| 金川 宗一郎 | 1451034: M, 1回目発表 | コンピューティング・アーキテクチャ |
|
title: [Paper Introduction]A Program Differencing Algorithm for Verilog HDL
abstract:During code review tasks, comparing two versions of a hardware design description using existing program differencing tools such as diff is inherently limited because existing program differencing tools implicitly assume sequential execution semantics, while hardware description languages are designed to model concurrent computation. This paper presents Vdiff, an instantiation of this position-independent differencing algorithm for Verilog HDL. I present Vdiff's algorithm and evaluation. language of the presentation:Japanese | ||
| 駒井 清顕 | 1451046: M, 1回目発表 | ユビキタスコンピューティングシステム |
|
title: Indoor positioning system with low cost sensors in a smart home.
abstract: Many studies for sensing environmental information, power consumptions and position information in the home to understand the living activity and provide services such as elderly monitoring have been conducted so far. Due to the cost of deploying sensing devices, however, these services have not yet penetrated in common household. In this study, we propose a low-cost indoor localization system for homes by using pre-installed sensors in household and sensors installed on smartphone equipped by inhabitants together. language of the presentation: Japanese 発表題目: スマートホームにおける低コストセンサを用いた屋内位置推定システムの提案 発表概要: 家庭内において環境情報や消費電力, 位置情報などをセンシングすることで住人の生活行動を推定し，見守りなどのサービスへ応用する研究が盛んに行われている．しかし, センシングデバイスの導入コスト等の問題から一般家庭へはほとんど普及が進んでいない．本研究では，家庭内に予め設置されたセンサと住人が所持するスマートフォンが搭載するセンサを組合せて用いることで家庭内での位置推定を低コストに実現するシステムを提案する. | ||
| 藤澤 和輝 | 1451089: M, 1回目発表 | ユビキタスコンピューティングシステム |
|
title: A system for supporting real-time curation of multiple real-time video streams
abstract: In this presentation, we present a system for supporting the curator (creator of contents) to create a content from multiple real-time video streams and SNS streams in real time. The system provides a function to score each of streams based on the value of the curator and show only high-score streams to the curator as well as a function to support the curator to efficiently switch the streams for output and deliver as contents. language of the presentation: Japanese 発表題目: 複数映像ストリームの実時間キュレーション支援システム 発表概要: 本発表では、キュレータと呼ばれるコンテンツの編纂者が、複数の映像ストリームやSNSストリームをリアルタイムに編纂する作業を支援するシステムを提案する。 提案システムは、複数の映像情報流や SNS情報流をキュレータの価値観に合わせてスコア付けし、上位のもの数件のみを選択表示する機能と、コンテンツとして出力・配信する情報流(一つないし少数)を次々に切替えていく作業を支援する機能を提供する。 | ||
| 伊藤 大史 | 1451010: M, 1回目発表 | インターネット工学 |
|
title: Paper Introduction: A Security Enforcement Kernel for OpenFlow Networks
abstract: OpenFlow is able to dynamically configure and manage networks on demand by defining flow rules on switches from a controller. However, there is a possibility to cause contradictions among flow rules on the switches. The states lead to misconfigurations and unexpected packet forwardings on the networks. In this paper, the authors proposed a software extension that provides role-based authorization and security constraint enforcement for the NOX OpenFlow controller. The extension provides detection and reconciliation of contradictions among each of the flow rules. In the evaluation, they measure computational overhead of FortNOX to solve the flow conflicts. I will give the details of this paper and my research plan. language of the presentation: Japanese | ||
| 岡本 拓也 | 1451029: M, 1回目発表 | ネットワークシステム学 |
|
title: Wireless
power transfer control method for parallel feeder-based wireless power transfer
system with plural of terminals abstract: Recently,
battery chargers using the wireless power transfer has appeared in the market,
and there are growing demands for wireless power transfer technologies. In this
study, I investigate a control scheme for the parallel feeder-based wireless
transfer system with two or more power reception nodes. Although the parallel
feeder-based scheme is capable of expanding the power transfer area, it
requires a control mechanism to supply power to two or more reception nodes
simultaneously. In this presentation, I propose a control scheme using
load-modulation for establishing the communication links among the transmitter
and two or more terminals. language of the presentation:Japanese | ||
| 伊原木 駿 | 1451012: M, 1回目発表 | ネットワークシステム学 |
title: Power distribution characteristics of the parallel feeder-based wireless power transfer systems with two or more mobile nodes abstract: I have investigated the power distribution characteristics of the parallel feeder-based wireless power transfer (WPT) system with two or more mobile nodes. The parallel feeder-based WPT system with single mobile node has already been proposed and the power transfer characteristics of the proposed scheme have been shown by the computer simulation and empirical experiment. When two or more mobile nodes exists over one parallel feeder, the existence of one mobile node could affect the other nodes. In this presentation, I have first proposed an equivalent circuit model corresponding to two mobile nodes. Then, I have investigated the power transfer characteristics for each mobile node.
| ||
| 遠藤 保 | 1451021: M, 1回目発表 | 環境知能学 |
|
title: Automatic detection of dementia by facial expression recognition
abstract: In this presentation,I talk about automatic detection of dementia by facial expression recognition with background, idea, purpose, method, and future work of this study. Recently, the population is aging rapidly, and the number of dementia patients has been increasing. Besides, dementia is progressive and irreversible disease. So, early detection is important. But, brcause of early detection of dementia is difficult, widely, easily, and quantitative screening tool are required. Then, I will develop a screening tool that is automatic detecting by facial expression recognition using talk videos. language of the presentation: Japanese 発表題目: 表情認識による認知症の自動検出 発表概要: 本発表では、表情認識による認知症の自動検出について、その研究の背景、着想に至った経緯、目的と手法、今後の予定について過去の関連研究を交えながら発表する。 近年、高齢化が急速に進展しており、認知症有病者数が増え続けている。 また、認知症は進行性で非可逆の疾患であり、早期発見が重要である。 しかし認知症の早期発見は容易ではなく、広域に、簡便に、定量的にスクリーニングできるツールが必要とされている。 そこで私は、高齢者の対話動画像から表情認識により認知症の自動検出を行うツールを開発することを目的とし、研究を行う。 | ||
| 東口 剛士 | 1451086: M, 1回目発表 | 環境知能学 |
|
title: Estimation of physical condition by the human body posture
abstract: In my presentation,I talk about an evaluation method of physical condition with background and previous research. Currently, many quantitative evaluation methods exist. But they are difficult to really be used in medical field, because they have many temporary results. Therefore I focus on posture of gait in the everyday life and evaluate the physical condition by the posture information for the purpose of the practical use. language of the presentation: Japanese 発表題目: 姿勢情報を用いた身体状態の評価 発表概要: 本発表では、姿勢情報を用いた身体状態の評価についての背景、既存研究をふまえた身体状態の評価手法を述べる。現在、身体状態を定量的に評価する方法が多く存在する。しかし、それらは暫定的な知見が多く、実際に医療現場に用いることは困難である。そこで、日常生活における歩容の姿勢情報に着目し、実用化を目的とした身体状態の評価を行う。 | ||
| 酒井 一樹 | 1451050: M, 1回目発表 | インタラクティブメディア設計学 |
|
title: Implementing dense camera tracking for DTAM
abstract: There are many dense reconstruction methods available. Recently, Dense Tracking and Mapping (DTAM), achieved compelling results. This method has great tracking performance, the only input device we need is monocular camera. We can obtain a dense 3D model and perform dense camera tracking simultaneously in real-time. I attempted implementing the dense camera tracking method from DTAM to understand the latest 3D reconstruction and camera tracking techniques and their problems. Finally I will show my implementation schedule. language of the presentation: Japanese 発表題目: DTAMのためのDense camera Trackingの実装に関する研究 発表概要: カメラ画像から三次元復元する技術の実用化が進む中、密な復元を行う手法の一つとしてDTAM（Dense Tracking and Mapping in Real-Time）がある。これは優れたトラッキング性能を持ち、単眼カメラのみでリアルタイムに密な三次元形状復元とカメラトラッキングを行う手法である。DTAMの手法を読み解き、三次元形状復元技術やカメラトラッキング技術に残る問題点を探る。また，本発表では今後の研究の方向性について述べる。 | ||
| 加藤 明彦 | 1451033: M, 1回目発表 | 自然言語処理学 |
|
title: Semantic Parsing for Time Expression
abstract: In order to understand the meaning of sentence, semantic analysis of time expression is important. But many of previous researches focus on specific application - Temporal Resolution (identify specific time range or point of time which corresponds to given time expression) -. Based on the above background, I will explain semantic analysis of time expression. The purpose is inference of meaning representation (logical form , Abstract Meaning Representation etc.) which can be applied to various applications. language of the presentation: Japanese 発表題目: 時間表現の意味解析 発表概要: 時間表現の意味解析は文の意味を理解する上で重要であるが、先行研究では時間表現の正規化(対応する時区間や時点の同定)という特定の応用に焦点を当てた研究例が多い。そこで本発表では各種の応用に適用可能な意味表現(論理式やAbstract Meaning Representation 等)の推定を目的とした意味解析について説明する。 | ||
| 澤井 裕一郎 | 1451054: M, 1回目発表 | 自然言語処理学 |
|
title: Document Classification using Support Measure Machines
abstract: Support Measure Machines are machine learning models that can classify distributions. They have been shown to be effective for image classification, but they cannot be applied to document classification naively by using sparse bag-of-words features. Instead, distributed representations of words can be effectively employed for the use of support measure machines in document classification tasks. In this research, I thoroughly investigate the combinations of parameters and types of word representations that lead to high classification performance of documents. language of the presentation: Japanese | ||
| 出島 遼 | 1451077: M, 1回目発表 | 計算システムズ生物学 |
|
title: The extraction of brain active information from the implantable brain imaging data
abstract: Blood flow in discrete active regions of the brain has been found to be increased locally, however, these are many unclear things. The study of functional localization of brain can be mearused by using fMEI and PET, but cannot measure a cerebral blood flow at the microvascular level. The measurement at the microvascular level of the brain was carried out by a newly developed device. The device is composed of a CMOS image sensor chip and nine LEDs for illumination. We obtained the speed of red blood cell, by using the absorbance characteristics of hemoglobin and observed changes in the neural activity. I introduce the related study analyzing microvascular brain blood flow spped and also our research. language of the presentation: Japanese | ||
| 岩本 昌大 | 1451016: M, 1回目発表 | ロボティクス |
|
title: Solving the uncanny valley problem by slightly changing appearance of an android robot.
abstract: In this study, we aim to realize friendly interaction with an android robot by slightly changing the robot’s appearance. To achieve this, we need to solve the uncanny valley problem, which expresses the phenomenon that humans feel uncanny when the human-like robot presents the aspect slightly different from those of humans. In this presentation, we clarify the reason of the uncanny valley problem and consider the method for solving the problem by attaching accessories to easily understand if an android robot is “robot.” language of the presentation: Japanese 発表題目:アンドロイドの外観の変更による「不気味の谷」問題の解消 発表概要: 本研究では、アンドロイドの外観に手を加えることによって、 より親しみやすい対話ができるアンドロイドの実現を目指す。 そのためには、人間に近い外観を持つロボットに微妙な非人間的な要素を 観察することによって人は非常に不気味に感じてしまうという 「不気味の谷」問題を解消する必要がある。本発表では不気味の谷の原因を具体化し、 アンドロイドを直感的にロボットだと認識できる外観にすることで 不気味の谷問題を解消する方法について考察する。 | ||
| 川名 雄樹 | 1451039: M, 1回目発表 | ロボティクス |
|
title: Improving Appearance Learning for Human Pose Estimation From Still Image
abstract: Estimation of human pose has wide range of application, such as surveillance and image search. In this research I consider the problem of estimating human pose from still image. Especially I focus on improving appearance modeling to mitigate incorrect pose estimation by addressing existing problems on learning appearance of body parts to avoid confusing detection of back ground image as valid body parts. language of the presentation: *** Japanese *** | ||
| 秋澤 翔 | 1451002: M, 1回目発表 | 数理情報学 |
|
title: Change-point detection for deterioration of a neurological disorder based on network dynamics
abstract: Recent study shows that the deterioration during the progression of many diseases is not necessarily smooth but abrupt. If we can predict such deterioration, the patients would have a better chance of a healthier life with appropriate treatment. In fact, there is a study that developed a method to detect early-warning signal indicating a sudden deterioration, based on a dynamical network biomarker. In this presentation, I will consider the effective measuring and analyzing method to predict the deterioration, focusing on subarachnoid hemorrhage which account for 4~6% of the cause of sudden death. language of the presentation: Japanese 発表題目: ネットワークダイナミクスに基づく神経疾患病態変化の検出 発表概要: 様々な疾病の進行は必ずしもスムーズなものではなく、健康状態から疾病状態への突然の遷移が起こりうることが近年の研究によって示されている。そのような突然の遷移を発生前に検出することができれば、より適切な投薬・治療が可能になり予後の改善に繋がる。実際にいくつかの疾病については、生体のネットワークダイナミクスに基づいた病態悪化検出のためのバイオマーカーが提案されている。本発表では、突然死の４〜６％を占めると言われる、くも膜下出血に焦点を当て、病態の悪化を検出するための計測・処理方法について提案する。 | ||
| 田村 真一 | 1451074: M, 1回目発表 | 数理情報学 |
|
title: Multigraph-based NNS methods for varying similarity criteria
abstract: In this presentation, I will suggest the flexible Nearest Neighbor Search (NNS) method based on multigraphs which is adjustable for varying similarity criteria. The NNS is essential on data search or recommendation and many methods have been suggested to improve the speed especially for large data. However, these methods have a fixed similarity criterion, which cannot be changed on-the-fly. To solve the problem, I will develop the method which uses multigraphs as indices. language of the presentation: Japanese 発表題目: 多重グラフを用いた複数類似度基準に対する最近傍探索手法 発表概要: 本発表では、類似度基準を調整できる、多重グラフを用いた最近傍探索手法を提案する。 最近傍探索はデータ検索やレコメンデーションなどにおいて重要であり、特に大規模な場合に探索を高速化する手法が研究されてきた。 しかしこれらの多くは類似度基準を固定した前処理が肝要であり、探索の都度その基準を調整することはできない。 本研究では、多重グラフを索引に利用することでこれを解決する手法を開発してゆく。 | ||