若手研究者海外派遣頭脳循環プログラム・報告講演会(Report Meeting of Brain Circulation Program)

若手研究者海外派遣頭脳循環プログラム・報告講演(Report Meeting of Brain Circulation Program)

日時：	平成28年2月10日(月)3限 (13:30 -- 15:30), 2016/02/10, Wednesday
場所(Location)：	L1
司会(Chair)：	高松淳(Jun Takamatsu)

講演者(Presenter)：	武富貴史(Takafumi TAKAFUMI), インタラクティブメディア設計学研究室
題目(Title)：	Activity report at University of Oulu, Finland: Camera pose estimation under dynamic focal length change
概要(Abstract)：	Augmented reality is a technology which can overlay virtual contents onto the real world. This technology can be used for real world human-computer interaction. To make augmented reality environments, a visual-SLAM (Simultaneous Localization and Mapping) technology is important because this technology doesn’t need any preparation for a target environment. However, previous technologies assume fixed focal length setting during the estimation process. In order to remove this limitation, we propose a method for handling focal length changes in the SLAM algorithm.

講演者(Presenter)：	高道慎之介(Shinnosuke TAKAMICHI), 知能コミュニケーション研究室
題目(Title)：	Activity report in Carnegie Mellon University
概要(Abstract)：	I report my activity in Carnegie Mellon University. I have worked on high-quality speech synthesis for improving Quality-of-Life (QoL). My talk presents my attempt to speech synthesis and my daily life.

講演者(Presenter)：	山口明彦 (Akihiko Yamaguchi), Carnegie Mellon University, Robotics Institute
題目(Title)：	Pouring Skills With Planning And Learning Modeled From Human Demonstrations
概要(Abstract)：	During the research in Carnegie Mellon University, I explored how to represent, plan, and learn robot pouring. This was a case study of a complex task that has many variations, and involves manipulating non-rigid materials such as liquids and granular particles. Variations of pouring we consider are the type of pouring, material, container shapes, initial poses of containers, and target amounts. The robot learns to select appropriate behaviors from a library of skills, such as tipping, shaking, and tapping to pour a range of materials from a variety of containers. The robot also learns to select behavioral parameters. Planning methods are used to adapt skills for some variations such as initial poses of containers. This presentation will show simulation and experiments on a PR2 robot. This case study is a step towards enabling humanoid robots to perform tasks of daily living.

講演者(Presenter)：	石井将大(Masahiro ISHII), 情報基盤システム学研究室
題目(Title)：	Modular arithmetic on the Kalray MPPA-256 processor for a fast and energy-efficient implementation of ECM
概要(Abstract)：	The Elliptic Curve Method (ECM) is one of the integer factorization algorithm. Now the number field sieve (NFS) the most effective algorithm to find a large integer. ECM is not enough to use for factorizing a large integer such an RSA modulus e.g. 768-bit integers. It is however important to exploit the more effective algorithm and enhance its performance since the ECM is used by NFS as a its subroutine. I joined the CARAMEL project team working on integer factorization and discrete logarithm in finite fields, algebraic curves for cryptography, and so on. We implemented the ECM on twisted Edwards curves on the Kalray MPPA-256 board which is a light-weight many-core processor with low power high performance computing. In this seminar, I would like to introduce how to implement the ECM on the memory-constrained device, and the experimental result of the ECM in comparison with ECM performances on other platforms. We achieved the extremely high power-performance of ECM implementation on the Kalray MPPA-256 processor.

講演者(Presenter)：	大倉史生(Fumio OKURA)，大阪大学産業科学研究所
題目(Title)：	Activity report in INRIA, France: Changing season in a photograph
概要(Abstract)：	I visited to REVES (REndering and Vitrual Environments with Sound) group in Inria Sophia-Antipolis, France. This group focuses on computer graphics, especially image-based rendering and texture synthesis techniques. I have worked for a novel image manipulation technique which is a combination of a two complementary image manipulation techniques, i.e. color and texture transfer. We propose the first algorithm that unifies color and texture transfer, outperforming both by leveraging their respective strengths. Our approach transfers large temporal changes between photographs, such as change of season.

講演者(Presenter)：	Sakriani Sakti, 知能コミュニケーション研究室
題目(Title)：	Context Awareness and Priority Control for Intelligent Transport Systems based on Automatic Speech Recognition
概要(Abstract)：	This report presents research work conducted during my stay at Robotics & Intelligent Transportation Systems (RITS) Team, INRIA Rocquencourt, France. First, I started my activity with a study on Intelligent Transportation Systems (ITS) research field, including the state-of-the-art technology and the existing problems. Bringing rapid assistance to motorists involved in a traffic accident is an important service to be provided by ITS. Existing proposals to automatic accident detection are based only on the vehicle's perception point of view which might not give an optimal solution. Here, I introduced situational awareness based on the “understanding” of conversational speech of drivers/passengers using an automatic speech recognition (ASR) system. Context-aware priority control and congestion control schemes are presented to ensure coexistence of ASR-triggered applications and cooperative awareness messages (CAM) in the IEEE 802.11p system. Application risk analysis and performance evaluations of ASR and V2X communications on ITS simulations are also carried out. Finally, I implemented end-to-end integration between NAIST ASR system with INRIA Cybercars, which enabled communicating with Cybercars through ASR. The video of the real field testing will also be shown.