Brain dynamic states analysis based on 3D convolutional neural network

Abstract

Drowsiness driving is one major factor of traffic accident. Monitoring the changes of brain signals provides an effective and direct way for drowsiness detection. One 3D convolutional neural network (3D CNN)-based forecasting system has been proposed to monitor electroencephalography (EEG) signals and predict fatigue level during driving. The limited weight sharing and channel-wise convolution were both applied to extract the significant phenomenon in various frequency bands of brain signals and the spatial information of EEG channel location, respectively. The proposed 3D CNN with limited weight sharing and channel-wise convolution has been demonstrated to predict reaction time (RT) of driving with low root mean square error (RMSE) through the brain dynamics. This proposed approach outperforms with the stateof-the-art algorithms, such as traditional CNN, Neural Network (NN), and support vector regression (SVR). Compared with traditional CNN and Artificial Neural Network, the RMSE of 3D CNN-based RT prediction has been improved 9.5% (RMSE from 0.6322 to 0.5720) and 8% (RMSE from 0.6217 to 0.5720), respectively. We envision that this study might open a new branch between deep learning application in neuro-cognitive analysis and real world application.

EEG data analysis with stacked differentiable neural computers

“…The related terminologies are also indicated as in Fig. 8, and a detailed description of the experiment is also contained in [21,22].…”

“…For data pre-processing, the procedures resemble the treatment as in [22]. However in [22] it works with graphs as shown in Fig. 4 (b), after manually adjusting the channel sequence for -axis.…”

“…4 (b) to (c) requires interpolation and extrapolation of original channel values, in this paper all the channels are retained to ensure the interpolating or extrapolating accuracies. It is still common to select some channels according to experience and knowledge to analyze as in [22]. The less picky on EEG data means more information can be kept during transformation.…”

Multilevel Feature Fusion With 3D Convolutional Neural Network for EEG-Based Workload Estimation

“…Several neuroscience researchers have found that the spatiotemporal dynamics of the EEG power spectrum over the scalp are strongly related to the mental workload [7]- [10]. Some studies have reported that the power of the theta (4-8 Hz) oscillations increases in the frontal lobe as the workload increases [7], [9]; the alpha (8)(9)(10)(11)(12) power is linked to idling [11] and cortical inhibition [12]; and the beta (12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30) power is increased in proportion to the workload [7]. The findings imply that the structure of the power spectrum over the frequency bands (theta, alpha, and beta) is closely related to different levels of the workload.…”

“…On the other hand, some studies have converted 1D EEG signals to 2D EEG images (EEG topographic map) to extract the spatial information from multi-channel EEG data over the scalp [16], [23]. In addition, 3D CNN has been utilized to simultaneously extract spectral and spatial information from spectral topographic maps across all frequencies [18], [21]. Recent studies on EEG decoding applications have shown that 3D CNN structures have superior performance in EEG decoding applications [18], [21], [28].…”