HUMAN ACTIVITY RECOGNITION BASED ON SMART CHAIR

• Main Authors: Fitri, 范琪
• Other Authors: Chien-Cheng Lee
• Format: Others
• Language: en_US
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/u8z9eb

碩士 === 元智大學 === 通訊工程學系 === 106 === This research aimed to monitor and classify daily activities for elderly people in their home. Due to the vast majority of elderly people are spending most of their time in a sitting position, a chair has a potential to be a huge source of information about their activity behavior. In order to figure it out, this research develops a smart chair with a simple and low-cost sensory system. This smart chair built from six pressure sensors mounted in a chair and a Raspberry Pi to collect the raw data. All the sensors are mounted inside the cushion, four sensors on the seat and two sensors on the backrest of the chair. When a subject sits in the chair, the system collects signal data from these pressure sensors and transmits to a server to analyze the data. These data will be used to analyze the activity of a subject. In this research, the smart chair will monitor and classify five different activities such as working on PC, eating, napping, coughing, and watching a movie. Moreover, these activities tested by eight different subjects. In addition, three different machine learning algorithms were used to recognize these activities which are random forest, extremely randomized trees, and support vector machine. The experimental results show that extremely randomized tree is the best classifier with classification accuracy above 98%. This smart chair was convenient and easily operated to monitor carefully elderly people daily activities at home.