| 麻生 栄樹 | 1451005: M, 1回目発表 | 自然言語処理学 |
| 三田 雅人 | 1451103: M, 1回目発表 | 自然言語処理学 |
|
title: Preposition Error Correction for English Learners abstract: Grammatical error correction is the task of automatically detecting and correcting erroneous word usage and ill-formed grammatical constructions in text. This task could be helpful for the people that are learning English as a second language. In particular, prepositions are challenging for learners because they can appear to have an idiosyncratic behavior which does not follow any predictable pattern even across nearly identical contexts. In this presentation, I will explain several approaches for the task, and then show my implementation schedule. language of the presentation: Japanese | ||
| 壷内 貴弘 | 1451076: M, 1回目発表 | インタラクティブメディア設計学 |
|
title:Basic verification for the realization of near-eye display based on the pinhole principle
abstract:Recently, research on augmented reality has been actively carried out by using the Optical See-Through HMD. Previous work is a categorized by its main design criteria. In this study, it is an object of the present invention the realization of all these four excellent Optical See-Through HMD. However, HMD with advantage in all four categories does not yet exist. In this presentation, we will describe the results and future challenges obtained by performing a survey of the viewing angle in a short distance display. language of the presentation: Japanese *** 発表題目: ピンホール原理に基づく近距離ディスプレイの実現に向けた基礎検証 発表概要: 近年、Optical See-Through HMDを用いた拡張現実感に関して多くの研究がなされている。 過去の研究は、“邪魔にならない”，“視野角の拡大”，“選択遮蔽”，“同時焦点深度”の4つに分類することができる。 しかし、4つの分類すべてに優位性を持つHMDはまだ存在しない。 本研究では、この4つのすべてに優れたOptical See-Through HMDの実現を目的としている。 本発表では、近距離ディスプレイにおける視野角の調査を行うことで得られた結果及び今後の課題について説明していく。 | ||
| 三原 基 | 1451104: M, 1回目発表 | 光メディアインタフェース |
|
title: A Study of Unnecessary Object Removing Using Light Field
In this presentation, we propose a method to synthesize an image without unnecessary object using light field camera. The input of our method is 4-dimentiona light field data acquired by light field camera. According to the acquired light field data, we set focal length for each pixel individually not to focus on the unnecessary object. As a result, unnecessary object become invisible, then, natural synthesized image without the object are generated. In the presentation, proposal method, experimental result, problems and future works are explained. language of the presentation: Japanese 発表題目: 光線空間を用いた不要物体の除去に関する研究 発表概要: 本発表では，光線空間情報をもとに，不要物体が除去された自然な合成画像を生成する手法について提案する． 提案手法では，光線空間カメラで取得した4次元の光線空間情報を入力とし，画素毎に独立した合焦位置の設定を可能とする． このとき不要物体に対応する画素には合焦させないよう設定することで，不要物体をぼかし，自然な合成画像の生成が可能となる． 発表では，提案手法の説明，実験結果，問題点，今後の展望等の説明を行う． | ||
| 田中 隆寛 | 1451070: M, 1回目発表 | 視覚情報メディア |
|
title: Image inpainting using an adaptive pattern
abstract: Image inpainting is a method that can remove undesired objects in an image and plausibly fill in the missing regions to enhance the utility value of images. Conventional methods of image inpainting define an energy function based on similarity between missing regions and the rest of the image, and search for patterns that are similar to the missing region and updates pixel values so that the energy function is minimized. However, when similar patterns do not exist in an image, a blurry result is generated in the conventional methods. This study proposes a method that uses an adaptive pattern in which partial pixels of a window are used for searching for similar patterns and updating pixel values. In this presentation, the overview of the proposed method and experimental results are shown. language of the presentation: Japanese 発表題目: 適応パターンを用いた画像修復 発表概要: 画像修復は，画像から不要な物体を除去し，除去された欠損領域を違和感なく修復することで画像の利用価値を高める技術である．従来の画像修復の手法は，欠損領域と他の領域間のパターン類似度に基づきエネルギー関数を定義し，定義されたエネルギー関数を最小化するように画素値の更新を行い，欠損領域を修復している．しかしこの手法では，画像内に類似パターンが存在しない場合は結果がぼけてしまう問題点がある．そこで本研究では，結果の品質の向上を図るために，ウィンドウの部分的な画素を利用する適応パターンにより類似パターンの探索および画素値の更新を行う手法を提案する．本発表では，提案手法の概要および予備実験結果を示す． | ||
| 高山 怜也 | 1451065: M, 1回目発表 | 知能コミュニケーション |
|
title: Automatic detection of cognitive impairment among elderly with prosodic and linguistic features
abstract: Nowadays, the number of cognitive impairment patient is increasing. Cognitive impairment related disease may also have the progressive feature so that the early detection of these disease is important for not only patients but also for their families to reduce their burden. However, implementation of screening test for cognitive impairment is need to be done by professionally skilled people to assure the reliability of each result and this make hard to take it by patients. So, we believe that if implementation of these kind of test through dialogue system without going to the hospital, make easy to take it for elderly and this may help the increase the early detection of cognitive impairment. From the prior research, they use prosodic and superficial linguistic features so that they did not consider the contents of its speech. In this research, we will use prosodic and deep level of linguistic features such as contents of speech to aim the establishment of early detection method of cognitive impairment. language of the presentation: Japanese 発表題目: 発言内容に着目した音声・言語特徴による高齢者の認知症状自動検出 発表概要: 近年、高齢者における認知症患者の増加が問題となっている。認知症状を引き起こす病には進行性という特徴を持つものもあり、患者やその家族の負担を軽減させるためにも、症状の軽い早期に病を発見、治療を開始することは重要である。しかしながら、認知症判断のために広く使われている認知症スクリーニングテストでは、結果の信頼性担保のため専門家による実施が必要不可欠となっており、患者がいつでも気軽にテストを受診できる状態とはなっていないのが現状である。そこで、対話システムを通じて、病院へ行くことなく同等レベルの信頼性を持つスクリーニングテスト実施が可能となれば、より多くの高齢者にテスト実施を促すことが可能となり、認知症の早期発見に貢献できると考える。先行研究では音声韻律・言語の様々な素性を利用する一方、表層的で発話内容を深く考慮していないのに対し、本研究では音声韻律特徴の利用に加え、発話の内容についても考慮することで、同じ内容の繰り返しや発話の内容のつじつまが合わないときの発見などを手がかりにした認知症早期発見手法の確立を目指す。 | ||
| 平田 鷹志 | 1451087: M, 1回目発表 | ロボティクス |
|
title: Manipulation based on learning of position concept of object
abstract: For life support robots, manipulation is one of the important tasks. In previous life support robots, when achieve manipulation, robots almost take advantage of shape and position of the object itself. However, objects have position concept between objects and elements of object, and between objects. Therefore, in this research, we focused on the position concept of objects, and by learning the position concept, we aim improvement of efficiency and diversification of manipulation. language of the presentation: Japanese 発表題目: 物体の位置概念の学習に基づくマニピュレーション 発表概要: 生活支援ロボットにおいて，マニピュレーションは重要なタスクの一つである．これまでの生活支援ロボットでは，マニピュレーションを実現する際に，物体そのものの形状や位置を活用することがほとんどであった．しかしながら，物体には物体内での要素の位置概念や物体間の位置概念がある．そこで，本研究では物体の位置概念に着目し，位置概念の学習によって，マニピュレーションの効率化，多様化を目指す． | ||
| 下村 環太朗 | 1451058: M, 1回目発表 | 数理情報学 |
|
title: Behavior of multiple eigenvalues in estimating parameter matrix of multivariate distribution
abstract: In multi-dimensional data analysis, the model expressed by parameter matrix is often used. It is known that the estimation for such parameter matrix is influenced by duplication of its eigenvalues. In this presentation, we will propose a new estimation for covariance matrix of multivariate normal distribution using maximum likelihood method with respect to eigenvalues as the method considering duplication of eigenvalues and show that the proposed method can increase accuracy. Also, we will refer to geometrical analysis for estimating parameter matrix as future work. language of the presentation: Japanese 発表題目: 多変量分布の行列パラメータの推定における固有値重複の影響 発表概要: 多変量のデータの解析において，しばしば行列をパラメータに持つモデルが使われる．このような行列パラメータを推定するとき，推定精度が固有値の重複に影響されることが知られている．本発表では，多変量正規分布における共分散行列の固有値の重複を考慮した推定として，固有値についての最尤法を用いた手法を提案し，この手法が推定精度を向上し得ることを示す．また，今後の研究として，行列パラメータ推定の幾何的解析についても触れる． | ||
| 井上 万実 | 1451011: M, 1回目発表 | 大規模システム管理 |
|
title: Analysis of Transaction Processing Time for Bitcoin System
abstract: Bitcoin has attracted considerable attention as a new virtual currency, and the population of Bitcoin users has grown faster than other virtual currencies. However, the processing speed of Bitcoin transactions is much lower than that of conventional payment systems such as VISA. In this research, we analyze the performance of transaction processing for Bitcoin system. We model the transaction processing mechanism of Bitcoin as a non-preemptive queueing system, evaluating the waiting times of high- and low-priority transactions. Numerical examples show how waiting times of both priority types of transactions are affected by the arrival rate of low-priority transactions. language of the presentation: Japanese | ||
| 柴田 耕太郎 | 1451056: M, 1回目発表 | 知能システム制御 |
|
title: Object Shape Estimation with Active Tactile Sensing
abstract: Since visual and tactile sensors have different merits and demerits, we could extract valuable information by fusing both sensors rather than using either visual or tactile sensor. Therefore, in this study, we consider object shape estimation using a camera and a robot hand with tactile sensors as an example of the visual-tactile sensor fusion problem, Then, we aim to efficiently estimate the shape of the object by active tactile sensing. In this presentation, we first introduce an existing method. Then we show our proposed that solves some problems in the existing method and simulation results. Finally, we describe our future work. language of the presentation: Japanese 発表題目: 能動的接触による物体の形状推定 発表概要: 視覚センサと触覚センサにはそれぞれ利点と欠点があり，うまく組み合わせることで，どちらか一方だけ使う場合よりも有益な情報を得ることができる．そこで本研究では，視触覚融合の例として，カメラと触覚センサを持つロボットハンドによる物体形状の推定について考える．そして能動的な触覚センシングによって，効率的に推定を行うことを目指す．本発表では，先行研究の推定手法，およびその手法の持つ問題を解決した提案手法とそれらのシミュレーション結果を紹介し，最後に今後の研究予定について述べる. | ||
| 舟木 義人 | 1451092: M, 1回目発表 | 知能システム制御 |
|
title: A study for optimal design of human-robot interaction with physical contact abstract: In recent years, the demand for nursing care and welfare of the robot has increased in the aging society. In particular, it is essential for the assistant robot to physically contact with the human and we have to adaptively adjust parameters for humans when we design the controller. It is necessary to search the optimal parameters with the physical contact. Since the evaluation of such parameters is "cost expensive", that is, needs to execute the robot action to the human, it is desirable to reduce the number of trials for the parameter search. In this presentation, I introduce some related works to this literature, and present my research plan. language of the presentation: Japanese 発表題目: 物理接触を伴うヒューマンロボットインタラクションの最適設計に関する研究 発表概要: 近年は高齢社会の中, 介護・福祉関連のロボットの需要が増加している. 特に, アシストロボットではヒトとの物理接触が不可欠であり, コントローラを設計する上で, ヒトに対して適応的にパラメータを調整しなければならない. 物理接触を伴い最適パラメータを探索しなければならないため, 網羅的に探索する方法も考えられるが, 危険を伴う物理接触においては可能な限り探索回数を少なくすることが望ましい. 本発表では, 関連研究についての紹介とその手法の検証を行い, 今後の方針を述べる. | ||
| 河村 一輝 | 1451041: 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 preview customer recommendation system to support real estate salespersons for a restaurant
abstract:This paper proposes a preview customer recommendation system to support real estate salespersons for a restaurant. Most customers who aim to start restaurant business currently begin to look for alternative real estate by a search engine site in the Web, however, a considerable gap is observed in the preference between the results by a search engine site and the final choice of real estate by the customer.We propose the system in order to fill the gap. language of the presentation:Japanese 発表題目:飲食店不動産営業を支援する内見顧客推薦システムの提案 発表概要:本研究では、飲食店向け不動産営業を支援する内見顧客推薦システムを提案する。飲食店開業を目指す顧客のほとんどが最初に検索サイトで不動産を探し始めるが、検索サイトで気に入る不動産と成約する不動産には大きなギャップが存在する。内見はこのギャップが埋まるプロセスに大きく関わっている。営業担当者は、担当の不動産を気に入る顧客を時間やコストを掛けて予測し、内見を行うよう促している。その際、営業のコストを削減するために、内見に意欲的な顧客をデータマイニングから推薦する。 | ||
| 大橋 亮太 | 1451027: M, 1回目発表 | ソフトウェア工学 |
|
title: Task Purpose Estimation in Software Development Based on Application Execution Histories
abstract: In this research, I propose a method to support Personal Software Process (PSP), which is a well known software process improvement framework for individual developers. The proposed method estimates developer's purposes (aims) from time-series data about developer's tasks, given by an execution history of software applications. The proposed method can help developers' process improvement as they become aware of how much time they spent on a specific aim (such as implementation and testing). language of the presentation: Japanese 発表題目: アプリケーション実行履歴に基づくソフトウェア開発の作業目的の推定 発表概要: ソフトウェア開発者個人に着目した能力向上・プロセス改善手法として，Personal Software Process (PSP)が知られている．本研究では，PSPにおける作業履歴の記録と作業目的に基づいた履歴の分析を支援するために，パソコン上のアプリケーション実行履歴から作業目的を推定する手法を提案する．提案手法により，実装，テスト等の作業目的に費やした時間を自動推定することが可能となり，開発者が自らの作業を振り返って改善に繋げることが期待される． | ||
| 河居 寛樹 | 1451035: M, 1回目発表 | ソフトウェア工学 |
|
title: What leads to a bad patch?
abstract: A changed source codes to fix a bug often reviewed by developers who have not changed them before the release. Because, the changed source codes should be confirmed correctness, readability to share high quality source codes among developers. Unfortunately, the 56-60% changes were requested to develop again. Some previous researches presented two factors for the rejection. 1) Bug reporter has not given the sufficient information to fix a bug for developers yet. 2) A developer has less expertise to fix the bug. In this study, we identify which bugs were likely to request changes again by these factors. I will introduce two previous studied in this presentation. Finally, I will describe my research schedule. language of the presentation: Japanese | ||
| 上村 恭平 | 1451018: M, 1回目発表 | ソフトウェア設計学 |
|
title:<paper introduction>Vdiff: a program differencing algorithm for Verilog hardware description language
abstract: Recently, developers pervasively use hardware description languages (HDLs) to develop hardware. Program differencing tools are frequently used during code review tasks. However, results of existing program differencing tools include a lot of non-semantic information. For example, changing location of the HDL code is not always affected by the behavior of circuit. This paper proposes an algorithm and a tool which detects and shows only semantic differential information for the Verilog which is a kind of HDLs. language of the presentation: Japan | ||
| 齋藤 雄輔 | 1451049: M, 1回目発表 | ソフトウェア設計学 |
|
title: A Tool for Supporting Pull Request Driven Software Development
abstract: The Pull Request Driven Development (PRDD) is one of the modern methodology for software development. PRDD strongly collaborates with distributed version control systems (DVCSs) such as Git and Mercurial and its hosting services. During PRDD, a developer creates a branch for each feature and send a pull request to the project manager when he/she start implementing a feature. While implementing, developer need to refine the history of development to make it more understandable. Developer also has to care about changes applied to the main branch. In order to achieve these objectives, developer needs to operate DVCS correctly for every situation. However, achieving those objectives is complicated and requires further knowledge of the DVCS. In this presentation, I will propose a tool for supporting PRDD on Git. Proposed tool supports developers by restricting the harmful and unused Git features for developers during PRDD. It also contains features for Git to solve a common problem which happens during PRDD. language of the presentation: Japanese | ||