ゼミナール発表

日時: 1月4日(水)3限(13:30-15:00)


会場: L1

司会: 爲井 智也
真木 勇人 1561019: D, 中間発表 知能コミュニケーション 中村 哲,池田 和司,戸田 智基,Sakriani Sakti,田中 宏季,Graham Neubig
title:Single-trial ERP Analysis Using Varying Covariance Modeling in the Time-Frequency Domain
abstract: Event-related potentials are usually analyed via synchronous averagingof multiple trials to cancel out background EEG activities which sacrifices information about each single-trial.
Hence, we developped a noise removal technique to realize single-trial ERP analysis using covariace matrix varying local Gaussian modelingin in the time frequency domain. We performed simulation study to validate our proposed method using a realistic head and source model and found our method estimate time cources of single-trial ERPs more acculately than compared convenional methods. We also found a method using spatial covariance matrices were better than independent one.
language of the presentation: Japanese
 
睫 優 1561014: D, 中間発表 自然言語処理学 松本 裕治☆,池田 和司,川人 光男,森本 淳

title: Common neural representations of anxiety from whole brain functional connectivity

abstract: Although we all know anxiety, we do not know whether our anxieties are the same each other, or at least having common elements among them.

Here, we investigated the common neural representations of anxiety among the different task conditions and participants’ populations in a unified analytical framework using functional connectivity magnetic resonance imaging (fc-MRI).

Through the anxiety provocation task, we found common anxiety related neural networks among participants. Then, using the resting state fc-MRI from the Human Connectome Project, we demonstrated that the task-independent brain also shared networks with the task-dependent anxiety. Furthermore, the overlapped networks among task-dependent and task-independent brain could detect the abnormal neural representations of patients with obsessive-compulsive disorder, which is characterized by the repeated anxiety-driven behaviours.

Our findings indicate the common neural representations of anxiety captured by whole brain functional connectivities across individuals, trait/state features and populations.

language of the presentation: Japanese

 
CHAN KAR LONG 1561022: D, 中間発表 ソフトウェア設計学 飯田 元,加藤 博一,市川 昊平,渡場 康弘,Duong Quang Thang

title: Improving the Accessibility of High-Quality Gaming using the technology of Cloud Gaming

abstract: The emerging Cloud Gaming Service provides highly accessible video gaming experience. However, compared to traditional gaming, there is still an obvious gap in term of delivered gaming quality because of the network constraints. Furthermore, the versatility of Cloud Gaming system also represents a potential platform to improve the accessibility of some novel gaming that is on trend. Therefore, in this research, from two aspects, we aim at improving the accessibility of high-quality gaming based on the technology of Cloud Gaming. For the first aspect, we propose a hybrid streaming method for improving the graphic quality delivered on cloud gaming. The proposed method utilizes both the rendering power at server side and client side to achieve the desired improvement. The performed quantitive evaluation shows that our proposed method achieve better graphic quality than traditional Cloud Gaming system as well as mitigating the server workload. As for the second aspect, since the novel VR gaming requires expensive local setup to maintain smooth gaming experience, we consider to apply cloud gaming technique to improve the accessibility. The main challenge is to maintain a fast responsiveness throughout the system, as VR gaming needs even more rapid response time than traditional gaming. For this reason, for applying the current Cloud Gaming system to VR gaming, we are planning to apply techniques such as head-motion prediction, eye tracking to efficiently mitigate the response time.

language of the presentation: English

 

会場: L2

司会: 横田 太
BROU WILLIAM-FABRICE 1551121: M, 2回目発表 ネットワークシステム学 岡田 実,安本 慶一,東野 武史,侯 亜飛,Duong Quang Thang
title: Inductive Power Transfer System Using Multiple Concatenated Parallel-Line-Feeder Segments
abstract: It has been known that the dimensions of the primary coil of an inductive power transfer (IPT) system need to be smaller than approximately one-tenth of a wavelength to avoid impact of standing wave. Recently, an IPT system of concatenated parallel-line-feeder segments has been proposed to overcome this limitation. Numerical analysis in literatures has showed that the system is capable of providing a stable output profile regardless of the length of the primary coil. In this study, the stability of output power of the system is verified via experiment systems at 13.56MHz.
language of the presentation: English
 
丸山 和輝 1551104: M, 2回目発表 ネットワークシステム学 岡田 実,佐藤 嘉伸,東野 武史,侯 亜飛,Duong Quang Thang
 
末長 和馬 1651061: M, 1回目発表 生体医用画像 佐藤 嘉伸

title: Analysis of muscle deformation before and after surgery in total hip arthroplasty

abstract:
3D surgical planning for computer-navigated total hip arthroplasty is complicated. Previously, we developed an automated preoperative planning method for THA based on statistical analysis of dataset of past patients and preoperative CT image of the current patient. However, the database currently consists of only preoperative data. It does not take account of the surgical outcome such as postoperative implant position and the patient's muscles. Planning considering postoperative data and muscle data is desirable. Therefore, in this study, we aimed an analysis of muscle deformation before and after surgery. In this presentation we report on the database we created. The database includes the shape of the muscle, the positional relationship of the bone before and after surgery, the implant planned preoperatively and used actually.
 

language of the presentation:Japanese

発表題目: 人工股関節全置換術における術前術後の筋肉変形解析

発表概要: 人工股関節全置換術における3D手術計画は容易でない.先行研究として術前のCT画像を入力とし多数の同症例の手術計画を学習データとして手術計画を自動立案する方法が開発された.しかし,現在データベースは術前データのみで構成されている.術後の実際に行われた内容や,患者の筋肉に着目していないという問題点がある.術後のデータや筋肉のデータも含めた計画立案が望ましい.そこで本研究では手術前後の筋肉の変形解析を目的とした 本発表では作成したデータベースについて報告する.筋肉の形状,術前術後の骨の位置関係,手術計画と実際に行われた手術のインプラントの位置や姿勢・サイズをデータベースとして構築した.

 

 
小原 一貴 1651031: M, 1回目発表 ネットワークシステム学 岡田 実
title:rainfall prediction using particle filter
abstract:Due to the development of phased array radar, the demand for realizing rainfall prediction system with high temporal resolution and spatial resolution is increasing. Therefore, assuming data transmission with an inexpensive internet line, I propose a prediction system with a small amount of data using a particle filter. In this presentation, I outline the proposed system and evaluate the change in prediction accuracy due to the change in particle number.
language of the presentation:Japanese
発表題目:パーティクルフィルタを用いた降雨予測システムの検討
発表概要:フェーズドアレイレーダが開発されることにより、高い時間分解能、空間分解能の降雨予測システム実現への要求が高まっている。そこで、安価なインターネット回線でのデータ伝送を想定し、パーティクルフィルタを用いた少ないデータ量での予測システムを検討する。本発表では提案システムの概要を説明し、粒子数の変化による予測精度の変化を評価する。
 
木戸 勇太 1651043: M, 1回目発表 ユビキタスコンピューティングシステム 安本 慶一
title:Proposal of smart cruet for support arranging seasonings quantity based on user's preference
abstract:Recently the number of people who watch a recipe site when cooking isincreasing. However, it takes a lot of labor to measure seasonings quantities according to the recipe. Alternatively, if a person having a little cooking experience uses seasonings by eye measure, this leads to the cooking failure. on the other hand , people that adjusting seasonings quantity based on preference exist. In the present study, I propose a system that adjusts the seasonings quantities in a recipe to user's preference.
language of the presentation:Japanese
タイトル:ユーザの嗜好を考慮した調味料分量調整支援に向けたスマート調味料入れの提案
概要:近年レシピサイトを見ながら料理する人が増加している. しかし,レシピ通りに調味料を毎回計量して料理することは手間や時間が掛かること、料理経験が浅い人が目分量で投入することは失敗に繋がる。 またレシピサイト通り料理しても自分の嗜好に合う様に少しアレンジする人は存在する.そこでレシピサイトのユーザの嗜好に合わせて調味料の分量をアレンジ(調整)し目分量が適量か否か教示を行うシステムを提案する.
 

会場: L3

司会: 侯 亜飛
白倉 尚貴 1651060: M, 1回目発表 ロボティクス 小笠原 司
title: sliding motion using a multi-copter with a lightweight manipulator
abstract: Multi-copters are expected to be active in rescue and inspection in the high place. Though aerial manipulation using multi-copters has being done, there are still some problems with stabilization control. Cleaning windows is one of the examples of aerial manipulation. Cleaning the windows needs to be done in high place, and conventional approaches which done by a specific machine and humans are expensive and dangerous. So, this study provides a system that cleans windows using multi-copters with a manipulator. Specifically, a sliding motion for cleaning windows is performed by one DoF manipulator mounted on a multi-copter. The multi-copter keeps the contact force constant by controlling the manipulator, and controls its orientation in order to keep the orientation of the manipulator. By making them each independently, this system becomes simple and could achieve the stable sliding motion.
language of the presentation: Japanese
発表題目: 軽量マニピュレータを搭載したマルチコプターによる倣い動作
発表概要: 災害救助や高所での点検などの分野において,マルチコプターの活躍が期待されている. マルチコプターによるマニピュレーションの研究も行われているが,マルチコプター自体の姿勢制御が困難なことが問題としてあげられる. マルチコプターによるマニピュレーションの例としては、窓の掃除がある. 窓の掃除は,高所での作業であり,従来の手法では,コストの大きさや作業者の転落の危険などが問題とされてきた. そこで,本研究ではマニピュレータを搭載したマルチコプターを利用して窓掃除を行うシステムを提案する. 具体的な方法としては,マルチコプターに1自由度のマニピュレータを搭載して窓掃除のための倣い動作を行う. 搭載した直動のマニピュレータを稼働させて窓との押し付け力を一定に保ち,これによって動いたマニピュレータの姿勢を保つためにマルチコプターの姿勢を制御する. 押し付け力と姿勢制御を独立に設計することで各々の制御が簡単になり,安定な倣い動作を生成することを狙う.
 
津田 香林 1651071: M, 1回目発表 光メディアインタフェース 向川 康博
title: Considering influence of specular reflection on light field segmentation
abstract: Light field segmentation is important in a field of computer vision.Previous work uses chroma components in color information.Although this work gives results with high accuracy, this is poor at segmentation scene having specular reflection.We propose using luminance added chroma components, and aim at to be robust for specular reflection.
language of the presentation:Japanese
 
河瀬 良亮 1651036: M, 1回目発表 大規模システム管理 笠原 正治
title: Analytical Model for Transaction-Confirmation Time of Bitcoin
abstract: When we transmit or receive bitcoin with someone, remittance information is included in a transaction. Transactions are grouped into a block, and the block is confirmed by the confirmation process called Mining. However, the block size is limited, and this causes a large transaction-confirmation time when the transaction-arrival rate increases. We analyze the transaction-confirmation time using a single-server queue with bulk service. Numerical examples show how the bulk size affects the transaction-confirmation time.
language of the presentation: Japanese
 
青山 泰枝 1651001: M, 1回目発表 計算システムズ生物学 金谷 重彦
title:Introduction of information scientific approach for allergic disease
abstract:Allergic diseases such as hay fever are one of the most popular diseases for Japanese. Allergic diseases occur when antibodies bind to allergic proteins. Originally, antibodies are for expelling parasites, viruses, and bacterial proteins that are enemies for the human body promptly. Whether this antibody attack harmless allergens has not been elucidated at this time. In this research, we are collecting the protein sequence of many known allergens parasites, viruses, and bacteria. I’m going to analyze these sequences to found important features for allergic disease.
language of the presentation:Japanese
タイトル:アレルギーに対する情報科学的アプローチの紹介
概要:花粉症を始めとしたアレルギー疾患は日本人にとって最もポピュラーな病気の一つである.アレルギーというのは抗体がアレルゲンであるタンパク質に結合することで起こる.本来,抗体というのは人体にとって敵となる寄生虫やウイルス,細菌のタンパク質を記憶し,速やかに駆逐するためのものである.この抗体がなぜ無害なアレルゲンを記憶・攻撃するようになるのかは現時点ではよく解明されていない.本研究では,現在多く公開されているアレルゲンのタンパク質情報と,抗体の本来の敵である寄生虫やウイルス,細菌のタンパク質情報を網羅的に解析し,抗体に認識されるようになるための特徴量を抽出することを目的とする.
 
安藤 瑞希 1651004: 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
title: A Study of an Augmented Reality for Supporting Rope Work
abstract: Recently, there are a lot of systems that support beginner's works using a computer. Since Augmented Reality(AR) can present intuitive information to real objects, it is expected to be applied to the task support system. However, the usefulness of augmented reality in complicated tasks has not been fully verified yet. In this study, we attempt to investigate how to support more complicated tasks such as rope work using AR. In this presentation, we introduce related studies that are typical papers in task support research.
language of the presentation: Japanese
発表題目: ロープワーク支援のための拡張現実感の検討
発表概要: 近年,初心者の作業をコンピュータを用いて支援するシステムが多く存在する.中でも拡張現実(AR)は作業をより効率的に支援するために生まれた技術であり様々な分野で利用されている.しかし,複雑なタスクにおけるARの有用性はまだ十分検証されていない.本研究ではロープワークのような複雑なタスクをARを用いて支援する方法を検討する.今回の発表では作業支援における代表的な論文を紹介する.
 
岩井 雄大 1651018: M, 1回目発表 知能システム制御 杉本 謙二
title: Design of prediction governors for nonlinear systems
abstract: In the conventional tracking control of dynamic system, it is assumed that the target value is known. However, an accurate target value may not be available in some cases. In this case, although it is necessary to predict the target value, but when it includes uncertainty, the control performance is deteriorated. In the previous research, a signal shaping mechanism called prediction governors has been proposed, and it is possible to approximate the ideal output by shaping a poorly predicted value. However, the prediction governors cannot be applied to nonlinear systems. Therefore, in this research, we propose nonlinear prediction governors that nonlinearly extends the class of prediction governors. In this presentation, we explain the design problem and derived solution of the nonlinear prediction governors. Moreover, evaluate performance by simulation.
language of the presentation: Japanese
発表題目: 非線形システムに対する予測ガバナの設計
発表概要: 従来の動的システムのトラッキング制御において,目標値が既知であることを前提としているが,正確な目標値が利用できない場合がある.このとき,目標値を予測する必要があるが不確かさを含む場合,制御性能が低下する.先行研究では,予測ガバナと呼ぶ信号整形機構が提案されており,精度の悪い予測値を整形して理想の出力に近づけることができる.しかし,予測ガバナでは非線形システムに適用することが出来ない.そこで,本研究では予測ガバナのクラスを非線形に拡張した非線形予測ガバナを提案する.本発表では,非線形予測ガバナの設計問題および導出解について説明する.また,シミュレーションを通して性能の評価を行う.