Recognition of Human Primitive Motions for the Fitness Trackers

Abstract

Principle elements of a healthy lifestyle and harmful risk factors caused to cardiovascular diseases are being described. Significance of diurnal monitoring of essential characteristics of trustworthy physical state assessment by a fitness tracker is being shown. It had been shown that assessment of registered physiological parameters has to be performed taking into account the personal data such as age, anthropometric data and level of recommended physical activity. The basic block diagram of the typical fitness tracker and its functions are being described. The overall factors such as sensor placement, type of physical activity and human primitive motion, impacting to the fitness trackers measurement accuracy are being analyzed. The two approaches of human primitive motion detection are being proposed. The first one is based on clusterization procedure of original multidimensional time series with 72,4% accuracy of true detection and the second one performs classification procedure of preprocessed multidimensional time series with 85,5% accuracy of true detection.