Heo, D., Kim, J., & Kim, S.-P. (2026) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2026.3685282
“…were designed to improve calibration e ffi ciency by transfer learning or fine-tuning a subject-independent pretrained model built from large-scale datasets [39], [40], [41], whereas we assumed a scenario in which only subject-specific data were available and thus did not rely on large-scale data.”
Hu, J., Xie, P., Wang, S., & Jiang, G. (2026) · IEEE Sensors Journal · doi:10.1109/jsen.2026.3690538
The robust frequency domain feature and hybrid CNN model for fatigue detection based on sEMG signal
Li, X., Zhao, W., Lin, Y., Teng, F., & Liu, W. (2026) · Computer Methods in Biomechanics & Biomedical Engineering · doi:10.1080/10255842.2026.2675492
Gao, Y., & Diniz, J. M. (2026) · Computer Methods in Biomechanics and Biomedical Engineering Imaging & Visualization · doi:10.1080/21681163.2026.2684109
Park, S., Han, Y. K., Ha, J., Mun, S., & Kim, L. (2026) · Journal of Computational Design and Engineering · doi:10.1093/jcde/qwag083
Metadata
- Authors
- Wang, H., Pei, Z., Xu, L., Xu, T., Bezerianos, A., Sun, Y., & Li, J.
- Affiliations
- Wuyi University (CN)
- University of Patras (GR)
- Zhejiang University (CN)
- University of Essex (GB)
- Venue
- IEEE Transactions on Instrumentation and Measurement
- Published
- 2021
- Volume
- 70
- Pages
- 1-12
- ISSN
- 0018-9456 (ISSN-L)
- 1557-9662
- OpenAlex
- W3138098462
- Cited by
- 35
- References
- 56
- Citations by year
- 2026: 8 · 2025: 6 · 2024: 8 · 2023: 7 · 2022: 5 · 2021: 1
- FWCI
- 2.76
- Type
- journal article
- Language
- en
- Open access
- closed
- Licence
- Locations
- IEEE Transactions on Instrumentation and Measurement (publishedVersion)
- Open Access at Essex (University of Essex) (acceptedVersion)
- Funders
- Special Projects in Key Fields Supported by the Technology Development Project of Guangdong Province (2020ZDZX3018)
- Special Fund for Science and Technology of Guangdong Province (2020182)
- Wuyi University and Hong Kong and Macao Joint Research and Development Project (2019WGALH16)
- Science Foundation for Young Teachers of Wuyi University (2018td01)
- Jiangmen Brain-like Computation and Hybrid Intelligence R&D Center ([2018]359, [2019]26)
- National Natural Science Foundation of China (61806149, 81801785)
- Basic and Applied Basic Research Foundation of Guangdong Province (2020A1515010991)
- Topics
- Life Sciences › Neuroscience › Cognitive Neuroscience › EEG and Brain-Computer Interfaces
- Physical Sciences › Computer Science › Human-Computer Interaction › Gaze Tracking and Assistive Technology
- Physical Sciences › Computer Science › Artificial Intelligence › Emotion and Mood Recognition
- Keywords
- P300 detection, brain-computer interface, multi-scale convolutional neural network, transfer learning, subject-specific modeling, information transfer rate, EEG classification
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.