A Latent-Diffusion-Based Inverse Model for Generating High-Resolution BPF Geometry Images
IEEE Microwave and Wireless Technology Letters · 2026-11 · 36 (11) · 1839-1842
DOI: 10.1109/lmwt.2026.3729977 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Microwave and Wireless Technology Letters, 2026-11. Verfasst von Ryota Maezaki und 2 weiteren Autor:innen. Die Arbeit nennt 23 Literaturangaben. Gefördert von: Japan Society for the Promotion of Science (JSPS) KAKENHI.
Abstract (Originalsprache)
This letter proposes an inverse model that generates high-resolution geometry images of microwave bandpass filters (BPFs) from coupling matrices by introducing a latent diffusion model (LDM). Recently, inverse models based on generative models have been studied as a new approach to microwave filter design, in which filter geometries are synthesized from desired frequency responses. However, existing inverse models either generate low-resolution filter geometry images or have not sufficiently demonstrated accurate BPF inverse design satisfying desired responses. To overcome this problem, the proposed inverse model consists of a variational autoencoder (VAE) and a diffusion model (DM). First, the VAE compresses BPF geometry images into a latent space. Next, the DM generates geometry images in the latent space. This approach allows us to generate high-resolution BPF geometry images, thereby enabling fine adjustment of resonator lengths and spacings on the generated images. As a case study, an inverse model for a third-order microstrip BPF is constructed to demonstrate the effectiveness of the proposed approach.
Förderung
- Japan Society for the Promotion of Science (JSPS) KAKENHI
Zitieren
APA
Maezaki, R., Ohira, M., & Deguchi, H. (2026). A Latent-Diffusion-Based Inverse Model for Generating High-Resolution BPF Geometry Images. IEEE Microwave and Wireless Technology Letters, 36(11), 1839-1842. https://doi.org/10.1109/lmwt.2026.3729977
Vancouver
Maezaki R, Ohira M, Deguchi H. A Latent-Diffusion-Based Inverse Model for Generating High-Resolution BPF Geometry Images. IEEE Microwave and Wireless Technology Letters. 2026;36(11):1839-1842. doi:10.1109/lmwt.2026.3729977
BibTeX
@article{Maezaki2026,
author = {Maezaki, Ryota and Ohira, Masataka and Deguchi, Hiroyuki},
title = {A Latent-Diffusion-Based Inverse Model for Generating High-Resolution BPF Geometry Images},
journal = {IEEE Microwave and Wireless Technology Letters},
year = {2026},
volume = {36},
number = {11},
pages = {1839--1842},
doi = {10.1109/lmwt.2026.3729977}
}
EUVIMED