Design And Implementation Of An Energy Efficiency Mobilephone Physical Activity Recognition System Based On Decision Tree

Activity recognition is an important part of health monitoring. The high penetration rate and easy to carry make the utilizing of smartphones to activity recognition very practical. But there is also some problems in current study, such as the low recognized accuracy, high power consumption and the phone must be fixed to a part of the body during monitoring.In order to solve these problems, in this paper, the Euclidean distance was introduced as a feature, and the classification algorithm is Decision Tree. It could efficiently work out the problem that sitting and standing were difficult to identify. Six volunteers' activity data was collected, the average recognition accuracy rate is 92.63%.In order to figure out the high power consumption during activity monitoring, the concept of duty cycle was introduced, which could manage the period that the accelerometer worked for data collection dynamically. The experiments prove that this approach could reduce the system power consumption of mobile phones.For purpose of solving the fixed position of smartphones problem, the insensitive feature, signal amplitude was used in this paper. At the same time, first-order regression coefficient and Interquartile range was also utilized as features before classification. These features make the system can detect the activity which has random placement of mobile phones. Through a relative experiment proved the effectiveness of the method.Finally, the system in this paper used Java language, the Eclipse development tool, and based on Android development platform. Design and implement an energy efficiency mobile phone activity recognition system which could be actually used.