스트레스 상태 측정을 위한 심전도 신호 QRS 검출 알고리즘

정우혁 ( Woo-hyuk Jung ),이동화 ( Dong-hwa Lee ),이희재 ( Hee-jae Lee ),김재호 ( Jae-ho Kim ),이다빛 ( David Lee ),이상국 ( Sang-goog Lee ) 한국정보처리학회 2014 한국정보처리학회 학술대회논문집 Vol.21 No.1

본 연구에서는 스트레스 상태 측정을 위한 심전도 신호 QRS 검출 알고리즘을 제안한다. 심전도 신호의 QRS 검출 과정은 4단계로 wavelet, moving average, squaring, threshold method로 구성된다. wavelet은 기저선 변동과 노이즈를 제거하고 moving average는 전체 신호를 부드럽게 하고 잔여 노이즈를 제거하며 squaring은 신호를 강조하는 역할을 한다. 마지막으로 threshold 기법을 이용해 검출간격을 설정하여 QRS를 검출하였다. 그 결과 Sensitivity는 99.54%, Positive Predictivity는 99.69%, Detection Error는 0.76%를 보였다. 또한, 피험자를 대상으로 게임을 이용해 스트레스 상태 변화에 대한 실험을 하였고, HRV 시간-주파수 파라미터를 분석함으로써 스트레스 상태 변화를 관찰할 수 있었다.

BCG 또는 CP-2가 생쥐의 막창자꼬리점막상피의 DNA 합성에 미치는 영향

정우혁(Woo-Hyuk Jung),안의태(E-Tay Ahn),박경호(Kyung-Ho Park),박대균(Dae-Kyoon Park),김명수(Myeong-Soo Kim),고정식(Jeong-Sik Ko) 대한해부학회 2006 Anatomy & Cell Biology Vol.39 No.5

이 실험에서는 Ehrlich 암세포를 샅부위에 이식한 후 결핵예방백신으로 이용되는 BCG와 황련(Coptis chinensis)과 파두(Croton tiglium)에서 추출한 생약제제인 CP-2를 투여하였을 때 생쥐 막창자꼬리 점막조직의 형태학적 변화와 DNA합성능을 비교 연구하였다. 실험동물로는 체중 25 gm 내외의 성숙한 생쥐(ICR계통)를 정상대조군과 실험군(암세포이식대조군, 암세포이 식후 BCG 투여군, 암세포이식후 CP-2 투여군)으로 구분하였으며 각 실험군당 5마리씩의 동물을 사용하였다. 실험군 동물들은 샅부위 피부밑조직에 각각 1×107의 Ehrlich암세포를 주사하여 이식하였다. BCG 투여군은 방광암 치료용으로 농축 건조하여 제조된 BCG를 25 g당 0.5 mL, CP-2군은 30 mg/kg을 0.5 mL씩 하루건너 한 번씩 피부밑조직에 주사하였다. 자기방사법적 관찰을 위해서는 모든 실험동물을 마지막 주사 후, 하루주기에 따른 DNA합성능의 변화를 최소화하기 위하여 다음날 오전 9시에서 10시 사이에 3H-thymidine 0.7 μCi/gm을 꼬리에 한번 정맥주사하고, 70분 후 도살하여 막창자꼬리조직을 떼어내어 10% formalin에 고정한 후 표본을 제작하였다. 관찰은 막창자꼬리조직이 세로로 잘 절단된 부위를 택하여 점막근육판을 따라 점막길이 3.5mm의 창자샘상피에 분포하는 3H-thymidine 표지세포수를 계수하였으며 일반조직 관찰을 위해서는 hematoxylin-eosin (H-E)염색을 시행하였다. 일반조직 관찰에서 암세포이식대조군, BCG 투여군 및 CP-2 투여군은 정상대조군에 비하여 고유판에서 림프구와 호산성백혈구가 비교적 많이 관찰된 것 이외에는 별다른 차이가 없었다. 자기방사법적 관찰에서 정상대조군, 암세포이식대조군, BCG 투여군 및 CP-2 투여군은 막창자꼬리 점막길이 3.5mm당 출현하는 은입자 표지세포수가 각각 362.2±56.12, 350.7±71.13, 265.8±27.08 및 241.3±53.29개 였다. 이상의 결과를 종합하여 보면 BCG와 CP-2는 막창자꼬리의 점막상피세포의 DNA합성능을 상당히 억제하나 형태학적인 손상은 거의 주지 않는 약품이라고 생각된다. This experiment was performed to evaluate the morphological responses of the appendicular mucosa of the mouse, inoculated with Ehrlich carcinoma cells in the inguinal area, following administration of BCG or CP-2 (Coptis chinensis-Croton tiglium extracts). Healthy adult ICR mice weighing 25 gm each were divided into normal and experimental groups (experimental control, BCG or CP-2 treated group). Each experimental group mouse was inoculated with 1×107 Ehrlich carcinoma cells subcutaneously in the inguinal area. From the next day after inoculations, 0.2 mL of saline, BCG (0.5 mL/25 g B.W.: 0.03×108~0.32×108 CFU) or CP-2 (30 mg/kg) were injected subcutaneously to the animals every other day, respectively. The day following the 7 th injection of BCG or CP-2, each mouse was injected with a single dose of 0.7μCi/gm of methyl-3H-thymidine (25 Ci/mmol, Amersham Lab., England) through tail vein. Seventy minutes after the tritiated thymidine injection, animals were sacrificed. The number of the labeled epithelial cells of the appendicular mucosae were observed and evaluated. On histological studies of the experimental control, BCG or CP-2 treated mice, general morphologies of the appendicular mucosae were similar. On autoradiographic study, number of the labeled cells of normal control, experimental control, BCG treated or CP-2 treated groups were 362.2±56.12, 350.7±42.65, 265.8±27.08 and 241.3 ±53.29, respectively. Above results show that BCG and CP-2 suppress the DNA synthetic activity of the epithelial cells of the appendix, but did not show any remarkable morphological alterations on the mucosae. These results suggest that BCG and CP-2 are ones of effective anticancer drugs for the cytostatic therapy.

스켈레톤 모델 기반의 포즈 인식 기술을 활용한 낙상 조기 대응 시스템

정우혁(Woo-hyuk Jung),이근재(Geun-jae Lee),배찬석(Chan-seok Bae),홍규량(Gyu-ryang Hong),권지현(Ji-hyun Kwon),유홍석(Hongseok Yoo) 한국컴퓨터정보학회 2023 한국컴퓨터정보학회 학술발표논문집 Vol.31 No.2

유 헬스케어 시스템의 심전도 분석 정확도 향상을 위한 R-peak 검출 방법

정우혁(Woo-Hyuk Jung),이희재(Hee-Jae Lee),김재호(Jae-Ho Kim),이다빛(David Lee),이상국(Sang-Goog Lee) 한국HCI학회 2013 한국HCI학회 학술대회 Vol.2013 No.1

IEEE 802.11 모바일 액세스 포인트의 전력절감을 위한 트래픽 적응방식의 알고리즘 연구

정우혁(Woo-Hyuk Jung),이승형(Seung-Hyong Rhee) 한국통신학회 2017 韓國通信學會論文誌 Vol.42 No.8

무 기준점 특징 추출을 이용한 개인 식별 방법

정우혁(Woo-Hyuk Jung),이상국(Sang-Goog Lee) 한국정보기술학회 2017 한국정보기술학회논문지 Vol.15 No.3

The present study proposes a method of human identification based on features extraction of non-fiducial point using ECG signal. In the pre-processing phase, data is acquired from one-lead, and unwanted noises are removed using wavelet. In the feature extraction phase, features are extracted by applying discrete cosine transform(DCT) to autocorrelation coefficient at intervals of 5 seconds from Normal Sinus Rhythm(NSR) and PTB Diagnostic Database(PTB), respectively. In the classification phase, the test data are matched into the most similar class by applying nearest neighbor(NN) method based on euclidean. As a result, the suggested method shows subject rate 100%, 95.24%, and windows rate 95.6%, 94.4% in NSR and PTB DB, respectively. The suggested method doesnt need extract data of accurate position as well as doesnt require many point features like fiducial point method. In addition, The results can be improved by efficiently eliminating unnecessary windows in advance.

기준점과 무 기준점 기술 결합 기반 특징을 이용한 심전도 개인 식별 연구

정우혁(Woo-Hyuk Jung),한미나(Mi-Na Han),이상국(Sang-Goog Lee) 한국정보기술학회 2017 한국정보기술학회논문지 Vol.15 No.5

The present study proposes ECG human identification using features based on combination of fiducial and non-fiducial technique. In the pre-processing phase, baseline wander and noises of ECG signal are removed using daubechies wavelet. In the feature extraction and selection phase, r peak is detected and the erroneously detected peak is removed. Thereafter, window is generated by grouping 4 r peaks, and features are extracted and selected using autocorrelation and discrete cosine transform. In the classification phase, human identification is performed using linear discriminant classifier. As a result, the proposed method shows subject rate 100%, 100%, and window rate 97.19%, 96.40% in NSR and PTB DB, respectively. The proposed method minimizes the error rate due to detection of the fiducial point boundary by extracting only one important fiducial point and compensates the non-uniform segment problem of the non-fiducial technique by setting the window based on the detected peak. In addition, discrimination test is performed on the extracted r peak to remove the erroneously detected peak in advance, thereby improving the overall identification rate.

Wavelet과 Threshold 기법을 이용한 ECG 신호 QRS 검출 알고리즘

정우혁(Woo-Hyuk Jung),이상국(Sang-Goog Lee) 한국정보기술학회 2013 한국정보기술학회논문지 Vol.11 No.12

In this paper, we present an algorithm for detecting the QRS of an Electrocardiogram(ECG) signal by using wavelet transform and a threshold method. The ECG signal is detected in two phases: the pre-processing phase and the decision phase. The pre-processing phase consists of the stages for the wavelet, moving average(MA), and squaring, whereas the decision phase consists of a single stage that detects the QRS. In the pre-processing phase, the wavelet removes baseline wander and noise. The MA removes the remaining noise of the signal that has gone through the wavelet to make the signal smooth, and squaring plays a role in strengthening the signal. In the decision phase, the threshold is used to set the interval before detecting the QRS. Consequently, the sensitivity was 99.54%, the positive predictivity was 99.69%, and the detection error was 0.76%. Furthermore, the quality of the signal in the data with a various noise was improved, and thus, the QRS was detected effectively.

부정맥 분류를 위한 최적 특징 결합 및 적합도 평가기법

정우혁(Woo-Hyuk Jung),이상국(Sang-Goog Lee) 한국정보기술학회 2017 한국정보기술학회논문지 Vol.15 No.10

This paper proposes arrhythmia classification method that improves the accuracy by combining the features of coefficients extracted from principal component analysis(PCA) and linear discriminant analysis(LDA) and using fitness evaluation technique. In preprocessing step, signal of noise is removed using band pass filter. In feature extraction and selection step, feature coefficients are extracted using daubechies to the preprocessed ECG signal. Each dimension is reduced by using PCA and LDA, and then extracted coefficients are combined to construct new features. In the classification step, the data is classified into four classes using the k-nearest neighbor, and the classified data is examined using the fitness evaluation technique. Finally, reclassification is performed on the data that does not fit the fitness evaluation. As a result, the sensitivity is 98.92%, the specificity is 99.74%, and the positive predictive value is 98.24%. The proposed method improves classification accuracy by constructing optimal feature through combination of coefficient and performing reclassification using fitness evaluation technique.

자기상관과 계층적 분류 모델을 이용한 부정맥 분류 연구

정우혁(Woo-Hyuk Jung),이희재(Hee-Jae Lee),이다빛(David Lee),이상국(Sang-Goog Lee) 한국정보기술학회 2018 한국정보기술학회논문지 Vol.16 No.1

This paper proposes arrhythmia classification that improves the accuracy by using autocorrelation representing temporal information and hierarchical classification model. In preprocessing, noise in electrocardiogram(ECG) is removed using band pass filter. Normalization and Moving multiplication are applied to balance the signal and enhance the peak data. In feature extraction, time information coefficients of ECG are extracted using autocorrelation, and features are selected by reducing redundant or unnecessary data using discrete cosine transform(DCT). Finally, classification is performed using cosine similarity and hierarchical classification model. As a result, sensitivity is 96.79%, specificity is 99.41%, and positive predictive value is 97.64%. The proposed method improves accuracy of arrhythmia classification by extracting few features of ECG signals and applying an efficient hierarchical classification model for multi classification.