Kyewon Kim

Prof. Kyewon Kim

information & communications engineering, Hoseo University, Asan, Chungnam, Republic of Korea
orcid

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
Verified
Journal Article
Open access
PDF file
XMXiangxu Meng
Xiangxu Meng
XLXin Liu
Xin Liu
Yi-Nan XuYi-Nan Xu
Yi-Nan Xu
YWYujing Wu
Yujing Wu
HLHang Li
Hang Li
Yihu XuYihu Xu
Yihu Xu
Accurate acoustic identification of marine mammals is vital for monitoring ocean health and human impacts. Existing methods often struggle with limited single-feature representations or suboptimal fusion of multiple features. This paper proposes an Evaluation-Adaptive Weighted Multi-Head Fusion Network that integrates CQT and STFT features via a dual-branch ResNet architecture. The model enhances intra-branch features using channel attention and adaptive weighting of each branch based on its validation accuracy during training. Experiments on the Watkins Marine Mammal Sound Database show that the proposed method achieves superior performance, reaching 96.05% accuracy and outperforming baseline and attention-based fusion models. This approach offers an effective solution for multi-feature acoustic recognition in complex underwater environments.

14 October 2024 in Electronics
Verified
Journal Article
Open access
PDF file
HLHojun Lee
Hojun Lee
BSBogeun Seo
Bogeun Seo
Kyewon KimKyewon Kim
Kyewon Kim
HKHaklim Ko
Haklim Ko
In this paper, we propose a phase-based pseudorange estimation and source localization algorithm to improve the localization performances of ultrashort baseline (USBL) systems. The conventional USBL system estimates pseudorange through cross-correlation of received chirps. However, due to the limited sampling rate, precise pseudorange cannot be estimated, resulting in low localization performance. The proposed method calculates the number of wavelengths between the transducer and receiver and detects the initial phase to estimate the pseudorange more accurately than the conventional method. It also improves localization performance by estimating and compensating for phase variations by underwater channels. The source localization performances of the proposed method and the conventional methods were compared and analyzed through computer simulations and ocean experiments, and the proposed method had better localization performances than the conventional methods.

1 January 2026 in IEEE Access
Verified
Journal Article
Open access
FMFanjin Meng
Fanjin Meng
HLHojun Lee
Hojun Lee
HKHaklim Ko
Haklim Ko
Kyewon KimKyewon Kim
Kyewon Kim

Verified
Journal Article
Open access
PDF file
YLYidan Liang
Yidan Liang
YCYu Cheng
Yu Cheng
YYYuxuan Yin
Yuxuan Yin
Yi-Nan XuYi-Nan Xu
Yi-Nan Xu
YWYujing Wu
Yujing Wu
Yihu XuYihu Xu
Yihu Xu
Reliable underwater visual perception plays a critical role in marine exploration, autonomous underwater vehicles, and subsea inspection. Nevertheless, underwater images are frequently degraded by severe color distortion, contrast reduction, and structural information loss due to wavelength-dependent light attenuation and scattering. Existing enhancement approaches either depend on fixed physical assumptions or require data-driven models, which often limits their adaptability across diverse underwater environments. To address these challenges, this paper presents a Physics-Guided Adaptive Bayesian Optimization (PG-ABO) framework for training-free underwater image enhancement. By combining physical degradation priors with image-specific parameter optimization, the proposed approach enables enhancement strategies to adapt dynamically to varying underwater conditions without relying on paired training data. The proposed framework incorporates an improved underwater dark channel prior with structure-aware transmission refinement to establish a physically consistent initial representation. Subsequently, a dual-path optimization strategy is constructed to generate complementary enhancement candidates by separately targeting color fidelity and structural perception. Bayesian optimization is employed to adaptively search enhancement parameters under the guidance of UCIQE and UIQM objectives. Furthermore, a robust multi-scale fusion strategy with anti-hotspot weight regulation is introduced, while a metric-aware refinement module is applied to enhance perceptual consistency and reduce enhancement artifacts. Experiments conducted on three underwater image datasets, including U45, UIEB, and LSUI, verify the effectiveness and robustness of the proposed framework across diverse underwater conditions. Consequently, the proposed method provides an adaptive, training-free, and physically motivated solution for underwater image enhancement under complex degradation conditions.

11 August 2026 in SSRN Electronic Journal
Verified
Preprint
Open access
YCYu Cheng
Yu Cheng
YLYidan Liang
Yidan Liang
YYYuxuan Yin
Yuxuan Yin
YXYinan Xu
Yinan Xu
HLHang Li
Hang Li
Yihu XuYihu Xu
Yihu Xu
Fixed-environment matched-field processing (MFP) is vulnerable when sound-channel-axis mismatch and array noise coexist. To address this problem, an energy–residual weighted joint source–environment matched-field localization method is proposed. Source range, source depth, and sound-channel-axis depth are incorporated into a unified search space, and candidates are evaluated using element-received energy, residuals after complex-gain alignment, conventional coherent matching, and weighted matching. To improve stability in the multimodal search space, the method combines multi-start genetic search, MFP-guided initialization, dynamic-neighborhood exploitation, local fine-grid refinement, low-fitness re-search, and posterior screening based on ambiguity score and peak contrast. Observation and replica fields are generated using a Munk sound-speed profile and a four-path image-source approximation. With a fixed 100-m sound-channel-axis mismatch and no added noise, the RMSEs of conventional MFP, standard GA-MFP, joint MFP, and ER-MPF are 417.61, 550.88, 2.52, and 33.08 m, respectively, showing that joint source–environment search is the primary mechanism for deterministic mismatch compensation. Under 5-dB circular complex Gaussian noise, energy–residual weighting reduces RMSE from 90.52 to 39.39 m and increases the 100-m hit rate from 93\% to 97\%. These results demonstrate improved robustness to environmental mismatch and noise under controlled simulation conditions.