웨어러블 컴퓨터 R&BD (Research and Business Development) 전략

이상국(Sang-Goog Lee) 한국정보과학회 2015 정보과학회지 Vol.33 No.11

동작 상상 EEG 분류를 위한 Ruelle-Takens 매립정리 기반 정규화 공통 공간 패턴

박상훈(Sang-Hoon Park),이다빛(David Lee),이상국(Sang-Goog Lee) 한국정보기술학회 2017 한국정보기술학회논문지 Vol.15 No.7

The Brain-Computer Interface (BCI) research is being actively pursued in the areas of signal processing, engineering, and rehabilitation medicine. Common Spatial Pattern (CSP) is very useful as a method for analyzing motor imagery Electroencephalogram (EEG) and is used in many studies. However, the CSP method shows poor performance in the Small Sample Setting (SSS) situation. In addition, noise and non-stationary in the EEG interfere with classification. In this paper, we propose the Ruelle-Takens embedding theorem based regularized common spatial Pattern(RTR-CSP) to overcome these problems. The proposed method consists of four steps. First, The obtained EEG is band-pass filtered in the 7~30Hz band. Second, we create a new signal matrix with time delay added. Third, regularized CSP is applied to signal. finally, the signal is classified through an ensemble. Compared with CSP, R-CSP and CSSP methods, the proposed method yielded 12.53%, 5.24% and 1.99% better performance, respectively.

Small Sample Setting 동작 상상 EEG 분류를 위한 필터 뱅크 기반 정규화 공통 공간 패턴 앙상블 기법 연구

박상훈(Sang-Hoon Park),이다빛(David Lee),이상국(Sang-Goog Lee) 한국정보기술학회 2017 한국정보기술학회논문지 Vol.15 No.4

For the last few years, BCI(Brain-Computer Interface) researches have been pursued in the fields of medicine and engineering. Specially, the CSP(Common Spatial Pattern) method was widely used because it is useful for motor imagery eeg classification. However, it is difficult to select the frequency range according to the subjects, and it shows low performance in the SSS(Small Sample Setting) situation. In this paper, we propose a filter bank based regularized CSP to solve the frequency range selection problem and the problem in SSS situation. The proposed method consists of four steps. First, the filter bank is used to divide the signal. Second, normalized CSP is applied to each signal. Third, the features are selected by the MIBIF(Mutual Information Based Individual Feature) method. Fourth, the classification and the ensemble is based on the selected feature. The proposed method showed better performance than did the conventional methods. Especially, it showed better performance in SSS situation.

필터 뱅크 정규화 공통 공간-스펙트럼 패턴을 이용한 동작 상상 뇌파 분류

박상훈(Sang-Hoon Park),이다빛(David Lee),이상국(Sang-Goog Lee) 한국정보기술학회 2017 한국정보기술학회논문지 Vol.15 No.10

Recently, the Brain-Computer Interface(BCI) research based on motor imagery Electroencephalogram(EEG) receive attention in various fields such as engineering, medicine, and industry. The various methods are proposed for motor imagery EEG classification. Among them, Common Spatial Pattern(CSP) method is very effective in classifying motor imagery EEG. Therefore, it is applied in many studies. However, the CSP includes several problems. First, it is highly dependent on the frequency band. Second, it performs relatively weak in environments with a small number of data. Third, it does not consider the non-stationary of EEG. In this study, we propose filter bank regularized common spatio-spectral pattern method for overcoming these problems. The proposed method regularizes the time delay signal and non-time delay signal obtained from the filter bank, and extracts the features. The extracted features are classified through the ensemble. The proposed method outperforms than CSP, R-CSP, CSSP and FBCSP methods.

동작 상상 EEG 분류를 위한 필터 뱅크 기반 정규화 공통 공간 패턴

박상훈(Sang-Hoon Park),김하영(Ha-Young Kim),이다빛(David Lee),이상국(Sang-Goog Lee) 한국정보과학회 2017 정보과학회논문지 Vol.44 No.6

최근, 동작 상상(Motor Imagery) Electroencephalogram(EEG)를 기반으로 한 Brain-Computer Interface(BCI) 시스템은 의학, 공학 등 다양한 분야에서 많은 관심을 받고 있다. Common Spatial Pattern(CSP) 알고리즘은 동작 상상 EEG의 특징을 추출하기 위한 가장 유용한 방법이다. 그러나 CSP 알고리즘은 공분산 행렬에 의존하기 때문에 Small-Sample Setting(SSS) 상황에서 성능에 한계가 있다. 또한 사용하는 주파수 대역에 따라 큰 성능 차이를 보인다. 이러한 문제를 동시에 해결하기 위해, 4-40Hz 대역 EEG 신호를 9개의 필터 뱅크를 이용하여 분할하고 각 밴드에 Regularized CSP(R-CSP)를 적용한다. 이후 Mutual Information-Based Individual Feature(MIBIF) 알고리즘은 R-CSP의 차별적인 특징을 선택하기 위해 사용된다. 본 연구에서는 대뇌 피질의 운동영역 부근 18개 채널을 사용하여 BCI CompetitionIII DatasetⅣa의 피험자 다섯 명(aa, al, av, aw 및 ay)에 대해 각각 87.5%, 100%, 63.78%, 82.14% 및 86.11%의 정확도를 도출하였다. 제안된 방법은 CSP, R-CSP 및 FBCSP 방법보다 16.21%, 10.77% 및 3.32%의 평균 분류 정확도 향상이 있었다. 특히, 본 논문에서 제안한 방법은 SSS 상황에서 우수한 성능을 보였다. Recently, motor imagery electroencephalogram(EEG) based Brain-Computer Interface (BCI) systems have received a significant amount of attention in various fields, including medicine and engineering. The Common Spatial Pattern(CSP) algorithm is the most commonly-used method to extract the features from motor imagery EEG. However, the CSP algorithm has limited applicability in Small-Sample Setting(SSS) situations because these situations rely on a covariance matrix. In addition, large differences in performance depend on the frequency bands that are being used. To address these problems, 4-40Hz band EEG signals are divided using nine filter-banks and Regularized CSP(R-CSP) is applied to individual frequency bands. Then, the Mutual Information-Based Individual Feature(MIBIF) algorithm is applied to the features of R-CSP for selecting discriminative features. Thereafter, selected features are used as inputs of the classifier Least Square Support Vector Machine (LS-SVM). The proposed method yielded a classification accuracy of 87.5%, 100%, 63.78%, 82.14%, and 86.11% in five subjects(“aa”, “al”, “av”, “aw”, and “ay”, respectively) for BCI competition III dataset Iva by using 18 channels in the vicinity of the motor area of the cerebral cortex. The proposed method improved the mean classification accuracy by 16.21%, 10.77% and 3.32% compared to the CSP, R-CSP and FBCSP, respectively The proposed method shows a particularly excellent performance in the SSS situation.

D2D 시선통신을 위한 사용자 인터페이스(User Interface) 연구

이희재(Lee Hee-Jae),김재호(Kim Jae-Ho),이다빛(Lee David),정우혁(Jung Woo-Hyuk),이상국(Lee Sang-Goog) 한국통신학회 2012 한국통신학회 학술대회논문집 Vol.2012 No.2

동작상상 EEG 기반 BCI 특징 추출 방법 연구

박상훈(Sang-Hoon Park),이다빛(David Lee),이상국(Sang-Goog Lee) 한국HCI학회 2016 한국HCI학회 학술대회 Vol.2016 No.1

서브 밴드 CSP기반 FLD 및 PCA를 이용한 동작 상상 EEG 특징 추출 방법 연구

박상훈(Sang-Hoon Park),이상국(Sang-Goog Lee) 한국정보과학회 2015 정보과학회논문지 Vol.42 No.12

뇌-컴퓨터 인터페이스는 사용자의 뇌전도(Electroencephalogram: EEG)를 획득하여 생각만으로 기계를 제어하거나 신체장애를 가진 사람에게 손 또는 발과 같은 신체를 대신하여 의사 전달 수단으로 사용될 수 있다. 본 논문에서는 동작 상상 EEG를 분류하기 위해 Sub-Band Common Spatial Pattern (SBCSP)를 기반으로 필터 선택을 하지 않는 특징 추출 방법에 대해 연구한다. 4~40Hz의 동작 상상 신호를 4Hz 대역마다 나눈 9개의 서브 밴드에 각각 CSP를 적용한다. 이후 Fisher"s Linear Discriminant (FLD)를 사용하여 도출된 값들을 결합한 FLD 점수 벡터에 차원 축소를 위한 Principal Component Analysis(PCA)를 적용하여 클래스 구분을 위한 최적의 평면에 특징을 투영한다. 데이터베이스는 BCI CompetitionⅢ dataset Ⅳa(2 클래스: 오른손・다리)를 이용하며, 추출된 특징은 Least Squares Support Vector Machine(LS-SVM)의 입력으로 사용된다. 제안된 방법의 성능은 10×10 fold cross-validation을 이용하여 분류 정확도로 나타낸다. 본 논문에서 제안하는 방법은 피험자 ‘aa’, ‘al’, ‘av’, ‘aw’, ‘ay’에 대하여 각각 85.29±0.93%, 95.43±0.57%, 72.57±2.37%, 91.82±1.38%, 93.50±0.69%의 분류 정확도를 보였다. The brain-computer interface obtains a user"s electroencephalogram as a replacement communication unit for the disabled such that the user is able to control machines by simply thinking instead of using hands or feet. In this paper, we propose a feature extraction method based on a non-selected filter by SBCSP to classify motor imagery EEG. First, we divide frequencies (4~40 Hz) into 4-Hz units and apply CSP to each Unit. Second, we obtain the FLD score vector by combining FLD results. Finally, the FLD score vector is projected onto the optimal plane for classification using PCA. We use BCI Competition III dataset IVa, and Extracted features are used as input for LS-SVM. The classification accuracy of the proposed method was evaluated using 10×10 fold cross-validation. For subjects ‘aa’, ‘al’, ‘av’, ‘aw’, and ‘ay’, results were 85.29 ± 0.93%, 95.43 ± 0.57%, 72.57 ± 2.37%, 91.82 ± 1.38%, and 93.50 ± 0.69%, respectively.

웨이블릿 변환과 주성분 분석 기반의 결합 특징 벡터를 이용한 동작 상상 EEG 분류

이다빛(David Lee),박상훈(Sang-Hoon Park),이희재(Hee-Jae Lee),이상국(Sang-Goog Lee) 한국정보기술학회 2017 한국정보기술학회논문지 Vol.15 No.4

The brain-computer interface is a communication channel that controls and manipulates mechanism by transmitting human thoughts from a human to a mechanism. To do this, it is necessary to extract appropriate electroencephalogram (EEG) features. In this paper, we propose a combined feature vectors based on wavelet transform and principal component analysis to extract features of EEG signals. The proposed method consists of three steps: In the first step, the wavelet transform is applied to extract feature vectors of the motor imagery EEG. In the second step, the principal component analysis was used to reduce the dimensionality of the feature vectors. In the third step, the nonlinear support vector machines was applied to classify left or right hand motor imagery EEG. The performance of the proposed method was evaluated in terms of accuracy through three different motor imagery EEG datasets. The results of this study showed that the proposed method achieved higher accuracy in the motor imagery EEG classification. The results of this study show that the proposed method improve the classification accuracy of some subjects by 0.4~4.3% compared to existing single feature methods, thus achieves a high classification accuracy of 82.8%.

GMM과 SVM을 이용한 움직임 상상 뇌파 분류에 관한 연구

이다빛(David Lee),이상국(Sang-Goog Lee) 한국정보기술학회 2013 한국정보기술학회논문지 Vol.11 No.7

The Brain-Computer Interface (BCI) use the Electroencephalogram (EEG) as a method to replace existing interface technologies. Therefore, BCI is the necessary technology not only for general users but also for disabled who can not move the muscle because of the nervous system problem and the senior who are uncomfortable with movement. In this paper, we propose a method which uses the support vector machine (SVM) with the support vector that is generated by the Gaussian mixture model (GMM) to classify movement imagery EEGs. This support vector was robust to noise and overfitting. The EEG classification consists of the feature extraction and the classification process. In the feature extraction process, the wavelet transform was used to decompose the EEG and extract the statistical feature of the wavelet coefficient. In the classification process, the expectation maximization (EM) algorithm was used to obtain the maximum likelihood estimation (MLE) of the distribution of statistical feature of the GMM. Lastly, we utilized the average of the estimated Gaussian distribution to generate the support vector. Using the proposed method, the classification accuracy of movement imagery EEGs was found to be 83.61%.