https://scholar.dgist.ac.kr/handle/20.500.11750/10149 Sat, 04 Apr 2026 09:03:25 GMT 2026-04-04T09:03:25Z https://scholar.dgist.ac.kr/handle/20.500.11750/59405 Title: A Novel Multi-parametric H∞ Filter Design Method for Imperfectly Reconstructed Lateral Vehicle Dynamics Author(s): Jin, Yongsik; Han, Seungyong Abstract: This paper proposes novel multi-parametric filtering problems for imperfectly reconstructed lateral dynamics of autonomous driving vehicles in the presence of disturbances. The primary objective of this study is to provide a theoretical filter design criterion for lateral vehicle dynamics where cornering stiffness is estimated. To achieve this goal, we establish a new condition to define stable regions for the cornering stiffness estimation error and formulate a multi-parametric filtering error system using a polytopic approach. Then, we present a new robust filter design condition in terms of linear matrix inequalities (LMIs), and it provides globally optimized solutions. In this formulation, the cornering stiffness estimation error is incorporated into the convex optimization problem by adding a constraint that ensures the stability criteria are satisfied. Finally, we demonstrate the effectiveness of the proposed approach by simulating a lateral vehicle dynamics model. Tue, 30 Sep 2025 15:00:00 GMT https://scholar.dgist.ac.kr/handle/20.500.11750/59405 2025-09-30T15:00:00Z https://scholar.dgist.ac.kr/handle/20.500.11750/59400 Title: MOSInversion: Knowledge distillation-based incremental learning in organ segmentation using DeepInversion Author(s): Kim, Jihyeon; Lee, Gyeongmin; Shin, Seung Yeon; Kim, Soopil; Park, Sang Hyun Abstract: Despite recent advancements in multi-organ segmentation (MOS) of medical images, existing models are limited in terms of extending their capability to unseen classes. Incremental learning has been proposed to enable models to learn new classes progressively, possibly using multiple datasets from different institutions. In this setting, models easily experience performance degradation on previously learned classes i.e., catastrophic forgetting. Although many methods have been proposed to mitigate this issue, applying them to medical imaging applications like multi-organ segmentation is not easy due to the large memory requirement when used for 3D medical data such as CT scans or the need for additional training of a generator for image synthesis. In this paper, we propose an incremental learning framework that leverages diverse synthetic images to retain the knowledge learned from previously seen data. We design MOSInversion to generate the synthetic images by utilizing a pre-trained model from the previous step. MOSInversion generates diverse images by using segmentation masks so that we can manipulate the shape, location, and size of organs. We evaluate our proposed method using three abdominal CT datasets (FLARE21, MSD, and KiTS19) and achieve state-of-the-art accuracy. Sun, 30 Nov 2025 15:00:00 GMT https://scholar.dgist.ac.kr/handle/20.500.11750/59400 2025-11-30T15:00:00Z https://scholar.dgist.ac.kr/handle/20.500.11750/59385 Title: FFT-기반 Partition Selection을 활용한 SP-MUSIC DOA 추정 성능 개선 Author(s): 김봉석; 김제석; 김중태; 김시형; 최락현; 김상동 Abstract: 본 논문에서는 FFT 기반 Partition 선택 기법을 활용한 효율적인 도래각 (DOA) 추정 방법을 제안한다. 기존 SP-MUSIC (Spectrum Partitioning-MUltiple Signal Classification) 방식은 전체 스펙트럼을 여러 Partition으로 나누고, 각 Partition 별로 각각의 MUSIC 알고리즘을 적용할 뿐 아니라, 신호의 세기가 임계값 이하인 Partition까지 처리하기 때문에 불필요한 연산이 요구된다. 제안된 기법은 FFT를 통해 스펙트럼 에너지 분포를 먼저 분석하고, 신호가 존재하는 Partition만 선별하여 MUSIC 알고리즘을 적용한다. 이를 통해 불필요한 연산을 줄이고 노이즈 영향을 최소화함으로써, 기존 SP-MUSIC 대비 연산 효율성 및 개선된 DOA 추정 성능을 달성하였다. 시뮬레이션 결과는 제안된 기법이 감소된 복잡도에도 불구하고 기존 SP-MUSIC과 유사한 성능을 달성함을 보인다. In this paper, we propose an efficient direction estimation (DOA) method using the FFT-based partition selection technique. The conventional SP-MUSIC (Spectrum Partitioning-MUltiple Signal Classification) method divides the entire spectrum into several parts and applies the MUSIC algorithm to each partition, resulting in unnecessary operations and problems of processing even parts that do not have a signal. The proposed technique first analyzes the spectral energy distribution through FFT and applies the MUSIC algorithm by selecting only parts with a signal. Through this, the computational efficiency and DOA estimation performance compared to the conventional SP-MUSIC were improved simultaneously by reducing unnecessary operations and minimizing noise effects. Simulation results demonstrate that the proposed method achieves performance comparable to SP-MUSIC despite its reduced computational complexity. Sun, 30 Nov 2025 15:00:00 GMT https://scholar.dgist.ac.kr/handle/20.500.11750/59385 2025-11-30T15:00:00Z https://scholar.dgist.ac.kr/handle/20.500.11750/59379 Title: Noise-Resilient Masked Face Detection Using Quantized DnCNN and YOLO Author(s): Choi, Rockhyun; Lee, Hyunki; Kim, Bong-Seok; Kim, Sangdong; Kim, Min Young Abstract: This study presents a noise-resilient masked-face detection framework optimized for the NVIDIA Jetson AGX Orin, which improves detection precision by approximately 30% under severe Gaussian noise (variance 0.10) while reducing denoising latency by over 42% and increasing end-to-end throughput by more than 30%. The proposed system integrates a lightweight DnCNN-based denoising stage with the YOLOv11 detector, employing Quantize-Dequantize (QDQ)-based INT8 post-training quantization and a parallel CPU-GPU execution pipeline to maximize edge efficiency. The experimental results demonstrate that denoising preprocessing substantially restores detection accuracy under low signal quality. Furthermore, comparative evaluations confirm that 8-bit quantization achieves a favorable accuracy-efficiency trade-off with only minor precision degradation relative to 16-bit inference, proving the framework's robustness and practicality for real-time, resource-constrained edge AI applications. Sun, 30 Nov 2025 15:00:00 GMT https://scholar.dgist.ac.kr/handle/20.500.11750/59379 2025-11-30T15:00:00Z