• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.12 no.6
• /
• pp.629-635
• /
• 2019

In Deep Learning models derivative is implemented by error back-propagation which enables the model to learn the error and update parameters. It can find the global (or local) optimal points of parameters even in the complex models taking advantage of a huge improvement in computing power. However, deliberately generated data points can 'fool' models and degrade the performance such as prediction accuracy. Not only these adversarial examples reduce the performance but also these examples are not easily detectable with human's eyes. In this work, we propose the method to detect adversarial datasets with random noise addition. We exploit the fact that when random noise is added, prediction accuracy of non-adversarial dataset remains almost unchanged, but that of adversarial dataset changes. We set attack methods (FGSM, Saliency Map) and noise level (0-19 with max pixel value 255) as independent variables and difference of prediction accuracy when noise was added as dependent variable in a simulation experiment. We have succeeded in extracting the threshold that separates non-adversarial and adversarial dataset. We detected the adversarial dataset using this threshold.

• Journal of the Institute of Convergence Signal Processing
• /
• v.10 no.1
• /
• pp.7-12
• /
• 2009

In this paper, we propose a directional feature extraction method for off-line handwritten numerals by using the morphological operations. Direction features are obtained from four directional line images, each of which contains horizontal, vertical, right-diagonal and left-diagonal lines in entire numeral lines. Conventional method for extracting directional features uses Kirsch masks which generate edge-shaped double line images for each direction, whereas our method uses directional erosion operations and generate single line images for each direction. To apply these directional erosion operations to the numeral image, preprocessing steps such as thinning and dilation are required, but resultant directional lines are more similar to numeral lines themselves. Our four [$4{\times}4$] directional features of a numeral are obtained from four directional line images through a zoning method. For obtaining the higher recognition rates of the handwrittern numerals, we use the multiple feature which is comprised of our proposed feature and the conventional features of a kirsch directional feature and a concavity feature. For recognition test with given features, we use a multi-layer perceptron neural network classifier which is trained with the back propagation algorithm. Through the experiments with the CENPARMI numeral database of Concordia University, we have achieved a recognition rate of 98.35%.

• The Journal of the Korea Contents Association
• /
• v.6 no.12
• /
• pp.1-11
• /
• 2006

Many research demonstrations and commercial applications have been tried to develop face detection and recognition systems. Human face detection plays an important role in applications such as access control and video surveillance, human computer interface, identity authentication, etc. There are some special problems such as a face connected with background, faces connected via the skin color, and a face divided into several small parts after skin region segmentation in generally. It can be allowed many face detection techniques to solve the first and second problems. However, it is not easy to detect a face divided into several parts of regions for reason of different illumination conditions in the third problem. Therefore, we propose an efficient modified skin segmentation algorithm to solve this problem because the typical region segmentation algorithm can not be used to. Our algorithm detects skin regions over the entire image, and then generates face candidate regions using our skin segmentation algorithm For each face candidate, we implement the procedure of region merging for divided regions in order to make a region using adjacency between homogeneous regions. We utilize various different searching window sizes to detect different size faces and a face detection classifier based on a back-propagation algorithm in order to verify whether the searching window contains a face or not.

• Journal of Biomedical Engineering Research
• /
• v.23 no.3
• /
• pp.197-205
• /
• 2002

The objectives of this Paper is to implement a diagnostic classifier of differential laryngeal diseases from acoustic signals acquired in a noisy room. For this Purpose, the voice signals of the vowel /a/ were collected from Patients in a soundproof chamber and got mixed with noise. Then, the acoustic Parameters were analyzed, and hierarchical neural networks were applied to the data classification. The classifier had a structure of five-step hierarchical neural networks. The first neural network classified the group into normal and benign or malign laryngeal disease cases. The second network classified the group into normal or benign laryngeal disease cases The following network distinguished polyp. nodule. Palsy from the benign laryngeal cases. Glottic cancer cases were discriminated into T1, T2. T3, T4 by the fourth and fifth networks All the neural networks were based on multilayer perceptron model which classified non-linear Patterns effectively and learned by an error back-propagation algorithm. We chose some acoustic Parameters for classification by investigating the distribution of laryngeal diseases and Pilot classification results of those Parameters derived from MDVP. The classifier was tested by using the chosen parameters to find the optimum ones. Then the networks were improved by including such Pre-Processing steps as linear and z-score transformation. Results showed that 90% of T1, 100% of T2-4 were correctly distinguished. On the other hand. 88.23% of vocal Polyps, 100% of normal cases. vocal nodules. and vocal cord Paralysis were classified from the data collected in a noisy room.

• Horticultural Science & Technology
• /
• v.32 no.6
• /
• pp.906-911
• /
• 2014

A new coreopsis cultivar, 'Uridream Pink', was developed by mutation breeding using gamma-ray irradiation. Rooted cuttings of 'Uridream 01', developed at Uriseed Company, were exposed to various gray ranges of gamma-rays from a $^{60}CO$ source for 24 h in 2009 at the Korea Atomic Energy Research Institute. Using gamma-rays with the range of 10-100 Gy, a chimeric mutant with pastel pink flowers was induced at 30 Gy from 'Uridream 01', which blooms red-purple flowers (Red-purple group, 59A). Clones that produced flowers with modified pink color were separated from the chimeric mutant and fixed by more than three rounds of cutting back from 2009 to 2010. The separated mutant clones with modified pink color were found to bloom flowers with marketable color and be free of any abnormal plant characteristics. The typical color of the flowers was pastel pink (Red-purple group, 67B), and the clones were registered as 'Uridream Pink' at the Korea Seed and Variety Service (plant variety protection number: 4410). Although the size of flowers and leaves of 'Uridream Pink' are smaller than those of 'Uridream 01', the number of flowers per plant is greater. 'Uridream Pink' can be used as both a pot and a garden plant because of its long blooming period from late spring to late autumn.

• Journal of the Korean Society for Nondestructive Testing
• /
• v.21 no.4
• /
• pp.398-405
• /
• 2001

The rivet joint has typical structural feature that can be initiation site for the fatigue crack due to the combination of local stress concentration around rivet hole and the moisture trapping. From a viewpoint of structural assurance, it is crucial to evaluate the size of crack around the rivet holes by appropriate nondestructive evaluation techniques. Lamb wave that is one of guided waves, offers a more efficient tool for nondestructive inspection of plates. The neural network that is considered to be the most suitable for pattern recognition has been used by researchers in NDE field to classify different types of flaws and flaw sizes. In this study, clack size evaluation around the rivet hole using the neural network based on the back-propagation algorithm has been tarried out by extracting some features from the ultrasonic Lamb wave for A12024-T3 skin panel of aircraft. Special attention was paid to reduce the coupling effect between the transducer and the specimen by extracting some features related to time md frequency component data in ultrasonic waveform. It was demonstrated clearly that features extracted from the time and frequency domain data of Lamb wave signal were very useful to determine crack size initiated from rivet hole through neural network.

• Journal of Korea Water Resources Association
• /
• v.31 no.1
• /
• pp.13-25
• /
• 1998

It is necessary to develop methodologies for the application of artificial neural network into hydrologic rainfall-runoff process, although there is so much applicability by using the functions of associative memory based on recognition for the relationships between causes and effects and the excellent fitting capacity for the nonlinear phenomenon. In this study, some problems are presented in the application procedures of artificial neural networks and the simulation of runoff hydrograph experiences are reviewed with nonlinear functional approximator by artificial neural network for rainfall-runoff relationships in a watershed. which is regarded as hydrdologic black box model. The neural network models are constructed by organizing input and output patterns with the deserved rainfall and runoff data in Pyoungchang river basin under the assumption that the rainfall data is the input pattern and runoff hydrograph is the output patterns. Analyzed with the results. it is possible to simulate the runoff hydrograph with processing element of artificial neural network with any hydrologic concepts and the weight among processing elements are well-adapted as model parameters with the assumed model structure during learning process. Based upon these results. it is expected that neural network theory can be utilized as an efficient approach to simulate runoff hydrograph and identify the relationship between rainfall and runoff as hydrosystems which is necessary to develop and manage water resources.

• Journal of Korea Water Resources Association
• /
• v.46 no.7
• /
• pp.697-706
• /
• 2013

In recent, Korea has faced on water quality management problems in reservoir and river because of increasing water temperature and rainfall frequency caused by climate change. This study is effectively to manage water quality for establishment of algal bloom forecasting models with artificial neural network. Daecheong reservoir located in Geum river has suitable environment for algal bloom because it has lots of contaminants that are flowed by rainfall. By using back propagation algorithm of artificial neural networks (ANNs), a model has been built to forecast the algal bloom over short-term (1, 3, and 7 days). In the model, input factors considered the hydrologic and water quality factors in Daecheong reservoir were analyzed by cross correlation method. Through carrying out the analysis, input factors were selected for algal bloom forecasting model. As a result of this research, the short term algal bloom forecasting models showed minor errors in the prediction of the 1 day and the 3 days. Therefore, the models will be very useful and promising to control the water quality in various rivers.

• Journal of Sensor Science and Technology
• /
• v.6 no.6
• /
• pp.476-482
• /
• 1997

A portable electronic nose system has been fabricated and characterized using an oxide semiconductor gas sensor array and pattern recognition techniques such as principal component analysis and back-propagation artificial neural network. The sensor array consists of six thick-film gas sensors whose sensing layers are Pd-doped $WO_{3}$, Pt-doped $SnO_{2}$, $TiO_{2}-Sb_{2}O_{5}-Pd$-doped $SnO_{2}$, $TiO_{2}-Sb_{2}O_{5}-Pd$-doped $SnO_{2}$ + Pd coated layer, $Al_{2}O_{3}$-doped ZnO and $PdCl_{2}$-doped $SnO_{2}$. The portable electronic nose system consists of an 16bit Intel 80c196kc as CPU, an EPROM for storing system main program, an EEPROM for containing optimized connection weights of artificial neural network, an LCD for displaying gas concentrations. As an application the system has been used to identify 26 carbon monoxide/hydrocarbon (CO/HC) car exhausting gases in the concentration range of CO 0%/HC 0 ppm to CO 7.6%/HC 400 ppm and the identification has been successfully demonstrated.

• Journal of Biomedical Engineering Research
• /
• v.18 no.4
• /
• pp.429-438
• /
• 1997

Mechanical valve is one of the most widely used implantable artificial organs of which the reliability is so important that its failure means the death of patient. Therefore early noninvasive detection is essentially required, though mechanical valve failure with thrombosis is the most common. The objective of this paper is to detect the thrombosis formation by spectral analysis and neural network. Using microphone and amplifier, we measured the sound from the mechanical valve which is attached to the pneumatic ventricular assist device. The sound was sampled by A/D converter(DaqBook 100) and the periodogram is the main algorithm for obtaining spectrum. We made the thrombosis models using pellethane and silicon and they are thrombosis model on the valvular disk, around the sewing ring and fibrous tissue growth across the orifice of valve. The performance of the measurment system was tested firstly using 1 KHz sinusoidal wave. The measurement system detected well 1KHz spectrum as expected. The spectrum of normal and 5 kinds of thrombotic valve were obtained and primary and secondary peak appeared in each spectrum waveform. We find that the secondary peak changes according to the thrombosis model. So to distinguish the secondary peak of normal and thrombotic valve quantatively, 3 layer back propagation neural network, which contains 7, 000 input node, 20 hidden layer and 1 output was employed The trained neural network can distinguish normal and valve with more than 90% probability. As a conclusion, the noninvasive monitoring of implanted mechanical valve is possible by analysing the acoustical spectrum using neural network algorithm and this method will be applied to the performance evaluation of other implantable artificial organs.