BROWSE
Detail
Conference Paper
A Robust Step Counting Algorithm for Walking Speed and Device Placement Using Accelerometer and Gravity Sensor
Cited - time in 
Share 
Share
- Authors
- Yongjin Kwon, Kyuchang Kang, Changseok Bae
- Issue Date
- 2014-12
- Citation
- International Conference on Internet (ICONI) 2014, pp.1-10
- Language
- English
- Type
- Conference Paper
- Abstract
- Regular physical activity is associated with one's overall health status. To monitor one's health status, a lot of studies have considered to access the amount of physical activity. One of the most effective methods is step counting that estimates the number of steps taken by a person. With the steep advances of smartphones, it is possible to exploit the smartphones for step counting instead of excess devices. Although there was a lot of attempts to develop pedometer algorithms using smartphone sensors, most of them suffered from degraded performances for slow walking speed and arbitrary device placements. This paper proposes a novel step counting algorithm that is robust to walking speed and device placement using accelerometer and gravity sensor in a smartphone. We point out that each step yields a significant change of accelerations in the direction of the gravity, thus transforming the signals into a sequence of one dimensional data along the gravity. We think that a max-min peak in the data occurs due to a step taken by a person. Hence we implement a step counting algorithm that consists of three max-min peak detection algorithms, which finds consecutive max-min peaks. The experiment results show that our algorithm outperforms the step counter sensor provided by Android API for slow walking (50-75 steps / min -1), while our algorithm is still comparable to the sensor for normal walking (100-150 steps / min -1), regardless of device placement. Especially our algorithm achieved 87.8% accuracy for the walking speed of 50 steps / min -1 , while the step counter sensor achieved only 33.3% accuracy.
- KSP Keywords
- Counting algorithm, Experiment results, Gravity sensor, Normal walking, Number of steps, One-dimensional, Peak detection algorithm, Physical activity, Slow walking, Smartphone sensors, Step counting