Browsing by Author "Huang, Yufei"

Now showing 1 - 4 of 4

EEG-based prediction of driver\'s cognitive performance by deep convolutional neural network
(2016-09)
Hajinoroozi, Mehdi
;
Mao, Zijing
;
Jung, Tzyy-Ping
;
Lin, Chin-Teng
;
Huang, Yufei
We considered the prediction of driver\'s 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. (C) 2016 Elsevier B.V. All rights reserved.
FEATURE EXTRACTION WITH DEEP BELIEF NETWORKS FOR DRIVER\'S COGNITIVE STATES PREDICTION FROM EEG DATA
(2015)
Hajinoroozi, Mehdi
;
Jung, Tzyy-Ping
;
Lin, Chin-Teng
;
Huang, Yufei
;
腦科學研究中心
This study considers the prediction of driver\'s cognitive states from electroencephalographic (EEG) data. Extracting EEG features correlated with driver\'s cognitive states is key for achieving accurate prediction. However, high dimensionality and temporal-and-spatial correlations of EEG data make extraction of effective features difficult. This study explores the approaches based on deep belief networks (DBN) for feature extraction and dimension reduction. Experimental results of this study showed that DBN applied to channel epochs (DBN-C) produces the most discriminant features and the best classification performance is achieved when DBN-C is applied to the time-frequency and independent-component analysis transformed EEG data. The results suggested that DBN-C is a promising new method for extracting complex, discriminant features for EEG-based brain computer interfaces.
Predicting EEG Sample Size Required for Classification Calibration
(2016-01-01)
Mao, Zijing
;
Jung, Tzyy-Ping
;
Lin, Chin-Teng
;
Huang, Yufei
;
腦科學研究中心
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 17.4 %. 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.
Target Classification in a Novel SSVEP-RSVP Based BCI Gaming System
(2019-01-01)
Nayak, Tapsya
;
Ko, Li-Wei
;
Jung, Tzyy-Ping
;
Huang, Yufei
;
生物科技學系
Recently game-based brain-computer interface (BCI) systems using electroencephalography (EEG) has been gaining popularity, providing a sophisticated experience to its users. Here we present such a novel hybrid system based on rapid serial visual presentation (RSVP) in conjunction with steady-state visual evoked potentials (SSVEP). Based on a matching computer game Jewel Quest a game is designed wherein a sequence of jewel images containing rare targets (< 3%) in an RSVP paradigm is presented on a display at four distinct locations each flickering at different rates (4, 5, 6 and 7 Hz). A score is awarded upon successful detection of target image from neural signals. During real-time implementation to achieve higher classification speeds, EEG signals were epoched at the onset of each image, creating a high degree of class overlap and imbalance. Given these challenges in our EEG datasets, we present classifiers that can classify single-trial EEG epochs at the onset of target image presentation accurately. Initial results from 14 subjects indicate Hidden Markov Model (HMM) with Dirichlet emission probabilities provide similar to 1% higher, on average, the area under the precision-recall curve (AUC-PR) compared to the ensemble technique Bagging, commonly used to handle class imbalance.