Evaluation of foot contact event detection algorithms

System and method for analyzing force sensor data, US provisionals 63/291,424 and 63/315,847

Authors: Blades, S¹, Marriott, H¹

1 Kinetyx

Highlight 

Kinetyx developed a proprietary foot contact event detection algorithm which reads data from the sensory insole to accurately determine foot contact events.

Abstract

The Kinetyx sensory insole system is specifically designed to measure lab quality data without the constraints of laboratories or lab equipment. This system has been designed specifically for researchers who are interested in continuous capture of biomechanical signals such as forces at foot ground interactions, pressure distribution, and inertial signals. This continuous stream of data allows for temporal running gait metrics to be examined to help guide training, monitor injuries, or influence future innovation. The analysis of temporal gait metrics requires an accurate method of determining foot contact events (foot contact and foot off) for each stride. Using data read from the Kinetyx sensory insole, Kinetyx developed a proprietary foot contact event detection algorithm to provide these events to users. This algorithm was validated using data collected from an instrumented treadmill (Bertec®) across different slopes and speeds (N=20). The Kinetyx proprietary foot contact event detection algorithm showed mean absolute errors of only 13.6 ms and 9.1 ms for foot contact and foot off, respectively, when compared to a standard 20 N ascending and descending force threshold on the Bertec treadmill. This paper illustrates that from the Kinetyx proprietary foot contact event detection algorithm, Kinetyx is able to provide the means to produce laboratory quality temporal running gait metrics.

Background 

With the increasing availability of smart devices and wearable technology, many runners utilize the gait metrics provided by these technologies to better understand and guide their performance and training. Additionally, these running gait metrics can provide insights into injury rehabilitation or mobility research. Identification of foot contact events (FCEs) specifically, identification of the foot contact (FC) and foot off (FO) events permit the running gait cycle to be divided into phases allowing measurements and comparisons to be made across different running speeds and individuals [1]. In addition, correct identification of FCEs allows for the determination of temporal running gait metrics such as stride rate, ground contact time, and swing time. Accurate detection of FCEs is largely dependent on the algorithm used to process the measured signal as well as the sensitivity of the measurement equipment being used [2]. The most accurate method of determining FCEs currently employed by gait researchers is through the use of the vertical ground reaction force (vGRF) and is typically measured using in-ground force plates or force measuring treadmills [1; 2; 3; 4; 5]. However, detection of FCEs from vGRF requires the use of specialized lab based equipment which can be prohibitive to researchers and restricts measurements from fieldbase measurement scenarios. To address these limitations researchers have investigated the use of alternate technologies for their ability to accurately detect FCEs during running including the use of accelerometers, force sensing resistors (FSRs), and plantar pressure measurement systems (PPMS) [6; 7]. Among these technologies, PPMSs have the advantage of providing both kinematic and kinetic running gait data including the distribution of forces under the foot. This paper evaluates the Kinetyx proprietary FCE detection algorithm for PPMSs across multiple speeds and slopes and compares against the traditional VGRF methods on an instrumented treadmill.

Methods

Twenty recreationally active subjects participated in this study after providing written informed consent. The protocol was approved by the University of Calgary’s Conjoint Health Research Ethics Board (REB20-1734). Subjects ran on an instrumented treadmill (Bertec, Columbus, OH, USA) on level ground followed by uphill and downhill on a 10.5% grade. Each subject ran on level ground at 9.4, 10.8, 12.2 and 13.7 km/h, uphill at 9.4, 10.1, 10.8 km/h, and downhill at 9.4, 10.1, 10.8, and 12.2 km/h for one minute each. Subjects were not requested to run at faster speeds during non-level ground trials to avoid the potential of injury. Concurrent data was collected with a “gold standard” in shoe plantar pressure measurement system and the instrumented treadmill. FC events and FO events were computed for each stride using two methods. First, a common thresholding method was used on the instrumented treadmill with a 20 N ascending threshold crossing for FC and a 20 N descending threshold crossing for FO [8]. A value of 20 N was selected as the lowest possible threshold where signal noise due to belt vibration during swing phase could not be detected [9]. A novel FCE detection algorithm was developed using the Kinetyx sensory insoles. This FCE algorithm was developed specifically to be sensitive to signal onset and offset in different running conditions such as different speeds, grades and foot strike patterns. The mean absolute errors between the two methods for FCE detection were then computed to evaluate the validity of the Kinetyx proprietary FCE detection algorithm.

Results

The Kinetyx proprietary FCE detection algorithm estimated FCEs across multiple speeds and slopes for over 23,000 strides. This algorithm produced an average error of 13.6 ± 5.8 ms at FC and 9.1 ± 8.9 ms at FO across all speeds and slopes. For individual subjects, this error was as low as 7.8 ms at FC and 4.2 ms at FO (Figures 1, 2).

Foot contact event detection error by subject

Foot off event detection error by subject

Summary

The Kinetyx proprietary FCE detection algorithm had minimal error when compared to a 20 N ascending and descending force threshold, giving mean absolute errors of 13.6 ms and 9.1 ms for FC and FO, respectively. These errors represent a small portion of the stance phase in running where the average ground contact time was 265.9 ms for the participants of this study, calculated using the threshold method on the force sensing treadmill. This concedes an error of only 5% of average ground contact time at FC, and 3% at FO. The implementation of the Kinetyx proprietary FCE detection algorithm allows for an accurate method to detect FCEs in plantar pressure measurement systems. This would provide runners and researchers alike with precise FC and FO timings. These FCEs can be used to compute temporal running gait metrics such as stride rate, ground contact time, and swing time.

Acknowledgements

We would like to acknowledge the contributions of the High Performance Laboratory at the University of Calgary under the supervision of Dr. Benno Nigg for their assistance with data collection for this research.

References

[1] P. M. Mills, R. S. Barrett, and S. Morrison, “Agreement between footswitch and ground reaction force techniques for identifying gait events: Inter-session repeatability and the effect of walking speed,” Gait Posture, vol. 26, pp. 323––326, 2007.

[2] O. Tirosh and W. A. Sparrow, “Identifying heel contact and toe-off using forceplate thresholds with a range of digital-filter cutoff frequencies,” Journal of Applied Biomechanics, vol. 19, pp. 178––184, 2003.

[3] J. Leitch, J. Stebbins, G. Paolini, and A. B. Zavatsky, “Identifying gait events without a force plate during running: A comparison of methods,” Gait Posture, vol. 33, pp. 130—-132, 2011.

[4] R. E. Fellin, W. C. Rose, T. D. Royer, and I. S. Davis, “Comparison of methods for kinematic identification of footstrike and toe-off during overground and treadmill running,” Journal of Science and Medicine in Sport, vol. 13, pp. 646–650, 2010.

[5] J. Hausdorff, Z. Ladin, and J. Wei, “Footswitch system for measurement of the temporal parameters of gait,” Journal of Biomechanics, vol. 28, pp. 347––351, 1995.

[6] R. Mann, L. Malisoux, R. Brunner, P. Gette, A. Urhausen, A. Statham, K. Meijer, and D. Theisen, “Reliability and validity of pressure and temporal parameters recorded using a pressure-sensitive insole during running,” Gait Posture, vol. 39, pp. 455––459, 2014.

[7] R. Harle, S. Taherian, M. Pias, G. Coulouris, A. Hopper, J. Cameron, J. Lasenby, G. Kuntze, I. Bezodis, G. Irwin, and D. G. Kerwin, “Towards real-time profiling of sprints using wearable pressure sensors,” Computer Communications, vol. 35, pp. 650––660, 2012.

[8] J. C. Handsaker, S. E. Forrester, J. P. Folland, M. I. Black, and S. J. Allen, “A kinematic algorithm to identify gait events during running at different speeds and with different footstrike types,” Journal of Biomechanics, vol. 49, pp. 4128––4133, 2016. [9] B. D. Hendershot, C. E. Mahon, and A. L. Pruziner, “A comparison of kinematic-based gait event detection methods in a self-paced treadmill application,” Journal of Biomechanics, vol. 49, pp. 4146– 4149, 2016.

Appendix

The intellectual property contained in this document is protected by the following: System and method for analyzing force sensor data, US patent applications 63/291,424 and 63/315,847