• Proceedings of the KOSOMBE Conference
• /
• v.1996 no.05
• /
• pp.283-287
• /
• 1996

This paper described the estimator and eliminator far three kinds of artifacts in electrocardiogram. The most efficient estimation of baseline drift could be obtain in the cubic spline interpolation method with the PQ and TP segment which are considered to be isoelectric, from the experimental results obtained from the applied 4 types of algorithms. The time loss and distortion could be avoided with the aid of detection criteria by checking if baseline drifts exist or not. The AIEF proposed in this paper was verified as having the best removal performance with less distortion in the QRS complex through the comparison of 5 proposed algorithms. furthermore, the AIEF are most suitable far the ECG analyzer which was only needed relatively short time data due to the fast conversion into the stable state. The proposed parabolic filter with 11 points width was identified as having the best performance for the elimination of muscle artifacts. Also we could obtain 99.7% detection accuracy of spike component and minimize the error identifying QRS complex as spike.

• Journal of Korea Multimedia Society
• /
• v.10 no.2
• /
• pp.218-225
• /
• 2007

This paper presents a technique to detect a document skew that often occurs during document scanning. To correct a skewed document is essential for automatic processing system including character segmentation, character recognition and so on. The proposed algorithm can detect a skew angle exactly by searching characters baselines that have slant information of the document within a candidated area. To reduce processing time, we resized the image small and then established a ROI (region of interest) by morphology operations and connected components analysis. We compared our method with the existing method based on morphology operations and proved correctness and efficiency of the proposed algorithm through experiments and analysis with various kind of document images.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.11
• /
• pp.5546-5559
• /
• 2019

Facial action units (AUs) have recently drawn increased attention because they can be used to recognize facial expressions. A variety of methods have been designed for frontal-view AU detection, but few have been able to handle multi-view face images. In this paper we propose a method for multi-view facial AU detection using a fused multilayer, multi-task, and multi-label deep learning network. The network can complete two tasks: AU detection and facial view detection. AU detection is a multi-label problem and facial view detection is a single-label problem. A residual network and multilayer fusion are applied to obtain more representative features. Our method is effective and performs well. The F1 score on FERA 2017 is 13.1% higher than the baseline. The facial view recognition accuracy is 0.991. This shows that our multi-task, multi-label model could achieve good performance on the two tasks.

• Journal of Biomedical Engineering Research
• /
• v.17 no.2
• /
• pp.235-240
• /
• 1996

The automated ECG diagnostic systems that are odd in hospitals have low performance of P-wave detection when faced with some diseases such as conduction block. So, the purpose of this study was the improvement of detection performance in conduction block which is low in P-wave detection. The first procedure was removal of baseline drift by subtracting the median filtered signal of 0.4 second length from the original signal. Then the algorithm detected R peak and T end point and cancelled the QRS-T complex to get'p prototypes'. Next step was magnification of P prototypes with dispersion and detection of'p candidates'in the magnified signal, and then extraction of contextual information concerned with P-waves. For the last procedure, the CIR was applied to P candidates to confirm P-waves. The rule base consisted of three rules that discriminate and confirm P-waves. This algorithm was evaluated using 500 patient's raw data P-wave detection perFormance was in- creased 6.8% compared with the QRS-T complex cancellation method without application of the rule base.

• Journal of Sensor Science and Technology
• /
• v.22 no.5
• /
• pp.338-345
• /
• 2013

Premature ventricular contractions are the most common of all arrhythmias and may cause more serious situation like ventricular fibrillation and ventricular tachycardia in some patients. Therefore, the detection of this arrhythmia becomes crucial in the early diagnosis and the prevention of possible life threatening cardiac diseases. Most methods for detecting arrhythmia require pp interval, or the diversity of P wave morphology, but they are difficult to detect the p wave signal because of various noise types. Thus, it is necessary to use noise-free R wave. So, the new approach for the detection of PVC is presented based on the rhythm analysis and the beat matching in this paper. For this purpose, we removed baseline wandering of low frequency band and made summed signals that are composed of two high frequency bands including the frequency component of QRS complex using the wavelet filter. And then we designed R wave detection algorithm using the adaptive threshold and window through RR interval. Also, we developed algorithm to classify PVC using RR interval. The performance of R wave and PVC detection is evaluated by using MIT-BIH arrhythmia database. The achieved scores indicate average detection rate of 99.76%, sensitivity of 99.30% and specificity of 98.66%; accuracy respectively for R wave and PVC detection.

• Proceedings of the KSRS Conference
• /
• 2002.10a
• /
• pp.460-464
• /
• 2002

Multitemporal interferometric SAR has became an useful geodetic tool for monitoring Earth's surface deformation, generation of precise DEM, and land cover classification even though there still exist certain constraints such as temporal and spatial decorrelation effects, atmospheric artifacts and inaccurate orbit information. The Korea where nearly all areas are heavily vegetated, JERS-1 SAR has advantages in monitoring surface deformations and environmental changes in that it uses 4-times longer wavelength than ERS-l/2 or RADARSAT SAR system. For generating differential SAR interferogram and differential coherence image fer deformation mapping and temporal change detection, respectively, topographic phase removal process is required utilizing a reference inteferogram or external DEM simulation. Because the SAR antenna baseline parameter for JERS-1 is less accurate than those of ERS-l/2, one can not estimate topographic phases from an external DEM and the residual phase appears in differential interferogram. In this paper, we examined topographic phase retrieval method utilizing an external DEM. The baseline refinement is carried out by minimizing the differences between the measured unwrapped phase and the reference points of the DEM.

• Proceedings of the KSRS Conference
• /
• v.1
• /
• pp.200-203
• /
• 2006

Cloud detection algorithm is being developed as major one of the 16 baseline products of CMDPS (COMS Meteorological Data Processing System), which is under development for the real-time application of data will be observed from COMS Meteorological Imager. For cloud detection from satellite data, we studied two different algorithms. One is threshold technique based algorithm, which is traditionally used, and another is artificial neural network model. MPEF scene analysis algorithm is the basic idea of threshold cloud detection algorithm, and some modifications are conducted for COMS. For the neural network, we selected MLP with back-propagation algorithm. Prototype software of each algorithm was completed and evaluated by using the MTSAT-1R and GOES-9 data. Currently the software codes are standardized using Fortran90 language. For the preparation as an operational algorithm, we will setup the validation strategy and tune up the algorithm continuously. This paper shows the outline of the two cloud detection algorithm and preliminary test result of both algorithms.

• Korean Journal of Remote Sensing
• /
• v.22 no.5
• /
• pp.367-372
• /
• 2006

Cloud detection algorithm is being developed as primary one of the 16 baseline products of CMDPS (COMS Meteorological Data Processing System), which is under development for the real-time application of data will be observed from COMS Meteorological Imager. For cloud detection from satellite data, we studied two different algorithms. One is threshold technique based algorithm, which is traditionally used, and another is artificial neural network model. MPEF scene analysis algorithm is the basic idea of threshold cloud detection algorithm, and some modifications are conducted for COMS. For the neural network, we selected MLP with back-propagation algorithm. Prototype software of each algorithm was completed and evaluated by using the MTSAT-IR and GOES-9 data. Currently the software codes are standardized using Fortran90 language. For the preparation as an operational algorithm, we will setup the validation strategy and tune up the algorithm continuously. This paper shows the outline of the two cloud detection algorithms and preliminary test results of both algorithms.

• Structural Engineering and Mechanics
• /
• v.29 no.3
• /
• pp.311-325
• /
• 2008

Noting that damage occurrence of offshore jacket platforms is concentrated in two structural regions that are in the vicinity of still water surface and close to the seabed, a damage detection method by using only partial measurement of vibration in a suspect region was presented in this paper, which can not only locate damaged members but also evaluate damage severities. Then employing an experiment platform model under white-noise ground excitation by shaking table and using modal parameters of the first three modes identified by a scalar-type ARMA method on undamaged and damaged structures, the feasibility of the damage detection method was discussed. Modal parameters from eigenvalue analysis on the structural FEM model were also used to help the discussions. It is demonstrated that the damage detection algorithm is feasible on damage location and severity evaluation for broken slanted braces and it is robust against the errors of baseline FEM model to real structure when the principal errors is formed by difference of modal frequencies. It is also found that Z-value changes of modal shapes also play a role in the precise detection of damage.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.8
• /
• pp.4285-4299
• /
• 2019

Adversarial attack is a technique that causes a malfunction of classification models by adding noise that cannot be distinguished by humans, which poses a threat to a deep learning model. In this paper, we propose an efficient method to detect adversarial images using Gaussian process regression. Existing deep learning-based adversarial detection methods require numerous adversarial images for their training. The proposed method overcomes this problem by performing classification based on the statistical features of adversarial images and clean images that are extracted by Gaussian process regression with a small number of images. This technique can determine whether the input image is an adversarial image by applying Gaussian process regression based on the intermediate output value of the classification model. Experimental results show that the proposed method achieves higher detection performance than the other deep learning-based adversarial detection methods for powerful attacks. In particular, the Gaussian process regression-based detector shows better detection performance than the baseline models for most attacks in the case with fewer adversarial examples.