| Summary: | 博士 === 國立臺灣科技大學 === 電子工程系 === 106 === Since the concept of Intelligent Environment has been widely spread in recent years. Smart Home as the most concerned branch of Intelligent Environment, it is expected that smart home applications will be incorporated into people's lives to improve their quality of life, avoiding the decline in quality of life caused by irregular lifestyles in contemporary society. By investigating people's daily routines, medical experts will have a better understanding of their health conditions to further assist to improve their life quality. Previous research has indicated that the non-intrusive sensor can be used to collect participants' biomarkers, such as Heart Rate Variability (HRV), which can be used to recognize people's activities. Moreover, the indoor context information extracted from the intelligent environment has the potential to be used for daily activities recognition. In this dissertation, participants' real-life data on HRV signals and daily routines are collected. Further, data preprocessing methods are used to improve the data quality, and machine learning techniques are used to produce classification models to recognize participants' activities from HRV data. Several supervised learning methods are evaluated, and the results indicate the classification models can detect daily activities with high accuracy. An additional method of Leave-Individual-Out is used to evaluate whether there are common HRV patterns in participants' activities across different individuals. After that, we refer to the related research on the activities recognition using indoor context information, to build simulated data of indoor context information based on participants' real-life data and the past research studies. The simulation results show that the recognition accuracy of daily activities has improved after adding the simulated indoor context information.
|
|---|
