| 増井 詠一郎 | 1251093: M, 1回目発表 | 知能システム制御 |
|
title: Visual feedback control to tolerate camera misalignment of roll angle
abstract: Due to the development of image sensing technologies, visual feedback control is widely used today. To avoid the sensing errors, the camera calibration is inevitable in general. However it is rather difficult to maintain accurately calibrated situation against the environmental changes. Therefore, in this study, we consider controller design problems to tolerate small camera misalignments. Especially in this presentation, the camera misalignment in terms of the roll angle is considered. language of the presentation: Japanese 発表題目: カメラのロール角設置誤差を許容する視覚フィードバック制御 発表概要: 画像センシング技術の発達に伴い, 視覚フィードバック制御は広く用いられている. 一般的に視覚フィードバック制御を用いる場合には様々な画像のゆがみによって誤差が生じるため, 精度の高いキャリブレーションが要求される. しかし, 環境によっては人の衝突や再設置などの外的要因によって高精度のキャリブレーションを維持できない場合がある. そこで本研究では, ある程度カメラの設置誤差を許容できるような制御系設計を考える. 特に本発表ではカメラのロール角設置誤差についての検討を報告する. | ||
| 増田 有利子 | 1251094: M, 1回目発表 | 生命機能計測学 |
| 細川 友紀奈 | 1251092: M, 1回目発表 | 数理情報学 |
|
title: Brain activity patterns during perception, dreaming, and imagery
abstract: Since our recognition of external environment depends on visual information highly, understanding our visual experiences is an ultimate goal of neuroscience. Not only visual perception induced by explicit stimulus presentation, but also visual imagery and dreaming which are generated without stimuli compose our visual experiences. Several studies attempted to reveal commonality and difference of brain activity patterns between perception and imagery, and a research of our laboratory demonstrated the similarity of brain activity patterns in visual cortex between perception and dreaming. However, commonality of neural representations in visual cortex during imagery and dreaming, which are generated without stimuli, are not investigated yet. And also differences of brain activity patterns in whole brain network among perception, imagery and dreaming have never been investigated. Here, we’ll investigate the commonality and difference of neural representations among perception, imagery, and dreaming in the whole brain. We first measured brain activity during imagery using fMRI in addition to data during perception and dreaming, and then performed fMRI decoding analysis among the three datasets. As a result, our analysis of object category decoding within each dataset showed higher decoding accuracy than chance in higher visual cortex. This results suggest that there are distinct brain activity patterns for object category during each visual phenomenon. Additionally, decoders trained with brain activity data in higher visual cortex during imagery successfully predict the contents of dreaming. Taken tougher with previous studies, our results suggest that brain activity patterns for object categories are partly shared among perception, imagery and dreaming. We will further investigate the commonality and difference among the three phenomena in other brain areas in future. language of the presentation: Japanese | ||
| 大下 将宗 | 1251022: M, 1回目発表 | ロボティクス |
|
title: Walk Optimization for Multi-legged Robot in Unknown Rough Terrain abstract:In these years, many researchers are studying multi-legged robots. Since such robots have many legs, they can walk in stable; thus, they are expected to work in rough terrain. In this research, we aim to design a system that estimates the type of a terrain from its feature, then generates an optimal walking pattern to the terrain. language of the presentation: Japanese | ||
| 上山 芳隆 | 1251016: 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 Proposal of Incentive Mechanism based Gamification in Participatory Sensing.
abstract:In recent years, participatory sensing has been attracting attention. In participatory sensing, instead of deploying sensors in the target area , we can sense the area by mobile users with sensors like smartphone. Thus, incentives are required for users to participate in sensing, for example, a client pays monetary rewards for users. Furthermore, stronger incentive is required for users to participate in costly sensing, resulting in the higher reward paid by the client. In this study, we propose a psychological incentive mechanism in addition to existing monetary incentive to suppress reward rising. The proposed incentive mechanism utilizes Gamification as a psychological incentive. Gamification facilitates users to participate in sensing with pleasant feeling as if they would play games. In this presentation, I will define the reward minimization problem for participatory sensing introducing Gamification and show basic ideas to solve it. language of the presentation:Japanese 発表題目:ユーザ参加型センシングにおけるゲーミフィケーションに基づくインセンティブメカニズムの提案 発表概要:近年注目を集めているユーザ参加型センシングでは，センシング対象のエリアにおいて，モバイルユーザが持つスマートフォン等をセンサとして用いてセンシングすることで，エリア内に直接センサを設置することなくセンシングが可能である．しかしユーザをセンシングに参加させるためには，それに対するインセンティブを提供する必要がある．その方法の一つとして，センシング依頼者が金銭的な報酬をユーザに与えることが考えられる．しかし，ユーザを負荷の高いセンシングに参加させるためには，強いインセンティブが必要となり，依頼者の支払うべき報酬が高騰してしまう．そこで本研究では，金銭的インセンティブに加え，心理的インセンティブを導入したインセンティブメカニズムを提案し，報酬額の高騰の抑制を目指す．心理的インセンティブとして，ゲーミフィケーションを利用し，ユーザをゲーム感覚で楽しみながらセンシングに参加させることで，センシングに対する参加意欲を向上させる．本発表では，ゲーミフィケーションを取り入れたユーザ参加型センシングの報酬額最小化問題を定義し，それを解くための基本方針について述べる． | ||
| 早苗 駿一 | 1251048: M, 1回目発表 | コンピューティング・アーキテクチャ |
|
title: A Method of Low-Cost Logic Synthesis for Yield Improvement
abstract: Manufacturing yield is degrading by CMOS downscaling. Therefore, LSI design methods for redundancy are becoming important. Conventional methods for redundancy such as DMR (Double Modular Redundancy) and using FPGA (Field Programmable Gate Array) require large overhead of area, performance and power consumption. Thus, conventional methods have room for improvement. In this research, a recently proposed circuit model, PPC (Partially-Programmable Circuit), is focused. A PPC is obtained by replacing parts of conventional logic circuits with LUTs and adding some redundant wires beforehand. When a defect is detected, internal logic of LUTs are modified for avoiding the defect. Establishment of a design method for PPCs contributes to yield improvement with low-cost overheads. In this presentation, features and problems of PPCs are explained, and a solution for that is proposed. language of the presentation: Japanese 発表題目: 歩留まり改善のための低コストな論理設計手法 発表概要: 半導体の微細化に伴い，製造歩留まりが低下しており，LSI設計における冗長化設計手法の重要度が大きくなっている． 従来の冗長化手法であるDMR (Dual Modular Redundancy)やFPGA (Field Programmable Gate Array)を使用する手法では，面積や性能・電力のオーバーヘッドが大きく，改善の余地がある． 本研究では，近年提案されたPPC (Partially-Programmable Circuit)と呼ばれる手法に着目する．PPCは，設計の段階で従来の論理回路の一部をLUT (Look Up Table)で置き換え，あらかじめ冗長配線を加える．製造後に故障が検出された場合には，故障を回避するようにLUTの論理を再構成することで正常な動作を保証する．PPCを用いた設計手法の確立により，少ないオーバーヘッドで歩留まりの向上が期待できる．本発表では，PPCの特徴と現状の問題点を説明し，問題点の解決手法を提案する． | ||
| 松浦 正尚 | 1251095: M, 1回目発表 | 情報基盤システム学 |
|
title: Reliable Publish/Subscribe System for Sensor Network
abstract: With the development of sensor-networking technologies, worth of realtime sensor information, such as meteorological information, is increasing. Publish/Subscribe(Pub/Sub) mechanism is suited for deliver information in sensor network. However, reliability of data is not guaranteed in the existing system. Therefore, in this presentation, I present basic policy for solve the issue. language of the presentation: Japanese | ||
| 土江 康太 | 1251065: M, 1回目発表 | 計算メカニズム学 |
|
title: A study for group key distribution using LKH for sensor networks
abstract: Many network applications (e.g., commercial broadcasting on Internet, group conference, etc) are based on a group communication model. A group key makes group communication secure and efficient, but we need the rekeying of the group key when a member joins to the group and a member leaves from the group. This rekeying cost depends on the number of members, so this can be a serious problem when we consider scalable group applications. As a solution to the problem, a group management system which uses logical hierarchy tree is studied, and known as an LKH. Although LKH achieves scalable group key management, LKH mainly assumes applications in conventional network such as Internet but not in sensor networks. To apply LKH in sensor network environment, we propose a proactive redundancy rekeying scheme to solve some problems which arise when we apply LKH to sensor networks. language of the presentation: Japanese 発表題目: センサネットワークにおけるLKHグループ鍵配送に関する研究 発表概要: 現在、インターネットの有料放送やグループ会議といったグループ通信モデル型のアプリケーションが普及している。グループ通信モデルで暗号通信を行う際にはグループ鍵を用いる方式が効率的であるが、グループ鍵はセキュリティ上、メンバの参加、離脱毎に更新(rekeying)が必要となる。Rekeyingを効率的に行うためのグループ鍵管理方式としてLKHが知られているが、LKHはインターネットのような従来のネットワークを想定した手法であり、センサネットワークへの適用は考慮されていない。実際、センサネットワークは構成上、従来のネットワークモデルに対し、パケットロスによる再送コストが高くなることから、rekeyingによる通信量や処理時間が問題となる。そこで本発表では、これらの問題を解決するために冗長性のあるrekeyingを提案し、その効果をシミュレーションによって評価する。 | ||
| 神保 希美 | 1251054: M, 1回目発表 | 知能コミュニケーション |
|
title:Learning for communication with hearing loss using real-time hearing loss
abstract:People with hearing loss is not good at communication by speech-to-speech. So they usually use the hearing aid. However, the hearing aid is not always enough perform when noisy environments, conversation with many person, and voice characteristics. Therefore, education and awareness of hearing loss and hearing aid has been performed. To better understanding, I expect that it is effective to experience hearing loss system. In my study, I propose real-time hearing loss system. In this presentation, real-time hearing loss system based on the hearing aid processing. language of the presentation:Japanese | ||
| 中井 駿介 | 1251072: M, 1回目発表 | 音情報処理学 |
|
title:Blind speech extraction to construct speech archives of a conference
abstract: along with the popularization of recording system using signal processing with microphone array, it becomes possible recording discussion of conference easily. However, it is difficult to find an objective voice from big amounts of recording data. therefore We purpose constructing a system that can find an objective voice easily, and research a recording system fitting for speech retrieval. This system suppress the noise which cause recording the discussion in actual environment using signal processing with microphone array and extract an objective voice. In this way, it makes speech retrieval easy. language of the presentation: *** English or Japanese (choose one) *** 発表題目:ポスタ会議発表の音声アーカイブ構築を目的としたブラインド音声抽出 発表概要: 近年，マイクロホンアレー信号処理を用いた収音システムの普及に伴い， 会議やポスターセッション等におけるディスカッションを簡易に収録することが可能となった． しかし，収録した膨大な量のデータから目的話者の音声を見つけ出すことは容易でない． そこで，我々は音声データ内の目的話者音声を容易に検出できるシステムの実現を目指し， 音声検索に適した音声収録が可能な収音システムの研究を行っている． このシステムでは，実環境におけるディスカッションの収録時に，アレー信号処理により目的話者以外の雑音を抑圧し， かつ目的話者それぞれの音声の分離，抽出を行うことで，後段の音声検索処理をより容易にするものである． | ||
| 光瀬 智哉 | 1251041: M, 1回目発表 | 自然言語処理学 |
|
title: Prepositional Error Correction with Syntactic Language Model
abstract: In various tasks of natural language processing, n-gram language models are used frequently to compute probability of input sentences. However, since n-gram language models uses only linear and local dependencies, they often assign inappropriate probabilities to sentences, so sometimes we need language models using more rich information. In this research, I will propose language models which can capture long-distance dependencies and the application of them, for example, prepositional error correction. This time, I will report the evaluation of a syntactic language model which is proposed in Pauls and Klein (2012).` language of the presentation: Japanese | ||
| 中尾 聡志 | 1251073: M, 1回目発表 | 視覚情報メディア |
|
title: Recovering Texture by Fitting Primitive Shapes to Point Cloud abstract: Recently, point cloud data are easily obtained from widely available depth sensors. Several methods have been proposed for reconstructing the shape of a real object by generating a 3D mesh model based on such point cloud data. However, a generated 3D mesh model can be inaccurate due to the lack and errors of the point cloud data. Furthermore, it consists of enormous number of meshes, which makes its manipulation difficult. In this work, we propose a method for reconstructing the shape and recovering the texture by fitting geometric primitives to the point cloud data. The proposed method is robust against the lack and errors of the point cloud data, because it represents the shape by a small number of geometric primitives controlled by a few parameters. In addition, this method enables us to recover the occluded texture of the real object. In this talk, I present an outline of this study and introduce papers which are applied to this study. language of the presentation: Japanese 発表題目: 点群へのプリミティブ当てはめによるテクスチャ修復 発表概要: 近年, 深度センサの普及により, 現実物体の点群データを容易に取得できるようになった. その点群データから, メッシュモデルにより3次元形状を復元する手法が提案されている. しかし, 一般に点群データには遮蔽等による欠損や誤差があり, 正確な現実物体の形状復元ができない. さらに生成されたメッシュモデルは多数のメッシュで構成されるため, モデルが複雑となり, 修正や加工等が困難である. 本研究では点群データに対してテクスチャで構成されるプリミティブを当てはめることにより, 屋内環境の形状復元とテクスチャ修復を行う. この手法では点群に少数のパラメータで制御可能なプリミティブ図形(直方体, 円柱等)を当てはめることで点群の欠損や誤差に頑健なモデルを作成する. 加えて, プリミティブに適用するテクスチャの欠損部分の修復を行い, プリミティブモデルの品質を向上させる. 本発表では研究の概要を説明すると共に, 要素技術として取り上げる論文を紹介し, 今後の課題について報告する. | ||
| 丸山 拓起 | 1251100: M, 1回目発表 | ネットワークシステム学 |
|
title: A Study on Vehicle Speed Detection System using Leaky Coaxial Cable
abstract: LCX (Leaky CoaXial Cable) is installed as mobile communication facility inside the tunnel and as wireless communication equipment for emergency in the underground city. LCX is not used only for communication but also as wide-area sensor, and object detection technology research using LCX have been carried out. In this presentation, we propose a method that LCX which is installed inside the tunnel utilizes as Intelligent Transport Systems. First of all, we have investigated a method for detecting the vehicle speed. language of the presentation: Japanese | ||