engleski

Quality of Human Activities Measurement from Accelerometer Data of a Smartphone

In the past few years, usage of smartphones increased significantly because of their computational power, large screens, memory capacity, large number of embedded sensors and other features. The aim of this study is to explore, test and compare the quality of measuring human activities through acquisition of raw data from smartphone built-in accelerometer. In a series of experiments, acceleration measurements from different smartphones were compared to the accelerometer data from a custom made sensor node. The selected activities that were measured are: walking, walking up stairs, walking down stairs, brisk walking, running and sitting (as a state of inactivity). Signal processing was made in Matlab. After calculation and comparison of the magnitude of linear acceleration, Fast Fourier Transform (FFT) was calculated for each activity on 256 samples of the signal. Then the energy of each activity was calculated and compared. Results show that the recorded acceleration values and their variations of amplitude of linear acceleration for sitting is negligible, while the amplitudes for other activities (walking, walking up stairs, down stairs and running) are more significant. Also, the amplitudes of the running activity measured with the sensor node are considerably larger than amplitudes of all other activities (walking, walking up stairs and walking down stairs). However, signals measured with smartphone introduced an error into calculation of linear acceleration because at least one of accelerometer axis from smartphone accelerometers was saturated for activities of greater intensity (brisk walking and running). Measurements using the accelerometer of the sensor node have shown larger differences between amplitudes of linear acceleration and differences between energy for each activity and the results of recordings can be used for measurements and not for monitoring applications only.

human activity measurement ; accelerometer ; smartphone ; human motion

nije evidentirano