Activity monitoring using an intelligent mobile phone

Abstract

This research examines both the practicalities and feasibility of using a smart phone in the monitoring of gross daily activity, namely step counts. An Adaptive Step Detection (ASD) algorithm has been proposed and evaluated, based on where the phone is worn on the body. Experiments involved collection of data from a participant who wore two mobile phones (placed at difference positions) while walking on a treadmill at a controlled speed for periods of five minutes. A video recording and pedometer were used to independently record the number of steps in addition to a count by human observation. A step detection calibration factor was determined via a data driven approach, i.e, for each recording, a calibration factor was obtained by learning from two thirds of the acceleration data gleaned from the accelerometer within the smart phone. The remainder of the data was used to test the algorithm. The step counts from the acceleration sensor were validated by the video recordings, which were consistent with the pedometer and human observation. The results show that the step counts detected by the proposed algorithm achieved accuracy of 100% when the mobile phone was placed in the right thigh positions, and achieved above 95% accuracy when the mobile phone was placed in the right breast pocket, bag over right shoulder and right ankle.