Zijing Mao

EEG-based prediction of driver's cognitive performance by deep convolutional neural network

We considered the prediction of drivers cognitive states related to driving performance using EEG signals. We proposed a novel channel-wise convolutional neural network (CCNN) whose architecture considers the unique characteristics of EEG data. We also discussed CCNN-R, a CCNN variation that uses Restricted Boltzmann Machine to replace the convolutional filter, and derived the detailed algorithm. To test the performance of CCNN and CCNN-R, we assembled a large EEG dataset from 3 studies of driver fatigue that includes samples from 37 subjects. Using this dataset, we investigated the new CCNN and CCNN-R on raw EEG data and also Independent Component Analysis (ICA) decomposition. We tested both within-subject and cross-subject predictions and the results showed CCNN and CCNN-R achieved robust and improved performance over conventional DNN and CNN as well as other non-DL algorithms. Novel channel-wise deep convolutional neural networks for prediction of drivers cognitive states from EEG signals.Two different deep learning structures CCNN and CCNN-R are proposed.CCNN and CCNN-R achieved robust and improved performance over conventional DNN and CNN.

Prediction of driver's drowsy and alert states from EEG signals with deep learning

We investigate in this paper deep learning (DL) solutions for prediction of drivers cognitive states (drowsy or alert) using EEG data. We discussed the novel channel-wise convolutional neural network (CCNN) and CCNN-R which is a CCNN variation that uses Restricted Boltzmann Machine in order to replace the convolutional filter. We also consider bagging classifiers based on DL hidden units as an alternative to the conventional DL solutions. To test the performance of the proposed methods, a large EEG dataset from 3 studies of drivers fatigue that includes 70 sessions from 37 subjects is assembled. All proposed methods are tested on both raw EEG and Independent Component Analysis (ICA)-transformed data for cross-session predictions. The results show that CCNN and CCNN-R outperform deep neural networks (DNN) and convolutional neural networks (CNN) as well as other non-DL algorithms and DL with raw EEG inputs achieves better performance than ICA features.

A Deep Learning method for classification of images RSVP events with EEG data

In this paper, we investigated Deep Learning (DL) for characterizing and detecting target images in an image rapid serial visual presentation (RSVP) task based on EEG data. We exploited DL technique with input feature clusters to handle high dimensional features related to time - frequency events. The method was applied to EEG recordings of a RSVP experiment with multiple sessions and subjects. For classification of target and non-target images, a deep belief net (DBN) classifier was based on the uncorrelated features, which was constructed from original correlated features using clustering method. The performance of the proposed DBN was tested for different combinations of hidden units and hidden layers on multiple subjects. The results of DBN were compared with cluster Linear Discriminant Analysis (cLDA) and Support vector machine (SVM) and DBN demonstrated better performance in all tested cases. There was an improvement of 10 - 25% for certain cases. We also demonstrated how DBN is used to characterize brain activities.

Classification of non-time-locked rapid serial visual presentation events for brain-computer interaction using deep learning

Deep learning solutions based on deep neural networks (DNN) and deep stack networks (DSN) were investigated for classifying target images in a non-time-locked rapid serial visual presentation (RSVP) image target identification task using EEG. Several feature extraction methods associated with this task were implemented and tested for deep learning, where a sliding window method using the trained classifier was used to predict the occurrence of target events in a non-time-locked fashion.. The deep learning algorithms explored based on deep stacking networks were able to improve the error rate by about 5% over existing algorithms such as linear discriminant analysis (LDA) for this task. Initial test results also showed that this method based on deep stacking networks for non-time-locked classification can produce an error rate close to that achieved for time-locked classification, thus illustrating the power of deep learning for complex feature spaces.

BIMMER: a novel algorithm for detecting differential DNA methylation regions from MBDCap-seq data.

DNA methylation is a common epigenetic marker that regulates gene expression. A robust and cost-effective way for measuring whole genome methylation is Methyl-CpG binding domain-based capture followed by sequencing (MBDCap-seq). In this study, we proposed BIMMER, a Hidden Markov Model (HMM) for differential Methylation Regions (DMRs) identification, where HMMs were proposed to model the methylation status in normal and cancer samples in the first layer and another HMM was introduced to model the relationship between differential methylation and methylation statuses in normal and cancer samples. To carry out the prediction for BIMMER, an Expectation-Maximization algorithm was derived. BIMMER was validated on the simulated data and applied to real MBDCap-seq data of normal and cancer samples. BIMMER revealed that 8.83% of the breast cancer genome are differentially methylated and the majority are hypo-methylated in breast cancer.

Asynchronous control of unmanned aerial vehicles using a steady-state visual evoked potential-based brain computer interface

AbstractThe goal of this study is to design an asynchronous steady-state visual evoked potential (SSVEP) BCI system to enable control of an unmanned aerial vehicle (UAV) with multiple commands. An SSVEP-based BCI system with six different flickering frequencies was constructed to realize six actuation commands for UAV control. In addition, asynchronous control was achieved by including a detection of the ‘idle’ brain state using a novel likelihood ratio test and the hover command was implemented for the idle state. Offline recording was conducted to evaluate the detection accuracies and a game-like online experiment was also conducted to assess the online performance of the proposed system. Forty-two subjects participated in offline recordings to evaluate the detection accuracy of commands as well as detection of the ‘idle’ state. An average error rate of 15% was obtained for detecting the six commands, whereas an average error rate of 23.06% was obtained for differentiating commands from idle brain state...

Prediction of Human Performance Using Electroencephalography under Different Indoor Room Temperatures

Varying indoor environmental conditions is known to affect office worker’s performance; wherein past research studies have reported the effects of unfavorable indoor temperature and air quality causing sick building syndrome (SBS) among office workers. Thus, investigating factors that can predict performance in changing indoor environments have become a highly important research topic bearing significant impact in our society. While past research studies have attempted to determine predictors for performance, they do not provide satisfactory prediction ability. Therefore, in this preliminary study, we attempt to predict performance during office-work tasks triggered by different indoor room temperatures (22.2 °C and 30 °C) from human brain signals recorded using electroencephalography (EEG). Seven participants were recruited, from whom EEG, skin temperature, heart rate and thermal survey questionnaires were collected. Regression analyses were carried out to investigate the effectiveness of using EEG power spectral densities (PSD) as predictors of performance. Our results indicate EEG PSDs as predictors provide the highest R2 (> 0.70), that is 17 times higher than using other physiological signals as predictors and is more robust. Finally, the paper provides insight on the selected predictors based on brain activity patterns for low- and high-performance levels under different indoor-temperatures.

Deep Transfer Learning for Cross-subject and Cross-experiment Prediction of Image Rapid Serial Visual Presentation Events from EEG Data

Transfer learning (TL) has gained significant interests recently in brain computer interface (BCI) as a key approach to design robust predictors for cross-subject and cross-experiment prediction of the brain activities in response to cognitive events. We carried out in this.aper the first comprehensive investigation of the transferability of deep convolutional neural network (CNN) for cross-subject and cross-experiment prediction of image Rapid Serial Visual Presentation (RSVP) events. We show that for both cross-subject and cross-experiment predictions, all convolutional layers and fully connected layers contain both general and subject/experiment-specific features and transfer learning with weights fine-tuning can improve the prediction performance over that without transfer. However, for cross-subject prediction, the convolutional layers capture more subject-specific features, whereas for cross-experiment prediction, the convolutional layers capture more general features across experiment. Our study provides important information that will guide the design of more sophisticated deep transfer learning algorithms for EEG based classifications in BCI applications.

Prediction of temperature induced office worker's performance during typing task using EEG

Understanding how indoor environment affects office workers performance and developing methods to predict human performance in changing indoor environment have become highly important research topic bearing significant economic and sociological impact. While past research groups have attempted to find predictors for performance they do not provide satisfactory predictions. We conduct in this paper a study to predict human performance by developing a regression model using neurophysiological signals collected from electroencephalogram (EEG), during simulated office-work tasks under different indoor room temperatures (22°C and 30°C). We found that using brain power spectral densities (PSD) from EEG as predictors provides the higher R2 than predictors using skin temperature or heart rate by approximately over 3 folds. Finally, we showed that the predictor using EEG is more robust than regression models using skin temperature and heart rate. Our work shows the potential of using brain signals to accurately predict human office work performance.

Predicting EEG Sample Size Required for Classification Calibration

This study considers an important problem of predicting required calibration sample size for electroencephalogram EEG-based classification in brain computer interaction BCI. We propose an adaptive algorithm based on learning curve fitting to learn the relationship between sample size and classification performance for each individual subject. The algorithm can always provide the predicted result in advance of reaching the baseline performance with an average error of 17.4i¾?%. By comparing the learning curve of different classifiers, the algorithm can also recommend the best classifier for a BCI application. The algorithm also learns a sample size upper bound from the prior datasets and uses it to detect subject outliers that potentially need excessive amount of calibration data. The algorithm is applied to three EEG-based BCI datasets to demonstrate its utility and efficacy. A Matlab package with GUI is also developed and available for downloading at https://github.com/ZijingMao/LearningCurveFittingForSampleSizePrediction. Since few algorithms are yet available to predict performance for BCIs, our algorithm will be an important tool for real-life BCI applications.