| dc.description.abstract | A smart home system capable of automatically learning and recognizing daily in-home activities of a resident would have many potential applications in real lives. This thesis presents the result of the survey analyst to study Vietnamese attitude and consumer behavior toward Smart home application, separate by locations, gender, age and other demographics information. In the mean time a training framework and description of a smart home simulation system are presented. The framework uses surveys as input datasets and outputs models in the form of activity sequences taking place in a single-resident home on a daily basis. Active learning technique is applied for training process (mining and clustering) and some modifications in learning algorithms is proposed to improve training cost and efficiency. A new algorithm is applied for the mining process to optimize mining expenses for the activity recognition problem. Through applied examples, it has been proven that the new algorithm named Closed Frequent Pattern Growth (CFPGrowth) used in this research work helps to reduce mining cost due to ignoring filtering process when compared to related approaches using Apriori or FPGrowth algorithm. Besides, activity models created in this way constitute smart contexts in recognizing and controlling in-home activities in real time inside Vietnamese Smart Homes Gathering the research Data This research used questionnaire as a primary data for analyze and evaluation. The survey is design in both language English and Vietnamese, adapt for future reference as well as easy to interact with Vietnamese community. Research Results In summary, my thesis contributes on the following topics:
An overview on Vietnam smart home market trend and consumer behavior and Vietnamese awareness toward smart home and smart home application
List of suggested Smart Home application for Vietnamese base on survey results
Propose several business Models for Vietnam Smart Home market
A training framework of in-home activity recognition is proposed. In this framework, a new mining algorithm (CFPGrowth) is used and a new method of calculating neighborhood radius for clustering is presented.
- xii -
Evaluate the training framework via survey datasets with the participant of Vietnamese people about their daily in-home activity sequences. Each training result reflects an in-home activity sequence of a Vietnamese or Vietnamese group.
A simulation system is implemented for automatically recognizing and controlling in real time all activities in a Vietnamese Smart Home (vSmartHome)
Keywords: Activity Recognition and Control, Smart Homes, Sensor Systems, Data Mining, Pattern Discovery, Vietnam smart home market | en_US |
