ゼミナール発表

日時: 5月30日(水)3限 (13:30-15:00)


会場: L1

司会: 米田
齊藤 康廣 1061202: D, 中間発表 松本 健一, 藤川 和利, 門田 暁人
title: Evaluation of Non Functional Requirements in a Request For Proposal (RFP)
abstract: In the beginning of a contracted based software development project, the RFP is provided by a software user company and used as an initial system requirements specification to ask software developer companies to propose their technical plans to fulfill the requirements. In this process, it is very important to evaluate the quality of the RFP to make sure that basic user requirements are written enough. Especially, non-functional requirements (NFRs) are important since the system architecture greatly depends on the NFRs such as response time and security issues. My study proposes a simple evaluation model of NFRs included in the RFP, mainly focusing on the user maintenance and operation issues. This model consists of NFR categories, NFR metrics, description level grading and weight to each NFR. As a case study, RFPs of 29 projects were evaluated by the proposed model. As a result, we confirmed that the model could identify poorly-written NFR aspects in the RFP, which need refinement before asking the developer company for a proposal.
language of the presentation: Japanese
 
Sukhanov Paul 1051207: M, 2回目発表 池田 和司, 松本 健一, 川人 光男
title: Neural Decoding of Moving Object Position
abstract: The primate visual system is often characterized as being composed of two distinct processing streams; a dorsal stream, responsible for representing the spatial position of visual objects, and a ventral stream, responsible for representing category information of visual objects. However, the fineness of spatial information for objects in the different visual areas has not been well characterized. Furthermore, in order to interact with the world purposefully, it is necessary to combine object spatial information and identity information together, and how the brain accomplishes this is also not well understood. Therefore, the current study attempted to 1. quantify the level of object spatial information contained across different visual areas, and 2. Investigate whether a combined, explicit "object identity and position" representation exists in any visual area. To achieve this, we used the method of neural decoding to learn a regression model between human neural activity measured with functional MRI, and the x,y positions of moving objects displayed to subjects during an fMRI experiment. By analyzing the accuracy of predictions from different visual areas and the generalization between different object categories, we were able to confirm that higher level visual cortex is able to encode object position highly accurately, with predicted trajectories for single objects achieving up to 0.7 correlation with the true trajectories. However, the current method was not able to identify a region of interest that simultaneously encoded object position and category, as prediction accuracies remained largely independent of the identity of the object used for training the regression model.
language of the presentation: English
 
Aisuwarya Ratna 1051208: M, 2回目発表 井上 美智子, 中島 康彦, 畠山 一実, 米田友和
title: Timing Optimization of Seed Ordering and Selection for High Quality Delay Test

abstract: Seed ordering and selection method has been proposed by Yoneda et al, 2010. This method select seeds and obtains significant reduction in seed count under SDQL constraint. However, this work still considered time consuming when applied to larger circuit. We address computation time problem for seed ordering and selection in high quality delay test. And proposed a method that can effectively reduce timing cost.

language of the presentation: English
 

会場: L2

司会: 原(直)
明戸 俊英 1151201: M, 1回目発表 音情報処理学
title: Evaluation of sound field reproduction method integrating multipoint controlled wavefront synthesis and wave field synthesis
abstract: Multichannel sound field reproduction system is aimed at reproducing spatial characteristic of the original sound field perfectly in human perception. To realize the system in practice, we have proposed a method integrating multipoint controlled wavefront systhesis (MCWS) and wave field synthesis (WFS). In this system, MCWS provides the accurate sound reproduction at the sweet spots around the user’s ears, whereas WFS can reproduce the approximated sound field over a large area, which contributes robustness against the user’s head movement. Therefore, it is essential to introduce the best WFS method for our hybrid system. In this presentation, we propose to use SDM (Spectral Division Method), and compare SDM and the conventional WFS method to clarify the efficacy of SDM, especially in the case of focusing source reproduction.
language of the presentation: Japanese
 
Arno in Wolde Luebke 1151205: M, 1回目発表 インタラクティブメディア設計学
title: Fast Surface Reconstruction of Particle-Based Fluid Representations using the Bilateral Grid
Abstract: Due to the advent of general purpose computing on GPUs and efficient particle-based fluid representations, physics-based real-time fluid animation is on the verge of becoming feasible. However, efficient and accurate surface reconstruction from particle clouds currently poses a challenge to these methods.Recently, so called screen space methods have gained popularity to address this issue. These methods do not rely on polygonization, instead they create a depth map of the fluid which in turn is used for subsequent rendering. In this presentation shortcomings of these approaches are discussed and a data structure from image processing - the bilateral grid - is introduced to resolve these issues.
language of the presentation: English
 
REINOSO CHISAGUANO DIEGO JAVIER 1151208: M, 1回目発表 ネットワークシステム学
title: Computational cost reduction of the detection process of MIMO-OFDM with ESPAR antenna receiver
abstract: MIMO-OFDM with ESPAR(Electronically Steerable Passive Array Radiator) antenna receiver can improve the bit error rate performance compared to the conventional MIMO-OFDM systems. This scheme uses the V-BLAST(Verticall-Bell Laboratories Layered Space-Time) algorithm for the detection process. However, due to the big size of the channel matrix, the computational cost is too high to be implemented. In this presentation a proposed solution based on a QR decomposition detection algorithm is introduced.
language of the presentation: English
 
金 周 1151212: M, 1回目発表 インタラクティブメディア設計学
title: An indoor tracking system for mobile devices using a bulit-in camera and sensors in tablet-based Augmented Reality
abstract: Markerless Augmented Reality(AR) systems need a user's location information to display 3d objects. AR systems on mobile devices are dependent on image processing to calculate location in indoor situations. This system uses a built-in camera, an accelerator and a gyroscope to calculate positions without other devices in indoor locations in which GPS(Global Positioning System) signals cannot be received.
language of the presentation: Japanese
 

会場: L3

司会: 池田(篤)
田中 陽士 1151203: M, 1回目発表 環境知能学
 
HUA HANYIN 1151210: M, 1回目発表 環境知能学
title: Human Pose Estimation on Still Image
abstract: Human pose estimation is a conparatively new research theme in computer vision field. In this presetation, two previous researches of part-based modeling and recognition of human pose will be introduced. The basic idea is to represent a humen body model by a collection of parts arranged in a deformable configruation. In one research, the application of shape context descriptors and discriminatively trained AdaBoost classifiers will be introduced. Finally, future research objectives will be introduced.
language of the presentation: Japanese
 
朴 建 1151214: M, 1回目発表 ロボティクス
title: Activating communications between human and robot by optimizing a behavior policy
abstract: In recent years, communication robots have been developed and offered for sharing a human's daily life, which require communication ability for communicating with human. To activate the communication with human, robots need to implement a suitable behavior policy for each person. However, former robots are not communicating with each person because they only considered communicating with people in general. In this study, the goal is for activating communications with human by optimizing a behavior policy of robots for each person. And robots will be able to activate communications with human by implementing a suitable behavior policy for each person.
language of the presentation: Japanese