ゼミナール発表

日時: 11月16日(月)3限(13:30-15:00)


会場: L1

司会: Tran Thi Hong
ADLIZAN BIN IBRAHIM 1651128: M, 1回目発表 インターネット工学 小笠原 司☆

title: New Security Architecture for IoT Network

abstract: We explain the notion of security architecture for Internet of Things (IoT) based on software-defined networking (SDN). In this context, the SDN-based architecture works with or without infrastructure, that we call SDN-Domain. This work describes the operation of the proposed architecture and summarizes the opportunity to achieve network security in a more efficient and flexible with SDN. An overview of existing SDN security applications were discussed and tackles its issues, presenting a new IoT system’s architecture. In this paper we considered the network access control and global traffic monitoring for ad-hoc networks. Finally, we point out architectural design choices for SDN using OpenFlow and discuss their performance implications.

language of the presentation: English

 
森 諒介 1651112: 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: [Paper Introduction] Haptic Dissection of Deformable Objects using Extended Finite Element Method
abstract: I will introduction a research paper titled “Haptic Dissection of Deformable Objects using Extended Finite Element Method”. The haptic device has a function that feedbacks the state of touching the object to the user. Recently, in the medical field, the device is used in surgery to cut an object by handling devices like their own hands. On the other hand, the surgery system requires not only these haptic devices but also operation environment to construct this system. Therefore this system is expensive, and this system is available in the limited institute. In this study, author constructed real-time cutting simulation system by using the general haptic device and to cutting down computational cost. This system generates the deformable object by using XFEM framework to simulate the cutting operation. In this presentation, I’ll introduce the technology that is used to build this system.
language of the presentation: Japanese
発表題目: ハプティックデバイスを用いた拡張有限要素法による変形可能オブジェクトの切断シミュレーション
発表概要: 触覚デバイスは物体に触れた状態を使用者に疑似的にフィードバックする機構を持つ.現在,医療分野ではデバイスを自分の手のように扱い物体の切断を行う手術等で利用されている.ただし,このような触覚デバイスだけでなく動作環境もデバイス用に構築されているためシステム全体が高価となり,限られた機関でしか使用できない. 今回紹介する論文では,これらの問題を回避するため安価な触覚デバイスを用い,計算コストを下げることで,リアルタイムの切断シミュレーションができる環境の構築を行った. 切断シミュレーションでは,拡張有限要素法(XFEM)のフレームワークを用いて変形可能なオブジェクトを生成する. 本発表では,このリアルタイムで切断シミュレーションを行う環境を構築するために利用される技術について紹介する.
 
松井 琢朗 1651098: 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: Fence Removal Based on Curvelet Transform for Diminished Reality
abstract: Diminished reality is a technique for visually removing real objects from captured images. Previously, pre-captured background based methods and image inpainting based methods have been proposed to generate diminished images. In contrast to the previous methods, we propose a method for achieving diminished reality using curvelet transform. Especially, we focus on fences in input images. Fences exist in many situations such as zoo and baseball field. In this presentation, we will show the initial results of curvelet based fence removal.
language of the presentation: Japanese
発表題目: Curvelet変換を用いた網状物体の除去による隠消現実感
発表概要: 隠消現実感は取得した画像から不要な物体を視覚的に除去する技術である.これまでに,除去後の背景を事前に計測しておく手法や画像修復に基づく手法が提案されている.本研究では,フェンスのような網状物体の除去を目的とし,これらの物体を自動で検出,除去することが可能な手法を提案する.提案手法では,周波数変換の一種であるCurvelet変換を用いて,物体の検出及び除去を行う.本発表では,Curvelet変換を用いた網状物体の除去手法とその結果を示す.
 
森田 達弥 1651114: 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: Implementation and Evaluation of Day-Care Report Generation System

abstract: In this research, I propose a semi-automatic day-care report generation system which can monitor movements/activity of senior citizens in daycare centers. The proposed system estimates multiple areas where senior citizens are located with the BLE beacon, by utilizing RSSI of the Bluetooth radio wave. Also, the accelerometer implemented in the tag estimates the acrivity of the elderly. The information of the estimated area and activity is stored in a server with time stamp. The server generates the daycare report based on it. In order to evaluate the proposed system, I have deployed my system in a day-care center: Ikoi-no-ie26. Evaluation result in Ikoi-no-ie26 showed that our system estimated the subject's present area with F-measure: 80.6% and activity with F-measure: 73.8% and generated the day-care report.

language of the presentation: Japanese

 
URIGUEN ELJURI PEDRO MIGUEL 1651132: M, 1回目発表 ロボティクス 小笠原 司
title:Re-arranging tasks in a unstructured environment with a humanoid robot
abstract: The use of robots in our daily life is becoming something more common, but there are still problems when the robot does not have the information about the environment and needs to execute a task such as pick and place and object. In this study we propose the use of a humanoid robot in an unstructured environment, where the robot will find the objects and determine the final positions based on the objects properties and categories.
language of the presentation: English
発表題目: *** この部分を発表題目に ***
発表概要: *** この部分を発表概要に ***
 
宮田 明裕 1651104: M, 1回目発表 光メディアインタフェース 向川康博
title: BRDF measurement using X-Slit
abstract: Objects' appearances can be reconstructed to observe the reflectance from enormous pairs of incident and outgoing direction (BRDF : Bi-directional Reflectance Distribution Function). While there are many previous works to find pairs, it is necessary to prepare expensive equipments. So we propose a novel isotropic dense BRDF sampling method using Crossed-slit(called X-Slit), which consists of only two slits. We simulate finding pairs and evaluate the proposed method.
language of the presentation: Japanese
発表題目: X-Slitを用いたBRDF計測
発表概要: 物体の見え方は, 光源方向と観測方向の変化に対する反射特性BRDF(Bi-directional Reflectance Distribution Function)を測定することで再現することが可能である. これに関し多くの手法がこれまで提案されてきたが, 測定に必要とされる装置が高価であるという欠点がある. 本発表では効率的なBRDF計測手法として, X-Slitと呼ばれる二つだけのスリットを使用して密にBRDFを計測できる手法を提案し, シミュレーションで評価する.
 

会場: L2

司会: 大和 勇太
吉田 拓弥 1651124: M, 1回目発表 大規模システム管理 笠原 正治
title: Analysis of large scale graph data on corporate social activities
abstract: Demand for analyzing large scale graphs such as transportation networks, social networks, and communication networks and so on is increasing. In this presentation, we present the results of analyzing data representing corporate social activities. This data has a graph structure in which nodes correspond to companies, or individuals, and edges correspond to relations between nodes. In our analysis, we measure the graph size and the number of connected components of the graph. Also, we calculate the betweenness centrality of nodes, which is an indicator of a node's centrality in a network, to discover nodes corresponding to key people. Furthermore we discuss difficulties of large scale graph analysis.
language of the presentation: Japanese
発表題目: 企業活動に関する大規模グラフデータの解析
発表概要: 交通ネットワークやソーシャルネットワーク,通信ネットワークなど,大規模グラフの解析に対する需要が高まっている. 本発表では,企業の社会的活動を表したデータを解析した結果について発表する. このデータは,企業・個人などをノードとし,ノード間の関係を属性付きエッジとするグラフ構造になっている. 解析として,このグラフの連結成分の大きさと個数を測る. またキーパーソンとなるノードを発見するために,それぞれのノードに対して媒介中心性と呼ばれる値の計算を行う. 媒介中心性とは,あるノードに対して,そのノードを通る経路が多いほど,ネットワークにおける重要度が高いとする指標である. 併せて,大規模グラフ解析の難しさについても述べる.
 
米澤 拓也 1651127: M, 1回目発表 情報基盤システム学 藤川 和利
title: Analysis of sensor data obtained from bus and Study of estimation state of vehicle
abstract: In passenger transport business,it is important to know the state of the vehicle from the point of view for service management and safety management.Currently,when service administrator grasp service state,drivers manually tell the state in real time and manually record the daily report.But,these operations have become a big burden for drivers and service administrators.To resolve this problem and achieve operational efficiency, we are trying to analyze sensor data obtained from bus for estimating state of bus.In this presentation,I will describe on the sensor data analysis progress for the purpose of bus state estimation.
language of the presentation: Japanese
 
YANG FAN 1651133: M, 1回目発表 数理情報学 池田 和司
title:Insect-plant Predation Data Analysis
abstract: Compared to relational data from other fields, insect-plant predation data is more difficult to be obtained, as biologists need to spend tremendous time and energy on observation. It poses the problem: how to utilize a very limited number of data to do analysis. In this study, we are going to cluster insect-plant predation and analyze its connection with bio-taxonomy by Infinite Relational Model, which models the relational data by considering groups instead of individuals, so that common properties are enriched by grouping similar individuals together.
language of the presentation: English

 
SINGH MADHURYA 1651131: M, 1回目発表 生体医用画像 佐藤 嘉伸

title: Prostate cancer area precise estimation by integration of MR images and intraoperative biopsy information

abstract: A precise localization of a cancer from a medical image is a major challenge for a minimally invasive therapy which aims to preserve healthy tissues as much as possible and to remove tumors completely. In an area of urology, magnetic resonance imaging (MRI) is used for image diagnosis of a prostate cancer because MRI can give a very clear picture of the prostate. There are some studies for an automated segmentation of the prostate cancer from a multi-parametric MRI with a machine learning technic. However an intensity in MRI is not standardized likes CT images, there is fear that a learned classifier from a dataset of one facility can not suitably work for an image from other facilities. Therefore, we propose an automated segmentation method which is robust to difference of an imaging condition by combining MP-MRI data and a biopsy. The biopsy which is a procedure taking tissue from a patient by needle is definite diagnosis but very sparse than MRI. We believe that the combination of the biopsy and MRI leads more correct segmentation. In this presentation, I talk about current progress and remaining tasks.

language of the presentation: Japanese

 
柳田 智也 1651115: M, 1回目発表 知能コミュニケーション 中村 哲
title: Incremental text to speech system for simultaneous speech translation system
abstract: Speech translation system consists of three components: automatic speech recognition (ASR), machine translation (MT), and text to speech synthesis (TTS). In traditional manner, ASR starts after the speaker has spoken the whole sentence, then perform translation and synthesis sentence-by-sentence. Standard TTS requires linguistic information of the full sentence. As spoken speech such lectures can be very long, this method can cause a significant delay. Simultaneous speech translation therefore attempts to translate speech in real time before the speaker has spoken the whole sentence. To deal with this task, several studies propose to construct incremental TTS (ITTS), in which the system can synthesize speech without having linguistic information of full sentence. However, the performance is still much lower than standard TTS. Furthermore, there is not yet exist ITTS in Japanese. The objective of this research is to improve the current ITTS technology, and I will focus on developing Japanese ITTS. In this talk, I will introduce the existing studies, my research plan, preliminary experiments and future works.
language of the presentation: Japanese
発表題目: 同時通訳システムのためのインクリメンタルテキスト音声合成システム
発表概要: 音声翻訳システムは、自動音声認識(ASR)、音声翻訳(MT)、テキスト音声合成(TTS)の3要素から構成される。従来の手法において、自動音声認識は、話者が全ての発話を話す前に認識し始め、それから、翻訳と音声合成が文ごとに実行される。通常のTTSは、文全体の言語情報を必要とし、講義のような口頭発話が長い文章の場合、この方法では遅れを発生させる。同時通訳システムは、それゆえ、話者が全文を話す前にリアルタイムに翻訳を行うことに挑戦するものである。この課題を取り扱うため、複数の研究者がIncremental TTS(ITTS)の実装を提案している。ITTSは文全体の言語情報が不明な場合に音声を合成できるシステムである。しかしながら、ITTSの性能はTTSより低い。更に、日本のITTSはまだ実装されていない。そこで、本研究の目的は、現在のITTSの性能を改善することであり、現在は、日本語ITTSの実装に注力している。今回、研究計画及びITTSの関連研究、事前実験と研究の進捗状況について発表する。
 
長村 佳歩 1651027: M, 1回目発表 知能コミュニケーション 中村 哲
title: Joint Optimization of Speech Recognition and Machine Translation
abstract: Speech translation is composed of three modules: automatic speech recognition (ASR), machine translation (MT), and text-to-speech synthesis (TTS). In most cases, ASR is trained and optimized to minimize word error rate. But, the best speech recognition result does not always provide an optimum solution for MT. Existing studies proposed joint optimization of hidden Markov model (HMM) based ASR and phrase-based MT, and showed to improve translation results. In this research, we propose to extend the existing ASR-MT joint optimization with neural network framework, in order to further improves the translation accuracy. In this talk, I will present existing approaches, research outline, current progress and future work.
language of the presentation: Japanes
発表題目: 音声認識と機械翻訳の同時最適化
発表概要: 音声翻訳システムは音声認識と機械翻訳,テキスト音声合成の3つのモジュールによって構成されている.たいていの場合,音声認識部は単語誤り率が最小になるように最適化されている.しかし,そうして認識された結果が機械翻訳部の理想とする入力とは限らない.そのため,先行研究では隠れマルコフモデルベース音声認識器とフレーズベース機械翻訳器の同時最適化を提案し,翻訳精度の向上を確認した.本研究では,翻訳精度を向上させるために,既存の音声認識器と機械翻訳器の同時最適化をニューラルネットワークによるフレームワークに拡張することを提案する.また,本発表では関連研究及び現在の進捗状況,今後の研究計画を報告する.
 

会場: L3

司会: Juntao Gao
中才 恵太朗 1651080: M, 1回目発表 ソフトウェア工学 松本 健一
title: Analysis of Donations in OSS: A Case Study of Eclipse Project
abstract: Although development activities, such as submitting patches and working with bug reports, are common contributions in open source software (OSS) projects, making donations is also an important contribution. Some OSS development projects are actively collect donations by preparing some benefits for donors to promote donation. In this research, we study Eclipse project to analyze donations. We analyzed donor lists and release dates, then found the followings; (1) benefits can be motivations for donors, (2) although the number of developers is small in all donors, they donated more than others, and (3) new releases are triggers of donations, but bugs can affect the amount of donations.
language of the presentation: Japanese
発表題目: OSSプロジェクトEclipseにおける寄付の分析
発表概要: オープンソースソフトウェア(OSS)への一般的な貢献として開発活動(パッチ投稿,バグ報告など)があるが,運営団体への寄付も重要な貢献である.OSS開発プロジェクトによっては,寄付を促進させるために特典を用意するなど,寄付収集に積極的である.本研究は,効果的な寄付の収集方法を明らかにするため,著名なOSSプロジェクトであるEclipseを対象として調査を行った.寄付者のリストとリリース状況を分析した結果,(1)特典が寄付への動機づけとなっていること,(2)全体的な寄付者のうち開発者の割合は少ないが開発者の寄付額は開発者でないものより大きかったこと(3)リリース日には寄付が増えるがバグ数が多いと寄付が落ち込むこと,がわかった.
 
ZUO YONG 1651134: M, 1回目発表 自然言語処理学 松本 裕治
title: Semantic Role Labeling with Predicate Similarity Analysis
abstract: Semantic role analysis has the potential in wide range applications such as information extraction, question answering, machine transltion and summarization, and recurrent neural network(RNN) model is found useful for this task. However, most of related work apply the same network for all predicates. We propose a method training each predicates with separate network, add feeding sample with near predicates.
language of the presentation: Japanese
 
前田 雄大 1651096: M, 1回目発表 計算システムズ生物学 金谷 重彦
title: Creation of lung adenocalcinoma clustering model using TCGA data
abstract: Adenocarcinoma of the lung is a leading cause of cancer death worldwide. TCGA(The Cancer Genome Atlas) provides molecular profiling data of 230 patients of lung adenocarcinoma regarding messenger RNA, microRNA and DNA sequencing integrated with copy number, methylation and proteomic analyses. 230 patients belong to 3 types of lung cancer as follows: TRU(Terminal Respiratory Unit), PI(Proximal ミ Inflammatory) and PP(proximal ミ Proliferative). The purpose of this research is to create a clustering model of cancer types by statistically analyzing these data and to use it for molecular target based treatment. Furthermore, by combining our results with that of a previous study on image based clustering model, we want to further improve the accuracy of classification and clarify the mechanism of lung cancer.
language of the presentation:Japanese

 
RATIH HIKMAH PUSPITA 1651130: M, 1回目発表 ネットワークシステム学 岡田 実
title: Compressed Sensing based Channel Estimation Algorithm for MIMO-OFDM System
abstract: For orthogonal frequency division multiplexing (OFDM) system, the channel estimation has an important role in determining the quality of the data transmission from transmitter to receiver. The channel estimation usually utilizes known pilot symbols at known positions to collect channel information and to estimate the channel status for these pilot positions. Then the channel information at data positions will be calculated using interpolation based methods. In addition, due to that the channel information in time-domain has the sparsity property, channel estimation using compressed sensing algorithms can achieve higher correctness of channel status but using small number of pilots than that of conventional interpolation based methods if the pilot positions can be random which is difficult for real OFDM systems. This research will randomly select known pilots to benefit the compressed sensing algorithms. The results provides show that using proposed compressed sensing algorithm, we can get better bit error rate (BER) performance than that of interpolation based OFDM system with a large reduced computational complexity.
language of the presentation: English

 
吉田 翔 1651122: M, 1回目発表 ネットワークシステム学 岡田 実

title: Optical Repeater for Next Generation Digital Terrestrial Television Broadcasting Signal Using Radio over Fiber

abstract: In the next generation digital terrestrial television broadcasting (DTTB), higher transmission rate than the current system is required to realize super high-vision broadcasting. Higher-order digital modulation and multiple-input and multiple-output (MIMO) technique are considered as a promising candidate to increase the transmission data rate without increasing radio frequency bandwidth. However, these schemes mainly focus on outdoor fixed/mobile reception, and there is no report on evaluating performance for compensating radio dead zone such as underground city and inside tunnel area. This presentation shows an optical repeater system using radio over fiber and proposals for next generation DTTB system.

language of the presentation: Japanese

発表題目: RoFを用いた次世代地上デジタル放送波の光中継に関する検討

発表概要: 次世代地上デジタル放送では高画質な映像を伝送するために必要とされる伝送容量が大きくなる.そのために超多値変調やMIMO等を用いた伝送容量の拡大が図られている.また,地下街等の電波不感地へは地上デジタル放送波をRoF(Radio over Fiber)によって光中継するシステムが導入されている.本研究ではワンセグ受信に焦点を当て,現在の光中継システムの構成を活かしつつ次世代地上デジタル放送波を受信するシステムについて提案を行う.