日時: 1月15日(火)3限 (13:30-15:00)

会場: L1

司会: 橋本健二
木下 修司 1251036: M, 1回目発表 計算メカニズム学
title: Granstrom's new development paradigm for efficient unit testing
abstract: We often create some "components" and compose them to construct a large software system. In that case, how to separately test each of the components that are dependent on one another is a big issue. One way is to prepare "mock objects" instead of real components. But if we don't create them properly, we cannot run unit tests correctly. To solve this problem, I will review Granstrom's paper that proposes a new development paradigm using "dependently typed programming language".
language of the presentation: English
発表題目: 効率的な単体テストを実現するGranstrom氏の新しい開発スタイル
発表概要: 大規模なソフトウェアを構築する際、ソフトウェアを「コンポーネント」という単位に分割して開発する方式は一般的である。その際には「他のコンポーネントに依存して動くコンポーネントを、どのようにテストするか」ということが問題になる。「モック」と呼ばれるテスト用のコンポーネントを作成することで対処するのが一般的だが、このモックが誤った動作をしている場合、テストが正常に実行できない恐れがある。このようなテストに関する問題の根本的な解決策として、Granstrom氏が論文で提唱する、「依存型(dependent type)付きプログラミング言語」を利用した新しい開発スタイルを紹介する。
西井 雅俊 1251080: M, 1回目発表 ディペンダブルシステム学
title: An experiment of burn-in time reduction based on parametric test analysis
abstract: The burn-In test method is a common testing method for screening out unreliable parts of microchips/semiconductors. The Burn-In-Test is widely applied by semiconductors. However, the method has some disadvantages. The burn-in test is very expensive in aspects of time (long burn-in periods) and testing equipment. Another disadvantage is the fact that the Burn-In-Test itself can destroy the microchip. This paper shows how to improve the long test time of the burn-in tests with techniques from the field of data mining. It is possible to identify a large portion of all parts that do not require long burn-in time, potentially providing significant cost saving (total 42.5% off).
language of the presentation: Japanese
発表題目: [論文紹介]パラメトリックテスト解析に基づくバーンイン時間縮小実験
発表概要: バーンインとは,出荷前の半導体デバイスに対して熱及び電圧といった負荷を与える事により,初期不良を事前に検出する手法である.バーンインは初期不良検出に非常に有効であるためよく用いられているものの,高コストである点・バーンイン自体がデバイスの劣化を導くといった問題があげられる.こういった背景から,バーンイン時間の縮小が求められている.そこで本論文は,データマイニング手法を用いることによってバーンイン時間の縮小を試みる実験を行った.結果,高いテストの質を保ちつつ,全体で42.5%のバーンイン時間の縮小を達成した.
五島 卓哉 1251042: M, 1回目発表 情報基盤システム学
title: Introduction to paper virtual machine integration approach for reducing power consumption
abstract: Equivalent to approximately 1.3% of the total power consumption of the entire world, the power consumption of data centers has been increasing still. Therefore, it is important to integrated onto fewer physical machines to virtual machines as much as possible. In this paper, I introduce the framework called pMapper, have been proposed for how to reduce power consumption while maintaining the SLA(service level agreement). I show several algorithms used by the framework. With the introduction of the paper in this presentation, I will show the direction of future research.
language of the presentation: Japanese

浦田 大地 1251020: M, 1回目発表 ソフトウェア設計学
title: [Paper Introduction] An Empirical Study of the Effect of File Editing Patterns on Software Quality
abstract: The authors of this paper analysed the relationship between file editing patterns and future bugs( i.e., bugs reported after developers' changes) through some case studies. File editing patterns describe some actions performed by software developers during coding; for their identification, some metrics are proposed. According to the results of their case studies, the potential and the number of future bugs in the file increase when developers follow one or more file editing patterns. The authors attempt to draw the attention of development teams and managers to monitor file editing patterns in order to reduce the risk of having future bugs; however, authors did not mention anything about the cause of such patterns. In this presentation, I introduce this paper and describe my research plan, which focus on the empirical analysis of the relationship between file editing patterns and the architecture or some metrics of the design of a software.
language of the presentation: Japanese
発表題目: [論文紹介] An Empirical Study of the Effect of File Editing Patterns on Software Quality
発表概要: 本稿では,File Edition Pattern (後述するファイルの編集パターン)と Future Bug (開発者が変更を加えた後に報告されたバグ)との関連を実証的研究により明らかにしている.File Editing Patternとは,ソフトウェア開発者がコーディング中にとる特定の行動を示すものであり,本稿で提案されたメトリクスによって判別される.開発者がいずれかあるいは複数のFile Editing Patternに従ってコーディングを行った場合,従わなかった場合に比べてFuture Bugの発生率および件数が増加することが本稿のケーススタディより明らかとなった.ただし,本稿では開発者がFile Editing Patternに従う原因には言及しておらず,マネージャおよび開発チームに対するFile Editing Patternへの注意喚起にとどまっている.本報告では,本稿の紹介に加えて研究計画について述べる.研究では,開発者がFile Editing Patternに従う原因はソフトウェアの設計にあると仮定し,ソフトウェアアーキテクチャや設計メトリクスとFile Editing Patternとの関係を分析する予定である.

会場: L2

司会: 中島悠太
宮崎 良 1251104: M, 1回目発表 インタラクティブメディア設計学

title: [Paper Introduction]Quick Viewpoint Switching for Manipulating Virtual Objects in Hand-Held Augmented Reality using Stored Snapshots

abstract: Recently, augmented reality(AR) technology is used in various situations. Especially, a method to manipulate the virtual object that is represented in real environment has been extensively studied. In this paper,by setting a new perspective snapshot taken in the AR environment, they proposed a method that can be performed intuitively switching perspective. They also developed a technique to manipulate the virtual object interaction using its perspective. In this presentation, I introduce the paper and present research plan.

language of the presentation: Japanese

発表題目: [論文紹介]Quick Viewpoint Switching for Manipulating Virtual Objects in Hand-Held Augmented Reality using Stored Snapshots

発表概要: 近年,拡張現実感技術は様々な場面で利用されている.拡張現実感技術を利用する上で,現実空間に表現された仮想物体を操作する手法は広く研究されている.その研究の一つとして本論文が挙げられる.本論文では,AR空間内で撮影したスナップショットを新たな視点とすることで、視点のスイッチングを直感的に行うことのできる手法を提案している.また,その視点を使って仮想物体を操作するインタラクション技術を開発している.本発表では,本論文の紹介に加え,今後の研究計画を発表する.

橋塚 和典 1251082: M, 1回目発表 環境知能学

title: Teaching system of solving Rubik's Cube by Augmented Reality

abstract: In recent years, there is a need for training on specific tasks applications by Augmented Reality.At the present day, we commonly use the method to display the object information and task procedures by Augmented Reality.However, there is a need for training which is assumed when there is no Augmented Reality.Therefore, we propose a system to efficiently impress the user with a specific tasks by Augmented Reality.We make a system that assumes the solution of a Rubik's Cube by reason of the difficulty of tracking and work.In this presentation, we introduce the papers of tracking which is a requirement of Augmented Reality.

language of the presentation: Japanese

野口 悠 1151081: M, 1回目発表 視覚情報メディア
title: 3-D Modeling System Using a Hand-Held Video Camera, Microfacet Billboarding
abstract: Recently, the opportunity that we see 3-D models of real objects in popular devices is increasing. For example, consumer game machines and cellular phones. For this reason, users who have no special skills and devices want to create 3-D models of real objects easily. In this study, we propose a system that users can create high-quality 3-D models easily. The proposed system uses billboard.
language of the presentation: Japanese
発表題目: 微小面ビルボーディングを用いたハンドヘルドビデオカメラによる3次元モデリングシステム
発表概要: 近年,家庭用ゲーム機,携帯電話等,一般に広く普及している機器上で実物体の3次元モデルを目にする機会が増えている.これにより,モデル生成に関する特殊な技能や機器を持たない一般の利用者が手軽に簡単に実物体の3次元モデルを作成したいという需要が高まっている.そこで,本研究ではビルボードを用いることで,写実性の高い3次元モデルを手軽に作成できるシステムを提案する.
今井 弦 1251014: M, 1回目発表 ネットワークシステム学
title: Particle Filter Assisted RFID Tag Location Method for Surgery Support System
abstract: I propose a particle filter assisted RFID (Radio Frequency Identification) tag location method for surgery support system. It is difficult to locate early stage small tumors during the intraoperative period. Ms. Takahata in my Lab has already proposed an RFID-based tumor location system, which implanted miniature RFID tags in the vicinity of the target tumor. The operator search for the RFID tag using the sensor antenna. Although the proposed method can find the tag location exactly, it requires a skilled operator. In order to simplify the location procedure. The proposed particle filter assisted RFID tag location system is capable of identifying the tag location with simple motion of the sensor antenna. Computer simulation result shows that the proposed method can provide the precise estimate of the location of the tag.
language of the presentation: Japanese

会場: L3

司会: Graham Neubig
小西 毅幸 0951055: M, 1回目発表 自然言語処理学
杉山 昂太郎 1251055: M, 1回目発表 知能コミュニケーション
title: Spoken dialogue interface to elicit the easy-to-recognize speech style
abstract: In current speech recognition systems, there are speaking styles which are easy to recognize or not. When people experience false recognition, they will try different speech styles to make the recognition succeed. The style at this time is clear, slow, and emphasizes a syllable in many cases. However, such a speech style is difficult to recognize for general speech recognition systems. In this research, we try to solve this problem by entrainment in human-machine communication. More specifically, the purpose of this study is to elicit a speech style that makes the user's speech easier to recognize, by placing restrictions on a user's utterance through system action.
language of the presentation: Japanese
発表題目: 受理しやすい発話様式を引き出す音声対話インタフェース
発表概要: 現在の音声認識システムには認識しやすい発話様式と認識しにくい発話様式が存在する。人々が誤認識を経験すると、認識を成功させるために異なる発話様式を試みる。この時の発話様式は、明瞭でゆっくりとしていて音節を強調したものであることが多い。しかし、そのような発話様式は一般の音声認識システムにとって認識が困難とされている。本研究では、人間とシステムの対話におけるエントレインメントによりこの問題の解決を試みる。具体的には、システムが利用者の発話に制約を課すことにより、利用者から認識が成功しやすい発話様式を引き出すことを目的とする。
有吉 智貴 1251006: M, 1回目発表 知能コミュニケーション
title: Impression of moving image by emotion recognition using EEG
abstract: There is service to evaluate and diagnose the visual media such as TV commercials in advance. Method of evaluation of this service is only subjective evaluation. However, we focus electroencephalogram and emotion in this study and purpose is to evaluate pictures or movies. We will classify emotions into values of valence and arousal in this study against previous research that emotion expressions such as anger, disgust, happy, surprise, sad, and fear were used in. We propose to adopt Self-Assessment Manikin scales using values of valence and arousal and improve Neural Network for classifying.
language of the presentation: Japanese
発表概要:今日, テレビコマーシャルのような映像メディアに対して事前に評価・診断を行うサービスがあり、このサービスの評価は主観評価である。しかし本研究は脳波と感情に着目し、動画像の評価することを目的とする。先行研究でのanger, disgust, happy, surprise, sad, そしてfearといった感情表現に対し, 本研究ではarousal, valenceの強度で感情の識別を行う。それらの値を用いる主観評価であるSelf-Assessment Manikin評定の採用、そしてクラス分けに用いるニューラルネットワークの改善を提案する。
西村 駿人 1251081: M, 1回目発表 自然言語処理学
title: Comparison of Morphological Analysis for Japanese Micro-blog's Post
abstract: When solving the task of Japanese natural language processing application, it is necessary to do word segmentation and part of speech tagging. Part-of-speech and morphological analyzer is used to do word segmentation and part of speech tagging. However, It is difficult to solve the unknown morpheme problem and Informal representation problem when we analyze social media text. In this presentation, I will talk about result of analyze for twitter texts and explain my research plan.
language of the presentation: Japanese
山崎 龍一 1251111: M, 1回目発表 音情報処理学
title: Multiple Moving speakers tracking based on particle filter using multiple microphone arrays
abstract: As a user-interface of smartroom, hands-free speech recognition is desirable. However, if more than one speaker speak simultaneously, there is an association problem between speakers and their speech. In a lot of cases, we need location information of speakers for sound source separation processing. Therefore, in this study, we try to track multiple moving speaker based on particle filter using multiple microphone arrays. In this presentation, I will explain research outline, progress and future research plans.
language of the presentation: Japanese