• Journal of Broadcast Engineering
• /
• v.23 no.6
• /
• pp.855-865
• /
• 2018

Sound event detection is one of the research areas to model human auditory cognitive characteristics by recognizing events in an environment with multiple acoustic events and determining the onset and offset time for each event. DCASE, a research group on acoustic scene classification and sound event detection, is proceeding challenges to encourage participation of researchers and to activate sound event detection research. However, the size of the dataset provided by the DCASE Challenge is relatively small compared to ImageNet, which is a representative dataset for visual object recognition, and there are not many open sources for the acoustic dataset. In this study, the sound events that can occur in indoor and outdoor are collected on a larger scale and annotated for dataset construction. Furthermore, to improve the performance of the sound event detection task, we developed a dual CNN structured sound event detection system by adding a supplementary neural network to a convolutional neural network to determine the presence of sound events. Finally, we conducted a comparative experiment with both baseline systems of the DCASE 2016 and 2017.

• 제어로봇시스템학회:학술대회논문집
• /
• 1991.10b
• /
• pp.1705-1711
• /
• 1991

On-line fault detection and diagnosis has an increasing interest in a chemical process industry, especially for a process control and automation. The chemical process needs an intelligent operation-aided workstation which can do such tasks as process monitoring, fault detection, fault diagnosis and action guidance in semiautomatic mode. These tasks can increase the performance of a process operation and give merits in economics, safety and reliability. Aiming these tasks, series of researches have been done in our lab. Main results from these researches are building appropriate knowledge representation models and a diagnosis mechanism for fault detection and diagnosis in a chemical process. The knowledge representation schemes developed in our previous research, the symptom tree model and the fault-consequence digraph, showed the effectiveness and the usefulness in a real-time application, of the process diagnosis, especially in large and complex plants. However in our previous approach, the diagnosis speed is its demerit in spite of its merits of high resolution, mainly due to using two knowledge models complementarily. In our current study, new knowledge representation scheme is developed which integrates the previous two knowledge models, the symptom tree and the fault-consequence digraph, into one. This new model is constructed using a material balance, energy balance, momentum balance and equipment constraints. Controller related constraints are included in this new model, which possesses merits of the two previous models. This new integrated model will be tested and verified by the real-time application in a BTX process or a crude unit process. The reliability and flexibility will be greatly enhanced compared to the previous model in spite of the low diagnosis speed. Nexpert Object for the expert system shell and SUN4 workstation for the hardware platform are used. TCP/IP for a communication protocol and interfacing to a dynamic simulator, SPEEDUP, for a dynamic data generation are being studied.

• Journal of the Korean Society for Nondestructive Testing
• /
• v.37 no.2
• /
• pp.91-98
• /
• 2017

Recently, composite materials have been mainly used in the main wings, ailerons, and fuselages of aircraft and rotor blades of helicopters. Composite materials used in rapid moving structures are subject to impact by hail, lightning, and bird strike. Such an impact can destroy fiber tissues in the composite materials as well as deform the composite materials, resulting in various problems such as weakened rigidity of the composite structure and penetration of water into tiny cracks. In this study, experiments were conducted using a 2 kW halogen lamp which is most frequently used as a light source, a 2 kW near-infrared lamp, which is used for heating to a high temperature, and a 6 kW xenon flash lamp which emits a large amount of energy for a moment. CFRP composite sandwich panels using Nomex honeycomb core were used as the specimens. Experiments were carried out under impact damages of 1, 4 and 8 J. It was found that the detection of defects was fast when the xenon flash lamp was used. The detection of damaged regions was excellent when the halogen lamp was used. Furthermore, the near-infrared lamp is an effective technology for showing the surface of a test object.

• Journal of Internet Computing and Services
• /
• v.15 no.6
• /
• pp.117-124
• /
• 2014

Cluttered background is a major obstacle in developing salient object detection and tracking system for mobile device captured natural scene video frames. In this paper we propose a context aware feature vector selection model to provide an efficient noise filtering by machine learning based classifiers. Since the context awareness for feature selection is achieved by searching nearest neighborhoods, known as NP hard problem, we apply a fast approximation method with complexity analysis in details. Separability enhancement in feature vector space by adding the context aware feature subsets is studied rigorously using principal component analysis (PCA). Overall performance enhancement is quantified by the statistical measures in terms of the various machine learning models including MLP, SVM, Naïve Bayesian, CART. Summary of computational costs and performance enhancement is also presented.

• Journal of Korean Society of Transportation
• /
• v.20 no.5
• /
• pp.163-174
• /
• 2002

Speeding has been reported as one of the major causes for fatal traffic accidents in Korea. The resolution against this dangerous speeding comes to make the automated speed enforcement system an enforcement tool. The speed detection device, which measures speeds of each incoming vehicles using double loop sensors, requires high accuracy. The object of this study is to develop an accurate speed measurement algorithm using double loop detectors. Some important findings are summarized as follows: 1) It was found that speed measurement errors are caused by scanning rate, distance of two loops, irregular vehicle trajectories, multiple vehicles in detection zone. 2) A proposed algorithm using two signal set proved to reduce variance as well as mean of speed measurement. 3) A proposed filtering algorithm was effective to filter irregular driving vehicles and multiple vehicles in detection zone. A comprehensive field test of developed algorithm resulted in significant improvement of speed measurement accuracy.

• Journal of the Institute of Electronics Engineers of Korea TC
• /
• v.41 no.12
• /
• pp.1-7
• /
• 2004

We have developed a 24GHz side-looking vehicle detection radar. A 24GHz front-end module and a novel signal processing algorithm have been developed for speed measurement and size classification of vehicles in multiple lanes. The system has a fixed antenna and FMCW processing module. This paper presents the background theory of operation and shows some measured data using the algorithm. The data shows that measured velocity of the passing vehicle is within the accuracy of 95% in single lane and the velocity of the vehicles in two lanes is within the accuracy of 90% by using variable threshold estimation. The classification of vehicle size as small, medium and large has been measured with 89% accuracy.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.17 no.4
• /
• pp.571-576
• /
• 2022

Due to the epochal development of the unmanned technology, the importance of LDT(: Location Detection Technology), which accurately estimates the location of a user or object, is dramatically increased. TOA(: Time of Arrival), which calculates a location by measuring the arrival time of signals, and TDOA(: Time Difference of Arrival) which calculates it by measuring the difference between two arrival times, are representative LDT methods. Based on the signals received from three or more base stations, TOA calculates an intersection point by drawing circles and TDOA calculates it by drawing hyperbolas. In order to improve the radio shadow area problem, a huge number of repeaters have been installed in the urban area, but the signals received through these repeaters may cause the serious error for estimating a location. In this paper, we propose an efficient location estimation technique using the signal received through the repeater. The proposed approach estimates the location of MS(: Mobile Station) employing TOA and TDOA methods, based on signals received from one repeater and two BS(: Base Station)s.

• Korean Journal of Remote Sensing
• /
• v.38 no.6_4
• /
• pp.1925-1934
• /
• 2022

Monitoring nuclear activity for inaccessible areas using remote sensing technology is essential for nuclear non-proliferation. In recent years, deep learning has been actively used to detect nuclear-activity-related small objects. However, high-resolution satellite imagery containing small objects can result in class imbalance. As a result, there is a performance degradation problem in detecting small objects. Therefore, this study aims to improve detection accuracy by analyzing the effect of the ratio of small objects related to nuclear activity in the input data for the performance of the deep learning model. To this end, six case datasets with different ratios of small object pixels were generated and a U-Net model was trained for each case. Following that, each trained model was evaluated quantitatively and qualitatively using a test dataset containing various types of small object classes. The results of this study confirm that when the ratio of object pixels in the input image is adjusted, small objects related to nuclear activity can be detected efficiently. This study suggests that the performance of deep learning can be improved by adjusting the object pixel ratio of input data in the training dataset.

• Journal of Digital Convergence
• /
• v.10 no.5
• /
• pp.223-232
• /
• 2012

Recently, the leakage of personal information and privacy piracy increase. The victimized case of the malicious object rapidlies increase. Most of users use the windows operating system. Recently, the Windows 7 operating system was announced. Therefore, we need to study for the intrusion response technique at the next generation operate system circumstances. The accident response technique developed till now was mostly implemented around the Windows XP or the Windows Vista. However, a new vulnerability problem will be happen in the breach process of reaction as the Windows 7 operating system is announced. In the windows operating system, the system incident event needs to be efficiently analyzed. For this, the event information generated in a system needs to be visually analyzed around the time information or the security threat weight information. Therefore, in this research, we analyzed visually about the system event information generated in the Windows 7 operating system. And the system analyzing the system incident through the visual event information analysis process was designed and implemented. In case of using the system developed in this study the more efficient accident analysis is expected to be possible.

• Journal of the HCI Society of Korea
• /
• v.2 no.2
• /
• pp.35-43
• /
• 2007

This paper presents a system for image matching and recognition based on image feature detection and description techniques from artificial satellite photographs. We propose some kind of parameters from the varied environmental elements happen by image handling process. The essential point of this experiment is analyzes that affects match rate and recognition accuracy when to change of state of each parameter. The proposed system is basically inspired by Lowe's SIFT(Scale-Invariant Transform Feature) algorithm. The descriptors extracted from local affine invariant regions are saved into database, which are defined by k-means performed on the 128-dimensional descriptor vectors on an artificial satellite photographs from Google earth. And then, a label is attached to each cluster of the feature database and acts as guidance for an appeared building's information in the scene from camera. This experiment shows the various parameters and compares the affected results by changing parameters for the process of image matching and recognition. Finally, the implementation and the experimental results for several requests are shown.