25 December 2020 in IEEE Transactions on Instrumentation and Measurement
Electroencephalography (EEG) signals have been proven to be one of the most predictive and reliable indicators for estimating driving fatigue state. However, how to make full use of EEG data for driving fatigue detection remains a challenge. Many existing methods include a time-consuming manual process or tedious parameter tunings for feature extraction, which is inconvenient to train and implement. On the other hand, most models ignore or manually determine EEG connectivity features between different channels, thus failing to thoroughly exploit the intrinsic interchannel relations for classification. In this article, we introduce a new attention-based multiscale convolutional neural network-dynamical graph convolutional network (AMCNN-DGCN) model, aiming to conquer these two issues in a unified end-to-end model. AMCNN-DGCN starts with attention-based multiscale temporal convolutions to automatically learn frequency filters to extract the salient pattern from raw EEG data. Subsequently, AMCNN-DGCN uses dynamical graph convolutional networks (DGCNs) to learn spatial filters, in which the adjacency matrix is adaptively determined in a data-driven way to exploit the intrinsic relationship between channels effectively. With the temporal–spatial structure, AMCNN-DGCN can capture highly discriminative features. To verify the effectiveness of AMCNN-DGCN, we conduct a simulated fatigue driving environment to collect EEG signals from 29 healthy subjects (male/female = 17/12 and age = 23.28±2.70 years) through a remote wireless cap with 24 channels. The results demonstrate that our proposed model outperforms six widely used competitive EEG models with high accuracy of 95.65%. Finally, the critical brain regions and connections for driving fatigue detection were investigated through the dynamically learned adjacency matrix.
22 March 2021 in IEEE Transactions on Instrumentation and Measurement
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.