| 築地原 里樹 | 1361009: D, 中間発表 | 小笠原 司, | 杉本 謙二, | 高松 淳, | 山口 明彦 | |
|
title: Whole-body coordinated movement for humanoid robots
abstract:Service robots are needed in daily life environments, especially to enable people to live alone in a self-sufficient and independent way and to keep a comfortable environment. In such environment the robot needs to perform various types of motions to complete its tasks. We focus on generating whole-body motions for humanoid robots in real time. In this presentation, we introduce a dual-arm controller and a walking algorithm that are implemented in the humanoid robot HRP-4. We discuss the execution time of a coffee-serving task using the proposed method. language of the presentation: Japanese | ||||||
| 杉垣 彰教 | 1351059: M, 2回目発表 | 小笠原 司, | 杉本 謙二, | 高松 淳, | 池田 篤俊 | |
|
title: Contribution of Somatic Sensation to Learning of Dynamic Manipulation
abstract: In human movement control, the sensation plays a critical roles to control the motion. In particular in dynamic manipulation to use the finger, we consider that human learn movement for the purpose of the task to be use feedback based on tactile and muscle senses.In this study, we investigate how roles tactile and muscle senses play to learning of dynamic manipulation.This presentation describes hypothesis to learning of sensory integration to spin a top, the experiment method and the simultaneous measurement system for EMG, tactile and muscle senses not to interfere natural manipulation. language of the presentation: Japanese | ||||||
| 米田 純 | 1351043: M, 2回目発表 | 安本 慶一, | 杉本 謙二, | 荒川 豊, | 玉井 森彦 | |
|
title: Intuitive appliance control system based on high-accurate indoor positioning system abstract: In our home, the increase of appliances causes the increase in the number of remote controls. In order to reduce the number of remote controls, there exists multifunctional learning remote controls. However, their too many buttons may confuse users. On the other hand, a gesture based control system has been recently proposed. However, it requires users to perform complicated gestures for selectingappliances or for turning on the machines, every time they want to control. In this paper, we propose an appliance control system which has advantage of both remote controls and gesture recognition. In our system, a user can select the target device by just pointing the appliance with a remote control. To achieve this, we utilize a high-accuracy indoor positioning system embedded in our smart house. After selecting the device, the user can control the appliance with the buttons displayed on the remote control for the selected device. Through an experiment, we evaluated the proposed system and confirmed that the system is enough applicable for the daily use. 発表題目: 高精度屋内位置情報を利用した直感的な家電操作手法の提案と実装 発表概要: 家庭内において，家電が増えるにつれリモコンの数が増え，同時に機能も増えるため，そのボタンの数が増加するという問題がある，一台のリモコンで複数の家電を操作できる学習リモコンは，リモコンの数を減らすことは可能であるが，家電の選定をボタン操作で行う点や，ボタンの多さから，操作の煩雑さの十分な解決には至っていない．また，直感的な入力方法として，ジェスチャ認識による家電操作法が挙げられるが，ボタン操作のような素早い操作を行うことは難しい．そこで本研究では，ジェスチャ認識とリモコンの長所を兼ね備えた新しい家電操作手法を提案する．提案システムは，高精度に屋内位置情報が得られることを前提とし，リモコンで家電を指すという直感的な動作で制御対象の家電を選択する．その後は，選択された家電に応じて液晶に表示されるリモコンを用いて家電を制御する．提案システムの構成と実装したシステムについて報告する．そして実装したシステムを実験により評価し，十分に実用的なシステムであることを確認した． | ||||||
| 角森 唯子 | 1351070: M, 2回目発表 | 中村 哲, | 小笠原 司, | 戸田 智基, | Sakriani Sakti, | Graham Neubig |
|
title: Deception detection with features during interaction, comparison between Japanese and English and analysis of questions
abstract: In recent years, many studies were accomplished about deception detection, and achieve a certain result in English. On the other hand, it reported that there are difference on characteristics between cultures or languages. So, we collect Japanese corpus including deception, besides compare features to detect deceptions effectively. Furthermore, it is necessary to ask questions to draw deceptions effectively. We analyze questions to detect deceptions effectively in this corpus. language of the presentation: Japanese | ||||||
| HOLLAND MATTHEW JAMES | 1351119: M, 2回目発表 | 池田 和司, | 松本 裕治, | 久保 孝富 | |
|
title: Location robust parameter estimation in probabilistic wind speed forecasting abstract: Forecasts of short-term wind speeds (up to 5 hours ahead) have numerous applications, including turbine control systems at wind farms, power supply planning by regional grid operators, and extreme weather early warning systems. Fully parametric probabilistic models of future wind speed have been shown to be effective, and while based on previous work in the literature, the selection of a particular distribution is relatively straightforward given a particular location at which to forecast, there remain many model elements, including the selection of predictive parameter models and determination of a loss metric for dynamic parameter estimation, which are non-obvious, may be location-dependent, and for which the previous research offers no insights. In our work we focus on all aspects of this problem, though the focus of this talk will be on the determination of an optimal loss metric. We propose a novel approach using a metric requiring only the first two moments and a location-adaptive method of initializing parameters. Using wind observations from climatically and geographically diverse sites across Japan, we establish the superiority of our proposed approach in terms of both minimal forecast error and optimal model fit, thereby establishing for the first time a method which is location robust, scales well in terms of forecast horizon, and is readily used on all standard predictive distributions. language of the presentation: Japanese | |||||
| 濱田 龍之介 | 1361010: D, 中間発表 | 池田 和司, | 松本 裕治, | 久保 孝富 | |
|
title: Analyzing syntactic rules of Bengalese finches' songs using beta process HMM
abstract: We analyzed syntactic rules of Bengalese finches' songs by an unsupervised fashion. At first the songs of Bengalese finches are symbolized to discrete-valued state sequences, by using one of Bayesian nonparametric approaches, a beta process hidden Markov model (BP-HMM). Symbolization of the songs are confirmed to be valid by investigating whether the symbolized state sequences are consistent with human annotated syllable sequences of the songs. Next correlation coefficients of state transition probability matrices are calculated between all pairs of the matrices, and a hierarchical clustering analysis is conducted by using the coefficients calculated. The results suggest that our analysis of the syntactic rules using BP-HMM can give a new insight to development of syntax. language of the presentation: Japanese | |||||
| 尾上 栄浩 | 1351026: M, 2回目発表 | 伊藤 実, | 松本 裕治, | 楫 勇一, | 関 浩之 |
|
title: Query Evaluation on XML Documents Compressed by Tree Grammars
abstract: XML documents often contain same repetitive substructures and various compression methods have been proposed. In this presentation, we propose an evaluation method for a query without uncompressing a document. We assume the compression based on SLCFTG. A query is given by a deterministic bottom-up tree automaton (DBTA) . We also give the experimental results to show the effectiveness of the proposed method. language of the presentation: Japanese 発表題目: 木文法による圧縮XML文書に対する問合せ処理 発表概要: XML文書は同じ構造が頻出することが多いため，様々な圧縮法が提案されている．本発表では，SLCFTGによる圧縮XML文書を対象とし，圧縮文書に対する問合せを解凍を行わず効率的に行う手法を提案する．問い合わせモデルには決定性ボトムアップ木オートマトン(DBTA)を用いる．また，提案手法を実装し，実際のXML文書を用いた評価実験の結果について述べる． | |||||