| 宮村 純真 | 1651105: M, 1回目発表 | ソフトウェア設計学 | 飯田 元☆ |
|
title:Classification of problem detection for incomplete software requirements
abstract:Software problem existing in spacecraft software can significantly hinder the success of mission. As for now, there are countermeasures for some problems, but there remains other that are not addressed yet. Up to this point, there problems as well as the corresponding counter measures. In this presentation, the method of the arrangement is introduced. In addition, alongside with the presenting the existing word, important terms, are also explained. In the presented work, main definitions as well as key terms related to dependability are documented. language of the presentation: Japanese | |||
| 尼子 琢朗 | 1651002: M, 1回目発表 | ロボティクス | 小笠原 司 |
|
title: Creation of a gripping behavior database taking into consideration hand and object states
abstract: In recent years, it is expected that human grip information will be used in a wide range of fields such as design and robot arm manipulation. In this research, I create a data set of grasping motion applicable to the real environment by measuring the hand shape, grasping force, grasping position and object shape. In this presentation, I introduce relevant researches on grasping attitude and give an approach to each measurement. language of the presentation: Japanese | |||
| 磯部 良太 | 1651012: M, 1回目発表 | 環境知能学 | 萩田 紀博 |
|
title: Research investigation which reduce of autonomous driving sickness using vection based on augmented reality in autonomous driving
abstract: Along with vehicles are converted autonomous driving, Vehicles are installed video head-up display. As a result, in addition to the problem of motion sickness that was concerned about before, problems of VR also occurred. It is expected that the comfort of passengers will be further impaired. Then, We aim to alleviate acceleration stimulation by vection based on AR technology for reducing motion sickness. In this presentation, I describe the background leading to the idea and the principle of motion sickness, VR sickness and the past approach. . language of the presentation: Japanese | |||
| 井上 剛 | 1651015: M, 1回目発表 | 自然言語処理学 | 松本 裕治 |
|
title: Fine-grained Arabic Part-of-Speech Tagging
abstract: Arabic language has complex inflection system to express its rich grammatical categories. This complexity increases possible part-of-speech variations, while it offers meaningful information for sysntactic parsing. The state-of-the-art tagger uses analyze-and-disambiguate approach using morphological analyzer, however, the analyzer does not always provide correct analyses for out-of-domain words. We propose a model for fine-grained Arabic part-of-speech tagging without the use of morphological analyzer. In this report, we present related research along with task description and our ongoing research in progress. language of the presentation: Japanese | |||
| 磯 颯 | 1651011: M, 1回目発表 | 自然言語処理学 | 松本 裕治☆ |
|
title: Tweet location estimation using Neural Network
abstract: The amount of location information available on Twitter is fewer than 0.5\% that makes the pursuit of many applications challenging, if not impossible. We propose the location estimation method from tweet data using Neural Network. language of the presentation: Japanese | |||
| 幾谷 吉晴 | 1651007: M, 1回目発表 | ソフトウェア工学 | 松本 健一 |
|
title: Identification of Difference in Brain Activity Between Experts and Non-Experts During Program Comprehension
abstract: Programming education is still a big issue even nowadays. However, little is known about how an expert acquire programming expertise. It is required to reveal how experts comprehend programs to improve effects of education. Recenlty several researchers investigated program comprehension process using brain activity measurement. To identify the special characteristics of experts, we use functional magnetic resonance imaging to compare brain activities of experts and non-experts during program comprehension. language of the presentation: Japanese | |||
| 野口 淳 | 1651085: M, 1回目発表 | 計算システムズ生物学 | 金谷 重彦 |
|
title:Gene expression analysis of alkaloid synthesis in Opium Poppy based on the metabolic pathway database
abstract:It is known that Papaveraceae synthesize various alkaloids, such as morphine, codine, papaverine, noscapine, which are useful for pharmaceuticals. In this study, we investigate the activity of the genes involved in the secondary metabolic pathway of alkaloid synthesis by comparing transcriptome of opium poppy (Papaver somniferum), which synthesis morphine, a mutant of P. somniferum that does not produce morphine, and another related species (P. orientale), that neither synthesize morphine but codeine. We will analyze the difference of expression of enzymes based on the database of the alkaloid synthesis pathways. language of the presentation: Japanese 発表題目:代謝パスウェイ情報に基づくケシのアルカロイド合成遺伝子発現解析 発表概要:ケシ科 (Papaveraceae)の植物は二次代謝物として、モルヒネ、コデイン、パパバリン、ノスカピンなど、医薬品として用いられるアルカロイド類を生合成することが知られている。本研究では、モルヒネを生産するケシ（Papaver somniferum）とモルヒネを生産しない遺伝子改変ケシ、モルヒネを生産しない近縁種であるオニゲシのトランスクリプトームを比較することで、アルカロイド類の合成経路に関連する遺伝子の働きを代謝パスウェイをもとにして明らかにすることを目的とする。 | |||
| 大野 大志 | 1651024: M, 1回目発表 | 光メディアインタフェース | 向川 康博 |
|
title: BRDF sampling from real object by multi-view images
abstract: In this study, we develop a simple method to reconstruct BRDF from real object without any special devices and complex setup by using multi-view images. At first, we reconstruct 3D shape of the target object and estimate camera and light directions from multi-view images. Next, because real object is composed by several materials, we classify them in order to perform BRDF sampling on each material. After that, we find the pairs of incoming and outgoing light directions which are extracted from MERL BRDF database as the optimal sampling directions and reconstruct its BRDF. language of the presentation:Japanese | |||
| 池谷 信寛 | 1651009: M, 1回目発表 | 光メディアインタフェース | 向川 康博 |
|
title: Improving vein pattern visualization using the spectral reflection ratio
abstract: In this study, we perform visualization vein pattern with high accuracy compared with the conventional method. Visualization of vein pattern is an important technique used for assisting surgery in the medical field and vein authentication in the authentication field. However, since the current vein visualization technology uses only a single wavelength, there is a problem in accuracy. It is empirically known that more accurate information can be obtained by using multiple wavelengths compared with only a single wavelength. Therefore, we measure the spectral reflection ratio using a hyperspectral camera to obtain the multiple wavelengths information. To specify the optimum combination of wavelengths and weighting for high-precision visualization, we perform vein visualization through regularized regression from the spectral reflection ratio. language of the presentation: Japanese | |||
| 堺井 大地 | 1651050: M, 1回目発表 | モバイルコンピューティング | 伊藤 実 |
|
title: Examination of encoding method enabling dimming in visible light communication using rolling shutter effect
abstract: A dedicated module was required for conventional visible light communication. Recent studies have realized visible light communication using LED illumination and mobile phone camera by using rolling shutter effect. However, these existing studies have problems such as poor energy efficiency because of dimming by voltage. Therefore, in this research, we will consider coding method which enables dimming without changing voltage. language of the presentation: Japanese | |||
| 雨森 千周 | 1651003: M, 1回目発表 | ユビキタスコンピューティングシステム | 安本 慶一 |
|
title: Simplified HRQOL Measurement Method for Healthy Life Expectancy Prediction
abstract: In recent years, average life expectancy has been extended. However, an increasing number of people will not have a Healthy life expectancy that can keep pace with a longer lifetime. Health-Related Quality of Life (HRQOL) is regarded as being strongly related to Healthy life expectancy, so it may be possible to predict Healthy life expectancy by continuously measuring HRQOL. However, this relation has not been quantified, and it may not seem so obvious.Therefore, we are trying to predict Healthy life expectancy from HRQOL. However, the existing HRQOL measurement methods rely on questionnaire surveys, which take much labor and time. Consequently, we are studying a simplified method of HRQOL measurement using a wearable device. language of the presentation: Japanese | |||
| 山野 龍佑 | 1651118: M, 1回目発表 | コンピューティング・アーキテクチャ | 中島 康彦 |
|
title: Advances in CGRA-based accelerator EMAXV focusing on local memory
abstract: As the aging progresses, society needs automation of Care-support. Machine learning, such as Deep Convolution Neural Network (DCNN) is expected in a care system to automate. However, it is difficult to achieve an aim in conventional CPU, GPU, FPGA when we paid more attention to electricity limitation and calculation resources. Therefore we focus on CGRA. CGRA still have room for improvement, because CGRA does not have built-in memory. We are developing EMAXV which is a pipeline memory incorporating arithmetic unit and behaves as CGRA base accelerator. We aim at mounting to ASIC finally. language of the presentation: Japanese | |||
| 森 剛基 | 1651111: M, 1回目発表 | 視覚情報メディア | 横矢 直和 |
|
title: Depth Map Inpainting Based on Compensation for Difference in Depth
abstract: When capturing depth maps with time of flight cameras, we have problems that we often cannot measure accurate depths because of the low reflection of light and the resolution of the captured depth map is low. To address these problems, patch-based depth map inpainting methods and super-resolution methods based on sparse coding have been proposed. In this presentation, I introduce a paper that proposes to improve the depth inpainting and super resolution by using not only depth maps but the texture images captured by cameras. The paper demonstrates the improvement of the proposed method by comparing results with those obtained by some conventional methods. In addition, I report my improvement of depth map inpainting that uses relative depth values rather than absolute ones. language of the presentation: Japanese 発表題目: 傾斜や奥行きの差に影響を受けない深度マップの修復法 発表概要: Time of Flightカメラを使った深度マップの取得において，材質の反射率などによって正確な深度マップの取得ができないことや，深度マップの解像度が低いという問題が存在する．これらの問題を解消するために，パッチベースによる深度マップの修復や，スパースコーディングによる深度マップの超解像アルゴリズムが提案されている．本発表で紹介する論文では，深度マップのほかにカメラから取得したテクスチャ画像を使用することでこれらの手法の精度を向上させる．いくつかの既存の手法と比較したところ，本手法による精度の向上が見られた．また，本論文を参考に深度マップの値を相対的な値とすることで，さらなる修復精度の向上を目指し，実験を行ったので，これについても報告する． | |||
| 細見 直希 | 1651094: M, 1回目発表 | 知能コミュニケーション | 中村 哲 |
|
title: Deception detection based on deep neural network and emotion related features
abstract: In recent years, research on automatic deception detection from speech by machine learning methods has attracted attention. Amiriparian et al. (Amiriparian, 2016) show that the detection accuracy improves by using predicted scores derived from emotion dimensions which analyzed by the emotional speech analyzer during learning of deceptive speech data. In our research, we focous on the fact that the emotion of speech is effective for deception detection and propose a method of learning emotion-related feature directly uing a classifier called a deep neural network which is excellent in identification of high dimensional data. | |||
| 田中 宏昌 | 1651069: M, 1回目発表 | 知能コミュニケーション | 中村 哲 |
|
title: Determination of the sequence of messenger RNA for improving protein
translation efficiency
abstract: Improving protein translation efficiency of messenger RNA (mRNA) is mportant for the development of new medicine. The aim of this study is to determine the mRNA sequence that improves the protein translation efficiency of Arabidopsis thaliana mRNA. As a fundamental step for our study, we have constructed several predictors for predicting PR values, which indicate the protein translation effectiveness, from given mRNA sequences. We developed the predictors by linear regression, random forest regression, gradient boosting regression, convolution neural network and long short-term memory network, then the random forest regression showed the best accuracy. language of the presentation: Japanese 発表題目: タンパク質翻訳効率向上のための messenger RNA の配列決定 発表概要: messenger RNA(以下mRNA)のタンパク質翻訳効率を向上させる事は， 新薬の開発の重要な要素である．本研究はシロイヌナズナのmRNAを用いて， タンパク質翻訳効率を向上させるmRNA配列の決定を目的としている．そのための足がかりとして， 与えられたmRNA配列から，タンパク質翻訳効率を測る指標であるPR値を予測するための学習器を構築した． 線形回帰・ランダムフォレスト回帰・勾配ブースティング回帰・畳み込みニューラルネットワーク・ long short-term memoryネットワークを用いて学習機を構築した場合， ランダムフォレスト回帰が最も良い精度を発揮した。 | |||
| 今林 亘 | 1651017: M, 1回目発表 | 知能システム制御 | 杉本 謙二 |
|
title: Addressing Outliers and Model Error Effect in Feedforward Learning Control
abstract: In non-contact-type sensors like cameras, signals may sometimes show outliers temporarily.Feedback control is sensitive to outliers while feedforwad control is expected to be less sensitive to the control performance. On the other hand, feedforward control is sensitive to modeling errors. This study aims at addressing outliers in these aspects and test the tolerance to modeling errors by means of simulation. language of the presentation:Japanese 発表題目: フィードフォワード学習制御の外れ値対策とモデル誤差 発表概要: カメラなどの非接触型センサでは，信号が一時的に外れ値を示すことがある．外れ値に対して，フィードバック制御は敏感であるが，プラントの逆モデルに基づくフィードフォワード学習制御では，制御性能の影響を小さくできるものと期待できる．一方でフィードフォワード制御はモデル誤差に敏感である．本研究の目的は，これらの観点から外れ値への対応を検討し，シミュレーションによってモデル誤差に対する耐性を検証することである． | |||
| 氏原 樹 | 1651019: M, 1回目発表 | ネットワークシステム学 | 岡田 実 |
|
title: Theoretical analysis on maximum efficiency of MIMO inductive power transfer system
abstract: On inductive power transfer (IPT) system, the use of Multiple Input Multiple Output (MIMO) is useful for improving the tranmission efficiency and reducing the leaked electromagnetic field. However, cross-copling among transmitters or among receivers decreases the tranmission efficiency. In this study, we derive the maximum efficiency of an arbitrary number of transmitters/receivers and invalidating the effect of cross-copling, by theoretical analysis. In this presentation, we report the theoretical formula, and the method to achieve the maximum efficiency using input currents of transmitters and the load impedance of the receivers. language of the presentation: Japanese 発表題目: 磁界結合型ワイヤレス給電システムにおける複数送受電器間の最大効率の理論解析 発表概要: 磁界結合型ワイヤレス給電システムにおいて複数の送電器または複数の受電器を用いることは，電力伝送の自由度や給電効率といった観点から有用であると考えられる．一方で，送電器間あるいは受電器間にクロスカップリングが発生し，電力伝送効率が低下する問題がある．こうした問題を解決するような，複数の送受電器を持つシステムの理論解析はこれまで行われていない．本研究では，任意の数の送受電器を持つ磁界結合型ワイヤレス給電システムについて，クロスカップリングの影響を無効化し最大伝送効率を達成する手法を，理論解析によって導く．本発表ではこれまでに導出した理論式および，送電器の入力電流と受電器の負荷インピーダンスによって最大効率を達成する手法について報告する． | |||