ゼミナール発表

日時: 12月21日(水)3限(13:30-15:00)


会場: L1

司会: 崔 恩瀞
大神 勝也 1651022: M, 1回目発表 ソフトウェア工学 松本 健一
title: Dynamic Program Visualization in 3D Space
abstract: Software visualization is a useful technique to understand complex software and identify problems. Compared to the 2D visualization, the visualization in 3D space is expected to present more information. In this study, we propose a technique of dynamic program visualization in real time. We implemented this technique as an Eclipse plugin, which targets Java programs. It can visualize Java programs in real time while using debug system (e.g. breakpoint, step execution) of existing Integrated Development Environments.
language of the presentation: Japanese
発表題目: 3次元空間におけるプログラムの動的可視化
発表概要: プログラムの可視化は,複雑なプログラムに対する理解や問題個所の抽出を容易に行うために有用な技術である.特に,3次元空間における可視化は2次元空間の場合より多くの情報を同時に表現することが可能であるため,プログラム実行の際に得られる複雑な情報の可視化が期待できる.本研究では,プログラムの動的情報を3次元空間上でリアルタイムに可視化する手法を提案した.提案手法は Java プログラムを可視化の対象としており,統合開発環境 Eclipse のプラグインとして実装した.既存の統合開発環境が備えているブレークポイントやステップ実行といったデバッグ機能を使用しながらリアルタイムに可視化できることが本ツールの特徴である.
 
松田 義貴 1651099: M, 1回目発表 知能コミュニケーション 中村 哲
title: Measuring Quality of in crowdsourcing
abstract: In crowdsourcing, estimating quality of results processed by workers is important, because not all workers are experts for the tasks. The workers do not always choose the right answer. In this situation, it is necessary to estimate the workers' abilities, the probability that the workers choose the correct answer, for earning the correct answers. In existing quality estimation methods, the workers' abilities have been treated as constant values. However, workers' abilities may improve as they work, and several workers may work irresposibly due to declining their concentration. To solve this issue, we try to measure the transition of workers' abilities to improve the quality estimation accuracy of products. In this presentation, we will report the results of analysis of the workers' behaviors, and find whether the workers' abilities change or not for every task.
language of the presentation: Japanese
発表題目: クラウドソーシングにおけるワーカの能力の変化を考慮した品質推定
発表概要: クラウドソーシングでは専門家でない不特定多数の人に作業を依頼するため、ワーカによる成果物の品質推定が重要な課題となっている。ワーカは必ずしも正しい答えを選ぶとは限らないが、その中で正しい答えを得るためには、ワーカが正しい答えを選ぶ確率、つまり作業能力を推定する必要がある。既存手法では、ワーカの作業能力は一定であるものとして扱われてきた。しかし、作業を行うにつれて作業能力が向上するワーカや、集中力が切れ適当に作業を行うワーカなどが存在する。そこで本研究では、このようなワーカの作業能力の変化に着目し、作業時点での作業能力を推定することにより成果物の品質推定精度が向上することを目的とする。今回の発表では、作業能力が変化するワーカが存在するかどうかの調査結果を報告する。
 
本田 将大 1651095: M, 1回目発表 知能コミュニケーション 中村 哲
title:Detecting Emotion Regulation using Functional Near-Infrared Spectroscopy
abstract: The purpose of this research is to detect emotion regulation using Functional Near-Infrared Spectroscopy(NIRS). A form of emotion regulation defined as the conscious inhibition of emotional expressive behavior while emotionally aroused. In dialogue, there is a problem that speakers regulate their emotions and detailed information on speakers can not be collected. Therefore, I propose recognition technique of emotion regulation to conduct more detailed dialogue. In this presentation, we will present related studies on detecting emotion regulation and the progress of preliminary experiments and research.
language of the presentation:Japanese
 
寺西 裕紀 1651075: M, 1回目発表 自然言語処理学 松本 裕治
title: Disambiguation Coordination Boundary using Neural Network
abstract: Coordination, common phenomena in Natural Language, is still a difficult problem in disambiguating even for state-of-the-art parsers. We propose dependency-based constraints on coordination spanning, and a model for identifying boundaries based on similarity between conjuncts. Extracting similarity feature from phrases, we use Convolutional Neural Network, which is widely used for classification tasks of sentence pairs.
language of the presentation: Japanese
発表題目: ニューラルネットワークを使った並列句範囲の曖昧性解消
発表概要: 自然言語における並列構造は文中に現れるありふれた現象であるが、最高精度の構文解析器においても曖昧性を解消することが困難な問題のひとつとなっている。本研究では並列句の依存構造に着目して並列句の範囲に制約を与え、並列句の内部構造の類似度に基いて範囲の同定を行う手法を提案する。句の対における類似構造の特徴を抽出するために、近年文対の分類タスクにおいて広く使用されている畳み込みニューラルネットを用いる。
 
城戸 孝士郎 1651042: M, 1回目発表 計算システムズ生物学 金谷 重彦

title: The development of a non-contact ECG Monitor system and its use in Detecting abnormal electrocardiograms during sleep

abstract: Holter, a medical device for 24-hours electrocardiogram (ECG) measurement, is usually used for abnormal ECG examination such as arrhythmia. However, holter monitoring has the disadvantages such as the necessity of cutaneous electrodes, offline signal process. There are also reports that ECG measurement routinely for a longer time (over 14 days) may help to find out more abnormal ECGs so as to provide more physiological information for clinical evaluation. Therefore, in this research, we aim to develop a system that can measures ECG continuously during sleep, but without attaching any electrode to skin. Further, this system should process the ECG signal and detects common abnormal ECGs in real-time.

language of the presentation: Japanese

 
稲光 勇人 1651013: 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: Investigation of the influence on users of AR connected with the real
abstract: Recently, a lot of games using AR have come out in the world. However, there are not many examples of application of strongly linked AR to entertainment with real world objects. Also, there are few games using AR based on the real world environment. Therefore, in this research, we aim to clarify the influence of AR object strongly connected with the real world on users.
language of the presentation: Japanese
 

会場: L2

司会: 油谷 曉
鈴木 裕貴 1651065: M, 1回目発表 環境知能学 萩田 紀博
title: Survey about continuity and contents of Conversations of Communication Robots for Elderly
abstract: In recent years, Social isolation of elderly living alone has become a Japan problem. Elderly living alone has a lack of opportunities to communicate on a daily basis, causing depression and dementia, inviting loneliness and suicide. In order to solve this problem, it is an important task to increase the opportunities for communication to elderly people. Therefore, we aim to provide continuous speech opportunities to elderly people using spoken communication robots. In this presentation, I will describe the content of utterance necessary especially for continuous dialogue.
. language of the presentation: Japanese
発表題目: 高齢者向けの雑談ロボットの発話内容と継続性に関する研究調査
発表概要: 現在日本では独居高齢者の社会的な孤立が問題となっている。独居高齢者は日常的にコミュニケーションを取る機会が不足しており、これは鬱や認知症のを引き起こし、孤独死や自殺を招く。この問題を解消する為に高齢者にコミュニケーションの機会を増やすことが重要な課題となっている。そこで、音声対話ロボットを用いて継続的な発話機会を高齢者に提供することを目的として研究を行う。本発表では特に継続的な対話に必要な発話内容について述べる。
 
河端 洋人 1651039: M, 1回目発表 ネットワークシステム学 岡田 実
title: Theoretical analysis on maximum effciency of multiple-transmitter inductive power transfer system
abstract: In this study, we derive the maximum power transmission effciency (RF-to-RF efficiency) in the case of cross-coupling between transmitters in a wireless power transfer system using an arbitrary number of transmitters. We model the system with the Z-parameter and derive the maximum efficiency, the optimum load impedance and each transmitter currents based on the first order necessary condition. We confirm the validity of analytical solution by electromagnetic field simulation.
language of the presentation: Japanese
 
千住 琴音 1651066: 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: An examination of increase establish of trip with vertial trip in one-way carsharing
abstract: Recently, one-way carsharing service that allows a user to drop the car off at the destination has been started. However, it doesn't become widespread because it can't assure the existence of both an available car at the start place and an vacant parking slot at the destination place. At this moment, the company moves cars to certain places manually like bicycle sharing system.However, it is so costly. In order to solve this problem in one way car-sharing service, we propose to involve a future potential user to move cars. We assume that there exist some requested one way trips. Our method inserts a new trip among the existing trips for maximizing the number of succeeded trips. This trip is called "a virtual trip". We assume that the virtual trip can be realized by changing a future schedule or asking future potential users who want to get some rewards. In this presentation, we explain the effectiveness of the proposed method through case study in surrounding area of our university.
language of the presentation: Japanese
発表題目: 乗り捨て型カーシェアリングにおける仮想的なトリップを用いたトリップ成立数増加の検討
発表概要: 近年,移動先で車を乗り捨てるワンウェイ方式カーシェアリングのサービスが開始されつつある. しかしながら,出発地に利用できる車があることと目的地に駐車できる空きスペースがあることの両方が 保証できないため,普及が遅れている. 現在のサービスでは,運営会社は自転車シェアリングのように手動で, ある場所に車を動かすが,それではコストがかかってしまう. そこで本研究では,ワンウェイ方式カーシェアリングサービスにおける問題を解決するために, 潜在利用者に車を動かしてもらう手法を提案する. 本手法では,片道トリップの需要があると仮定し,あらかじめユーザの予約しているトリップに "仮想トリップ"という新たなトリップを追加することで,トリップ成立数最大化を実現する. 仮想トリップは,カーシェアリングのスケジュール変更や報酬を得たい潜在利用者への依頼により実現できる. 本発表では,本学付近でのケーススタディを通して,提案手法の有効性を述べる.
 
友近 圭汰 1651077: M, 1回目発表 ロボティクス 小笠原 司
title: Musculoskeletal Hodel of Human Hand Based on its Joint Structure and Development of Robot Hand Based on it
abstract: Numerous robot hands have been developed so far. However, there is not yet anything that can be able to realize all of human dexterity. In order to develop a dexterous robot hand, it is necessary to understand the movements of the hand at the muscle level. Then, to construct a musculoskeletal model of human hand based correct joint structure and aim for development of robot hand with skill from knowledge obtained by analysis using that model. In this presentation, I will describe the background leading to the idea and the policy and approach to robot hand development.
language of the presentation: Japanese
発表題目: 手の正確な筋骨格モデルの構築とそれに基づいたロボットハンドの開発
発表概要: これまでに数多くのロボットハンドが開発されてきた.しかしながら,人の器用さのすべてを実現できていると言えるものは今もまだ存在していない.手の巧みな動作が可能なロボットハンドを開発するためには,手の動きを筋レベルで理解する必要があると考える.そこで,人の手の関節構造の正しい筋骨格モデルを構築し,そのモデルを用いた解析で得られた知見から巧みさを持つロボットハンドの開発を目指す.本発表では,着想に至った背景とロボットハンド開発までの方針とアプローチについて述べる.
 
坪田 一総 1651072: M, 1回目発表 光メディアインタフェース 向川 康博
title: Scattering analysis based on wavelength response for translucent objects
abstract: Translucent objects cause scattering phenomenon by incident ray. In previous studies, there are analysis of internal structure and material estimation about translucent objects by using difference of light transpose. In these studies, light attenuation by distance and light spread by degree are analyzed. Light transpose change by also wavelength. But it has not researched much in light scattering studies. So, we aim to analyze translucent objects by considering the relation of light scattering between wavelengths of light. We experimented about observing three kind of translucent objects that are different of optical density. In this presentation, we will state light spread is different by each of wavelengths in low optical density object.
language of the presentation: Japanese
 
西田 知弘 1651083: M, 1回目発表 大規模システム管理 笠原 正治
title: On group formation of secondary users for collaborative spectrum sensing in cognitive radio networks with multiple primary users
abstract: In cognitive radio networks, it is important for secondary users (SUs) to correctly detect whether the corresponding primary user (PU) has been using its own spectrum. Collaborative spectrum sensing (CSS) among SUs can improve probability to detect the PU signal compared with non-collaborative spectrum sensing, where each SU detects the signal by itself. Saad et al. have been proposed group formation of SUs for CSS in single-PU cognitive radio networks. Since there may be more than one PU in actual systems, we propose group formation of SUs for CSS in multi-PUs cognitive radio networks. Through simulation experiments with a two PUs scenario, we show the fundamental characteristics of the proposed scheme.
language of the presentation: Japanese

 

会場: L3

司会: 能地 宏
田中 雄大 1651070: M, 1回目発表 モバイルコンピューティング 伊藤 実

title:An Observation of transporting goods with DTN when a disaster strikes

abstract: Due to the Great East Japan Earthquake and the Kumamoto Earthquake, various parts of Japan suffered serious damage. After the occurrence of the disaster, large disruption occurred in the transportation of relief supplies, and problems occurred in which goods were not sufficiently transported to evacuation centers. The purpose of this research is to propose a logistic model for accurately and quickly transporting goods to evacuation centers on the assumption that supply of supplies right after the disaster is insufficient. In this presentation, we report suggested methods,problems to be solved through related research and future tasks.

language of the presentation: Japanese

発表題目: DTNを利用した災害発生時の物資輸送の検討

発表概要:東日本大震災や熊本地震によって日本各地は甚大な被害を受けた.災害発生後,救援物資の輸送においては大きな混乱が生じ,避難所への物資の輸送が十分になされない問題が発生した.本研究では,災害直後の物資の供給が不足することを想定し,避難所への物資の輸送を的確に,そして迅速におこなうためのロジスティック・モデルの提案を目的とする.本発表では,関連研究を通した課題と提案手法ならびに今後の課題について報告する.

 
木村 一統 1651044: M, 1回目発表 情報基盤システム学 藤川 和利
title: Method for Network Switching with SDN to Support Investigation of Malware
abstract: On security incident response, isolation of malware infected host and investigation of the damage situation of malware activity are needed. However, it is difficult to carry out them at the same time, when the malware has function to detect change of network state and change activity. In this presentation, I introduce the paper. The paper suggested the method to isolate malware infected host avoiding being detected by malware with prediction of malware's activity and safe network switching using SDN(Software Defined Network). The paper evaluated false-positive rate and false-negative rate of the prediction method with communication log of malware and found it useful in proposed situation.
language of the presentation: Japanese

 
沼田 椋太郎 1651084: M, 1回目発表 数理情報学 池田 和司☆
title: an influence of mental rehearsing to motor performance and brain activity
abstract: Human being has an innate ability to learn some new motor skills in trainings.The aims of this study are to reveal an influence of mental rehearse to brain activities during motor task and to examine a novel EEG based motor training method to enhance motor skill in short term and without repeating over and over again. As the methods, separating subjects into two of groups there asked to perform two type of tracking tasks. One is a simple repeat task, the other is combined with mental rehearse. In each group, EEG is measured during executing tasks. Comparison of result of each group will show an influence of mental rehearse to real behavior. Simultaneously, to discover the changes of brain activities that involved a progress of motor skill learning, we analyze EEG during motor task. In this presentation, I’ll present a result of simple repeat task.
language of the presentation: Japanese
 
天満 勇介 1651076: M, 1回目発表 生体医用画像 佐藤 嘉伸
title: Kinematic analysis of forearm rotation by 2D-3D registration using low-dose biplane fluoroscopy.
abstract: In orthopedics, visualization and quantification of bone kinematics is often useful for diagnosis and treatment planning. In case of forearm (radius and ulna), 2D kinematic analysis using fluoroscopy and 3D analysis using CT images have been developed.However,conventional methods do not provide temporal resolution sufficiently high for the analysis of a quick movement such as dislocation . In this study, we propose a method that achieves high temporal resolution and minimal invasiveness in the analysis of forearm rotation kinematic by 2D-3D registration using biplane low-dose fluoroscopy and CT image. In this presentation, we report the results of accuracy verification using a bone phantom for comparing 3 different calibration methods and the results of 3D kinematics estimated from the clinical fluoroscopy images.
language of the presentation:Japanese
発表題目: 低線量biplane透視を用いた2D-3Dレジストレーション法による前腕回旋動態解析
発表概要: 整形外科の診断において, 骨の3次元的な運動の可視化と定量化が必要とされている.これまでに前腕(橈骨・尺骨)の動態解析では,透視画像による2次元的な解析もしくは複数姿勢で撮影されたCT画像による3次元的な解析が行われてきた.しかし,従来法では脱臼動作のような瞬間的に大きな位置の変化が起こる症例には適用できない問題点がある. 本研究では高時間分解能かつ低侵襲な3次元動態の推定を目的とした,2方向の低線量透視動画像とCT画像から2D-3Dレジストレーション法による前腕回旋動態解析を行う.今回の発表ではキャリブレーション手法の比較のための模擬骨を用いた精度検証結果と高線量の透視動画像から推定した3次元動態の結果を報告する.
 
小澤 裕斗 1651028: M, 1回目発表 知能システム制御 杉本 謙二
title:Sample-efficient Reinforcement Learning using Gaussian Process
abstract:In this research, we investigated an approach to improve sample efficiency by learning state transition model using Gaussian process for EM policy search which is one method of reinforcement learning. As a result of simple simulation experiments, we confirmed that learning is more stable than normal EM policy search when the number of sample data is small.
language of the presentation:Japanese
発表題目:ガウス過程を用いたサンプル効率の高い強化学習
発表概要:本研究では、強化学習の一手法であるEM方策探索について、ガウス過程を用いて状態遷移モデルを学習することで、サンプル効率を向上させるアプローチを検討した。 簡単なシミュレーション実験の結果、サンプルデータ数が少ないときに、通常のEM方策探索よりも学習が安定することを確認した。
 
睨 雄太 1651068: M, 1回目発表 知能システム制御 杉本 謙二
title: Light environment control based on addition and subtraction of light
abstract: In the conventional light environment control system, the illuminance distribution was controlled only by turning on / off the lighting devices. This is a method based on addition of light. On the other hand, in this research, we focused on not only addition of light but also subtraction. And we built a system that consists of multiple lighting devices and windows that can partially block off the light. In order to make the actual illuminance distribution closer to the target illuminance distribution, this system controls lighting pattern of the illumination and light blocking pattern of the windows.
language of the presentation: Japanese
発表題目: 光の加減算に基づく光環境制御システムの構築
発表概要: 従来の光環境制御では,複数の照明機器の点灯・消灯により照度分布を制御していた.これは,光の加算に基づく手法である.これに対して,本研究では,光の加算だけでなく減算にも注目する.具体的には,複数の照明機器と部分的に遮光が可能な窓から構成されるシステムを構築する.そして,照明の点灯パターンと窓の遮光パターンを制御することで,実際の照度分布を目標の照度分布に近くすることを目指す.