| 立花 巧樹 | D, 中間発表 | ユビキタスコンピューティングシステム | 安本 慶一, 藤川 和利, 諏訪 博彦, 松田 裕貴 |
|
title: Proposal of a method for collecting types and locations of litter using smart watches and sensor-equipped tongs
abstract: Presentation summary: In order to support an urban design approach to prevent littering, which has developed into a social problem, it is necessary to comprehensively collect garbage type and location information. Therefore, I propose a system that allows garbage pickers to collect information on littering simply by picking up garbage. This research consists of three functions: a litter recognition function to collect information on littering, a motivation improvement function for users to continue using the system, and a data utilization function to prevent littering. I will explain in detail the active acoustic sensing for litter recognition function and the sound feedback for motivation improvement function, which are the progress made in the doctoral course. language of the presentation: Japanese 発表題目: スマートウォッチとセンサ装着型トングを用いたポイ捨てごみの種別・位置収集手法の提案 発表概要: 社会問題に発展しているごみのポイ捨てを未然防止する都市デザイン的アプローチを支援するためには，ごみの種別・位置情報を網羅的に収集することが必要である．そこで私は，ごみ拾いする人がごみ拾いするだけでポイ捨てごみの情報を収集できるシステムを提案する．本研究はポイ捨ての情報を収集するためのポイ捨てごみ認識機能，ユーザが継続的にシステムを利用するためのモチベーション向上機能，ポイ捨てを未然に防ぐためのデータ活用機能の3機能で構成されている．博士後期課程での進捗である，ポイ捨てゴミ認識機能のアクティブ音響センシングとモチベーション向上機能の音フィードバックを詳細に説明する． | |||
| 後藤 逸兵 | M, 2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, 藤川 和利, 諏訪 博彦, 松田 裕貴 |
|
title:Proposal and Evaluation of a Pedestrian Flow Estimation Method Considering Direction of Movement in Streets Using BLE
abstract:Analysis of pedestrian flow is indispensable for urban planning and optimization of commercial facilities. Currently, camera-based pedestrian flow analysis methods are widely adopted, but they come with high costs and privacy concerns. As a new approach to address these issues, numerous methods using BLE (Bluetooth Low Energy) for analyzing pedestrian flow have been proposed. However, most previous studies have mainly focused on the density of the area, without giving adequate attention to the direction of movement. To bridge this gap, our study proposes a pedestrian flow estimation method considering the direction of movement using two BLE sensors. In the proposed method, time-series data obtained from the two sensors is used to create differences between them and clustering is performed. The number of BLE addresses belonging to each cluster is then used as a feature to estimate the pedestrian flow. To validate the effectiveness of our method, experiments were conducted on Midosuji in Osaka Prefecture, evaluating pedestrian flow in 5-minute intervals. As a result, while the average MAE without clustering was 20.8, our proposed method with clustering yielded an average MAE of 18.5. language of the presentation:Japanese 発表題目:BLEを用いた街路における移動方向を考慮した人流推定手法の提案と評価 都市計画や商業施設の最適化に際し、人流解析は不可欠である。現在、カメラベースの人流解析手法が広く採用されているが、これに伴う高いコストやプライバシーの問題が浮き彫りになっている。これらの問題を解決する新たなアプローチとして、BLE（Bluetooth Low Energy）を活用した人流解析方法が多数提案されている。しかし，多くの先行研究は空間の混雑度に重点を置いており、移動方向の分析は十分に行われていない。このギャップを埋めるため、本研究では移動方向を捉えるために２台のBLEセンサを用いた移動方向を考慮した人流推定手法を提案する。提案手法では，2台のセンサから得られる時系列データを基に、センサ間の差分を作成しクラスタリングを行い，各クラスタに所属するBLEアドレス数を特徴量として人流を推定する。本手法の有効性を評価するため、大阪府の御堂筋で実験を行い，5分単位の人流を評価した。その結果、クラスタリングを用いない場合は平均MAEが20.8となり，クラスタリングを用いた提案手法では平均MAEは18.5という結果が得られた。 | |||
| 伊勢田 氷琴 | M, 2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, 藤川 和利, 諏訪 博彦 |
|
title: Investigation of a Non-contact Daily Activity Recognition System Using Frequency-Shifted WiFi Backscatter Tags
abstract: In recent years, various activity recognition techniques have been proposed with an aim to be implemented in monitoring systems for elderly individuals living alone and health support applications. This paper presents a method and system for recognizing daily household activities by utilizing WiFi backscatter tags attached to furniture, fittings, and everyday items. The proposed system employs frequency-shifted backscatter tags and a software-controlled WiFi device called SDWiFi, capable of detecting shifts in the WiFi signal's frequency. In the proposed method, the frequency shifts caused by tag movement associated with everyday activities are detected by the SDWiFi receiver, based on wireless signals transmitted from the SDWiFi transmitter. This allows for non-contact recognition of household activities derived from information about the order in which household objects (such as furniture, fittings, and daily necessities) are moved.To investigate the feasibility of the proposed method, a preliminary experiment was conducted at the NAIST Smart Home, testing various positional relationships and angles between the SDWiFi transmitter and receiver, and the WiFi backscatter tags. By placing the SDWiFi transmitter and receiver 1-2 meters apart, and varying the tag position from near the transmitter to near the receiver, it was found that when the tag position was close (about 25 cm) to either the transmitter or receiver, recognition accuracy ranged between 77.6% to over 100%. However, when the distance between the transmitter and receiver was 2 meters, and the distance between the transmitter/receiver and the tag was about 1 meter, the accuracy dropped to approximately 46.3%. Furthermore, the angle between the straight line connecting the tag and the transmitter-receiver pair also significantly affected accuracy, resulting in a range between 46.3% to 87.6%. Lastly, when tags were attached and moved on the sink faucet lever, kitchen drawer, and refrigerator door, the recognition accuracies were 92%, 24.6%, and 52% respectively.From these findings, while there is a current limitation that tags must be placed near the transmitter or receiver, the feasibility of the proposed method, based on appropriate positioning and angling of the SDWiFi transmitter, receiver, and tags, was demonstrated. In future studies, there are plans to design tags optimized for the movement of furniture and appliances, as well as conduct activity recognition experiments involving participants. 発表題目: 周波数シフト型WiFiバックスキャッタータグを用いた非接触生活行動認識システムの検討 発表概要: 独居高齢者の見守りシステムや健康支援アプリケーションへの応用を目指し、様々な行動認識手法が提案されている。本論文では、家具や建具、生活用品に取り付けた WiFi バックスキャッタータグを用いて、家庭内の日常生活行動を認識する手法及びシステムを提案する。提案システムでは、周波数シフト型バックスキャッタータグと、受信した WiFi 信号の周波数シフトを捉えることができるソフトウェア制御型 WiFi デバイスである SDWiFi を用いる。提案手法では、SDWiFi の送信機から伝送される無線信号に対し、生活行動に伴うタグの動きが与える周波数シフトを、SDWiFi の受信機で検知し、家のオブジェクト(家具・建具・日用品など)がどのような順序で動かされたのかといった情報から生活行動を非接触に認識する。提案手法の実現可能性を調査するために、NAIST スマートホームにおいて、SDWiFi の送信機と受信機、WiFi バックスキャッタータグの様々な位置関係や角度に対し、タグの動きを検出可能かどうか予備実験を行った。1–2m離して SDWiFi の送信機・受信機を設置し、タグを送信機に近い場所から受信機に近い場所まで変更した結果、タグの位置が送信機あるいは受信機に近い(25cm 程度)時に、77。6–100%以上の精度で認識できた。一方で、送受信機間の距離が 2m、送信機・受信機とタグの位置が距離が 1m 程度の時は、精度が 46。3%程度まで悪化した。また、タグと送信機-受信機を結ぶ直線の間の角度によっても、精度が 46。3–87。6%まで大きく変化することが確認された。最後に、シンク水栓のレバー、キッチンの引き出し、冷蔵庫のドアにタグを取り付けて動かしたところ、認識精度は、それぞれ、92%、24。6%、52%となった。以上より、現状では、送信機・受信機の近くにタグを設置しなければならないという制約は存在するものの、SDWiFi の送受信機およびタグの位置関係や角度を適切に設定することによる提案手法の実現可能性が示された。今後、家具・家電の動きに最適化したタグの設計及び被験者を入れての行動認識実験を行う予定である。 | |||
| LIN CHING-YUAN | M, 2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, 中村 哲, 諏訪 博彦, 松田 裕貴 |
|
title: Estimating stress and anxiety level at home using audio-visual data during communicating with smart speaker
abstract: Mental health is a subject of significant concern in today's society. In comparison to physical health issues, problems related to mental well-being are often more challenging to detect and address. For instance, the fatigue from overtime work among office workers or late-night study habits among students may potentially lead to mental health issues. These problems typically result from prolonged accumulation and are not easily noticeable in their early stages. The purpose of this study is to analyze the stress and anxiety levels of people at home everyday. It utilizes audio-visual data from each of the subject's 40-second interactions with the smart speaker multiple times throughout the day. This data collection method allows for the capture of facial expressions, voice feature, and heart rate without requiring users to wear any additional sensors, thus reducing the burden on the users and aligning with the natural flow of their daily lives. On the other hand, we utilized the results of a nine-item questionnaire called the Depression and Anxiety Scale (DAMS), filled out after the video recording as the labels. DAMS is suitable for assessing short-term changes in stress and anxiety. language of the presentation: English | |||