| JU YUAN | 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, | Dipanita Chakraborty |
|
title: Improvement of Misalignment Tolerance in 85kHz DWPT Systems using Meta-Transmission Line Structure
abstract: In Dynamic Wireless Power Transfer (DWPT), misalignment between transmitting and receiving coils due to vehicle movement causes a decrease in the coupling coefficient, leading to reduced power transfer efficiency. This paper proposes a novel approach to improve misalignment tolerance for an 85kHz DWPT system. Specifically, a meta-transmission line structure is introduced by placing a "reflection coil" between adjacent transmitter coils to optimize the electromagnetic coupling characteristics. We evaluated the system using a dynamic power transfer model where the receiver coil moves within a range of -400 mm to +400 mm. The results show that the proposed method effectively suppresses the degradation and fluctuation of the coupling coefficient during movement compared to conventional methods. This indicates the potential to maintain stable and high-efficiency power transfer over a wide range even during vehicle movement. language of the presentation: English | ||||||
| LI ZHIDONG | 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, | Dipanita Chakraborty |
|
title: Radio-over-Free Space Optics based Inter-Satellite Links for Ultra-Dense Satellite Swarms as a Distributed Phased Array
abstract: Future space missions envision ultra-dense satellite swarms composed of thousands of pico-satellites operating in close proximity (e.g., 10 cm spacing) to function as a single large-aperture distributed antenna. This research proposes a Radio-over-Free Space Optics (RoFSO) architecture to distribute Radio Frequency (RF) signals and local oscillator references among the satellites. Unlike traditional RF inter-satellite links, the proposed optical system leverages narrow beam divergence to minimize inter-node interference in high-density formations. Furthermore, this study numerically analyzes the impact of optical signal degradation on the array's performance. Specifically, simulation results reveal that optical beat noise generated during the photonic mixing process induces phase jitter, which poses a fundamental limit to the beamforming efficiency of the distributed antenna system. language of the presentation: English | ||||||
| 河崎 伸太朗 | M, 2回目発表 | ヒューマンロボティクス（デジタルヒューマン学） | 和田 隆広☆, | 多田充徳, | 丸山翼 | |
|
title: A Study on the Evaluation of Joint Rotation in Pitching Motions for Pose Estimation Using an Event Camera
abstract: In monitoring daily life, it is necessary not only to classify postures but also to continuously track subtle changes in physical condition. Human Mesh Reconstruction (HMR) is effective for capturing detailed information on whole-body posture, and the Mean Per Joint Position Error (MPJPE) is widely used to evaluate its accuracy. However, since MPJPE treats errors in joint position and mesh geometry as a composite measure, it does not clearly evaluate joint rotation itself. Therefore, this study focuses on evaluating joint rotation to complement conventional metrics. Additionally, we use event cameras, which output luminance changes—a feature considered useful for non-contact monitoring and privacy protection. In this study, we selected the throwing motion—which involves significant changes in joint angles and allows for angle-based motion phase segmentation—as the evaluation target and constructed a synthetic event dataset. Furthermore, we aim to analyze joint rotation errors based on the results of event-based HMR and identify rotational information that is effective for distinguishing motion states. language of the presentation: Japanese 発表題目: イベントカメラを用いた姿勢推定における投球動作の関節回転評価の検討 発表概要: 日常生活のモニタリングでは、姿勢分類に加え、身体状態の微小な変化を継続的に把握することが求められる。人体メッシュ復元(HMR)は全身姿勢における詳細の把握に有効であり、その精度評価として，関節位置誤差(MPJPE)が広くも散られている。一方、MPJPEは関節位置とメッシュ形状の誤差が複合的に扱われるため、関節回転そのものの評価は明確でない。そこで本研究では、従来指標を補完する関節回転評価に着目する。また、非接触性、プライバシー保護に有用とされている輝度変化を出力するイベントカメラを用いる。今回は、関節角度の変化が大きく、角度に基づく動作フェーズ分割が可能な投球動作を評価対象とし、合成イベントデータセットを構築した。さらに、イベントベースHMRの結果から関節回転誤差を解析し、動作状態の識別に有効な回転情報を明らかにすることを目的としている。 | ||||||
| 荒深 健伍 | M, 2回目発表 | 光メディアインタフェース | 向川 康博, | 和田 隆広, | 藤村 友貴, | 北野 和哉 |
|
title: Material Classification Using Low-Temporal-Resolution Multi-Zone SPAD Sensor
abstract: Material information is important for scene understanding. The objective of this work is to classify plastic materials with similar appearances using a non-contact method. With an 8x8 zones SPAD sensor capable of detecting single photons, we measure the slight time delay of light occurring inside the material and utilize it as a feature of materials. Through real-world experiments, we demonstrated that it is possible to classify multiple materials in simple scenes, even with low-time-resolution measurements of approximately 100 ps. language of the presentation: Japanese 発表題目: 低時間分解能な多画素SPADセンサによる材質分類 発表概要: 物体の材質に関する情報は、シーンの理解において重要である。本研究では、外観の類似したプラスチック材質を非接触で分類することを目的とする。単一光子を検出可能な8x8画素のSPADセンサを用いて、材質内部で生じる光のわずかな時間遅れを計測し、材質固有の特徴として利用する。実環境実験を通じて、100ps程度の低時間分解能な計測でも単純なシーンであれば複数種類の材質が分類可能であることを示した。 | ||||||
| 越間 龍之介 | M, 2回目発表 | 光メディアインタフェース | 向川 康博, | 和田 隆広, | 藤村 友貴, | 北野 和哉 |
|
title: Spectral Analysis for Tomb Murals Using a Foundation Model Approach
abstract: This study aims to apply a foundation model to spectral images of tomb murals to achieve digital restoration of damaged sections. Previously, analysis of tomb murals has been conducted using physics-based methods or autoencoders. However, this approach faced challenges due to the scarcity of training data inherent to cultural heritage data. This research aims to achieve more accurate and natural restoration of damaged areas by leveraging the general visual knowledge acquired by the foundation model through extensive pre-training and the detailed color information contained in spectral images. language of the presentation: Japanese 発表題目: 古墳壁画における基盤モデル適用による分光壁画解析 発表概要: 本研究は，基盤モデルを古墳壁画の分光画像に適用し，損傷部分のデジタル修復を実現することを目指す．従来，古墳壁画の解析は，物理モデルベースの手法やオートエンコーダを用いて行われてきた．しかし，文化財特有のデータ希少性による学習データの不足という課題を抱えていた．本研究では，基盤モデルが大規模な事前学習を通じて獲得した汎用的な視覚知識と分光画像が持つ詳細な色彩情報から，損傷部分をより高精度かつ自然に復元することを目的とする． | ||||||
| 鈴木 涼太 | M, 2回目発表 | 生体画像知能 | 大竹 義人, | 向川 康博, | Soufi Mazen, | Gu Yi |
|
title: Development of a Musculoskeletal Shape Prediction Method from Non-contact Three-dimensional Body Surface Scan Data
abstract: This study aims to estimate the shapes of bones and muscles from body surface scan data. The ultimate goal is to develop a non-invasive and practical method for estimating internal anatomical structures for health management applications. As a preliminary study, a Statistical Shape Model (SSM) was constructed for the pelvis, femurs, and the bilateral gluteus maximus, gluteus medius, and gluteus minimus. Musculoskeletal shapes were then estimated by fitting the SSM to the body surface geometry. In future work, the target anatomy will be extended to the entire lower limb by improving the quality of the segmentation data used for SSM construction. Since errors in automatic segmentation are expected to affect both the quality of the shape model and the prediction accuracy, we will refine the existing dataset and retrain the segmentation model to obtain more accurate segmentation results. This is expected to improve the accuracy of musculoskeletal shape prediction. language of the presentation: Japanese 発表題目: 非接触三次元体表面計測データからの筋骨格形状予測手法の開発 発表概要: 本研究では，体表面計測データから骨および筋肉の形状を推定することを目的とする。これにより，健康管理において非侵襲的かつ簡便に内部形状を推定できる手法の実現を目指す。 予備実験では，骨盤，大腿骨および左右の大殿筋・中殿筋・小殿筋を対象として統計的形状モデル（Statistical Shape Model: SSM）を構築し，体表面形状とのフィッティングによる筋骨格形状推定を行った。 今後は，対象部位を下肢全体へ拡張するため，SSM構築に用いるセグメンテーションデータの高品質化に取り組む。自動セグメンテーションの誤りは形状モデルの品質や推定精度に影響を及ぼすと考えられるため，既存データセットの修正および再学習を通して高精度なセグメンテーションモデルを構築し，より高精度な筋骨格形状推定を目指す。 | ||||||
| RUHIYAH FARADISHI WIDIAPUTRI | D, 中間発表 | ヒューマンＡＩインタラクション | Sakriani Sakti, | 渡辺 太郎, | 大内 啓樹, | Faisal Mehmood, | Bagus Tris Atmaja |
|
title: *** Structural Ambiguity Resolution in Indonesian–English Speech-to-Text Translation by Utilizing Prosodic Information ***
abstract: *** Most speech-to-text translation (ST) systems overlook structural ambiguity in the source speech, often resulting in inaccurate or ambiguous translations. To address this issue, this study introduces the first structural ambiguity-free ST system, which not only translates spoken language into text in another language but also resolves structural ambiguity by leveraging prosodic information from the input speech, thus providing non-ambiguous translations. Our contributions include: 1) developing the first structural ambiguity-free ST corpus for Indonesian-to-English translation and 2) proposing structural ambiguity-free ST by modifying the cascaded and end-to-end approaches of standard ST to leverage prosodic information in producing structural ambiguity-free output. Experimental evaluations demonstrate that our best-performing system achieves strong translation quality (BLEU 42.57, chrF++ 59.88, COMET 79.09) while effectively disambiguating input speech according to the intended interpretation with an accuracy of 85.17%. *** language of the presentation: *** English *** | |||||||