Investigation of a Deep-Learning Based Brain–Computer Interface With Respect to a Continuous Control Application

Abstract

The task of a deep neural network (DNN) as a component of a brain-computer interface (BCI) is to analyze the measured EEG data and to recognize the neural signal patterns characteristic of a given motor movement. Our studies are intended to investigate the use of such a DNN in continuous control applications where the EEG signals need to be interpreted continuously, as well as to gain insights into the learned neural patterns. Method: We examined EEGNet, a commonly referenced Convolutional Neural Net (CNN) trained with trials from the BCI Competition IV 2a EEG data set. In addition to the impact of the size and temporal location of the trials on classification accuracy, we examined the contributions and temporal behavior of known neural patterns and their impact on system's response time and the amount of time for which the mental state was detected. Results: Because the optimal temporal positions of the trials are different for the neural patterns involved, we introduced cropped training in which the DNN is trained using trials with different temporal positions. This enabled the DNN to learn the neural patterns in the 0-8 Hz frequency range that are important for a short response time and the patterns in the 8-30 Hz frequency range that are important for determining state duration. Conclusion: Cropped training is essential for achieving a good response time, which could be improved by ~200 ms, as well as for a good state duration, which could be increased from 150 ms to 1.75 s.

Electroencephalogram Analysis Method to Detect Unspoken Answers to Questions Using Multistage Neural Networks

“…Consequently, the present study incorporates a comparative analysis between CNN and DNN models. Moreover, Lawhern et al [46], Zhu et al [55], and Strahnen et al [54] employed EEGNET, whereas Zhu et al [53] and Lim et al [52] employed EEGNET for SSVEP. Despite our study not focusing on MI-EEG or SSVEP, we included EEGNET as a comparative model owing to its high performance in EEG analysis and specialization in artificial intelligence-driven EEG analysis.…”

“…For instance, Lu et al [7] and Sturm et al [8] employed deep neural network (DNN) models, while Yang et al [6], Amin et al [9], Tabar et al [10], Alazrai et al [50], and Moussa et al [51] favored convolutional neural network (CNN) models. Furthermore, EEGNET was employed by Lawhern et al [46], Zhu et al [55], and Strahnen et al [54]. These models have demonstrated superior classification accuracy in comparison to other methods, including DNNs.…”

Motor Imagery Classification Based on EEG Sensing with Visual and Vibrotactile Guidance

“…Hou et al [46] achieved an accuracy of 93.06% by using graph convolutional neural networks (GCNs) in classifying MI of the right hand, left hand, fists, and feet. Strahnen and Kessler [47] used a deep neural network (DNN) and achieved up to 80.7% accuracy in classifying MI of cyclic opening/closing of the left or right fist. On the other hand, Lee et al [48] used a channel-wise variational autoencoder CNN to classify data from Ofner et al [39] and achieved up to 60% accuracy in classifying the elbow extension class against other different MIs of the same limb.…”