Presented at the
23rd Annual Meeting of the

American Society of Biomechanics
University of Pittsburgh
October 21-23, 1999


Susan M. Bowley1, Gregory A. Breit2, and Robert T. Whalen1
Musculoskeletal Biomechanics Laboratory, NASA Ames Research Center, Moffett Field, California
2Qualcomm, Inc., San Diego, California
Email: Web:


The NASA GRF Activity Monitor is used to autonomously measure vertical ground reaction forces (GRFz) in human subjects during normal daily activity (Whalen et al., 1993; Breit and Whalen, 1994). The capacitance insole sensor (E.Q., Inc.) has measurement errors during dynamic human activity loading. Bowley et al., 1998 described a non-linear calibration method to correct for these errors. The objective of this study was to examine accuracy of the activity monitor in measuring peak GRFz during walking and running following non-linear corrections. A second objective was to determine accuracy in measuring contact time (tc) and stride period (T) with the activity monitor.


Twenty-three human subjects (12 females, Average Age = 42.6 13.3 yrs; 11 males, Average Age = 40.4 11.5 yrs) were asked to perform a series of walking (Range: 0.637 to 2.58 m/s) and running (Range: 1.43 to 6.71 m/s) gait cycles at a range of speeds. Tests were done over a calibrated force plate (AMTI) mounted outdoors and flush with a sidewalk. This study was approved by the NASA Ames Research Center Institutional Review Board.

Subjects wore the NASA GRF Activity Monitor during gait testing, with the capacitance insole sensor in their right shoe. No control was made for the type of shoe each individual wore, although most wore tennis shoes. The insole sensor was calibrated using our non-linear calibration method prior to performing the walking and running cycles (Bowley et al., 1998).

The NASA GRF Activity Monitor containing onboard sampling and filtering algorithms (Breit et al., 1994) was carried in a fanny pack. Sampling rate for the activity monitor was 100Hz and 200Hz for the force plate. Force plate and activity monitoring data were collected autonomously. Data between monitoring systems were synchronized by the time of occurrence of a single body weight (BW) measurement event at the start of testing. Ground contact was determined from a threshold of 10% BW at initial footfall and 20% BW at lift-off on the falling edge. Impulse momentum equations were used to estimate T from force plate data.


The filtered GRFz peaks, contact time and stride period stored in the activity monitor compared well with force plate values (Figs 1-3 and Table 1).

Figure 1: Comparison of corrected insole sensor filtered peaks and force plate peak GRFz generated for walking and running in 23 human subjects. Mean RMS error over all peaks was 0.185 BW (N = 980, R^2 = 0.929).

Figure 2: Comparison of insole and force plate contact times for walking and running at a range of speeds in 23 human subjects. Mean RMS error over all peaks was 0.036 seconds (N = 504, R^2 = 0.978).

Figure 3: Comparison of insole and force plate stride periods for walking and running at a range of speeds in 23 human subjects. Mean RMS error was 0.079 seconds (N = 504, R^2 = 0.898).

Table 1: Results in RMS error found by comparing insole and force plate peak GRFz, contact time (tc), and stride period (T).


Using these data we intend to develop further correction methods to account for the observed error in insoles sensor values filtered and stored in the NASA GRF Activity Monitor. We are also interested in simplified activity monitoring techniques using tc and T as predictions and/or corrections on stored GRF data (Breit and Whalen, 1997).


Whalen et al. (1993). ASME BED, 26:535-538.
Breit and Whalen (1994). ASB, 231-232.
Bowley, Breit and Whalen (1998). NACOB '98, 169-170.
Breit and Whalen (1997). Med. Sci. Sports Exerc., 29(4):540-547.


We thank all our volunteers from the NASA Ames Research Center who generously agreed to participate in this study. NASA Grant #199-26-12-35 supported this work.