Ching-Lieh Li

Prof. Ching-Lieh Li

orcid

Research Focus & Metrics

Academic Career Length

23Years (2004-Present)

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

Representative Publications

View All Publications
16 April 2004 in Radio Science
Verified
Journal Article
Open access
PDF file
CLChing-Lieh Li
Ching-Lieh Li
SCShao‐Hon Chen
Shao‐Hon Chen
CYChih‐Ming Yang
Chih‐Ming Yang
Chien-Ching ChiuChien-Ching Chiu
Chien-Ching Chiu
This paper presents a computational approach to the imaging of a partially immersed perfectly conducting cylinder by the steady state genetic algorithm. A conducting cylinder of unknown shape scatters the incident transverse magnetic (TM) wave in free space while the scattered field is recorded in free space. Based on the boundary condition and the measured scattered field, a set of nonlinear integral equations is derived and the imaging problem is reformulated into an optimization problem. An improved steady state genetic algorithm is employed to search for the global extreme solution. Numerical results demonstrate that, even when the initial guess is far away from the exact one, good reconstruction can be obtained. Numerical results also show that the reconstructed image in the free space area is better than that in the medium area.

27 November 2024 in IET Microwaves, Antennas & Propagation
Verified
Journal Article
Open access
LTLung‐Fai Tuen
Lung‐Fai Tuen
CLChing-Lieh Li
Ching-Lieh Li
YCYu-Jen Chi
Yu-Jen Chi
CCChien‐Ching Chiu
Chien‐Ching Chiu
PCPo Hsiang Chen
Po Hsiang Chen
This paper presents a systematic antenna design methodology that integrates machine learning, leveraging the progressive growth technique of Progressive Growing of GANs (PGGAN) to generate images of various dual‐band PIFA‐like antenna structures. The process involves using data augmentation methods to generate 4180 antenna samples. In the latent space, the authors employ Latin Hypercube Sampling to maintain diversity and combine it with the Hough Transform to enhance the edge features of the antennas while providing labelling functionality. This labelling method strengthens the relationship between antenna frequency and wavelength characteristics. The paper clearly outlines the design process, starting from the simulation of two types of single‐frequency PIFA‐like antennas (2.45 and 5.2 GHz, respectively) to the completion of PGGAN's generation task, resulting in a novel dual‐band Wi‐Fi PIFA‐like antenna structure. The measurement results of the dual‐band antennas exhibit excellent consistency with the simulation results.

5 November 2024 in Electronics
Verified
Journal Article
Open access
PDF file
Chien-Ching ChiuChien-Ching Chiu
Chien-Ching Chiu
CLChing-Lieh Li
Ching-Lieh Li
Po-Hsiang ChenPo-Hsiang Chen
Po-Hsiang Chen
HCHung-Ming Cheng
Hung-Ming Cheng
HJHao Jiang
Hao Jiang
This paper introduces a novel approach for reconstructing microwave imaging by combining the Whale Optimization Algorithm (WOA) with deep learning techniques. In it, electromagnetic waves are used to illuminate inhomogeneous dielectric objects in free space, and the scattered field is recorded. Due to the highly nonlinear nature of microwave imaging, the WOA is first employed to calculate an initial guess from the measured scattered field of dielectric objects. This step significantly reduces the training complexity for machine learning. Subsequently, the initial guess provided by the WOA is fed into a U-Net to accurately reconstruct the microwave image. Numerical simulation results indicate that the combination of the WOA and machine learning outperforms traditional methods under varying noise levels, enhancing the precision and effectiveness of the reconstruction process. In detail, the RMSE can be reduced 4–10% for dielectric constant distribution from 1 to 2.5 and SSIM can be increased about 30% for most cases.

12 February 2025 in Sensors
Verified
Journal Article
Open access
PDF file
Chien-Ching ChiuChien-Ching Chiu
Chien-Ching Chiu
CLChing-Lieh Li
Ching-Lieh Li
Po-Hsiang ChenPo-Hsiang Chen
Po-Hsiang Chen
YLYen-Chun Li
Yen-Chun Li
ELEng Hock Lim
Eng Hock Lim
U-Net with the iterative modified contrast scheme (IMCS) is proposed to solve inverse scattering problems (ISPs) in half-space. IMCS is an innovative inversion technique that utilizes contrast functions to improve the visibility of target regions and reconstruct the internal structure of objects. In contrast to applying IMCS alone, our proposed method improves the detection of contrast boundaries, enhancing noise immunity as well as increasing the structural similarity (SSI) through deep learning with U-Net. We compare the numerical results for 200-iteration IMCS and U-Net with 3-iteration IMCS, and it is found that the accuracy of reconstructed images can be improved a lot by U-Net with the 3-iteration IMCS architecture. In addition, even in the case of large Gaussian noise, the reconstruction is still good with our proposed method.

22 November 2025 in Electronics
Verified
Journal Article
Open access
PDF file
Lung-Fai Lung TuenLung-Fai Lung Tuen
Lung-Fai Lung Tuen
CLChing-Lieh Li
Ching-Lieh Li
YCYu-Jen Chi
Yu-Jen Chi
PCPo-Han Chen
Po-Han Chen
This study proposes a data-efficient fine-tuning strategy for multi-band antenna synthesis using a Wasserstein Auxiliary-Guided Progressive Growing GAN (WAG-PGGAN). Starting from a pretrained 512 × 512 dual-band PIFA-like generator trained on 4180 samples at 2.45/5.2 GHz, we introduce three 3.5-GHz wideband seeds augmented to 836 images (new:legacy ≈ 1:5) and fine-tune only the highest-resolution stage on the combined 5016-image corpus. A Hough-transform-based edge-enhancement module with an edge-aware loss preserves conductor boundaries and strengthens frequency–geometry correlation. Across n = 8 fabricated prototypes, all achieve |S11| < −10 dB and collectively span 1.86–5.83 GHz; measured total efficiencies are 52–87% (e.g., 73.6% @ 2.68 GHz, 66.7% @ 3.56 GHz, 69.0% @ 5.83 GHz), with radiation patterns consistent with simulation. The method retains prior 2.45/5.2 GHz performance while adding 3.5-GHz wideband behavior using ≤ 17% new data (836/5016), demonstrating effective transfer from small datasets. On an RTX 3060 Ti, inference is ≈ 3 s/design after ~192 h of training. Simulation–measurement agreement confirms that fine-tuned WAG-PGGAN yields high-resolution, physically valid multi-band antennas with reduced data and computational cost.

24 February 2023
Verified
Conference Paper
Open access
CLChing-Lieh Li
Ching-Lieh Li
BKBo-Lin Ke
Bo-Lin Ke
YCYu-Jen Chi
Yu-Jen Chi
In this paper, the application of LSTM and XGBoost models for rainfall prediction of a local area is investigated. The raw weather data, provided by the Central Weather Bureau of Taiwan government, for the Tamsui district of New Taipei City from 2001 to 2020 is utilized and cleaned, which is organized into three datasets: 1) the dataset of weather data for the same month only, 2) the dataset of weather data for the same month associated the two months before and after, respectively, and 3) the dataset of weather data for each month associated the whole year. The execution details show that LSTM and XGBoost have their own applicable months, respectively, and 1) for the rainfall forecast of each month, it is not necessary to use all the weather data of 12 months. In this regard, computing resources can be reduced, while the prediction results are even more accurate; 2) In 2020, the local rainfall pattern of nearly six different months do exhibit a time shift phenomenon; 3) There is no magic time sequence that is suited for all months.