Automatic Human Daily Activity Segmentation Applying Smart Sensing Technology

Human daily activity segmentation utilizing smartphone sensing technology is quite new challenge. In this paper, the segmentation method combining statistical model and time series analysis is designed and implemented. According to designed partition procedure, real measured accelerometer datasets of human daily activities are tested. The segmentation performance of sliding window autocorrelation and minimized contrast algorithms is analysed and compared. Experiments demonstrate the effectiveness of this proposed automatic human activity separation method focusing on the application of mobile sensor. As the properties of signal, mean, variance, frequency and amplitude are all useful features on the case of motion sensor-based human daily activity segmentation. In the end, the suggested work to improve the developed partition model is presented.