• Annual Conference of KIPS
• /
• 2022.05a
• /
• pp.667-669
• /
• 2022

2차원 이미지를 통한 자세 추정의 경우 관절이 겹치거나 가려져 있는 등의 인식 저해 요소로 인하여 자세 추정 정확도가 감소하는 한계가 있다. 본 논문에서는 GAN을 통해 2차원 이미지를 3차원으로 증강한 뒤 자세를 추정하는 기법을 제안한다. 제안하는 방법은 2차원 이미지의 평면좌표 값에서 GAN을 통해 노이즈 벡터 z축 값과 피사체에 투영되는 빛의 방향 값을 반영한 3차원 이미지를 만든다. 이러한 이미지 합성 과정을 거친 후 DeepLabCut을 사용해 관절 좌표를 추출하고 자세 추정 및 분류를 진행한다. 이를 통해 2차원에서의 자세 추정 정확도 향상을 기대할 수 있으며, 향후 이를 기반한 이상행동 탐지 분야에서 적용할 수 있다.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2022.01a
• /
• pp.99-101
• /
• 2022

이상 행동을 탐지하는 딥러닝 기반 검지 시스템은 동영상 기반 데이터로부터 움직임을 보이는 객체를 추적하고 그 객체의 행동을 분석하여 정상적인 행동 범위를 벗어나는 패턴을 보이는 영역을 이상으로 탐지한다. 특히 생성적 적대 신경망(GAN)과 광학 흐름 추정(Optical flow estimation) 기법을 활용하여 움직임에 대한 특징 정보를 추출하고 이를 학습하여 행동 패턴에 대한 모델링을 수행한다. 모델 학습 및 테스트에 활용되는 데이터셋의 해상도가 낮거나 이상 행동을 표현하는 특징 정보가 부족할 경우 최종 모델 성능에 부정적 영향을 미치게 되며, 특히 광학 흐름이 표현하는 이동량 측면에서 차이가 크게 나지 않는 이상 객체의 경우 탐지가 정확하게 이뤄지지 않는다. 본 연구에서는 동영상 프레임에서 나타나는 객체의 평균 종횡비를 구하고 정상적인 비율을 벗어나는 객체에 대해서 이상 행동을 취하는 샘플로 처리하는 후처리단 모듈을 제안하여 최종적인 모델 성능을 향상시키는 방법을 고안한다.

• International Journal of Computer Science & Network Security
• /
• v.23 no.11
• /
• pp.133-141
• /
• 2023

This work take deeper analysis of Adaptive Moment Estimation (Adam) and Adam with Weight Decay (AdamW) implementation in real world text classification problem (DGA Malware Detection). AdamW is introduced by decoupling weight decay from L2 regularization and implemented as improved optimizer. This work introduces a novel implementation of AdamW variant as AdamW+ by further simplifying weight decay implementation in AdamW. DGA malware detection LSTM models results for Adam, AdamW and AdamW+ are evaluated on various DGA families/ groups as multiclass text classification. Proposed AdamW+ optimizer results has shown improvement in all standard performance metrics over Adam and AdamW. Analysis of outcome has shown that novel optimizer has outperformed both Adam and AdamW text classification based problems.

• International Journal of Computer Science & Network Security
• /
• v.24 no.4
• /
• pp.44-50
• /
• 2024

This paper offers an overview of matrix formation and calculation techniques within the framework of General Linear Models (GLMs). It takes a sequential approach, beginning with a detailed exploration of matrix formation and calculation methods in regression analysis and univariate analysis of variance (ANOVA). Subsequently, it extends the discussion to cover multivariate analysis of variance (MANOVA). The primary objective of this study was to provide a clear and accessible explanation of the underlying matrices that play a crucial role in GLMs. Through linking, essentially different statistical methods, by fundamental principles and algebraic foundations that underpin the GLM estimation. Insights presented here aim to assist researchers, statisticians, and data analysts in enhancing their understanding of GLMs and their practical implementation in diverse research domains. This paper contributes to a better comprehension of the matrix-based techniques that can be extended to GLMs.

• Journal of Broadcast Engineering
• /
• v.19 no.3
• /
• pp.362-371
• /
• 2014

Blur variation caused by camera de-focusing provides a proper cue for depth estimation. Depth from Defocus (DFD) technique calculates the blur amount present in an image considering that blur amount is directly related to scene depth. Conventional DFD methods use two defocused images that might yield the low quality of an estimated depth map as well as a reconstructed infocused image. To solve this, a new DFD methodology based on infocused and defocused images is proposed in this paper. In the proposed method, the outcome of Subbaro's DFD is combined with a novel edge blur estimation method so that improved blur estimation can be achieved. In addition, a saliency map mitigates the ill-posed problem of blur estimation in the region with low intensity variation. For validating the feasibility of the proposed method, twenty image sets of infocused and defocused images with 2K FHD resolution were acquired from a camera with a focus control in the experiments. 3D stereoscopic image generated by an estimated depth map and an input infocused image could deliver the satisfactory 3D perception in terms of spatial depth perception of scene objects.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.20 no.3
• /
• pp.499-505
• /
• 2016

In this Paper, we propose a body-area localization technique based on WUSB (Wireless USB) over WBAN (Wireless Body Area Networks) protocol required for wearable computer systems. The proposed localization algorithm is executed on the basis of WUSB over WBAN protocol at each sensor node comprising peripherals of a wearable computer system. To increase the accuracy of input information through various body motions in wearable computer systems, a new localization technique with high precision must be developed. To achieve the goal, This paper proposes a combined TDoA/FDoA/AoA (Time Of Arrival/Time Difference Of Arrival/Angle Of Arrival) localization technique with more than four WUSB over WBAN devices to estimate body-area location accurately. The combined TDoA/FDoA/AoA technique reduces 10mm in location estimation errors comparing with a combined TDoA/FDoA technique. This performance enhancement in location error reduction can be ignored at other systems but is meaningful results in body-area localization-based communications.

• Journal of Korea Society of Digital Industry and Information Management
• /
• v.16 no.2
• /
• pp.27-33
• /
• 2020

Backward compatibility is one of the key issues for radio equipment supporting IEEE 802.11, the typical wireless local area networks (WLANs) communication protocol. For a successful packet decoding with the backward compatibility, the frame format detection is a core precondition. This paper presents a novel frame format detection method based on a deep learning procedure for WLANs affiliated with IEEE 802.11. Considering that the detection performance of conventional methods is degraded mainly due to the poor performances in the symbol synchronization and/or channel estimation in low signal-to-noise-ratio environments, we propose a novel detection method based on convolutional neural network (CNN) that replaces the entire conventional detection procedures. The proposed deep learning network provides a robust detection directly from the receive data. Through extensive computer simulations performed in the multipath fading channel environments (modeled by Project IEEE 802.11 Task Group ac), the proposed method exhibits superb improvement in the frame format detection compared to the conventional method.

• Proceedings of the KSME Conference
• /
• 2007.05a
• /
• pp.1103-1108
• /
• 2007

This paper describes a robust model-based roll state estimator for application to the detection of impending vehicle rollover. The roll state estimator is based on a 2-D bicycle model and a roll model to estimate the maneuver-induced vehicle roll motion. The measurement signals are lateral acceleration, yaw rate, steering angle, and vehicle speed. Vehicle mass is adapted to obtain robust performance of the estimator. Computer simulation is conducted to evaluate the proposed roll state estimator by using a validated vehicle simulator. It is shown that the roll state estimator shows robust performance without exact vehicle mass information.

• Journal of Electrical Engineering and Technology
• /
• v.1 no.1
• /
• pp.35-41
• /
• 2006

This paper presents a new numerical algorithm for fault location estimation and for faults recognition based on the synchronized phasors. The proposed algorithm is based on the synchronized phasor measured from the synchronized PMUs installed at two-terminals of the transmission lines. In order to discriminate the fault type, the arc voltage wave shape is modeled numerically on the basis of a great number of arc voltage records obtained by transient recorder. From the calculated arc voltage amplitude it can make a decision whether the fault is permanent or transient. The results of the proposed algorithm testing through computer simulation are given.

• Journal of the Korean Institute of Telematics and Electronics S
• /
• v.35S no.1
• /
• pp.68-79
• /
• 1998

In this paper, we propose a model-based FDI(fault detetion and isolation) algorithm to detect and isolate fault in a linear system. The proposed algorithm is gased on an HFC(hydrid fault classifier) which consists of an FCART2(fault classifier by ART2 neural network) and an FCFM(fault classifier by fault models) which operate in parallel to isolate faults. The proposed algorithm is functionally composed of three main parts-parameter estimation, fault detection, and isolation. When a change in the system occurs, the estimated parameters go through a transition zone in which errors between the system output and the stimated output and the estimated output cross a predetermined thrseshold, and in this zone the estimated parameters are tranferred to the FCART2 for fault isolation. On the other hand, once a fault in the system is detected, the FCFM statistically isolates the fault by using the error between ach fault model out put and the system output. From the computer simulation resutls, it is verified that the proposed model-based FDI algorithm can be performed successfully to detect and isolate faults in a position control system of a DC motor.