Variability in the Length and Frequency of Steps of Sighted and Visually Impaired Walkers Sarah J. Mason, Gordon E. Legge, and Christopher S. Kallie
Abstract: The variability of the length and frequency of steps was measured in sighted and visually impaired walkers at three different paces. The variability was low, especially at the preferred pace, and similar for both groups. A model incorporating step counts and step frequency provides good estimates of the distance traveled. Applications to wayfinding technology are discussed.
Sighted people often estimate distances estimates of distance should take into ac- by counting their steps (for example, to count changes in walking speed. Third, it approximate the dimensions of a large is often impractical to count steps accu- room). Such estimates implicitly rely on rately, given the concurrent cognitive de- the consistency of the length of one’s mands of other activities. Ramsey, steps. In wayfinding situations, where es- Blasch, Kita, and Johnson (1999) found timates of distance may be useful, people that visually impaired cane users de- who are visually impaired (that is, those creased the length of their stride when who are blind or have low vision) may not they had to concurrently complete an au- make use of step counting for several rea- ditory task or anticipate a simulated sons: First, it is possible that the length of drop-off. Finally, the practical value of steps is less consistent for visually im- estimates of distance relies on a map that paired people because of the use of mo- converts these distances into useful infor- bility devices. Second, since walking mation about the layout of the environ- speed affects the length of steps, accurate ment. In an unfamiliar environment, a person who is visually impaired rarely has access to a suitable map. In a familiar Many thanks to members of the Minnesota environment, he or she may conceivably Laboratory for Low Vision Research for their memorize a cognitive map on the basis of assistance, especially Rudrava Roy for writing the software program interfaced to the pe- step counts (for example, Dr. Walker’s of- dometer system and MiYoung Kwon for help fice door is on the left, 30 steps down the in preparing the manuscript. We also thank corridor from the main entrance). Scott Kalpin for his helpful discussion and Currently, there is wide interest in the for supplying the computer-readable pedome- development of assistive technology for ter. This research was funded by Grant wayfinding by people who are visually EY02857 from the National Institutes of Health and Grant H133A011903 from the impaired. It has been demonstrated that National Institute of Disability and Rehabili- signals from global positioning satellites tation Research. (GPS) can be incorporated into wayfind-
©2005 AFB, All Rights Reserved Journal of Visual Impairment & Blindness, December 2005 741 ing technology for use outdoors (Loomis, In principle, such technology would Golledge, & Klatzky, 1998; Loomis, take care of two of the problems just Golledge, Klatzky, Speigle, & Tietz, listed regarding step counting for the 1994). For instance, GPS-based software nonvisual estimation of distance: The has been developed by the Sendero technology, rather than the pedestrian, Group and runs on the BrailleNote re- would count the steps and would convert freshable braille notetaker (by Pulse Data the step counts into an estimated position HumanWare). This technology updates a on a map of the building. This article re- position continuously, so there is no need ports on the results related to the two re- to estimate distances by counting steps. maining issues: How accurately can dis- Because GPS signals are usually not tances be estimated from step counts, and available indoors, another method for how is this accuracy affected by a per- tracking a pedestrian’s position is neces- son’s walking pace? sary. Stable landmarks, such as the loca- Prior research has found that the tion of the elevators, or special-purpose length of the stride of healthy, sighted markers, such as the Talking Signs In- adults remains fairly consistent over frared Communications System (by Talk- time, especially at an individual’s pre- ing Signs), may provide information on ferred walking pace. The participants in one’s position at key points in a building. Sekiya, Nagasaki, Ito, and Furuna’s Technology for estimating distance and (1997) study walked on a flat walkway direction from these landmarks could with a freely chosen step rate at five dif- then be useful. As part of a larger project ferent self-regulated speeds. The length on indoor wayfinding technology at the of their strides was consistent (variable Minnesota Laboratory for Low Vision error < 4 centimeters, or 1.6 inches), par- Research at the University of Minnesota, ticularly at the preferred walking speed we have been exploring technologies for (variable error < 2.5 centimeters, or 0.9 estimating the distance and direction inch). traveled by pedestrians who are visually Although Danion, Varraine, Bonnard, impaired. A straightforward method for and Pailhous (2003) did not find that the estimating distance would be to count preferred walking speed necessarily mini- steps with a computer-readable pedome- mized variation in the length of strides, ter. If the step counts could be reliably they found that human gait was consis- converted into distance estimates and if tent. Their participants walked on a the pedometer was interfaced to a treadmill at 25 combinations of fre- computer-readable digital map of a quency and step length (imposed by the building, then a pedestrian’s current po- experimenter). The variability in the sition in the building (computed as an length of strides (expressed as a coeffi- estimated distance from a known land- cient of variation, SD/M)was generally mark) could be determined. (More re- under 3%. Stride length was found to be fined technology would be necessary to most consistent at the frequency of 1Hz estimate the direction of travel. In build- (60 strides per minute), more variable for ings, however, direction is often highly shorter strides, and more consistent for constrained by the layout of corridors.) longer strides.
742 Journal of Visual Impairment & Blindness, December 2005 ©2005 AFB, All Rights Reserved To our knowledge, no study on vari- tween the placements of one foot), we ability in the length of steps has been car- chose to count steps (the distance in the ried out using participants who are visu- forward direction between alternate feet). ally impaired. One could hypothesize that We chose this measure to maintain con- the lack of visual cues may increase vari- sistency with those parts of the experi- ation in a visually impaired individual’s ment that used the pedometer, which walking pattern. Nakamura (1997) found counted steps, not strides. that the length of the stride of adults who Experiment 2 was conducted two are blind who walked at their preferred months later, with four of the participants pace was significantly shorter than that from Experiment 1. The participants were of sighted adults, possibly because of the again asked to walk at a slow, preferred, need to integrate nonvisual input to walk and fast pace, but this time on walkways safely. Considering that Danion et al. that were 20, 40, and 80 feet long. Experi- (2003) found sighted walkers’ shorter ment 2 had two aims. The first was to de- strides to be more variable in length, termine how reproducible each partici- there is a possibility that the short strides pant’s mean step length was (especially at found in people who are visually im- the preferred pace, at which a pedometer paired may also show increased variabil- would be calibrated). The second was to ity. Studies on veering (for example, Guth compare the accuracy of two methods of & LaDuke, 1995) have concluded that estimating the distance traveled. Previous people who are visually impaired often research has found that the relationship have difficulty maintaining a straight- between length and frequency of steps is ahead direction when walking; perhaps linear (Inman, Ralston, & Todd, 1981; attaining a consistent step length may Danion et al., 2003). Instead of calibrat- also prove difficult for them. ing a pedometer with an individual’s pre- Experiment 1 consisted of 18 partici- ferred step length at the preferred pace, a pants (6 sighted and 12 visually im- more accurate way of estimating the dis- paired) walking on a flat walkway 80 feet tance traveled may be to calibrate by mea- long. Eight trials were performed at each suring the linear relationship between the of three paces (slow, preferred, and fast), length and frequency of steps for an indi- from which variability in the length and vidual. Distance would then be estimated frequency of steps was calculated. Each from a linear equation that takes step fre- participant then walked an additional quency into account in estimating step eight trials at his or her preferred pace length. For the four participants who re- wearing a computer-readable pedometer, turned for Experiment 2, we calculated to test the pedometer’s accuracy. Com- their mean preferred-pace step length bining each participant’s individual vari- from Experiment 1 and the linear equa- ability in step length with the pedometer’s tion that best represented the relationship error rate gave us an estimation of the ac- between the length and frequency of steps curacy of a pedometer-based system for for each trial from Experiment 1. We then estimating walking distances. used both methods to estimate the dis- Although prior studies, cited earlier, tance traveled in Experiment 2, to com- have measured strides (the distance be- pare the accuracy of each method.
©2005 AFB, All Rights Reserved Journal of Visual Impairment & Blindness, December 2005 743 Our overall aim was to compare the ally impaired (with Snellen acuities rang- variability in the length and frequency of ing from 20/180 to total blindness). The steps between participant groups and to visually impaired participants consisted answer the question, “Do participants of a younger group (Participants V1–V6, who are visually impaired tend to show mean age 27.3 years) and an older group greater variability in walking patterns?” (Participants V7–V12, mean age 56.5 Other comparisons also interested us: years). Four used dog guides, 4 used Would the young visually impaired par- canes, and 4 used no mobility devices. All ticipants show more or less variability the visually impaired participants were than the older visually impaired partici- independent and confident travelers (see pants, and would the type of mobility de- Table 1 for characteristics of all the par- vice (dog guide or white cane) used by the ticipants). participants be an influencing factor? DESIGN Experiment 1: Method The dependent variables that were mea- PARTICIPANTS sured were the time taken to walk the 80- Of the 18 participants, 6 had typical vi- foot length of hallway and the number of sion (Snellen acuity 20/30 or better) and a steps taken. The independent variable, mean age of 24.5 years, and 12 were visu- manipulated by the experimenter, was the
Table 1 Characteristics of the sighted (S) and visually impaired (V) participants. Mobility Participant Age Gender device Acuity Diagnosis Age of onset Sighted S1 25 M NA 20/20 NA NA S2 19 M NA 20/20 NA NA S3 23 M NA 20/20 NA NA S4 23 F NA 20/20 NA NA S5 25 F NA 20/20 NA NA S6 32 F NA 20/30 NA NA Visually impaired V1 30 M Dog 20/1500 Leiber’s disease Birth V2 32 F Cane No vision Congenital glaucoma Birth V3 26 F None 20/180 Congenital aniridia Birth V4 24 F Cane 20/1500 Optic nerve hypoplasia Birth V5 26 M Dog 20/1200 Congenital glaucoma Birth V6 26 F Dog No vision Diabetic retinopathy with glaucoma 18 V7 53 F None 20/400 Optic atrophy Birth V8 62 F Cane No vision Retinopathy of prematurity Birth V9 57 F Cane No vision Retinopathy of prematurity Birth V10 57 F Dog No vision Retinopathy of prematurity Birth V11 56 M None 20/1000 Corneal vascularization secondary 6 to Stevens-Johnson syndrome V12 54 M None 20/200 Macular degeneration 42
Note: NA = not applicable.
744 Journal of Visual Impairment & Blindness, December 2005 ©2005 AFB, All Rights Reserved pace of walking: the participant’s pre- the participants were asked to keep as ferred pace, slow pace (72 steps per closely as possible to the beat, the experi- minute), and fast pace (144 steps per menter emphasized that maintaining a minute). To choose these particular natural walking style was more impor- paces, we started with Danion et al.’s tant than was staying exactly with the (2003) slowest and fastest frequencies metronome. It was suggested to the par- (76.8 and 151.2 steps per minute) and ticipants that while walking at the slow then modified them after testing them pace, they should imagine strolling with the visually impaired members of through a park and while walking at the our laboratory, reducing each slightly. fast pace, they should imagine walking The independent participant variables quickly to catch a bus. The participants that were used for the analysis included were urged to walk only at a speed that age (younger versus older group), visual felt safe to them. If they did not feel com- status (sighted versus visually impaired), fortable walking at the fast pace set by the and type of mobility device used (dog metronome, it was turned off, and they guide, cane, or none). walked as quickly as they felt comfort- able. (Note that the participants matched Procedure the metronome frequency fairly well; the An overview of the study was given to the mean slow pace maintained by the partic- participants, followed by an acuity test ipants was 76.04 steps per minute, while (when applicable). The participants read the mean fast pace was 136.5 steps per and signed a consent form that was ap- minute. Two of the participants who were proved by the University of Minnesota In- visually impaired asked to have the stitutional Review Board for the Use of metronome turned off at the fast pace.) Human Subjects, outlining the possible Twenty-four trials were completed by risks of participating in the study. An 80- each participant; 8 at each of the three foot track was marked with electrical tape paces. Before each set of 8 trials began, in a quiet hallway (Elliott Hall, University each participant was asked to walk up of Minnesota). The participants were and down the hallway at the relevant taken to the start point of the track and speed to become accustomed to the re- stood with their toes on the start line. For quired pace. If the participant used a mo- each trial, timing began when a participant bility device in his or her daily life, the began walking and stopped as he or she participant used the same device during crossed the tape marking the end of the the study. The experimenter walked a few track. The experimenter also recorded the paces behind, giving the participant number of steps taken to walk the track. plenty of room to walk in a natural man- Three walking speeds were used during ner. If anyone approached the participant the trials: the participant’s preferred pace, during the trial, the participant was slow pace (72 steps per minute), and fast stopped and the trial was restarted. pace (144 steps per minute). For the slow After these trials were completed, a and fast paces, a metronome set the computer-readable pedometer was clipped tempo for walking and remained audible onto the participant’s waistband at the hip for the duration of the trials. Although and connected via a cable to a laptop com-
©2005 AFB, All Rights Reserved Journal of Visual Impairment & Blindness, December 2005 745 puter that was carried by the experi- recorded as 2 feet. When we refer to the menter. The cable was long enough so that means and standard deviations of step the participant’s walking pace should not lengths, we are referring to the means and have been influenced by the experimenter’s standard deviations of these estimates presence. Eight trials were performed at across a number of 80-foot trials (eight the participant’s preferred walking pace: trials in Experiment 1), not to measure- four with the pedometer clipped onto the ments of individual steps within trials. participant’s right hip and four with it Similarly, the frequency of steps was clipped onto the participant’s left hip. Two computed as the number of steps taken variables were measured: the number of to walk 80 feet, divided by the total time. steps taken to walk the 80-foot track, as Second, variability for each participant recorded by the pedometer, and the num- was calculated as a coefficient of variation ber of steps taken to walk the same track, (SD/M for the eight trials of each recorded by the experimenter. (The results pace) and is expressed as a percentage. The of numerous pilot studies convinced us percentage variability was found for each that the experimenter accurately and con- participant (within-subject variability), sistently counted the number of steps and then the overall means for each group taken by the participants.) were calculated using the appropriate par- ticipants. Third, note that the lengths of MATERIALS steps are stated in feet, and the frequencies The pedometer we used was an AME of steps are stated in steps per minute. Micro Pedometer (V1.0), manufactured by Fourth, comparisons among the follow- Advanced Medical Electronics Corpora- ing groups were made using an indepen- tion in Minneapolis, interfaced via a serial dent samples t-test: sighted and visually connection to a laptop computer carried impaired; sighted and young visually im- by the experimenter. A software program paired; young visually impaired and older written in Python script (by Rudrava Roy, visually impaired; men and women; and University of Minnesota) kept track of the dog users, cane users, and those who used pedometer’s step counts and converted no mobility device. An alpha level of 0.007 these counts to an online estimate of the was selected, on the basis of a Bonferroni distance traveled by the pedestrian. This correction of a p-value of .05 for seven information was displayed on the laptop’s comparisons. Pairwise tests of this sort screen for the experimenter. were conducted, rather than an omnibus analysis of variance (ANOVA), because we RESULTS OF EXPERIMENT 1 did not have a crossed design suitable for The reader should keep the following an ANOVA. points in mind when reviewing the re- sults. First, the length of an individual’s Mean lengths and frequencies of steps step on a given trial was computed as the The male participants had significantly distance walked (80 feet in Experiment longer step lengths at the fast pace (M = 1) divided by the number of steps taken. 2.86 feet) than did the female participants For instance, if the participant took 40 (M = 2.44 feet, t [16] = 3.402, p < .005). steps to travel 80 feet, the step length was There were no other significant group dif-
746 Journal of Visual Impairment & Blindness, December 2005 ©2005 AFB, All Rights Reserved Table 2 Variability in the length and frequency of steps at three paces: Coefficient of variation (percentage). Walking pace Variability Slow: M (SE) Preferred: M (SE)Fast: M (SE) Step length 2.74 (0.31) 1.94 (0.13) 2.02 (0.12) Step frequency 1.96 (0.17) 2.05 (0.18) 2.42 (0.19)
Note: Variability is expressed as the percentage coefficient of variation (SD/M). N = 18. ferences in either the length or frequency age. Error rates are shown for both the of steps. sighted and visually impaired partici- pants plus an overall average. No signifi- Variability in the length and frequency of steps cant differences in error rates were found Taking the mean of all 18 participants, we between the sighted and visually im- found that the variability in both the paired groups or among the dog users, length and frequency of steps was low (see cane users, and those who used no device Table 2 for the means and standard er- (when the pedometer was worn on the rors). The variability in step length was the right hip or left hip or for the overall aver- lowest at the preferred pace (M = 1.94%, age). The error rates indicate an overesti- SE = 0.13%), while the variability in step mation because the pedometer picked up frequency was the lowest at the slow pace extraneous movement. Note that the pe- (M = 1.96%, SE = 0.17%). When we com- dometer was more accurate when it was pared the variability among the groups, we worn on the right hip. We will return to found no significant differences. We also this difference in the Discussion section. compared the variability between the sighted and functionally blind participants Relationship between length (that is, removing the participants with and frequency of steps low vision from the analysis) and again For the four participants who later returned found no significant difference. for Experiment 2, the step length (x- axis) and step frequency (y-axis) for each Accuracy of the pedometer trial were plotted, plus the straight line that Table 3 displays the percentage of step- best fit the data. The linear equations de- detection errors (error rates) from the pe- scribing the lines were used in Experiment 2 dometer when it was worn on the right as an alternative method to predict the dis- hip and the left hip and an overall aver- tance traveled. Figure 1 displays the graphs,
Table 3 Accuracy of the pedometer (percentage). Pedometer location Participant group Right hip Left hip Mean Sighted group 4.55 7.90 6.23 Visually impaired group 2.70 13.79 8.24 Combined mean for both groups 3.31 11.82 7.57
Notes: Values show the percentage overestimate of the pedometer’s step count compared to the experimenter’s step count. N = 18.
©2005 AFB, All Rights Reserved Journal of Visual Impairment & Blindness, December 2005 747 linear equations, and correlation coeffi- to participate. Two were sighted (Partici- cients for each of the four participants. pants S3 and S5), and two were visually (Note that the r2 value is low for Participant impaired (V9 and V11). S5 because she unexpectedly decreased the The same hallway was used in this ex- length of her step at the fast pace.) periment as in Experiment 1. Three tracks were marked on the floor: 20 feet, 40 feet, Experiment 2: Method and 80 feet. The participants were asked PARTICIPANTS AND PROCEDURE to walk these tracks at one of three paces: Four participants from Experiment 1 re- slow, preferred, or fast. This time, the turned after approximately two months metronome was not used to set the slow
Figure 1. Graph depicting the relationship between the length and frequency of steps for the four participants in Experiment 2.
748 Journal of Visual Impairment & Blindness, December 2005 ©2005 AFB, All Rights Reserved and fast paces; the participants chose From these data, we were able to assess their own interpretation of these paces. In whether the percentage of the variability the 36 trials that were completed, each in the length of steps was constant with combination of distance and speed was the distance or whether it increased or de- repeated four times. Trials were arranged creased as the participants walked far- in six blocks; each block was at one of the ther. An inspection of the data suggested three speeds and included two trials at a trend toward decreasing variability with each of the three distances. The six blocks increasing distance. However, an were arranged in a different order for each ANOVA showed that the variability in participant so as to minimize the effects of step length did not vary significantly with order. As in Experiment 1, the experi- distance. We conclude that for our partic- menter recorded the number of steps and ipants and for the range of distances that the time taken to walk the track. we tested, there was no systematic change in the percentage of variability across dis- RESULTS OF EXPERIMENT 2 tances. Stated another way, the variability Repeatability of the length and frequency in the number of steps required to cover a of steps over time given distance scales in proportion to the The mean percentage changes (from Ex- distance. periment 1 to Experiment 2, calculated as absolute differences) in the length and fre- Use of a linear equation to predict quency of steps at the three paces for the the distance traveled four participants were calculated. The Two methods were used to calibrate for overall mean percentage change for the en- the length of steps in estimating the dis- tire group was as follows. At the slow pace, tance traveled in the Experiment 2 trials the percentage change was comparatively on the basis of data that were recorded high (step length: 8.5%, frequency: 23.3%). for each participant in Experiment 1. At the preferred and fast paces, the per- Calibration Method 1 (simple). The centage changes were lower: preferred pace mean preferred step length from Experi- (length: 5.4%, frequency: 5.2%) and fast ment 1 was multiplied by the number of pace (length: 5.5%, frequency: 6.7%). steps taken in each trial in Experiment 2 to estimate the distance that was traveled. Variability in step length Calibration Method 2 (linear). A lin- As in Experiment 1, the variability in the ear equation was calculated for each par- length of steps was calculated as a coeffi- ticipant that best represented the rela- cient of variation (SD/M). The overall tionship between the length and variability in the mean step length (plus frequency of steps for each trial in Ex- SE)for each walking pace is listed here periment 1. Using this equation, the fre- for the 20-, 40-, and 80-foot distances, re- quency of steps in each trial in Experi- spectively: slow pace (4.1% [1.600], 3.1% ment 2 was used to select the appropriate [0.11], and 2.9% [0.007]); preferred pace step length. This step length, together (4.6% [0.016], 5.8% [0.013], and 3.4% with the number of steps taken, was used [0.006]); and fast pace (3.2% [0.018], 2.0% to estimate the distance that was trav- [0.011], and 1.9% [0.010]). eled.
©2005 AFB, All Rights Reserved Journal of Visual Impairment & Blindness, December 2005 749 Table 4 and Figure 2 compare the per- and frequency of steps for the partici- centage errors in the calculation of dis- pants who were visually impaired and to tance using both methods, with Method 2 compare the results to those for the (linear) depicted by triangles in Figure 2, sighted participants. Would the visually showing the smallest percentage error at impaired participants show greater vari- all the walking paces. ability in their walking patterns? Would any subgroup within the visually im- Discussion paired group show marked differences Our primary aim was to study the length from the rest of the group?
Figure 2. A comparison of the accuracy of distance prediction using Method 1 (simple) and Method 2 (linear) for the four participants in Experiment 2 at each of the three walking paces. The diagonal lines represent a perfect prediction of distance; the closer the symbols are to this line, the more accurate the prediction.
750 Journal of Visual Impairment & Blindness, December 2005 ©2005 AFB, All Rights Reserved Table 4 was low. For step length, our results mir- Mean percentage error in estimates of distance rored those of Sekiya et al. (1997) for for the four participants in Experiment 2. sighted participants. In our study, the Walking pace least variability was found at the pre- Method Slow Preferred Fast ferred pace (M = 1.94%). However, little Method 1 (simple) 25.84 14.38 7.97 variability was found at either the slow Method 2 (linear) 5.84 5.58 5.59 or the fast pace (M = 2.74% and 2.02%, Note: Values show the comparison between Methods respectively). Considering that the par- 1 and 2 at three walking paces. ticipants were allowed to walk freely at their preferred pace, variability of less MEASURES OF THE LENGTH than 2% would seem to indicate that AND FREQUENCY OF STEPS human walking is highly consistent in a The only significant difference between controlled indoor environment. Like- the participant groups for either the wise, the overall variability in the fre- length or frequency of steps was gender. quency of steps was low, with the least The men had significantly longer step variability found at the slow pace (M = lengths at the fast pace, probably because 1.96%) and increasing slightly at the pre- they were generally taller and could ex- ferred (M = 2.05%) and the fast (M = tend their step length at faster frequen- 2.42%) paces. Although the participants cies, whereas the women reached a step- walked under controlled conditions to length ceiling. Unlike our findings, ensure safety, walking in a corridor Nakamura (1997) found that the blind (rather than on a treadmill or in a virtual participants in his study had shorter environment) helped create as natural an stride lengths and slower walking speeds environment as possible from which to than did the sighted participants. He interpret these results. suggested that the “slow, guarded gait” of the blind participants might have been COMPARING VARIABILITY BETWEEN due to anxiety that was associated with THE GROUPS walking without their customary devices. No significant differences between the We found no significant difference be- groups were found. We acknowledge tween the stride length of the visually that our participant groups were small impaired and sighted participants, possi- and that Type II errors could be possi- bly because all the participants in our ble. We did note that those who used dog study were confident travelers who were guides showed slightly more variability using their usual mobility devices. at the slow pace that appeared to con- fuse the dogs. However, since no signifi- VARIABILITY IN THE LENGTH cant differences were found, we conclude AND FREQUENCY OF STEPS that under these experimental condi- Taking the mean for all the participants tions, the participants who were visually (both those who were sighted and those impaired walked as consistently as did who were visually impaired) into ac- the sighted participants and that their count, we found that the variability in mobility devices did not appear to affect both the length and frequency of steps the variability.
©2005 AFB, All Rights Reserved Journal of Visual Impairment & Blindness, December 2005 751 ACCURACY OF THE PEDOMETER nology. We recognize that these results The pedometer manufacturer indicated apply to the particular pedometer that we that the pedometer should be accurate to used for our study and might not general- within 10%, and our results support this ize to other pedometer designs. Also, if claim. When we combined the results this style of pedometer was incorporated from both the visually impaired and into a travel device, it would be necessary sighted groups and for the measures to discount spurious counts that were taken from both hips, we found that the registered when the participants turned overall accuracy was 7.57%. For 15 of the or shifted their weight from foot to foot. 18 participants, the pedometer gave a This might be done by some automatic more accurate step count when it was detection scheme or by giving the user a worn on the right hip. Although uncon- Pause button. firmed, there is a possibility that the cable extending from the pedometer to the lap- REPEATABILITY OF THE LENGTH top was more prone to interference from AND FREQUENCY OF STEPS OVER TIME the participant’s swinging arm when the How stable are the characteristics of walk- pedometer was worn on the left hip be- ing over time? We addressed this question cause the cable extended from the right by bringing back four participants after side of the laptop. If we used data col- two months to take part in Experiment 2. lected from the right hip only, accuracy At the preferred pace, the percentage improved to 3.31%. Since step-count er- changes for both the length and frequency rors were due to the pedometer picking of steps from Experiment 1 to Experiment up extraneous movement, errors would 2 were slightly higher than 5%, higher than give an overestimation in the steps taken the estimated variability of about 2% from and therefore the distance traveled. For Experiment 1. However, even in the trials example, in the 80-foot hallway, a 3.31% in Experiment 2, we noted somewhat overestimation would give a pedometer higher within-subject variability for the readout of approximately 82.6 feet. Indi- length of steps at the preferred pace: a vidual variability in the length of steps at mean of 4.6% in Experiment 2 versus a the preferred pace (M = 1.94%) over the mean of 1.94% in Experiment 1. This dif- same distance would result in an over- or ference was probably due to differences in underestimation of approximately 1.6 the experimental design; in Experiment 1, feet. Combining the pedometer error all the trials of a single pace were con- with individual variability over 80 feet ducted as a block, whereas in Experiment would still place an individual within ap- 2, the trials of different walking paces proximately 4.5 feet of his or her target were interspersed. Considering that each destination. In most cases, this accuracy trial would exert some influence on the would be adequate to find the correct next, it is not surprising that individual door to a room on a long corridor. This variability in the length and frequency of simple example, derived from our data, steps increased. However, over the 80-foot shows that estimates of distance, based hallway, even a 5% error rate would result on pedometer tracking, could be a useful in only a 4-foot over- or underestimation component of indoor wayfinding tech- of the distance traveled.
752 Journal of Visual Impairment & Blindness, December 2005 ©2005 AFB, All Rights Reserved VARIABILITY IN THE LENGTH OF STEPS at the slow and fast paces. Our results OVER DIFFERENT DISTANCES from Experiment 2 showed that this was During Experiment 2, we collected data the case. We conclude that a system that from three different distances and walk- records both step counts and step fre- ing paces and could calculate whether the quency, together with a linear calibration variability in the length of steps was con- model, makes substantially better predic- stant with distance or whether it in- tions of the distance traveled than does a creased or decreased. As was noted in the system that relies on step counting alone. Results of Experiment 2, it appears that variability decreased slightly as the dis- Conclusions tance increased, although an ANOVA Our results support those of prior stud- showed that the variability in the length ies, indicating that human gait (as de- of steps did not vary significantly with scribed by the length and frequency of distance and therefore that the error in steps) appears to be consistent across tri- this case was generally a fixed percentage als and days, typically close to a 5% vari- of the distance that was walked. So if the ability or less. We also conclude that, in percentage error is constant with distance general, persons who are visually im- (absolute error proportional to distance), paired show no greater variability than we would expect approximately an 8-foot do sighted persons when walking in a error for a trip of 160 feet in a long hall- controlled indoor environment, such as a way. corridor. Finally, using the linear rela- tionship between the length and fre- USE OF A LINEAR EQUATION TO PREDICT quency of steps provides an improved THE DISTANCE TRAVELED method for estimating the distance that is Before an individual uses a pedometer to traveled. estimate the distance traveled, he or she must calibrate it with his or her average References (preferred) step length. When walking at Danion, F., Varraine, E., Bonnard, M., & a preferred pace, this step length may Pailhous, J. (2003). Stride variability in provide a viable figure from which to esti- human gait: The effect of stride frequency mate distance when it is multiplied by the and stride length. Gait and Posture, 18, number of steps taken. However, when 69–77. walking at a slow or a fast pace, this fig- Guth, D., & LaDuke, R. (1995). Veering by blind pedestrians: Individual differences ure would over- or underestimate the dis- and their implications for instruction. Jour- tance traveled because of the variation of nal of Visual Impairment & Blindness, 89, the length of steps with the frequency of 28–37. steps. Considering that most participants Inman, V. T., Ralston, H. J., & Todd, F. showed a strong linear relationship be- (1981). Human walking. Baltimore, MD: tween the length and frequency of steps, Williams & Wilkins. Loomis, J. M., Golledge, R. G., & Klatzky, we wondered whether accounting for fre- R. L. (1998). Navigation system for the quency in the step-length calibration blind: Auditory display modes and guid- would provide a substantially better indi- ance. Presence: Teleoperators and Virtual cator of the distance traveled, especially Environments, 7, 193–203.
©2005 AFB, All Rights Reserved Journal of Visual Impairment & Blindness, December 2005 753 Loomis, J. M., Golledge, R. G., Klatzky, Sekiya, P. T., Nagasaki, H., Ito, H., & Fu- R. L., Speigle, J. M., & Tietz, J. (1994). Per- runa, T. (1997). Optimal walking in terms sonal guidance system for the visually im- of variability in step length. Journal of Or- paired. Proceedings of the First Annual thopaedic & Sports Physical Therapy, 26, ACM/SIGGAPH Conference on Assistive 266–272. Technologies (pp. 85–91). New York: Asso- ciation for Computing Machinery. Nakamura, T. (1997). Quantitative analysis Sarah J. Mason, B.A., research assistant, University of Minnesota; 75 East River Road, Minneapolis, of gait in the visually impaired. Disability MN 55455; e-mail:
The New York Institute for Special Education Founded in 1831 as The New York Institute for the Education of the Blind
During the 2006/2007 school year, The New York Institute for Special Education will be celebrating its 175th Anniversary. Through its doors have been many leaders in the field of blindness and other disabilities. If you are a former student, parent, employee, intern, student teacher, consultant, or were involved in any other capacity we would like to hear from you.
Drop us a note via email or regular mail, and let us know what you are doing now, how and when you were involved with the school, and most importantly, what The New York Institute means to you. Please send correspondence to [email protected], or via regular mail to NYISE, 999 Pelham Parkway, Bronx, NY 10469, Attn: 175th Anniversary. Please be sure to include your name, address, telephone number (s), email address and also the years and positions involved at the Institute. Thank you and we hope to hear from you.
754 Journal of Visual Impairment & Blindness, December 2005 ©2005 AFB, All Rights Reserved