智慧家庭 (Attentive Home) 研究群
研究成果

 
回首頁
學術成果
技術成果
 


Technological Achievements (技術研發成果)

Persuasive System for Postoperative Eldercare

Year: 2010

Contributors(s): 顏羽君, 楊宗翰, 鄭一中, 葉慧文, 翁茂源, 何宇傑



Today, healthcare technology applying in hospital is vigorous development. One important task is to monitor the elderly people’s daily activities. In this work, we implement a non-obtrusive sensing way to automatically inference the current activity performed by the target patient. Recognized data will be stored into database and further analyze to report data. Caregivers such as nurses and doctors can access the report on website. We expect this mechanism will help caregivers to understand their patient more detailed. In addition, we want to see the capability of this system to improve the activity level of the patient. The project is now in laboratory stage and will be deploying in National Taiwan University Hospital.


智慧服務盒

Year: 2010

Contributors(s): 張馨之, 廖峻鋒



根據報告顯示(資策會產業情報研究所2009數位家庭安全生活需求調查),台灣消費者使用數位家庭服務的主要疑慮是產品價格(77.6%),而目前智慧居家系統多成套布建於高級住宅區。本系統針對一般市售家電進行整合,與家中多種感測器、使用者活動辨識引擎和服務發現及組合模組共同運作,根據目前使用者活動及環境狀態提供適時適地的服務,除此之外,使用者可以從市面上購買家電動態加入家中智慧服務,達到舒適、安全、便利及低成本的目標。


Multi-modal Leaving-Bed Events Detection and its Applications

Year: 2010

Contributors(s): 顏羽君, 李俊毅, 翁茂源, 鄭一中, 楊宗翰



Fall detection is an important task for elderly people. Human caregivers must pay special attention to the elders whenever they are going to leave bed. However, it is labor intensive when the time scaling to long term care. In this work, we develop a new approach to automatically detect the leaving bed activity. In this system, we fuse the information coming from laser sensor and the pressure mat. In addition, once the system detects leaving activity, timely reaction will be triggered to remind the elder being careful.


Experimental Dataset for Multiple-Resident Activity Recognition

Year: 2008-2010

Contributors(s): 徐國鐘



We used many kinds of sensors to collect sensory data in a smarthome. There was two residents living for a long period in this environment and they performed 18 activities of interest. During the experiment, the order of activities were not be constrained, but no concurrent activities. The participants we recruited were 8 males with engineer background and 6 females with other background.


A Clinical Guideline Execution Engine for Healthcare in Smart Homes

Year: 2008-2009

Contributors(s): 廖峻鋒, 張馨之, 林欣芸, 劉勇成



本系統目的為開發一結合GLIF與智慧家庭整合平台的執行系統,藉以提供最佳的居家照護服務。 醫護專家首先透過Protege平台,依照護者習慣背景與疾病特徵彼此之間的關係,建立符合GLIF之醫療指引。在受照護者的家庭中將裝設具情境感知 與推論能力的整合平台,根據家感測器所偵測到照護者的生理資訊和習慣以及周圍環境因素,經過整合後的資訊,做出適當的照護與警訊;此外,智慧 家庭整合平台透過監控系統蒐集執行指引後的行為回饋至醫療人員,可有效幫助醫療人員隨時進行監控與追蹤照護者的健康狀態,危及時,第一時間利用 傳訊系統傳送簡訊通知相關單位。智慧家庭整合平台能有效的減少照護者負擔與降低突發事件產生,提昇家庭照護品質。


Activity Recognition / Preference Recognition Engine for Smart Living Spaces

Year: 2008-2010

Contributors(s): 葉慧文, 顏羽君, 陸敬互, 陳意函, 何育誠,

We deploy many kinds of sensors(e.g., pressure, motion, reed switch and current) in a real smart home for collecting sensory data. Then we extract context information of resident from these data. By exploiting the information, we construct a activity recognition model to infer the resident's activities or behaviors. In the purpose of targeting more comfort for the resident, we use the feedback of her/him to build a preference recognition model to learn and adapt to her/his habits.


Ambient Intelligence Compliant Objects (AICOs)

Year: 2008-2009

Contributors(s): 陸敬互, 陳意函, 徐國鐘, 葉慧文, 何育誠

In order to improve the performance of activity recognition under device uncertainties in a dynamic environment, we propose making use of WSN technologies and a non-obtrusive deployment strategy to enhance regular household objects (such as an appliance, or a furniture, etc.) into various Ambient Intelligence Compliant Objects (AICOs). An AICO is an enhanced household object whose design helps secure WSN sensors from unexpected damage via a seamless (or non-obtrusive) deployment strategy, which can further prevent device-related dynamic events regarding device breakdown and malfunction.


支援快速開發高彈性智慧居家服務的應用程式框架

Year: 2008-2009

Contributors(s): 廖峻鋒, 詹景逸

情境感知在智慧居家服務應用程式已成為最重要的核心功能,情境感知系統和傳統桌上型及Web應用程式最大的不同處在於這類系統通 常不需要使用者刻意介入即可主動依據感測器所感知的環境資訊提供適宜服務。本研究提出了一個支援開發情境感知功能的智慧居家服務的 應用程式框架,基於服務導向的概念,此框架透過服務邏輯與執行環境分離的方式來加速開發流程,開發人員可以依組態設定彈性調整服務所需的執行環境。 我們基於此框架已成功開發了多個典型的智慧居家服務並將其佈署於二個不同實驗屋。

Publications:

詹景逸, 廖峻鋒, 傅立成, 陳俊杉, "支援快速開發高彈性智慧居家服務的應用程式框架," 「智慧生活科技區域整合中心計畫」暨「前瞻優質生活環境計畫」聯合成果發表會論文集, 台北, 台灣,2009.



具Web能力之智慧化居家感測整合服務平台

Year: 2008-2009

Contributors(s): 鍾亞文, 廖峻鋒

本系統的主要目的為基於OGC Sensor Web Enablement (SWE) Web 服 務架構提供未來智慧型家庭一個無線感測網路應用平台,並透過Web 軟體介面提 供無線感應網路系統在管理、控制、資料存取等功能。本研究案將著重於透過SWE 將智慧型家庭中的無線感測網路所得的資訊與外界進行分享及整合。


Skype / PSTN / 家電控制系統之整合與應用

Year: 2008-2009

Contributors(s): 柳沃辰

本系統主要目的為智慧型家庭影音電話整合應用情境之實現,包含家電設備發出訊息時(警告/告知),由PC主動Call Out 到 所設定的電話號碼,接通後用TTS以語音的方式告知訊息、結合語音辨識系統達到Call out / 接聽及使用者打電話回家藉由語音提示介面,提示使用者操作家電等功能。


E-Food System: Context-aware Cooking Support for a Smart Kitchen

Year: 2008-2009

Contributors(s): 盧佳言

本系統主要目的為智慧型廚房食物管理及電子食譜之應用實現,包含以RFID偵測食物及其進出時間,及食譜合成與提供功能。


Floor-AICO (無線壓力地板)

Year: 2007

Contributors(s): 陸敬互

Since piezoelectric straps or pads can detect exerted pressure on the floor and can easily be located out of sight, they become good candidate sensors to embed in an AICO for human location detection. We have also prototyped a smart floor block (called a “floor-AICO”) to detect residents’ current locations for assisting activity recognition. The advantage of implementing floor-AICOs is threefold: 1) These blocks are easy and flexible to deploy. 2) their accuracy and cost can be determined by the size of each mat as well as the total cost considerations of the floor. One could even deploy some dummy blocks, which have no sensors or wireless sensor nodes, under furniture or appliances. 3) These floor-AICOs make various interactive or interoperable scenarios feasible.


State-based Speech Recognition System for Domestic Use

Year: 2007-2008

Contributors(s): 黃崇恩

本系統針對現有的家用語音系統做分析討論並提出可改善的空間及改善的實作方法。現行在智慧家庭中使用的語音逐字辨識法將所有需要辨識的指令全放在一起, 當電腦接收到語音訊息時會逐一去比對每個指令,再選出最有可能的結果。如果每次辨識都必須去比對所有指令,則指令一多辨識錯誤率也會相對提高,畢竟目前沒有100%準確的辨識引擎, 過多的選擇可能讓正確結果被其他錯誤結果蓋過。如同一單選題,當我們只能大概分配選擇各個選項的機率時,選項一多則選到正確答案的機率也會降低。 因此,我們將現有的指令作樹狀排列,有些指令必須在開啟某樣電器之後才有作用,在非必要時將不會用到的指令收起來, 將有助於避免不必要的辨識錯誤,進而提升辨識成功率。


可攜式UPnP AV多媒體播放系統

Year: 2006-2007

Contributors(s): 廖峻鋒

隨著技術的發展,消費性電子產品 (Consumer Electronics, CE) 漸具有比以往更強大的計算能力,而由這些消費性電子產品所構成的家庭網路的概念也逐漸成型。 UPnP 協定是一個泛用的裝置服務模型,讓各項不同平台與實作技術的電子產品具有互通性。 UPNP AV Architecture 是以 UPNP 架構為基礎, 針對影音的傳輸而制訂的一套標準。 UPNP AV Architecture 的目標是要能夠支援各種傳輸通訊協定以及內容格式,讓這些內容在各項影音裝置間自動傳送,而不需要經過控制點 (out-of-band transfer) 。本系統的主要目的為在 OSGi 平台上實現 UPnP AV 的元件包模組 (bundle) ,並且可透過可攜式裝置在透過無線網路做遠端控制。



                                                                                                                                                                                                                                   

 
 
Copyright (C)2010 NTU CSIE All rights reserved.