在動態智慧家庭環境下具調適性且情境感知之行為辨識

每個家庭對於情境感知的應用都有各自獨特的考慮因素，居住者當下活動乃是其中一項重要的參考資訊，因為智慧家庭可以藉此資訊即時提供貼心服務。此外，一個真實的智慧生活環境本質上會隨著時間動態進行變遷，並衍生各式的改變，包括感測器電池耗盡、因操作過當導致誤觸動或零件失靈、居住者喜好/行為改變、以及系統組態必須隨時間進行升級與替換等等。若是一個智慧家庭的服務提供僅僅仰賴一隨時可能過時且固定不變的活動辨識系統，上述的各式變化可能會影響服務提供的品質並導致不愉快的用戶觀感，因此，增加系統的調適性和可靠度可因應居住者多變的需求與環境的變遷。為了探討上述的挑戰，本文目標是建立一協同式且可擴充的活動辨識基礎架構並透過可調適的模型更新機制來進行情境感知的活動辨識，並因應智慧家庭的特質提出一更可靠且以人為中心的方法來收集所有相關資料（包括來自居住者的即時回饋和環境的訊息）。透過此設計，各種智能模組（其包括利用無線感測網路技術所開發之人本智慧元件，致動器，智慧家電以及具調適性的推理與學習模型等）可以相互協同合作來增進情境（特別是正在進行的活動）預估的精確度，並可相互彌補各個智能模組先天上不足的缺失，以改善系統在動態環境下的辨識效能。此外，此設計更有利於智能元件之升級與替換以因應智慧家庭環境未來可能的各式變遷並進而提升系統的適應性與實用性。

關鍵字

行為辨識；可適性模型學習機制；日常活動辨識之基礎架構；環境智能相容元件；情境感知；動態環境；概念漂移

並列摘要

Every home has its own unique considerations for context-aware applications, and residents’ on-going activities are among the key contextual information since a smart home can rely on the information to provide attentive services whenever necessary. Ad-ditionally, in a real smart living space, the environment is dynamic in nature, which leads to various changes over time (such as sensor malfunction/breakdown, unintentional sensor triggering, preference/behavior changes of a resident, and changes in system configuration due to model upgrade, etc.) and would render a smart-home system built under the assumption of a static environment outdated; thus, such changes may later cause unreliable service-provision and unpleasant user-experience. Therefore, increasing the adaptability and reliability of a smart home will better fulfill real human needs and thus improve system practicality. In order to address the above challenges, the goal of this work is to realize adaptive context-aware activity recognition built on a flexible infrastructure; in addition, the system gathers all related information (from both residents and the environment) in a more reliable and human-centric way. By taking advantages of the flexible and cooperative interactions among the integrated smart components (which include enhanced smart objects using wireless sensor networks, ac-tuators, appliances, and adaptive learning/inference models) via the proposed infra-structure, the work will increase the accuracy of context estimation (especially on-going activities) and compensate the limitation of each smart component to improve resultant performance and robustness in a dynamic smart-home environment. Furthermore, the proposed approach enables more flexible upgrade and replacement of smart components to respond to various inevitable changes inherent in a home environment to improve the overall adaptability and practicality.

並列關鍵字

Activity Recognition ； Adaptive Model-Learner ； ADL (Activity of Daily Living) Infrastructure ； Ambient Intelligence Compliant Objects (AICO) ； Context-Aware ； Dynamic Environment ； Concept Drift

參考文獻

[1] J. Hightower, A. LaMarca, and I. Smith, “Practical Lessons from Place Lab,” IEEE Pervasive Computing, pp. 4-6, 2006.

[3] L. Liao, D. Fox, and H. Kautz, “Extracting Places and Activities from GPS Traces Using Hierarchical Conditional Random Fields,” The International Journal of Robotics Research, vol. 26, pp. 119-125, 2007.

[4] D. H. Wilson and C. Atkeson, “Simultaneous Tracking and Activity Recognition (STAR) Using Many Anonymous, Binary Sensors,” in Pervasive 2005, 2005, pp. 62-79.

[8] M. Philipose, K. P. Fishkin, M. Perkowitz, D. J. Patterson, D. Fox, H. Kautz, and D. Hahnel, “Inferring Activities from Interactions with Objects,” IEEE Pervasive Computing, vol. 3, pp. 50-57, 2004.

[9] D. J. Patterson, D. Fox, H. Kautz, and M. Philipose, “Fine-Grained Activity Recognition by Aggregating Abstract Object Usage,” in Ninth IEEE International Symposium on Wearable Computers, 2005. , 2005, pp. 44-51.

國際替代計量

在動態智慧家庭環境下具調適性且情境感知之行為辨識

全文下載

主題瀏覽