| YU ZIHAO | M, 1回目発表 | 数理情報学 | 池田 和司☆, | 松本 健一, | 川鍋 一晃(客員）, | 森本 淳(客員）, | 福嶋 誠 |
|
title: *** Extracting microstates from EEG and their corresponding dynamics using dynamic mode decomposition *** abstract: *** In recent years, EEG microstate analysis has gained popularity as a tool to characterize Spatio-temporal dynamics of large-scale electrophysiology data. It has been used in a wide range of EEG studies and the discovered microstates have been linked to cognitive function and brain diseases. One way to make sense of such dynamic networks is with modal decomposition. A relatively new modal decomposition method is DMD(Dynamic Mode Decomposition). DMD was developed initially to study experiments and simulations in the fluid mechanics community, where it was introduced to reduce very high-dimensional dynamic data into relatively few coupled spatial-temporal modes. Recently, DMD has been generalized as an eigen-decomposition of approximately linear operators, making it easier to apply to a larger class of datasets. The main purpose of this work is by using DMD to extract microstates from brain signals (specifically EEG) and their corresponding dynamics.*** language of the presentation: *** English *** | |||||||
| PHAN-UDOM PURIT | M, 1回目発表 | ソフトウェア設計学 | 飯田 元, | 松本 健一, | 市川 昊平, | 高橋 慧智 | |
|
title: Maintainability Model Improvement Based on Open Source Framework Empirical Study abstract: Maintainability has existed as one of characteristics of software quality. Today, many software, such as open-source frameworks, see longer lifespan and huge number of users. This trend comes to reflect the significance of maintenance work made on the system. Existing works have introduced methods to quantify the level of maintainability through computable functions, code metrics and formula. However, these approaches do not take into account the actual maintenance activities occurred, or how the maintainability should align with developers’ preference. This research aims to improve the existing maintainability model by analyzing the code changes in the contribution and adjust the model’s composition accordingly. From the preliminary analysis, it has been found that there are various changes in source code, contributing to various facets of the software. The expected contribution of this work is to deliver a more accurate maintainability model, as well as the result of analyzing the changing pattern of maintainability in open-source framework projects. language of the presentation: English | |||||||
| 柿本 和希 | D, 中間発表 | ソフトウェア設計学 | 飯田 元 ☆, | 松本 健一, | 片平 眞史（客員）, | 石濱 直樹（客員）, | 高井 利憲（客員） |
| 佐藤 真梨子 | D, 中間発表 | 計算システムズ生物学 | 金谷 重彦, | 松本 健一, | 小野 直亮, | Md. Altaf-ul-Amin, | 黄 銘 |
|
title: Determination of Optimum Number of Constitution on Japanese participants by Using CCMQ-J and Estimate Constitution Score by Healthcare Questionnaire
abstract: Chinese Medicine Questionnaire (CCMQ-J) consists of 60 independent questions and can determine nine physical constitutions called subscales. Since CCMQ-J was developed initially in China, we considered the application of CCMQ-J in Japan. The purpose of this study is to determine the optimal number of Japanese subscales and to understand the characteristics of subscales. As a result, it was suggested CCMQ-J can classify Japanese individuals into five different groups. In addition, most questions of CCMQ-J were relevant to the sub-health (as known as intermediate health and disease, correspond to ‘mibyou’ in Japan). We tried to explain CCMQ-J by another existing healthcare questionnaires, which consists of 254 questions related to disease, social factors, mental factors, dietary habits, and the sleeping state. Consequently, CCMQ-J score could be associated with public healthcare questions. language of the presentation: Japanese | |||||||
| 川瀬 寛也 | M, 1回目発表 | インタラクティブメディア設計学 | 加藤 博一, | 清川 清, | 神原 誠之, | 藤本 雄一郎 |
|
title: Eyes-free Text Entry for Augmented Reality Using Finger Gesture
abstract: AR Head-Mounted Displays (ARHMDs) are currently under active development and are expected to become widespread. However, in order to use ARHMDs as a general-purpose interactive information terminal, existing text input methods for HMDs have some issues to be solved. In this research, we propose an eye-free text input method using hand gestures on ARHMDs to improve the accuracy and input speed, which have been the issues of existing text input methods. Currently, we are implementing a keypad with haptic feedback and eye-free using HoloLens 2. language of the presentation: Japanese | ||||||
| FANG YU | M, 2回目発表 | インタラクティブメディア設計学 | 加藤 博一, | 清川 清, | 神原 誠之, | 藤本 雄一郎 |
|
title: Self-supervised Pedestrian Depth Estimation with Geometric Constraint abstract: Recently, self-supervised monocular depth estimation achieve the close performance to the supervised one, which ease the effort of collecting a large number of annotations in practicial application. However, these depth estimation is designed to simply predict all pixels without the assist of the semantic information, such as pedestrians and cars. To better estimate the depth of the pedestrians, we noticed that some imaging geometry constraint can help. We extended the self-supervised depth estimation baseline with a geometry aware module to imporve the depth prediction for pedestrians. The qualitive results on KITTI dataset show the necessity of injecting geometric constraint into the current baseline. language of the presentation: English | ||||||
| ZHANG RENJIE | M, 2回目発表 | インタラクティブメディア設計学 | 加藤 博一, | 清川 清, | 神原 誠之, | 藤本 雄一郎 |
|
title: Learning the Stable Feature for Cloth Changing Person Re-identificaiton abstract: Person re-identification (ReID) aims to recognize instances of the same person contained in multiple images taken across different scenes. Existing methods for ReID tend to rely heavily on the assumption that both query and gallery images of the same person have the same clothing. Unfortunately, this assumption actually limits the use of ReID in daily scenes. To tackle the re-ID problem in the context of clothing changes, we propose a novel framework by splitting this problem as a retrieval problem as well as a verification problem. We also proposed a new model called Stable Feature Verification Network (SFVN) for extracting those stable features from the image. Besides, our SFVN can automatically leverage the priority of different images regions for different image pairs. The quantitative results on the VC-Cloth dataset confirms the robustness and superiority of our approach against several state-of-the-art approaches. language of the presentation: English | ||||||
| ZHOU HANGYU | D, 中間発表 | インタラクティブメディア設計学 | 加藤 博一, | 清川 清, | 神原 誠之, | 藤本 雄一郎 |
|
title: *** Virtual Reality Public Speaking Training to Reduce Anxiety and Improve Skills ***
abstract: *** Public speaking may be a rare but most valued part of social communication, however, public speaking anxiety would make it more challenging. This research propose a virtual reality(VR) public speaking training system combining simulation training, feedback and reflection, to improve speaking skills while reducing anxiety. For the reflection part, a VR reflection method using VR record is considered. The experimental study result shows that individuals with low self-confidence on public speaking skills or negative self-bias could benefit more from VR reflection. For the feedback part, we focus on the affect of the feedback timing and want to make a dynamic method for feedback timing. *** language of the presentation: *** English *** 発表題目: *** この部分を発表題目に *** 発表概要: *** この部分を発表概要に *** | ||||||