木村 圭吾 | M, 2回目発表 | ネットワークシステム学 | 岡田 実 | 林 優一 | 東野 武史 | CHEN Na |
title: Optimizing Secondary RF Power to Reduce Leakage Magnetic Field in Dynamic Wireless Power Transfer
In recent years, Dynamic Wireless Power Transfer (DWPT) for electric vehicles (EVs) has gained significant attention. While DWPT offers a promising solution for EV charging, reducing the leakage magnetic field is critical for safety and system performance. Previous studies have focused on coil design and compensation circuits to address this issue. In this study, we propose optimizing the secondary coil's RF power supply to minimize leakage magnetic fields. The theoretical model derived through mathematical equations was implemented in simulations, demonstrating effective field reduction. This research aims to enhance the safety and efficiency of DWPT systems through this optimization approach. language of the presentation: Japanese | ||||||
德永 凜 | M, 2回目発表 | ネットワークシステム学 | 岡田 実 | 林 優一 | 東野 武史 | CHEN Na |
title: Coverage expansion for multi-LPWA in
Radio-on-Fiber transmission with modulation power
allocation
abstract: The low power wide area (LPWA) is a promising wireless access for Internet-of-Things (IoT) devices. The analog radio-on-fiber (A-RoF) technique is one of the resolutions for decreasing the radio dead zone which enables us to transfer the different kinds of LPWA air interface compared to the homogeneous multi-hop system. This paper introduces the modulation power allocation for heterogeneous LPWA wireless network in analog radio-on-fiber. The proposed method enables us to improve the signal to inter-modulation distortion power ratio with keeping the total optical modulation index. The wireless coverage expansion is also numerically evaluated in addition to the EVM improvement. language of the presentation: Japanese | ||||||
中村 康一郎 | M, 2回目発表 | ネットワークシステム学 | 岡田 実 | 林 優一 | 東野 武史 | CHEN Na |
title: Anomaly detection of RoF MIMO link using machine learning abstract: In recent years, base station configurations using Radio over Fibers(RoF) have been considered. Although the advantage of RoF is its simple configuration, a method for detecting anomalies in devices such as semiconductor lasers, fiber-optic transmission lines, photodetectors, and electric amplifiers has not been discussed so far. In this presentation, we focus on the gain degradation of transmission lines and propose a method to detect anomalies using machine learning techniques. language of the presentation: Japanese | ||||||
浦上 大世 | D, 中間発表 | ネットワークシステム学 | 岡田 実 | 林 優一 | 東野 武史 | CHEN Na |
title: A Study on Beam Control of Reconfigurable Intelligent Surface
abstract: Reconfigurable intelligent surfaces (RISs) have recently been considered as the countermeasures for the coverage hole problem in the fifth- and sixth-generation (5G/6G) mobile communication systems. In this presentation, we propose a collaborative learning framework, named self-enhanced multi-task and split federated learning (SM-SFL), for joint channel semantic reconstruction and beamforming for RIS-aided cell-free (CF) systems. Simulation results show that the proposed framework can achieve better channel semantic reconstruction, higher SE with accurate beam selection, negligible computation overhead, and highly efficient multi-device cooperation. language of the presentation: Japanese | ||||||
森 和真 | M, 2回目発表 | インタラクティブメディア設計学 | 加藤 博一 | 清川 清 | 神原 誠之 | 藤本 雄一郎 |
title: A method to visualize uncertainly of state recognition in AR step-by-step tutorials
abstract: Augmented reality (AR)-based step-by-step tutorials are gaining attention as an effective tool to improve user work efficiency in the manufacturing and educational sectors. There is a function that recognizes the current step and automatically displays the next step for efficient tutorials. However, even with the latest object detection technology, misrecognition is likely to occur when recognizing visually similar components, which may hinder the accurate progression of the tutorial. In this study, we clarify how to display information including state recognition ambiguity in AR step-by-step tutorials. language of the presentation: Japanese | ||||||
末原 和樹 | M, 2回目発表 | 生体医用画像 | 佐藤 嘉伸 | 加藤 博一 | 大竹 義人 | SOUFI Mazen |
title: Bone density estimation using low-dose imaging system (EOS Imaging System)
abstract: Musculoskeletal diseases such as osteoporosis are typical examples of musculoskeletal diseases that cause bone fractures. Since the progression of these conditions can seriously affect daily life, regular examination of musculoskeletal health is very important, and a method that can accurately diagnose these conditions with low radiation exposure is expected. Unlike conventional radiography, EOS can simultaneously capture images from two directions (frontal and lateral) in a standing position. Existing studies have estimated bone mineral density (aBMD: area bone mineral density, vBMD: volumetric bone mineral density) and muscle mass using a single X-ray image, but no study has been conducted using EOS images. In this study, we aim to estimate bone density for EOS images by fine tuning an AI model trained on X-ray images with frontal EOS images, and to evaluate its accuracy using bone density and muscle mass measured from CT images. Experiments were conducted using CT and EOS images of 77 patients undergoing hip arthroplasty who did not include bilateral implants before surgery. As a result, we confirmed that the Pearson correlation coefficient between the estimated proximal femur bone mineral density and the correct value is 0.928, and that the correlation is high even from EOS images taken at low doses. language of the presentation: Japanese | ||||||
GOURINE SANAA AMINA | M, 2回目発表 | 生体医用画像 | 佐藤 嘉伸 | 加藤 博一 | 大竹 義人 | SOUFI Mazen |
title: Automated musculoskeletal segmentation and analysis of torso CT images in a large-scale database
abstract: Musculoskeletal (MSK) segmentation in CT scans is important for diagnosing diseases like sarcopenia by assessing muscle quality. Previous research mainly automated segmentation of the lower body, but torso segmentation was limited to 2D-specific spine level or a 3D segmentation performed on a few Torso structures and not the whole structures. The creation of large databases of these images helps validate models and analyze body features, but the lack of ground truth data complicates measuring model accuracy. A useful method involves using predictive uncertainty to show how confident models are in their predictions. My research uses state-of-the-art models (2D Bayesian UNet and 3D nnUNet) on a dataset of 20 cases and 21 torso structures to study this uncertainty. I also looked at how muscle features, like volume and fat content, relate to demographic details (sex, age) in a large 3D CT scan database. Lastly, I will briefly demonstrate my current progress and future tasks. language of the presentation: English | ||||||