A Gyroscopic Data based Pedometer Algorithm Sampath Jayalath Department of Electrical and Computer Engineering Sri Lanka Institute of Information Technology New Kandy Rd, Malabe, Sri Lanka adasdjsampath@gmail.com Nimsiri Abhayasinghe Department of Electrical and Computer Engineering Sri Lanka Institute of Information Technology New Kandy Rd, Malabe, Sri Lanka nimsiri.a@sliit.lk Abstract—Accuracy of step counting is one of the main problems that exist in current Pedometers, especially when walking slowly on flat lands and performing different activities, such as climbing up and down stairs and walking on inclined planes. Although accelerometer based pedometers provide a reasonable accuracy when walking at higher speeds, the accuracy of them are not sufficient at slow walking speeds and performing different activities. This paper proposes a novel algorithm to detect steps using single-point gyroscopic sensors embedded in mobile devices. Preliminary analysis of data collected in different environments with the involvement of male and female volunteers indicated that gyroscope alone provides sufficient information necessary for accurate step detection. Algorithm was developed based on the gyroscopic data in conjunction with zero crossing and threshold detection techniques. The results proved that gyroscope based step detection algorithm provide a high accuracy when performing different activities and at slow paced walking. Keywords—Pedometer algorithms; gyroscopic data; single- point sensors; off-the-shelf devices; mobile applications; I. INTRODUCTION Modern medical researches highlight that pedometers support not only to physical body but mental activities of human beings to a greater extent[1],[2],[3]. Low cost pedometers help to improve the motivation of the walker [4], indoor navigation, activity recognition and for various applications in the field of health care. Pedometers can be used to detect steps from vertical acceleration of the human body. This works under two systems of mechanism. One is of mechanical based and other being of the electrical based accelerometers. Modern pedometers are generally based on MEMS (micro-electromechanical systems) accelerometer, mostly 1-axis, but the use of 2‐axis and 3‐axis accelerometers, gyroscopes and magnetometers improves precision and releases some utilization constraints e.g. positioning of the pedometer. The accuracy of these systems is at an acceptable level but not perfect due to various drawbacks [5]. Applications of pedometer are now upgraded and can be found in mobile devices. It is obvious with the application of pedometers to mobile devices, has now improved the standards of healthcare applications. The approaches of some pedometer algorithms proposed by researchers are discussed in the background section including their features and drawbacks. II. BACKGROUND S.E Crouter et al. [6] have compared the accuracy and reliability of 10 pedometers available in the market. These pedometers were based on mechanisms like, accelerometer, metal-on-metal and magnetic reed proximity switch. It is important to notice that all the testing was done at normal walking speeds. Their conclusion was that accuracy of pedometers was highly subjective upon the internal mechanism and sensitivity. But they have failed to measure the accuracy of pedometers when walking slowly and performing different activities like ascending and descending stairs. Comparative study with commercially available pedometers done by Jerome and Albright [7] has shown accuracies are poor with a minimum average absolute error value of 13%. Their conclusion was that none of the pedometers can be used for research purpose or general usage. Wasiq Waqar et al. [8] have developed a pedometer based on accelerometer for their “Indoor Positioning System” which consists of a preset threshold based peak detection method to identify a valid step and step cycle pattern detection method to discard invalid steps due to instantaneous readings of the accelerometer. It should be noted that the results of the pedometer may change with different individual walking patterns and speeds due to preset threshold. Melis Oner et al. [9] have implemented another step detecting algorithm for their “Early detection of the falling event system”. Step detection of this particular algorithm relies on the detecting peaks within a period in the data produced by the accelerometer sensor during walking. They were able to achieve higher accuracies during higher speeds of walking and with the mobile based pedometer placed fixed and loose in the pocket. However, their algorithms failed to count steps accurately during slow paced walking. Mi-hee Lee et al. [10] were able to achieve 99% accuracy in their portable acceleration sensor module with some advanced processing like FFT, Fuzzy C and statistical calculations. But they have agreed finally that their system doesn’t process data in real time, inability to measure steps during activities like ascending and descending stairs walking and need for an efficient device to carry out processing. A.M. Cavalcante et al. [11] have developed a pedometer to be used with their research on “Real-time indoor tracking on