• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.21 no.4
• /
• pp.972-983
• /
• 1996

This paper proposes a new technique for frequency offset correction for OFDM systems on a frequency selective fading channel. Frequency offset in OFDM introduces interchannel interference among the multiple subcarriers of OFDM signal. To compensate the interference, this paper describes an algorithm with two stages:acquisition and tracking. At both stages, the proposed algorithm oversamples the received OFDM signal to obtain a couple of demodulated symbol sets. At acquisition stage the frequency offset is reduced to half or less of the intercarrier spacings by matching the sign pattern of each element of the sets. Next, at tracking stage the frequency offset is corrected with a frequency detector which is controlled by the correlation of the two sets. It is shown that the proposed algorithm can correct the frequency offset in the event of uncertainty in the initial offset that exceeds one half of the intercarrier spacing. In addition, the proposed algorithm is robust to transmitted symbols and channel characteristics by using oversampled symbol sets.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2006.11a
• /
• pp.19-22
• /
• 2006

In this paper we present a vision based place recognition method which uses Bayesian method with feed back of image retrieval. Both Bayesian method and image retrieval method are based on interest features that are invariant to many image transformations. The interest features are detected using Harris-Laplacian detector and then descriptors are generated from the image patches centered at the features' position in the same manner of SIFT. The Bayesian method contains two stages: learning and recognition. The image retrieval result is fed back to the Bayesian recognition to achieve robust and confidence. The experimental results show the effectiveness of our method.

• Proceedings of the KIEE Conference
• /
• 1996.11a
• /
• pp.395-397
• /
• 1996

The existing CPC(Center Position Controller) has unstably performed because of dusts on reflection panel, CCD protector contamination due to high temperature in furnace or other parameters. The reason is that the existing CPC has a Z80 processor as a CPU and only performs low level image processing as a simple edge detector. So the improvement of control system through the development of robust edge detection algorithm overcoming changes of measuring environment is needed. For this, in this study we carefully analyze the image of the strip rolled in occasion that measuring environment is changing, develop the optimal edge detection algorithm to solve the problems, generate the control signal suitable for the existing CPC(Center Position Controller), and propose the capability of application to the actual environment.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.8 no.9
• /
• pp.3177-3195
• /
• 2014

In many moving object detection problems of an aerial video, accurate and robust stabilization is of critical importance. In this paper, a novel accurate image alignment algorithm for aerial electronic image stabilization (EIS) is described. The feature points are first selected using optimal derivative filters based Harris detector, which can improve differentiation accuracy and obtain the precise coordinates of feature points. Then we choose the Delaunay Triangulation edges to find the matching pairs between feature points in overlapping images. The most "useful" matching points that belong to the background are used to find the global transformation parameters using the projective invariant. Finally, intentional motion of the camera is accumulated for correction by Sage-Husa adaptive filtering. Experiment results illustrate that the proposed algorithm is applied to the aerial captured video sequences with various dynamic scenes for performance demonstrations.

• Journal of Sensor Science and Technology
• /
• v.30 no.5
• /
• pp.290-294
• /
• 2021

Experts have designed popular and successful model architectures, which, however, were not the optimal option for different scenarios. Despite the remarkable performances achieved by deep neural networks, manually designed networks for classification tasks are the backbone of object detection. One major challenge is the ImageNet pre-training of the search space representation; moreover, the searched network incurs huge computational cost. Therefore, to overcome the obstacle of the pre-training process, we introduce a network adaptation technique using a pre-trained backbone model tested on ImageNet. The adaptation method can efficiently adapt the manually designed network on ImageNet to the new object-detection task. Neural architecture search (NAS) is adopted to adapt the architecture of the network. The adaptation is conducted on the MobileNetV2 network. The proposed NAS is tested using SSDLite detector. The results demonstrate increased performance compared to existing network architecture in terms of search cost, total number of adder arithmetics (Madds), and mean Average Precision(mAP). The total computational cost of the proposed NAS is much less than that of the State Of The Art (SOTA) NAS method.

• Journal of information and communication convergence engineering
• /
• v.20 no.2
• /
• pp.73-78
• /
• 2022

Based on experimental measurements conducted on many different practical wireless communication systems, ambient noise has been shown to be decidedly non-Gaussian owing to impulsive phenomena. However, most multiuser detection techniques proposed thus far have considered Gaussian noise only. They may therefore suffer from a considerable performance loss in the presence of impulsive ambient noise. In this paper, we consider a large-scale multiuser multiple-input multiple-output system in the presence of non-Gaussian noise and propose a genetic algorithm (GA) based detector for large-dimensional multiuser signal detection. The proposed algorithm is more robust than linear multi-user detectors for non-Gaussian noise because it uses a multi-directional search to manipulate and maintain a population of potential solutions. Meanwhile, the proposed GA-based algorithm has a comparable complexity because it does not require any complicated computations (e.g., a matrix inverse or derivation). The simulation results show that the GA offers a performance gain over the linear minimum mean square error algorithm for both non-Gaussian and Gaussian noise.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2022.05a
• /
• pp.560-563
• /
• 2022

인공지능 기반의 딥페이크(Deepfakes) 기술이 사회적인 이슈로 대두되고 있다. 하지만 기존 딥페이크 탐지기는 sharpening, additive noise와 같은 간단한 이미지 변형만으로 탐지 우회가 가능한 문제점이 있다. 본 논문에서는 안티 포렌식에 강인한 딥페이크 탐지기를 개발하기 위해 이미지 편집 도구 기반의 안티 포렌식 데이터셋을 생성하고 적대적 학습을 수행하는 방법을 제안한다. 실험 결과를 통해 안티 포렌식에 취약한 기존 딥페이크 탐지기 성능이 제안한 적대적 학습 기법을 수행한 이후에 탐지율이 크게 개선된 것을 확인할 수 있었다.

• Journal of the Semiconductor & Display Technology
• /
• v.22 no.3
• /
• pp.161-165
• /
• 2023

One of the difficult works in an autonomous driving system is detecting road lanes or objects in the road boundaries. Detecting and tracking a vehicle is able to play an important role on providing important information in the framework of advanced driver assistance systems such as identifying road traffic conditions and crime situations. This paper proposes a vehicle detection scheme based on deep learning to classify and tracking vehicles in a complex and diverse environment. We use the modified YOLO as the object detector and polynomial regression as object tracker in the driving video. With the experimental results, using YOLO model as deep learning model, it is possible to quickly and accurately perform robust vehicle tracking in various environments, compared to the traditional method.

• Journal of Korean Society for Geospatial Information Science
• /
• v.16 no.2
• /
• pp.49-56
• /
• 2008

Image registration is one of the critical techniques of image mosaic which has many applications such as generating panoramas, video monitoring, image rendering and reconstruction, etc. The fundamental tasks of image registration are point features extraction and tracking which take much computation time. KLT(Kanade-Lucas-Tomasi) feature tracker has proposed for extracting and tracking features through image sequences. The aim of this study is to demonstrate the usage of effective and robust KLT feature detector and tracker for an image registration using the sequence image frames captured by UAV video camera. In result, by using iterative implementation of the KLT tracker, the features extracted from the first frame of image sequences could be successfully tracked through all frames. The process of feature tracking in the various frames with rotation, translation and small scaling could be improved by a careful choice of the process condition and KLT pyramid implementation.

• Journal of Digital Contents Society
• /
• v.17 no.6
• /
• pp.509-521
• /
• 2016

A novel Samples Balanced Genetic Algorithm combined with Extreme Learning Machine (SBGA-ELM) for Parkinson's Disease diagnosis and detecting bio-markers is presented in this paper. Proposed approach uses genes' expression data of 22,283 genes from open source ParkDB data base for accurate PD diagnosis and detecting bio-markers. Proposed SBGA-ELM includes two major steps: feature (genes) selection and classification. Feature selection procedure is based on proposed Samples Balanced Genetic Algorithm designed specifically for genes expression data from ParkDB. Proposed SBGA searches a robust subset of genes among 22,283 genes available in ParkDB for further analysis. In the "classification" step chosen set of genes is used to train an Extreme Learning Machine (ELM) classifier for an accurate PD diagnosis. Discovered robust subset of genes creates ELM classifier with stable generalization performance for PD diagnosis. In this research the robust subset of genes is also used to discover 24 bio-markers probably responsible for Parkinson's Disease. Discovered robust subset of genes was verified by using existing PD diagnosis approaches such as SVM and PBL-McRBFN. Both tested methods caused maximum generalization performance.