| 里内 匠 | 1251047: M, 1回目発表 | ネットワークシステム学 |
| 水上 雅博 | 1251101: M, 1回目発表 | 知能コミュニケーション |
|
title: Personal style translation system.
abstract: There are various feature of writers that appear in text (personal style) The ability to convert between one writer's style to another writer's style would be helpful for both readability and communicability. In this study we propose a conversion method for writing style. We will introduce a related work, that translates a spoken utterance into written sentences based on machine translation. Next, we describe a text translation system that translates from the one personal style to another personal style in a one-to-one fashion. Finally, we explain how to expand to many-to-many translation of writing style. language of the presentation: Japanese | ||
| FABIAN LORENZO BAYTION DAYRIT | 1251124: M, 1回目発表 | 視覚情報メディア |
|
title: Human pose estimation with invisible body part handling
abstract: Human pose estimation has plenty of applications, such as for natural user interface or in gaming. An existing approach to do this is through the use of a single depth camera. That is, it assigns body part labels to the pixels of the depth image, then estimates the joints from the labeled pixels. However, body parts often become temporarily invisible, due to occlusion or going outside the camera's field of view. Such invisible body parts lead to estimation failure or only partial estimation, because this approach does not handle them. In this presentation, I describe my ideas to solve this problem, which use machine learning techniques. language of the presentation: English | ||
| 白井 尊昭 | 1251053: M, 1回目発表 | 自然言語処理学 |
|
title: Multiclass Entity Set Expansion
abstract:Entity set expansion is a task extracting words related with user's input words. In this task, it is desirable that words extracted earlier are reliable. Some previous systems extract multiclass entity sets simultaneously and outperform single class systems by penalizing ambiguous words over multiclass. In this presentation, we show the experimental results of an existing method where extracting multiclass outperforms extracting single class. In addition, we point out the problems of existing methods and will propose the concept of graph-based approach. language of the presentation:Japanese | ||
| 原田 真秀 | 1251087: M, 1回目発表 | 音情報処理学 |
|
title: Evaluation of Musical-Noise-Free method using Non-Stationary noise
abstract: The conventional noise reduction method for speech enhancement has a critical problem that method causes tonal noise called musical noise. That noise causes by nonlinear processing and leads to a serious deterioration of sound quality. However it is found that musical noise and kurtosis have deeply relation and new noise suppression method that musical noise does not occur at all has been proposed. That method supposes that input noise is stationary and estimated noise is set up by VAD (voice activity detection). In this study, we evaluate this method using Non-Stationary noise and dynamic noise estimation. language of the presentation:Japanese | ||
| 内田 大貴 | 1251018: M, 1回目発表 | 数理情報学 |
|
title: Develop the simulation system for clothing assistant robot.
abstract: Robotic assistanc is needed for an aging society. However, there are few robots for assistance in practise. I propose the simulation system for clothing assistant robot and use this system to real assistant robot.I evaluate whether reinforcement leaning is faster by using initial trajectory of simulation system. language of the presentation: Japanese | ||
| 木村 孝広 | 1251037: M, 1回目発表 | ロボティクス |
|
title:Acquiring of habitual activity by tracking human behavior with mobile robot.
abstract:In recent years, there has been widespread use of mobile robots in the general public. However, they perform the same behavior for any user, thus these services are not intelligent. In this study, I propose that the robot acquires habitual activity by tracking human activity to provide different services for each user. language of the presentation:Japanese | ||
| 高井 雄紀 | 1251056: M, 1回目発表 | 生命機能計測学 |
|
title: Study of quantitative diagnosis of Alzheimer's disease using PET
abstract: Risk factor for the development of Alzheimer's disease in recent years has begun to be elucidated.β amyloid protein is accumulated in the brain prior to the onset of Alzheimer's disease as one of them.This protein has been able to be imaged by using a PET.However, quantitative diagnosis must be an arterial blood sampling, there are still problems such method is invasive.Therefore, the purpose of the study of the dynamics model to explore the behavior of similar data obtained in the arterial blood sampling. language of the presentation: Japanese 発表題目: PET装置を用いたアルツハイマー病の定量診断の検討 発表概要: 近年アルツハイマー型認知症の発症の危険因子が解明され始めてきた。その一つとしてアミロイドβ蛋白質はアルツハイマー病を発症する前から脳内に蓄積している。 この蛋白質はPET装置を用いることでイメージングすることに成功している。しかし、定量的に診断するには動脈採血をしなければならず、侵襲的な方法であるなど課題が残っている。 そこで、動脈採血で得たデータと同様の振る舞いをする動態モデルの模索を研究の目的とする。 | ||
| 松本 匡史 | 1251098: M, 1回目発表 | 知能システム制御 |
|
title: Feedforward Learning Control in Simplified 2 DOF Structure with Its Application to a 2-link Flexible Arm
abstract: Two-degree-of-freedom (2DOF) control is useful because stability is ensured by feedback and response is adjusted by feedforward, independently. But the 2DOF structure is complicated and feedforward control needs an accurate plant model. To solve these problems, we propose a feedforward learning control in a simplified 2DOF structure. We further try to apply this control scheme to a two-link flexible arm. language of the presentation: Japanese | ||
| 北口 智英 | 1151043: M, 2回目発表 | 横矢直和, 加藤博一, 佐藤智和, 河合紀彦 |
|
title: A survey of non-photorealistic rendering
abstract: Non-photorealistic rendering is an field of computer graphics that focuses on enabiling a wide variety of expressing styles for digital art. In this presentation, I will explain the researches, especially on techniques by which users can change some images like painterly arts that the users prepare. language of the presentation:Japanese | ||
| 伊藤 渓太 | 1251010: M, 1回目発表 | ディペンダブルシステム学 |
|
title: Paper introduction: Online Measurement of Timing in Circuits: for Health Monitoring and Dynamic Voltage & Frequency Scaling
abstract: State-of-the-art FPGA chips use most recent CMOS technology to meet high performance demands. However, this technology is at the front line to face some challenges such as manufacturing variations and reliability concerns. Transistor aging is one of the most important reliability concerns. It causes timing performance degradation, and induces errors. Therefore, in order to achieve reliable operation, it is important to perform tests and measurements for detection of failures and degradation. In this presentation, I will introduce the delay measurement method for FPGA proposed in the paper, and present my research plan. language of the presentation: Japanese | ||
| 山下 徹 | 1251112: 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: Estimating foodstuffs and remaining amounts in fridge using weight sensor
abstract: In this study, we propose a system which estimates existing foodstuffs and their remaining amount in a fridge for the user without opening the fridge’s door. Since the same fridge is shared among multiple users at home or office, it is difficult for each user to grasp foodstuffs and their residual amounts. Therefore, each user has to check what foodstuffs exist and how much the remaining amounts are by opening the fridge’s door before going shopping or starting cooking. In general, the user opens fridge’s door to check the existing stuffs, but it leads to increase of electricity consumption and waste of the time. In addition, some users might think that it is a troublesome task. To manage foodstuffs in a fridge, some management methods with RFID tags are developed. However, since those RFID-based methods require attaching tags to all of the stuffs, the costs are expensive. Therefore, we propose a method which estimates and manages the stuffs and their residual amounts with only weight sensor embedded in the shelves or side-pockets of the fridge. There are two problems to realize the proposed method: identifying each stuff and estimating its current amount using the weight sensor. We solve these problems by constructing the database retaining the weight information of foodstuffs, and calculating the weight difference between before and after each stuff is used. language of the presentation: Japanese 発表題目: 重量センサを利用した冷蔵庫内のモノの種類および残量推定法 発表概要: 本研究では，冷蔵庫を開くことなく，冷蔵庫内のモノの種類・残量をユーザに提示するシステムを提案する．家庭や職場において，冷蔵庫は複数人に利用されるため，冷蔵庫に入れられている食材や飲料などの種類・残量を一元管理することは難しい． そのため，買い物や料理の前に，（ユーザは）冷蔵庫内のモノを確認する必要がある． 一般的に，ユーザは冷蔵庫のドアを開き内部を確認するが，これは消費電力の増加や時間の浪費につながる． また，ユーザはこの作業を面倒だと感じるかもしれない． 既存研究として，RFIDタグを利用したモノの管理手法が存在するが，管理する物品全てにタグを貼りつける必要があるためコストが膨大になる． そこで本研究では重量センサのみを使用し，モノおよびその残量の推定・管理を行う手法を提案する． 本研究の課題として，（１）モノの識別，（２）残量の推定，などが考えられる．そこで，提案手法では（１）をモノの重量を保持するデータベースを作成し利用することにより，（２）を使用前と後の重量差を算出することで解決する． | ||
| 矢野 正人 | 1251110: M, 1回目発表 | 計算メカニズム学 |
|
title:
Computational methods for predicting G-quadruplexes in nucleic acid sequences
abstract: Nucleic acid sequences such as DNA and RNA with abundant guanines (G) can form stacked guanine tetrad motifs known as G-quadruplexes. G-quadruplexes are considered as important factors in gene regulation. Several computational methods have so far been developed to search G-quadruplex-forming regions in genome sequences using motifs for G-quadruplex. However, we still need to reduce false positives in such scans by incorporating some other information. In this presentation, we will introduce a paper "RNA folding algorithms with G-quadruplexes" that extends a standard method for RNA secondary structure prediction to incorporate G-quadruplexes, which is one of the existing studies with the above purpose. language of the presentation: Japanese 発表題目: 核酸配列におけるG四重鎖予測手法 発表概要: DNAやRNAなどの核酸配列でグアニン（G）に富む配列は四重鎖構造を取ることがある。 G四重鎖として知られるその構造は遺伝子の発現制御において重要な役割を果たしていると考えられている。 ゲノム配列中のG四重鎖領域探索問題について、これまで主として配列モチーフ検索に基づく研究が行われてきたが、我々はG領域以外の情報を新たに組み込むことで探索における擬陽性を減らす必要があると考えている。 本発表では、そのような方向性の１つを示している先行研究で、RNAの2次構造予測法をG四重鎖を予測するように拡張した論文「RNA Folding Algorithms with G-Quadruplexes」について紹介する。 | ||
| 小西 文章 | 1251043: M, 1回目発表 | ソフトウェア工学 |
|
title: Software Effort Prediction Using Log-log Regression with Sampling of Predictor Variables
abstract: Effort prediction have been in widespread use to control human resources in software projects. To predict development effort with high accuracy can prevent failures in software development project such as cost or deadline over. As one of the effort prediction method, there are model based effort estimation using software measurement data. However, effort estimation model cause a problem of over-fitting, if explainatory variables of the model increases in order to express diversity of projects. In this reserch, we introduce a technique to reduce over-fitting by appling random sampling of predictor variables to multivariate log-log regression model. language of the presentation: Japanese 発表題目: 説明変数のサンプリングを用いたlog-log重回帰モデルによるソフトウェア開発工数予測 発表概要:ソフトウェア開発プロジェクトにおいて，プロジェクト完遂に必要な人的資源を管理するために，工数予測が行われている． 精度良く開発工数を予測することで，開発費の超過や納期遅れというプロジェクトの失敗を未然に防ぐことが可能となる． 工数予測の一手法として，過去のソフトウェア開発プロジェクトの実績データに基づいて工数予測モデルを構築する手法がある． しかし、プロジェクトの多様性を表現できるようにモデルの説明変数を増やすと，オーバーフィッティングが発生する. そこで本研究ででは，オーバーフィッティングに考慮した説明変数のランダムサンプリングを用いたlog-log重回帰モデルによるソフトウェア開発工数予測手法を提案する. | ||
| 寺本 翼 | 1251068: M, 1回目発表 | 情報基盤システム学 |
|
title: Location Tracking Prevention in Low Traffic Density Environment for VANET
abstract: Various communication messeages in Vecular Ad-hoc Networks (VANET) can be observed and then used for tracking a vehicle location by adversary. Pseudonym is most commonly idea to prevent location tracking, however, it is difficult to prevent tracking if vehicle traffic density is low. In this presentation, we will propose location tracking prevention scheme with dummy node messeage in low traffic density environment. language of the presentation: Japanese | ||