Deep learning based on Batch Normalization for P300 signal detection

Liu, Mingfei; Wu, Wei; Gu, Zhenghui; Yu, Zhuliang; Qi, Feifei; Li, Yuanqing

doi:10.1016/j.neucom.2017.08.039

Cited by 206 publications

(137 citation statements)

References 14 publications

Supporting

Mentioning

131

Contrasting

Unclassified

Order By: Relevance

“…We record the information translate rates of datasets I and II in different algorithms. As shown in Figures 4 and 5, the ITR value of the PCA-CNN is higher than that of the other three algorithms (BN3 [28], CNN-1 [24], and SVM [21]) in the maximum repeat. In Figure 4, after the 7th repeat, the ITR value of PCA-CNN is higher than the other three algorithms (BN3 [28], CNN-1 [24] and SVM [21]).…”

Section: Resultsmentioning

confidence: 83%

“…In dataset II, there are 20 training characters and 30 test characters for each subject, each of which is repeated 10 times. Figure 3 presents the test accuracy rates of the proposed PCA-CNN and other classification methods in the literature, including CNN algorithms BN3 [28] and CNN-1 [24], and the traditional SVM algorithm [21], on datasets II of 10 subjects. The different color line in the figure records the results of all subjects that used different methods on each repeat.…”

Section: Resultsmentioning

confidence: 99%

“…where N represents the number of classes, P represents the accuracy rate of character recognitions, and T represents the time taken for character recognitions [24,28]. In dataset I, as each flash lasts for 100 ms followed by a pause of 75 ms (12 * (75 + 100) = 2100) and a pause of 2.5 s between each character epoch, T can thus be defined as:…”

Section: Discussionmentioning

confidence: 99%

“…LIU improved the convolutional neural network on the basis of Cecotti's algorithm which was named BN3 algorithm, taking the batch normalization (BN) layer and the dropout (DP) layer to deepen the network layers and overcome the problem of overfitting. The BN3 algorithm achieved good results in classification, but still need to improve the recognition accuracy when the number of experiments are reduced [28].…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

A Novel P300 Classification Algorithm Based on a Principal Component Analysis-Convolutional Neural Network

Wang

et al. 2020

Applied Sciences

View full text Add to dashboard Cite

Aiming at enhancing the classification accuracy of P300 Electroencephalogram signals in a non-invasive brain–computer interface system, a novel P300 electroencephalogram signals classification algorithm is proposed which is based on improved convolutional neural network. In the data preprocessing part, the proposed P300 classification algorithm used the Principal Component Analysis algorithm to not only remove the noise and artifacts in the data, but also increase the data processing speed. Furthermore, the proposed P300 classification algorithm employed the parallel convolution method to improve the traditional convolutional neural network framework, which can increase the network depth and improve the network’s ability to classify P300 electroencephalogram signals. The proposed algorithm was evaluated by two datasets (the dataset from the competition and the dataset from the laboratory). The results show that, in the dataset I, the proposed P300 classification algorithm could obtain accuracy rates higher than 95%, and achieve one of the best performances in four classification algorithms, while, in the dataset II, the proposed P300 classification algorithm can get accuracy rates higher than 90%, and is superior to the other three algorithms in all ten subjects. These demonstrated the effectiveness of the proposed algorithm. The proposed classification algorithm can be applied in the actual brain–computer interface systems to help people with disability in the daily lives.

show abstract

Section: Resultsmentioning

confidence: 83%

Section: Resultsmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

A Novel P300 Classification Algorithm Based on a Principal Component Analysis-Convolutional Neural Network

Wang

et al. 2020

Applied Sciences

View full text Add to dashboard Cite

show abstract

“…DCNN applications in EEG-based BCI research include mental imagery decoding, primarily for mobility applications [17], [18] and Event-Related Potential (ERP) detection, primarily for BCI speller applications [19], [20]. The first study to use DCNN in SSVEP classification [14] is implemented for a 5 class problem.…”

Section: Related Workmentioning

confidence: 99%

On the Relative Contribution of Deep Convolutional Neural Networks for SSVEP-Based Bio-Signal Decoding in BCI Speller Applications

Podmore

Breckon

Aznan

et al. 2019

IEEE Trans. Neural Syst. Rehabil. Eng.

View full text Add to dashboard Cite

Brain-computer interfaces (BCI) harnessing Steady State Visual Evoked Potentials (SSVEP) manipulate the frequency and phase of visual stimuli to generate predictable oscillations in neural activity. For BCI spellers, oscillations are matched with alphanumeric characters allowing users to select target numbers and letters. Advances in BCI spellers can, in part, be accredited to subject-specific optimization, including; 1) custom electrode arrangements, 2) filter sub-band assessments and 3) stimulus parameter tuning. Here we apply deep convolutional neural networks (DCNN) demonstrating cross-subject functionality for the classification of frequency and phase encoded SSVEP. Electroencephalogram (EEG) data are collected and classified using the same parameters across subjects. Subjects fixate forty randomly cued flickering characters (5 × 8 keyboard array) during concurrent wet-EEG acquisition. These data are provided by an open source SSVEP dataset. Our proposed DCNN, PodNet, achieves 86% and 77% offline Accuracy of Classification acrosssubjects for two data capture periods, respectively, 6-seconds (information transfer rate=40bpm) and 2-seconds (information transfer rate= 101bpm). Subjects demonstrating sub-optimal (<70%) performance are classified to similar levels after a short subject-specific training period. PodNet outperforms filter-bank canonical correlation analysis (FBCCA) for a low volume (3channel) clinically feasible occipital electrode configuration. The networks defined in this study achieve functional performance for the largest number of SSVEP classes decoded via DCNN to date. Our results demonstrate PodNet achieves cross-subject, calibrationless classification and adaptability to sub-optimal subject data and low-volume EEG electrode arrangements.

show abstract