| 仲 純平 | M, 1回目発表 | 光メディアインタフェース | 向川 康博, 加藤 博一, 舩冨 卓哉, 藤村 友貴, 北野 和哉 |
|
title: Verification of Robustness in Laser Speckle Authentication Focusing on
Distance between Image Sensor and Object
abstract: Artifact-metrics is a technology that measures the unique physical characteristics of an object. This technology makes it possible to identify individuals without the use of printed barcodes or serial numbers. Laser speckle authentication, one of the individual identification technologies, can identify objects with high accuracy due to the high independence and randomness of the speckle pattern. On the other hand, it is sensitive to changes in the speckle pattern caused by object misalignment. To make the system robust against object misalignment, the speckle pattern identification method needs to be improved. As a preliminary step, we focused on the distance from the image sensor to the object and verified the robustness of the system against changes in distance. language of the presentation: Japanese 発表題目: 撮像素子・物体間の距離に着目したレーザスペックル認証におけるロバスト性の検証 発表概要: 人工物メトリクスは、物体に固有の物理的特徴を測定する技術である。この技術を用いると、印字されたバーコードやシリアル番号を用いずに物体を個体ごとに識別することが可能になる。個体識別技術の一つであるレーザスペックル認証は、スペックル模様の高い独立性・ランダム性のために物体を高精度に識別できる。一方で、物体の位置ずれによるスペックル模様の変化にセンシティブであるという課題がある。システムを物体の位置ずれに対してロバストにするには、スペックル模様の識別方法を改善する必要がある。本研究では、その前段階として、撮像素子から物体までの距離に着目し、距離の変化に対するロバスト性を検証した。 | |||
| 藤本 悠太 | M, 1回目発表 | 光メディアインタフェース | 向川 康博, 加藤 博一, 舩冨 卓哉, 藤村 友貴, 北野 和哉 |
|
title: Hyperspectral Imaging by Spatial-Spectral Encoding
abstract: Although hyperspectral imaging is expected to be applied in various fields such as agriculture and medicine, complicated and costly optical system and time-consuming measurement are the current obstacles for these applications. Therefore, this study aims to propose a hyperspectral imaging method with spatially random spectral encoding and computational decoding on a simple optical system. Furthermore, we verified the principle of this research and demonstrated its effectiveness through simulation experiments. language of the presentation: *** Japanese *** 発表題目: 空間ー波長符号化による分光計測手法の提案 発表概要: 分光計測は農業や医学など、様々な分野で活用することが期待できるにも関わらず、現状は光学系が複雑でコストが高いことや一度の計測に時間がかかるというような問題がある。そこで本研究では、空間的にランダムな波長の符号化と計算による復号化を活用した分光計測を簡易な光学系で実現する方法を提案する。さらにシミュレーション実験を通して本研究の原理検証を行い、その有効性を実証した。 | |||
| 木村 江梨花 | M, 1回目発表 | サイバネティクス・リアリティ工学 | 清川 清, 加藤 博一, 内山 英昭, Perusquia Hernandez Monica, 平尾 悠太朗 |
|
title: Enhancing Avatar Embodiment Sensation in Virtual Reality through Priming
The Proteus effect, recognized for behavioral changes resulting from preconceived notions about avatars, has been identified in the context of virtual reality (VR). Furthermore, it is established that priming, involving instructional cues designed to reinforce preconceptions, can intensify the Proteus effect. Experiments involving priming based on avatar characteristics conducted with participants before their VR experiences have revealed observable behavioral changes in those who underwent such priming. Therefore, this study aims to explore the influence of priming avatars with characteristics different from pre-existing biases and to examine how such priming affects the Proteus effect. language of the presentation: *** English or Japanese (choose one) *** 発表題目: プライミングを用いたバーチャルリアリティにおけるアバター身体化感覚の強化 発表概要: アバターに対する先入観から引き起こされる行動変容でプロテウス効果が知られている。また、先入観を助長するような教示を行うプライミングによりプロテウス効果が増強されることが知られている。VR体験を行う前の被験者にアバター特性に従ったプライミングを行った実験では、プライミングを受けた被験者に行動変化が見られた。そこで、アバターの先入観とは異なる特性のプライミングを行い、プライミングがプロテウス効果に及ぼす影響を検討する。 | |||
| 堀 光希 | M, 1回目発表 | サイバネティクス・リアリティ工学 | 清川 清, 加藤 博一, 内山 英昭, Perusquia Hernandez Monica, 平尾 悠太朗 |
|
title: *** Improving Interaction of Pseudo-Haptics through Visual Cue Separation ***
abstract: *** One of the methods for providing haptic stimuli in virtual reality environments is Pseudo-Haptics, which manipulates tactile perception by appropriately changing visual feedback based on body movements. Pseudo-Haptics has the advantage of not requiring dedicated equipment as it can express tactile sensations through visual stimuli. However, there have been issues with Pseudo-Haptics, such as generating discomfort in the interaction and limited applicability due to altering the behavior and characteristics of virtual objects. In this study, we propose a method to separate the visual cues that induce Pseudo-Haptics from the behavior and characteristics of virtual objects and present them as a separate user interface (UI). The objective is to provide Pseudo-Haptics without compromising the quality of interaction. *** language of the presentation: *** Japanese *** 発表題目: *** 視覚的手がかりの分離によるPseudo-Hapticsのインタラクション向上 *** 発表概要: *** VR空間内で触覚刺激を提示する手法の一つに、Pseudo-Haptics（疑似触覚）が存在する。Pseudo-Hapticsは、身体の動きに応じて視覚的なフィードバックを適切に変化させることで触知覚を操作するものであり、視覚的な刺激によって触覚を表現することができるため専用の機器を必要としない利点がある。しかし、Pseudo-Hapticsはバーチャルな物体の挙動や特性を変化させるため、提示する際にインタラクションに違和感が生じるほか、応用できる状況が限定的であるといった問題があった。そこで本研究では、Pseudo-Hapticsを引き起こす視覚的手がかりをバーチャル物体の挙動や特性から分離してUIとして提示する手法を提案し、インタラクションの質を低下させることなくPseudo-Hapticsを提示することを目的とした。 *** | |||
| 森 和真 | M, 1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, 澤邊 太志 |
|
title: AR inspection work system for long-term use using head-mounted display
abstract: There are inspection work in the AR work support system to record equipment values and conditions. In actual work, the system is expected to be used for long periods of time, more than one hour. However, problems of eye strain and muscle fatigue are assumed for long hours of AR using a head-mounted display, and these problems have not yet been clarified. Therefore, this study proposes an AR inspection work system for long-time use with a head-mounted display by designing an eye-friendly UI that takes eye strain into account and examining input and operation methods that take muscle fatigue into account. language of the presentation: Japanese | |||
| GOURINE SANAA AMINA | M, 1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: ***Development of a Whole Body Muscle Segmentation Tool from CT data ***
abstract: *** Muscle segmentation plays a crucial role in diverse areas such as biomechanical analysis, which can help in understanding muscle function movement through the analysis of muscle size and geometry. Advances in deep learning have facilitated the development of automated segmentation techniques with high accuracy, which help in constructing high-fidelity biomechanical models. However, several factors could potentially impact the accuracy of this segmentation, such as the variability in imaging conditions related to different scanners or manufacturers, which have not been sufficiently investigated. Additionally, training data is usually collected from centers targeting only specific structures (partially labeled data). Our research aims to develop an automated tool for biomechanical analysis based on musculoskeletal segmentation in CT images. Given our lab's existing research on the hip and lower body muscles, our focus will now extend to include the muscles of the chest and abdomen while developing a segmentation tool that deals with different image qualities and exploits partially labeled data. Finally, I will briefly demonstrate my current progress and organized datasets for the upcoming steps. *** language of the presentation: *** English*** | |||
| EUNIKE ANDRIANI KARDINATA | M, 2回目発表 | 自然言語処理学 | 渡辺 太郎, 荒牧 英治, 大内 啓樹 |
|
Title: Constructing Indonesian-English Travelogue Dataset
Abstract: Research in low-resource language are often hampered due to the under-representation of how the language is being used in reality. This is particularly true for Indonesian language because there is a limited variety of textual dataset and majority were acquired from official sources with formal writing style. All the more for the task of geoparsing, which could be implemented for navigation and travel planning applications, such dataset are rare, even in the high-resource language such as English. Being aware of the need for a new resource in both languages for this specific task, we constructed a new dataset consisting both Indonesian and English from personal travelogue articles. Our dataset consists of 88 articles, exactly half of them written in each language. We covered both named and nominal expressions of four entity types related to travel: location, facility, transportation, and line. We also conducted experiments by training a classifier to recognise named entities and their nominal expressions. The results of our experiments showed a promising future use of our dataset as we obtained F1-score above 0.9 for both languages. Language of the Presentation: English | |||
| 郷原 聖士 | M, 1回目発表 | 自然言語処理学 | 渡辺 太郎, 荒牧 英治, 上垣外 英剛 |
|
title: Is an LLM a good teacher?
abstract: To improve students' comprehension, it is necessary to provide education tailored to individual learning levels. In instructional contexts such as language learning, it is crucial for educators to grasp the vocabulary known by each student. However, providing individualized instruction to all students is challenging due to time constraints for teachers. As a solution to this problem, using machine learning models to support student question-and-answer interactions is considered. Particularly, Large Language Models (LLM) can generate appropriate responses based on prompts even when creating training data is difficult. Therefore, automation of detailed instruction using LLM is expected. However, the extent to which LLM can provide detailed instruction as a substitute for instructors in question-and-answer scenarios is unclear. Hence, this study focuses on vocabulary to investigate the ability of LLM to understand users' vocabulary, aiming to promote the utilization of LLM in the field of education. language of the presentation: *** Japanese *** 発表題目: LLMは優れた教師になれるのか？ 発表概要: 学生の理解度向上には、個人の学習レベルに適した教育が必要である。また、言語学習などの指導において、教員は各学生が知っている単語を把握しておくことが重要である。ただし、教員が全学生に対して個別指導を行うことは時間的な制約から困難である。この問題の解決策として、機械学習モデルを使用して学生の質問応答をサポートする方法が考えられる。特に大規模言語モデル(LLM)は、訓練データの作成が難しい場合でもプロンプトで適切な回答を生成することができる。そのため、LLMを活用した細かな指導の自動化が期待される。しかし、LLMが指導者の代わりに質問応答できるとして、どこまで細かい指導ができるのかは分からない。そこで本研究では、教育分野におけるLLMの活用を促進するために語彙に焦点を当てて、LLMが持つユーザの語彙を理解する能力を調査する。 | |||
| 藤川 直也 | M, 1回目発表 | ソーシャル・コンピューティング | 荒牧 英治, 渡辺 太郎, 若宮 翔子, 矢田 竣太郎 |
|
title: A Loneliness Dataset for Qualitative Analysis
abstract: Loneliness affects both mental and physical health and is associated with mental disorders such as depression. Additionally, loneliness has economic implications, such as a decrease in productivity. Therefore, it is important to identify individuals struggling with loneliness and intervene appropriately. However, there is a lack of linguistic resources for qualitative analysis of loneliness. Hence, this study aims to create a general Japanese loneliness dataset for qualitative analysis. language of the presentation: Japanese | |||
| HAN PEITAO | M, 1回目発表 | ソーシャル・コンピューティング | 荒牧 英治, 渡辺 太郎, 若宮 翔子, 矢田 竣太郎 |
|
title: Semantic Structure Augmented Language Model for Cross-domain Event Causality Identification
abstract: Event Causality Identification (ECI) requires language model to understand implicit causal relation adequately with limited resources. To improve the preformance and relieve the low resources issue, we propose a retrieval augmented generation architecture utilizing semantic parser to decompose the implicit causal relations. Sine semantic structures have strong domain transfer ability, our method can also relieve the low resources issue. language of the presentation: *** English *** | |||