Secondary Logo

Journal Logo

Effect of Four Different Step Detection Thresholds on Nonmotorized Treadmill Sprint Measurement

Cronin, John B.1,2; Rumpf, Michael C.1,3

Journal of Strength and Conditioning Research: October 2014 - Volume 28 - Issue 10 - p 2996–3000
doi: 10.1519/JSC.0000000000000497
Technical Report

Cronin, JB and Rumpf, MC. Effect of four different step detection thresholds on nonmotorized treadmill sprint measurement. J Strength Cond Res 28(10): 2996–3000, 2014—The purpose of this study was to investigate the effect of 4 different step detection thresholds (10, 15, 20, and 30% body mass [BM]) on the kinetics and kinematics of a youth population sprinting on a Woodway nonmotorized treadmill (NMT). A total of 16 male youth athletes sprinted 30 m from a split start position. Of the 15 variables measured, significant differences (p ≤ 0.05) were found in the measurement of 5 kinematic (step length, vertical displacement, contact time, eccentric, and concentric time) and 2 kinetic (vertical and leg stiffness) variables between the 10 vs. 20 and 30% BM step detection thresholds. Contact time was also significantly different (12%) between 15 vs. 30% BM step detection thresholds. In terms of reliability, the 15 and 30% BM step detection thresholds were found the most stable across all variables (average coefficient of variation ∼6.0%). Given this information, a step detection threshold of 15% BM is recommended for quantifying kinematic and kinetic variables on a NMT, as this threshold seems to account for signal variability appropriately without compromising reliability.

1AUT University, Sport Performance Research Institute New Zealand;

2School of Exercise, Biomedical and Health Sciences, Edith Cowan University, Perth, Australia; and

3National Sports Medicine Programme, Excellence in Football Project, Aspetar, Qatar Orthopaedic and Sports Medicine Hospital, Doha, Qatar

Address correspondence to John B. Cronin,

Back to Top | Article Outline


Running is a fundamental movement pattern of human beings and its analysis therefore of great interest to many professions. A variety of equipment has been used to quantify the kinematics and kinetics associated with running, including accelerometers (8), cameras (10,11,15), and force plates (2,5). In terms of data collection with force plates, sampling frequencies of 200–2000 Hz (2,5), cutoff frequencies of 30 (10,14) or 50 Hz (12) and absolute thresholds of 10–150 N (5,10,14) or relative thresholds of ∼10–20% body mass (BM) (11), or 1% of maximal vertical force values (∼35 N) (15) have been used to detect and analyze steps in adults. It seems a great variety of sampling rates, filtering methods, and step detection thresholds have been used when analyzing running gait in adults (10,11,14). Furthermore, the step detection threshold (20 N vs. 5 and 10 N) has been shown to impact gait results (6) by delaying the detection of foot strike by 1 frame. The ability to detect the beginning and end of the stance phase (step detection threshold) seems an important parameter to consider when analyzing gait and running data, as the threshold chosen can have substantial effects on the kinematic and kinetic description of the movement of interest.

Because it is thought that youth and in general younger and less mature individuals present greater variability in motor skills (4), it would seem important to establish reliable protocols to analyze running patterns in a youth population. This is especially so for young children, before their individual adolescent growth spurt, their low BM possibly providing the greatest challenge to determine the optimal step detection threshold for the analysis of running kinematics and kinetics. Given that the nonmotorized treadmill (NMT) can provide mechanistic information around gait, this would seem of particular prognostic and diagnostic value to youth populations where early detection of gait issues and long-term monitoring of performance can be impactful on athlete development. However, no studies to the knowledge of these authors have investigated the step detection thresholds associated with running on a NMT, and this is especially the case in a youth population. The purpose of this study, therefore, was to determine the effect of different step detection thresholds on the sprint kinematics and kinetics of youth sprinting on a NMT.

Back to Top | Article Outline


Experimental Approach to the Problem

Sixteen athletes sprinted 30 m 3 times at 4 different step detection thresholds (10, 15, 20, and 30% BM) on an NMT. The kinematic and kinetic variables from the fastest 4 consecutive steps in which the highest treadmill velocity was achieved were chosen and averaged for each of the 3 trials. Within trial reliability was quantified using a coefficient of variation. A repeated measures analysis of variance (ANOVA) was used to determine significant differences (p ≤ 0.05) between step detection thresholds on the averaged data.

Back to Top | Article Outline


Sixteen athletes (age: 10.5 ± 1.50 years; height: 142.7 ± 6.60 cm; mass: 34.8 ± 5.20 kg) participated in the study. The athletes were categorized as physically active and trained a minimum of 2 times per week in their sport and represented their school or club at a regional and/or state level. The participants and their parents/legal guardians were informed about the potential risk involved with the study and gave written consent and assent to participate. The ethics committee of AUT University approved this investigation.

Back to Top | Article Outline

Data Acquisition and Analysis

Before any physical testing, anthropometric measurements were taken. Participants then received a familiarization session on the NMT (Force 3 Woodway, Wheil am Rhein, Germany), which consisted of standing, walking, and running at a self-chosen speed. The familiarization was also used as a warm-up phase (∼5 minutes) if the participants felt safe and demonstrated the ability to sprint maximally on the treadmill. If the participants were unable to run maximally without holding on to the frame of the treadmill, the data collection was postponed to another day. Otherwise, a series of warm-up sprints on the treadmill involving 3 five-second sprints preceded the testing. Three sprints of 30 m from a split start (1 leg in front of the other on the treadmill) position were then performed on the NMT. The participants were allowed to recover completely (4 minutes) after each trial.

Variables of interest were collected using a sampling rate of 200 Hz using the Pacer Performance software (Fitness Technology, Joondalup, Australia) and analyzed with a custom-designed Matlab (MathWorks, Inc., Natick, MA, USA) program. Each channel of data was filtered using a zero-lag low-pass Butterworth filter that optimized the cutoff frequencies based on a residual analysis (16). Vertical and horizontal forces and running velocities were measured directly through 4 individual vertical load cells that were mounted underneath the running surface. A load cell (Modell BS-500 Class III; Transcell Technology, Inc., Buffalo Grove, IL, USA) attached to a tether measured the subject's horizontal force, and 2 optical speed photomicrosensors in conjunction with a tachometer XPV7 PCB (Fitness Technology, Adelaide, Australia) measured belt velocity. All other variables were derived from the force plates, horizontal load cell, and/or photomicrosensor inputs. Variables were calculated as follows:

  • Step length: difference in calculated distance from end to start of step detection threshold between alternating foot contacts.
  • Step frequency: number of steps per second.
  • Contact time: difference in time of start and end of step detection threshold.
  • Eccentric time: time from start of step detection threshold to zero vertical velocity, which was derived from the integral of vertical acceleration. The vertical acceleration was obtained from the peak vertical force divided by BM after subtracting gravitational acceleration (3).
  • Aerial time: time outside the step detection threshold.
  • Concentric time: time from zero vertical velocity to end of step detection threshold.
  • Vertical displacement: vertical displacement was determined by double integration of the vertical acceleration (3) in the eccentric phase. Vertical acceleration was obtained from the peak vertical force divided by BM after subtracting gravitational acceleration (3).
  • Vertical stiffness: maximum ground reaction force during contact divided by the vertical displacement of the center of mass (7).
  • Leg stiffness: maximum vertical force (Fmax) divided by the peak displacement ([INCREMENT]) of the initial leg length (L) (9), calculated from standing height minus sitting height:

The peak displacement of the initial leg length was calculated as:

where v = running velocity (in meter per second), tc = contact time (in seconds), and Δyc = the vertical displacement (in meters) of the center of mass when it reached its lowest point during mid-stance. Dimensionless variables for vertical and leg stiffness were derived from further multiplying the vertical stiffness with the initial leg length and then dividing the product by the participants' BM and gravitational acceleration of 9.81 (7).

  • Peak horizontal power: peak horizontal force × running velocity.
  • Mean eccentric power: mean power during eccentric time.
  • Mean concentric power: mean power during concentric time.

Steps were identified when the force was greater (foot touchdown) or less than (toe-off) 10, 15, 20, or 30% of the participant's BM (Figure 1). Data were separated into the right and left steps using the data from the individual force plates of the NMT. The fastest 4 consecutive steps in which the highest treadmill velocity was achieved were chosen and averaged for each of the 3 trials. The average data of the 3 runs were used for further statistical analyses.

Figure 1

Figure 1

Back to Top | Article Outline

Statistical Analyses

Mean and SDs were used as measures of centrality and spread of data. A repeated measures ANOVA (SPSS, 18.0; IBM, Armonk, NY, USA) with Bonferroni post hoc contrasts were used to determine whether there was a significant difference in the 4 different thresholds for each variable of interest. An alpha level of p ≤ 0.05 was chosen as the criterion for significance. Within-trial reliability of the 3 trials at the 4 different thresholds was quantified using a coefficient of variation (CV = SD/mean × 100).

Back to Top | Article Outline


In terms of the kinetic variables, vertical and leg stiffness were significantly (p ≤ 0.05) lower (28.4–45.0%), whereas step length, vertical displacement, contact time, eccentric, and concentric time were greater (−14.3 to −29.7%) when the 10% and 30% BM thresholds were compared (Table 1). Horizontal force and power, vertical force, and eccentric and concentric power were not significantly affected (0.13–14%) by the various step detection thresholds.

Table 1

Table 1

Regarding the kinematic factors, vertical displacement and eccentric time were also significantly greater (−14.3 to −22.4%) when the 10 and 20% BM thresholds were compared. Contact time was the only variable that was significantly different (12%) between the 15 and 30% BM comparisons.

Averaged CVs across all variables for each threshold (10, 15, 20, and 30% BM) were 10.3, 6.10, 8.00, and 5.90%, respectively. When CVs for each variable were averaged over all thresholds, only power, eccentric power, vertical, and leg stiffness were greater than 10%. Generally, the CVs improved (i.e., lesser value) as the threshold of detection increased (Table 1).

Back to Top | Article Outline


As the step forces of the youth athletes in this study did not return to zero at the completion of a step cycle, the use of a step detection threshold that enabled all steps to be detected while having minimal effect on the kinematic and kinetic outputs needed to be determined. To these ends, step detection thresholds of 10, 15, 20, and 30% BM were compared to determine whether these thresholds affected the variables of interest. Five kinematic variables (step length, vertical displacement, contact time, eccentric, and concentric time) and 2 kinetic variables (vertical and leg stiffness) were significantly influenced by the magnitude of the step detection threshold. Most of the observed differences were between the 10 and 30% BM thresholds of detection. It seemed that the kinematic variables (Table 1) were more affected by these thresholds as compared with the kinetic variables, especially those kinematic variables incorporating time during the stance phase in their calculation. Intuitively, this is to be expected, as the larger the threshold/cutoff the greater the loss of data in the force-time signal, which has been reported previously by other researchers (6).

To the best of the authors' knowledge, this is the first investigation comparing different step detection thresholds while using an NMT, and therefore, comparisons to other studies (2,5,10,11,14) are problematic. Our participants' average BM was 34.8 (±5.20 kg), therefore, the mean step detection thresholds were 34.6, 51.9, 69.3, and 97.1 N for 10, 15, 20, and 30% BM, respectively. Therefore, all our threshold values were within the range of total values of 10–150 N reported by other researchers (2,5,10,11,14) while analyzing participants running over in-ground force plates.

An analytical goal of the CV being 10% or below has been chosen arbitrarily by some scientists, but the merits of this value as a measure of reliability are the source of conjecture (1). Nonetheless, CVs for most of the variables averaged over all thresholds were equal or less than 10%, the exceptions being power, eccentric power, vertical, and leg stiffness. This high variability with power measures is similar to other research findings (13) and attributed to the greater sensitivity of power as the product of force and velocity. When all the CVs within 1 threshold were averaged and compared between thresholds, the 15 and 30% BM step detection thresholds were found to have the lowest variability (CV = 6.10 and 5.90%, respectively).

Back to Top | Article Outline

Practical Applications

Five kinematic and 2 kinetic variables were significantly different between the 10 vs. 20 and/or 30% BM step detection thresholds in this youth population sprinting 30 m on an NMT. Fifteen and 30% BM thresholds were found the most stable across all the variables of interest. The 15% BM threshold seemed to be the most appropriate step detection threshold, as it accounted for signal variability appropriately without compromising reliability. The 15% BM threshold would seem the threshold of choice to reflect the “real” values of the measurement of the variables of interest.

Back to Top | Article Outline


1. Atkinson G, Nevill AM. Statistical methods for assessing measurement error (reliability) in variables relevant to sports medicine. Sports Med 26: 217–238, 1998.
2. Butler RJ, Davis IS, Hamill J. Interaction of arch type and footwear on running mechanics. Am J Sports Med 34: 1998–2005, 2006.
3. Cavagna GA, Franzetti P, Heglund NC, Willems P. The determinants of the step frequency in running, trotting and hopping in man and other vertebrates. J Physiol 399: 81–92, 1988.
4. Gerodimos V, Zafeiridis A, Perkos S, Dipla K, Manou V, Kellis S. The contribution of stretch-shortening cycle and arm-swing to vertical jumping performance in children, adolescents, and adult basketball players. Pediatr Exerc Sci 20: 379–389, 2008.
5. Karamanidis K, Arampatzis A. Mechanical and morphological properties of different muscle-tendon units in the lower extremity and running mechanics: Effect of aging and physical activity. J Exp Biol 208: 3907–3923, 2005.
6. Leitch J, Stebbins J, Paolini G, Zavatsky AB. Identifying gait events without a force plate during running: A comparison of methods. Gait Posture 33: 130–132, 2011.
7. McMahon TA, Cheng GC. The mechanics of running: How does stiffness couple with speed? J Biomech 23: 65–78, 1990.
8. Mercer JA, Dufek JS, Mangus BC, Rubley MD, Bhanot K, Aldridge JM. A description of shock attenuation for children running. J Athl Train 45: 259–264, 2010.
9. Morin JB, Dalleau G, Kyrolainen H, Jeannin T, Belli A. A simple method for measuring stiffness during running. J Appl Biomech 21: 167–180, 2005.
10. Morio C, Lake MJ, Gueguen N, Rao G, Baly L. The influence of footwear on foot motion during walking and running. J Biomech 42: 2081–2088, 2009.
11. Riley PO, Dicharry J, Franz J, Croce UD, Wilder RP, Kerrigan DC. A kinematics and kinetic comparison of overground and treadmill running. Med Sci Sports Exerc 40: 1093–1100, 2008.
12. Riley PO, Paolini G, Della Croce U, Paylo KW, Kerrigan DC. A kinematic and kinetic comparison of overground and treadmill walking in healthy subjects. Gait Posture 26: 17–24, 2007.
13. Tong RJ, Bell W, Ball G, Winter EM. Reliability of power output measurements during repeated treadmill sprinting in rugby players. J Sports Sci 19: 289–297, 2001.
14. Weyand PG, Sandell RF, Prime DN, Bundle MW. The biological limits to running speed are imposed from the ground up. J Appl Physiol (1985) 108: 950–961, 2010.
15. Williams KR, Cavanagh PR. Relationship between distance running mechanics, running economy, and performance. J Appl Physiol (1985) 63: 1236–1245, 1987.
16. Winter DA. Biomechanics and Motor Control of Human Movement. Hoboken, New Jersey: John Wiley & Sons, 2005.

kinematics; kinetics; speed; vertical and leg stiffness

Copyright © 2014 by the National Strength & Conditioning Association.