Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Chien-Ching Chiu has a total of 141 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    GAN with Self-Attention for Electromagnetic Imaging of Biaxial Objects

    by Chien-Ching Chiu, Po-Hsiang Chen, Yi-Hsun Chen, Hao Jiang

    Electromagnetic imaging of anisotropic objects is investigated using a deep learning-based framework combined with a Self-Attention (SA) mechanism. In this work, anisotropic scatterers are excited using both TM (Transverse Magnetic)- and TE (Transverse Electric)-polarized electromagnetic waves, and the resulting scattering responses are measured to reconstruct the electromagnetic image of the target. A Back Propagation Scheme (BPS) is implemented to generate a first-pass reconstruction of the dielectric distribution. Unlike our previous work on uniaxial objects, the present study extends the BPS formulation to biaxial anisotropic media, enabling the independent reconstruction of the permittivity components along the x-, y-, and z-directions. The reconstructed coarse profiles are subsequently refined through a Generative Adversarial Network with Self-Attention (GAN-SA) operations. Unlike conventional GAN approaches, the proposed framework enhances feature correlation across different spatial regions, allowing the network to preserve object boundaries and dielectric variations more effectively. Simulation results with 5% and 20% Gaussian noise reveal that the proposed framework achieves better reconstruction quality, improved image clarity, and stronger resistance to noise interference. In addition, the integration of the SA module significantly improves the reconstruction of anisotropic dielectric properties while reducing reconstruction artifacts. The proposed GAN-SA model contains approximately 5.2 million trainable parameters, requires approximately 20 min for training, and achieves an inference time of less than 1 s for each reconstruction. These results demonstrate the effectiveness and computational efficiency of the proposed framework for biaxial anisotropic electromagnetic imaging.

  • Open Access

    Application of Cascaded Convolutional Neural Network in Electromagnetic Imaging

    by Tsai-Hua Kang, Jian-Yu Lee, Tzu-Yin Lin, Chien-Ching Chiu, Po-Hsiang Chen

    A cascaded neural network framework is developed to solve the two-dimensional inverse scattering problem under transverse magnetic wave illumination. The framework employs a multi-stage deep learning architecture in which each sub-network iteratively refines the reconstruction from the previous stage. The framework enhances image resolution and structural accuracy. The cascaded design is adopted to address the inherent nonlinearity and ill-posedness of the problem, particularly when reconstructing high-contrast and geometrically complex dielectric objects. Numerical simulation results showed the effectiveness and superior performance of the framework in terms of structural similarity index and root-mean-square error when compared with a convolutional neural network. The framework’s cascaded architecture enables robust and high-fidelity reconstructions even under significant measurement noise.

  • Open Access

    Electromagnetic Inverse Scattering of Buried Conductors in Half-Space Using TE Waves and Deep Convolutional Neural Networks

    by Po-Hsiang Chen, Chien-Ching Chiu, Hsin-Chien Wen, Hao Jiang

    Electromagnetic Inverse Scattering (EMIS) is a key technology for reconstructing the geometry of buried conductors based on measured scattered fields. This research employs a Deep Convolutional Neural Network (DCNN) to address the intrinsic nonlinear characteristics in reconstructing perfect conductors in a half-space using transverse electric (TE) waves. First, TE electromagnetic waves are transmitted to illuminate the conductor, and the corresponding scattered fields are collected. Subsequently, the scattered field is fed into the DCNN to reconstruct the precise shape of the conductor. Considering that real-world measurements may contain noise, we add 5%, 10%, and 20% noise levels in the simulation. Numerical results for various conductor geometries, including shield, oval, arrow, heart, and four-petal shapes, demonstrate the reconstruction capability of the proposed method under the investigated Gaussian noise levels. These results demonstrate the feasibility of applying a DCNN to the reconstruction of buried conductors under TE-wave illumination.

  • Open Access

    Transverse Electric Inverse Scattering of Buried Conductors in a Slab Medium Using DSM and U-Net

    by Chien-Ching Chiu, Po-Hsiang Chen, Yen-Chen Chang, Eng Hock Lim

    This paper presents a DSM-initialized U-Net framework for reconstructing the shape of a perfectly conducting target buried in a slab medium from measured electromagnetic scattered fields. The conductive target is irradiated with Transverse Electric (TE) waves, followed by the acquisition of the resulting scattered fields. These scattered field data were first used for forward scattering calculations with the Method of Moments (MoM), followed by preliminary shape reconstruction using the DSM. The preliminary images received from the DSM were then input into a U-Net for imaging. Compared with the free space case, the slab medium condition limited the incident and receiving angles. Numerical results indicate that the standalone DSM provides only an approximate estimate of the conductor profile, whereas the proposed DSM–U-Net framework achieves average NRMSE values of 6.90% under 5% Gaussian noise and 8.95% when trained with 5% noise and tested with 10% noise.

  • Open Access

    Image Reconstruction by Frequency Extrapolation and Deep Learning in Three-Layer Medium

    by Chien-Ching Chiu, Po-Hsiang Chen, Guan-Jang Li, Eng Hock Lim

    This paper proposes a novel multi-frequency extended Deep Learning (DL) model for electromagnetic image reconstruction under Transverse Magnetic (TM) wave incidence in layered media, inspired by conventional microwave imaging techniques that combine nonlinear inversion algorithms with neural networks to improve reconstruction performance. The proposed framework adopts a two-stage neural network architecture. In the first stage, a Deep Residual Convolutional Neural Network (DRCNN) is employed to extrapolate multi-frequency scattered fields from single-frequency input data, thereby enriching the frequency-dependent scattering information available for reconstruction. Subsequently, the extrapolated multi-frequency scattered fields are fed into a Deep Convolutional Encoder–Decoder (DCED) network to reconstruct an accurate dielectric constant distribution within the imaging domain. To validate the effectiveness of the proposed approach, two representative comparison methods are considered: (1) a hybrid framework combining the Back-Propagation Scheme (BPS) with a Convolutional Neural Network (CNN), and (2) a framework integrating the Dominant Current Scheme (DCS) with a CNN. In both approaches, conventional inversion algorithms are first utilized to generate coarse initial reconstructions, which are subsequently refined by the neural network. Numerical simulations and experimental results show that the proposed multi-frequency extension model achieves lower reconstruction error and higher structural similarity than the reference methods. These results confirm the effectiveness and potential of the proposed framework for advanced electromagnetic imaging applications.

  • Open Access

    Hybrid-Input Deep Learning Framework for Electromagnetic Imaging of Anisotropic Objects

    by Chien-Ching Chiu, Po-Hsiang Chen, Chih-Yu Wang, Eng Hock Lim

    This paper presents a hybrid-input deep learning framework for electromagnetic imaging of anisotropic objects in a two-dimensional environment. The proposed method combines transverse magnetic (TM) and transverse electric (TE) illuminations to reconstruct the directional permittivity components. Backpropagation (BP) and the direct sampling method (DSM) are first applied to the measured scattered fields to generate complementary preliminary images. These images are then superimposed to form a physically informed hybrid input for U-Net reconstruction. Numerical results show that the proposed hybrid-input strategy improves reconstruction accuracy, structural consistency, and noise robustness compared with single-input baselines using BP or DSM alone. The results demonstrate the potential of the proposed method for anisotropic electromagnetic imaging.

  • Open Access

    Electromagnetic Imaging of Anisotropic Objects Using a Self-Attention Perceptual Generative Adversarial Network

    by Po-Hsiang Chen, Chien-Ching Chiu, Yang-Han Lee, Eng Hock Lim

    Reconstructing high-resolution images of anisotropic targets in microwave imaging remains a challenging problem due to the strong directionality of electromagnetic responses and the inherent nonlinearity of the inverse scattering process. To address these issues, we propose a novel Perceptual Generative Adversarial Network (PGAN) enhanced with a Self-Attention mechanism for anisotropic electromagnetic imaging. The perceptual loss encourages the preservation of high-level structural features, while the Self-Attention module enables the model to capture long-range dependencies and directional correlations that are critical in representing anisotropic material distributions. This joint architecture is trained to refine coarse permittivity estimates obtained from conventional Back-Propagation Schemes (BPSs). Numerical simulations and validation using measured experimental data demonstrate that the proposed method achieves improved reconstruction accuracy and structural similarity compared with the PGAN without SA and U-Net. In particular, PGAN with SA reduces the Root Mean Square Error (RMSE) by 15.1% and improves the Structural Similarity Index Measure (SSIM) by 3.8%, confirming its effectiveness in recovering fine-scale details and enhancing reconstruction quality. These results suggest that the proposed framework offers a promising solution for robust and high-resolution electromagnetic imaging in geophysical and remote sensing applications.

  • Open Access

    Indoor Localization for 6G Communication Systems Under Single Co-Channel Interference Using a Back Propagation Neural Network with Hybrid Self-Attention

    by Chien-Ching Chiu, Hung-Yu Wu, Po-Hsiang Chen, Chen-En Chao, Eng Hock Lim

    Recent advancements in digital signal processing and lightweight neural architectures have opened new possibilities for developing efficient and interference-resilient indoor localization systems suitable for next-generation wireless networks. This paper proposes an indoor localization system for 6G communication systems with single co-channel interference. Ray tracing technique is used to compute the frequency-domain channel state information (CSI). Next, CSI fingerprints are input into a lightweight back propagation neural network (BPNN) with channel-wise self-attention (CSA) and spatial self-attention (SSA) to improve the model’s resilience to interference and noise. Numerical results demonstrate that the attention-enhanced BPNN models significantly outperform the standard BPNN. In particular, CSA focuses on informative frequency channels and excels under spectral distortion caused by interference, while SSA puts emphasis on spatial features and shows superior performance in spatially stable environments, reducing RMSE by up to 20% in high interference scenarios and 10% in low interference scenarios, respectively. These findings validate the effectiveness of integrating attention mechanisms into neural localization frameworks, making them well-suited for next-generation 6G indoor positioning systems in interference-limited environments.


  • Open Access

    Inverse Reconstruction of Uniaxial Dielectric Objects in Slab Medium Using Deep Learning Techniques

    by Wei Chien, Chien-Ching Chiu, Po-Hsiang Chen, Guan Jang Li, Hao Jiang

    Electromagnetic imaging in a slab medium presents significant challenges due to complex wave reflections and refractions at the interfaces of different layers. Multiple scattering and interference increase ill-posedness and nonlinearity, degrading reconstruction accuracy and stability. Under transverse magnetic (TM) and transverse electric (TE) excitations, we compare the CNN-refined reconstructions based on the Back Propagation Scheme (BPS) and the Dominant Current Scheme (DCS) to solve the Electromagnetic Inverse Scattering (EMIS) problem. Numerical results demonstrate that our proposed method can accurately reconstruct buried objects of various sizes and positions, even in the presence of noise. In particular, the DCS-CNN framework yields superior reconstruction performance compared to the BPS-CNN approach, highlighting the advantage of integrating the DCS with DL for imaging in a slab medium. Overall, this work validates the feasibility and effectiveness of combining preliminary imaging with DL, offering practical potential for solving complex inverse scattering problems.

  • Open Access

    Two-Stage Multi-Frequency Deep Learning for Electromagnetic Imaging of Uniaxial Objects

    by Wei-Tsong Lee, Chien-Ching Chiu, Po-Hsiang Chen, Guan-Jang Li, Hao Jiang

    In this paper, an anisotropic object electromagnetic image reconstruction system based on a two-stage multi-frequency extended network is developed by deep learning techniques. We obtain the scattered field information by irradiating the TM different polarization waves to uniaxial objects located in free space. We input the measured single-frequency scattered field into the Deep Residual Convolutional Neural Network (DRCNN) for training and to be further extended to multi-frequency data by the trained model. In the second stage, we feed the multi-frequency data into the Deep Convolutional Encoder–Decoder (DCED) architecture to reconstruct an accurate distribution of the dielectric constants. We focus on EMIS applications using Transverse Magnetic (TM) and Transverse Electric (TE) waves in 2D scenes. Numerical findings confirm that our method can effectively reconstruct high-contrast uniaxial objects under limited information. In addition, the TM/TE scattering from uniaxial anisotropic objects is governed by polarization-dependent Lippmann–Schwinger integral equations, yielding a nonlinear and severely ill-posed inverse operator that couples the dielectric tensor components with multi-frequency field responses. Within this mathematical framework, the proposed two-stage DRCNN–DCED architecture serves as a data-driven approximation to the anisotropic inverse scattering operator, providing improved stability and representational fidelity under limited-aperture measurement constraints.