Analyzing the Transfemoral Amputee Gait using Inertial Sensors

Identifying Gait Parameters for Investigating the Symmetry of Gait - A Pilot Study

Katja Orlowski

1,2

, Harald Loose

, Falko Eckardt

, J

urgen Edelmann-Nusser

and Kerstin Witte

Department of Computer Science and Media, Brandenburg University of Applied Sciences, Brandenburg, Germany

Institute of Sport Science, Otto von Guericke University Magdeburg, Magdeburg, Germany

Keywords:

Gait Analysis, Gait Parameters, Mobile Sensors, Inertial Sensors, Symmetry.

Abstract:

The amputation of a lower limb is a drastic event and it completely changes the life of the person. Current

development of prosthesis is already advanced, but most of the affected persons suffer from changes in the

gait which are visible to the general public. The gait of transfemoral amputees was investigated in the labo-

ratory environment and is called asymmetric due to different facts: shorter step length, smaller velocity and

smaller cadence. The use of mobile inertial sensors can be supportive in the rehabilitation process of these

patients. That is why a pilot study is conducted to evaluate the gait of transfemoral amputees and compare

their gait parameters with those of the healthy subjects. The purpose of the investigation is to identify gait

parameters showing the asymmetric properties of the amputees gait. Eight parameters seem to be distinctive

and descriptive.

1 INTRODUCTION

The gait is an essential movement for getting from one

place to another and therefore inﬂuencing the daily

life of people. Consequently, human gait disorders,

such as lower limb amputations, affect the quality of

life. The familiarization with the prosthesis may take

years and strongly depends on the adaptation to the

prosthesis socket. Due to the fact that the human gait

is a complex motion involving a lot of muscles, it is

very important for amputee gait to improve the in-

teraction of the remaining muscles. As known from

the literature the transfemoral (TF) and -tibial (TB)

amputee gait is asymmetric and gait parameters devi-

ate from the normal gait (Murray, 1980; Goujou-Pillet

et al., 2008). The relation between swing and stance

phase differs signiﬁcantly from the well-known 40-

60-ratio (Perry, 2010); the cadence is much smaller

(p < .001), the stride length much shorter (p < .01)

and the walking speed much slower (p < .001) in

amputees’ gait (Murray, 1980). Goujou-Pillet et al.

(Goujou-Pillet et al., 2008) show that (1) amputees

have larger range of motion (ROM) in the upper body

the slower they walk, (2) the pelvic rotation is speciﬁc

and (3) the existing counter-rotation of the pelvic and

the shoulder girdles is less present than in normal gait.

The gait parameters in amputees gait are much more

variable than in normal gait. In the literature there are

various methods named for the assessment of the gait

symmetry. Tura et al. (Tura et al., 2010) investigated

the step and stride regularity using the unbiased au-

tocorrelation coefﬁcient and compares the results of

those coefﬁcients with deﬁned symmetry index (SI).

Petersen et al. (Petersen et al., 2010) use the abso-

lute symmetry index (ASI) and the butterﬂy symme-

try ratio (BSR) to decide whether the gait symmetry

is signiﬁcantly improved wearing different technolo-

gies of prosthesis. Ambrosini et al. (Ambrosini et al.,

2011) considered the swing and stance ratio in per-

centage of stride of both legs to assess the impact of

the 2-week training. The named parameters are calcu-

lated from autocorrelation function of the acceleration

or angular velocity measured by inertial sensors. Fur-

thermore, the authors reconstructed the kinematic sig-

nals of the opposite body side by employing the Prin-

cipal Component Analysis (PCA) and achieve good

results. Gouwanda (Gouwanda, 2012) compares the

normalized SI and the normalized cross-correlation

and can show that the gait of participants wearing a

knee or ankle brace is signiﬁcant asymmetric. Wang

and Low (Wang and Low, 2009) used the autocor-

relation to decide whether the gait is symmetric or

not based on the muscle activity measured during the

gait sequence. Other methods for investigating the

gait symmetry are the clustering (Jiang et al., 2010),

symbolization (Sant’Anna and Wickstr

om, 2010) or

258

Orlowski K., Loose H., Eckardt F., Edelmann-Nusser J. and Witte K..

Analyzing the Transfemoral Amputee Gait using Inertial Sensors - Identifying Gait Parameters for Investigating the Symmetry of Gait - A Pilot Study.

DOI: 10.5220/0005250802580263

In Proceedings of the International Conference on Bio-inspired Systems and Signal Processing (BIOSIGNALS-2015), pages 258-263

ISBN: 978-989-758-069-7

 2015 SCITEPRESS (Science and Technology Publications, Lda.)

classiﬁcation by neural networks (Wafai et al., 2013;

Wafai et al., 2014). The SI proposed by Robinson et

al. (Robinson et al., 1987) is most often mentioned in

the literature.

To evaluate the gait of a person different methods

are conceivable and generally accepted in the medi-

cal environment. Muro-de-la-Herran et al. (Muro-de-

la-Herran et al., 2014) gave an overview of various

methods. They differentiated between non-wearable

and wearable systems used in the clinical ﬁeld. They

emphasized the advantages of the mobile systems, but

also referred to possible drawbacks of sensor accu-

racy. Motion capture systems, such as Vicon, were

regarded as gold standard. They mentioned inertial

measurement units (IMU) as sensor types which are

the most popular in the wearable gait analysis (with

almost two-thirds of the total number of wearable sys-

tems). On this account and due to the simplicity of

the usage of IMUs in the motion analysis, our focus

is on the development of automatic methods for the

analysis and evaluation of human gait based on IMU

using those distributed by Shimmer Research / Shim-

mer Sensing

This paper presents common gait parameters deter-

mined by using two inertial sensors. The identiﬁca-

tion of characteristic features of the amputee gait is

based on the comparison of the gait parameters of TF

amputees with those of the normal gait.

2 METHODS AND MATERIALS

In this section the determination of gait parameters

based on the data measured using two inertial sensors

(acceleration and angular velocity) is presented. Fur-

thermore the conducted experiments are brieﬂy ex-

plained. A short description of the used sensors is

included in the subsection 2.2.

2.1 Gait Parameters

Using two inertial sensors attached to the lower shank

above the ankle, 20 gait parameters (as mean and std)

are automatically determined by various algorithms.

Only two of the three axes of the acceleration (hori-

zontal and vertical) and one of the three axes of the

angular velocity are necessary for the calculation of

the parameters. Most of them are based on the angu-

lar velocity about the sagittal plane (angVel

) and the

detection of the gait events - initial and terminal con-

tact (IC and TC). The gait events are characterized by

www.shimmersensing.com

two local negative peaks within the signal of the angu-

lar velocity of the sagittal plane. Others are calculated

by integrating the acceleration or angular velocity.

Further information about the used algorithms can

be found in former publications (Orlowski and Loose,

2013; Orlowski and Loose, 2014).

Features describing the gait symmetry are the step

length and the stance ratio (G

otz-Neumann, 2011).

The ﬁrst feature is not considered in this investigation.

The second gait symmetry feature is included and can

be analyzed by considering the parameter stance (St)

for both sides (left, right). For each of the determined

parameters the symmetry index based on Robinson et

al. (Robinson et al., 1987) is calculated using the fol-

lowing equation 1:

SI[%] =

right

− x

le f t

∗ (x

right

+ x

le f t

)

∗ 100 (1)

with x

right

and x

le f t

representing a gait parameter mea-

sured for the left and right side.

The gait is assumed symmetric if SI of one or

more parameters are close to zero. Consequently, the

greater the value deviates from zero in both directions

(positive and negative), the more asymmetric is the

gait. Due to the fact that the normal gait is not com-

pletely symmetric, it has to be deﬁned which devia-

tion is acceptable to call the gait as symmetric.

Based on the information given in the literature com-

bined with the conducted evaluation of the gait pa-

rameters of all subjects (healthy and amputees) eight

gait parameters are identiﬁed as distinctive and de-

scriptive for the amputees gait:

• midswing height (MH): mean of the amplitude

during the midswing (see ﬁgure 1)

• height difference between TC and IC (HD): mean

difference of the amplitudes of the IC and TC (see

ﬁgure 1)

• swing (Sw): percentage of swing (TC-IC) at the

gait cycle (IC-IC)

• stance (St): percentage of stance (IC-TC) at the

gait cycle (IC-IC)

• single limb support (SLS): percentage of single

limb support at the gait cycle (IC-IC)

• put-on-angle (PA): maximum angle at the IC

• cadence (C): steps per minute

• velocity (V): distance per minute

2.2 Experiments

The gait of six healthy subjects (see table 1) and two

patients (see table 2) was measured with two inertial

AnalyzingtheTransfemoralAmputeeGaitusingInertialSensors-IdentifyingGaitParametersforInvestigatingthe

SymmetryofGait-APilotStudy

259

sensor capturing the acceleration and angular veloc-

ity. The sensors were each attached laterally on the

lower shank above the ankle. The ”normal” gait of the

subjects (healthy adults) and patients (TF amputees)

were captured while walking (straight forward) be-

tween 10 and 20 meters at normal speed.

9-DoF sensors (Shimmer2r) were used, without

registering the magnetometer, though it is not in-

cluded in the experiments. The acceleration and angu-

lar velocity were measured with a frame rate of 102.4

Hz and a sensitivity of 6g for the accelerometer and

500 deg/s for the gyroscope.

The capturing was done using the software Multi-

Shimmer-Sync provided by the manufacturer and an

capturing software implemented in MATLAB



us-

ing the interface given by Shimmer Research. The

analysis and evaluation was exclusively done in

MATLAB



In the preliminary study two patients (TF am-

putees) participated in our experiments. As refer-

ence values the data (parameters and especially the

SI) from healthy adults was used. As the table 1 show

385 gait cycles of 33 gait sequences were included to

measure the reference values.

Table 1: Anthropometric data and information of the gait

cycles (GC) of the six healthy subjects. For each subject

four to seven gait sequences (GS) are registered.

Subject

No.

1 2 3 4 5 6

age 29 37 42 32 25 25

BMI 27 25 23 22 21 21

sex m m f f m f

GS 4 5 6 7 5 6

GC 37 45 85 83 45 90

Table 2: Anthropometric data and information of the gait

cycles of the two patients.

Patient No. 1 2

age 30 32

BMI 25 21

sex m m

GS 15 12

GC 93 72

prosthesis left right

3 RESULTS

After measuring the gait of the two patients, the stored

data of the accelerometer and gyroscope was analyzed

with the existing algorithms using the in-house devel-

opment. The gait cycle of TF amputees gait have the

characteristic events IC and TC. Therefore, the fur-

ther analysis could be conducted without adjustment

on the algorithms. The gait parameters were calcu-

lated for all datasets. Furthermore, all datasets of the

healthy subjects were analyzed. Consequently, for

each subject and each measurement a table including

the parameters for the left and right leg was created.

The symmetry was determined for each parameter us-

ing the SI value. Furthermore the mean values of the

parameters (left and right side) and the SI values are

calculated for all available gait sequences for each

subject (two patients, six healthy) and for the group

of healthy subjects and both amputees.

Figure 1: The angular velocity in the sagittal plane of the

sound side (blue solid) and the amputated side (red dotted)

with marked TCs (terminal contacts), midswings and ICs

(initial contacts).

The tables 3, 4 and 5 show the parameters identi-

ﬁed as descriptive and distinctive. Table 3 represents

the gait parameters of a normal gait. The values in ta-

ble 4 belong to a patient one with a TF prosthesis on

the left side based on 15 gait sequences. The values

in table 5 belong to a patient two with a TF prosthe-

sis on the right side based on 12 gait sequences. The

SIs in the tables represent the mean value of all gait

sequences using the absolute value of each SI of the

33 gait sequences due to the fact that the measured

subjects can have both a dominant right or left leg.

The calculation of the mean based on the raw SI val-

ues (positives and negatives), would falsify the results

because a positive value would compensate a nega-

tive one. In contrast to the SI which is unaffected by

the dominance of one leg, this is not applied to the

gait parameters presented in table 3. Consequently,

the gait parameters of the healthy subjects can not be

directly compared with those of the patients.

The main difference between the normal and

amputees gait is the relation of swing and stance

phase (Sw and St). The same applies to the ratio of

single limb support and stride (SLS). These three

parameters show a symmetric value for the normal

gait (SI ranges from 2.16 to 3.66), while the amputees

gait has larger deviations from zeros (SI ranges from

17.39 to 20.89 or 6.16 to 7.03). Furthermore the

velocity and the cadence differs clearly. The ﬁgure 1

BIOSIGNALS2015-InternationalConferenceonBio-inspiredSystemsandSignalProcessing

260

shows the angular velocity presenting the motion of

the lower shank in the sagittal plane. The gait events

midswing, IC and TC are marked for both sides

(blue: sound and red: amputated). Clearly visible

is the difference of the maximum amplitude during

the swing phase of the amputated and sound side.

Furthermore the TCs of the amputated side are much

smaller than those of the sound side.

Table 3: Selected gait parameters of the normal gait given

for the left and right side as well as the SI as mean values

(standard deviation) of 33 gait sequences of the six healthy

subjects. For calculating the mean SI of all healthy sub-

jects the absolute value of each SI is used. (MH - midswing

height, HD - height difference of TC and IC, Sw - swing, St

- stance, SLS - single limb support, PA - put-on-angle, C -

cadence, V - velocity).

Param unit left right SI [%]

◦

/s 305.47

(24.48)

317.79

(20.94)

9.77

(5.48)

◦

/s 21.24

(11.26)

16.11

(8.24)

57.23

(44.80)

Sw % 47.4

(1.01)

48.3

(1.28)

2.33

(1.53)

St % 52.6

(1.04)

51.7

(1.30)

2.16

(1.45)

SLS % 48.5

(2.26)

48.0

(1.22)

3.33

(3.95)

◦

12.03

(0.62)

12.88

(1.53)

11.10

(8.56)

steps

min

116.30

(3.23)

116.30

(3.23)

—

min

79.02

(5.56)

79.02

(5.56)

—

4 DISCUSSION

While the normal gait has a almost 50-50 ratio for

both body sides in our setting, the amputees gait has

this ratio on the affected side. The stance ratio on

the unaffected side partly shows a distribution much

closer to 40-60 %. In contrast to the common relation

of swing and stance phase of 40-60 % (Perry, 2010),

evaluating the gait of healthy subject we received al-

most a 50-50 % ratio. The determination of start and

end point of the swing and stance phase are two rea-

sons for that deviation. The investigation of the two

patients show that TF amputees have a larger stance

phase on the sound leg than on the amputated leg, as

given in the literature (Murray, 1980).

As described by G

otz-Neumann (G

otz-Neumann,

2011) the mean stride length, consequently the ve-

locity differs signiﬁcantly between the two groups of

Table 4: Selected gait parameters of the amputees gait (pa-

tient one) given for the amputated and sound side as well as

the SI as mean values (standard deviation) of 15 registered

gait sequences.

Param unit ampu-

tated

sound SI [%]

◦

/s 278.91

(3.50)

257.32

(4.56)

8.06

(1.51)

◦

/s 122.06

(4.19)

43.36

(6.96)

95.56

(13.06)

Sw % 52.5

(1.33)

43.1

(0.67)

18.68

(3.39)

St % 47.5

(1.07)

56.9

(0.72)

17.39

(2.69)

SLS % 42.1

(0.75)

51.9

(1.18)

20.89

(3.21)

◦

12.83

(0.31)

8.16

(0.28)

44.49

(3.21)

steps

min

104.02

(1.44)

104.02

(1.44)

—

min

62.75

(4.53)

62.75

(4.53)

—

Table 5: Selected gait parameters of the amputees gait (pa-

tient two) given for the amputated and sound side as well as

the SI as mean values (standard deviation) of 12 registered

gait sequences.

Param unit sound amputa-

ted

SI [%]

◦

/s 333.18

(8.44)

283.21

(9.38)

16.23

(1.42)

◦

/s 110.54

(4.49)

61.22

(8.48)

58.01

(11.66)

Sw % 45.5

(0.61)

48.4

(1.28)

6.16

(3.37)

St % 54.5

(0.52)

51.6

(1.18)

7.03

(2.84)

SLS % 48.5

(1.35)

45.5

(1.39)

6.93

(4.75)

◦

13.23

(0.47)

14.46

(0.96)

10.05

(5.47)

steps

min

109.56

(3.23)

109.56

(3.23)

—

min

62.84

(3.03)

62.84

(3.03)

—

subjects. The normal gait (normal speed) achieves a

velocity of about 79m/min, the amputees gait reaches

a velocity of about 63m/min. Due to the fact that the

velocity is a parameter which is not calculated sepa-

rately for each side, no symmetry value is derived.

The gait parameter put-on-angle is distinctive: in

normal gait the angle is always larger then ten degrees

AnalyzingtheTransfemoralAmputeeGaitusingInertialSensors-IdentifyingGaitParametersforInvestigatingthe

SymmetryofGait-APilotStudy

261

and depends on the walking velocity, within the am-

putees gait the angles never reaches ten degrees for

sound side. The put-on-angle of this subject is seen

as asymmetric due to its SI value of 44.49.

Further parameters which refer to an asymmetry

in the amputees gait are MH and HD. Both are partly

characteristic and can even be detected through vi-

sual inspection of the gyroscope data. Within the gait

cycle of the normal gait the difference of the ampli-

tudes at the events IC and TC are almost the same for

both legs (HD). Considering the amputees gait there is

an imbalance of that difference regarding the affected

and the unaffected body side (see ﬁgure 1). This is re-

ﬂected by the calculated SI value of patients one (nor-

mal 57.23 and amputee one 95.56 and two 58.01).

Regarding the parameter MH the deviation is much

smaller, but it is recognizable in the SI value for one

of the patients (normal 11.10 and amputee one 44.49

and two 10.05). Considering the results of the exper-

iments, there are three to ﬁve gait parameters for the

motion of the foot (lower shank) which show a differ-

ence and therefore descriptive for the assessment of

the symmetry. As the tables 4 and 5 show that the gait

of amputees is not comparable due to different facts:

duration of wearing prosthesis, length of the stump,

type of the knee joint and others.

5 CONCLUSIONS AND

OUTLOOK

The determination of the gait parameters based on the

algorithms is equal for both, the normal and the am-

putees gait. The symmetry of the gait is calculated for

each parameter using the symmetry index of Robin-

son et al. (Robinson et al., 1987). A group of eight

is chosen from the 20 parameters. The parameters

HD, Sw, and St (SLS) characterize the asymmetric gait

of TF amputees in comparison to the normal bipedal

gait. The gait parameters MH, PA, and V are useful

for the identiﬁcation of an amputees gait. These seem

to be descriptive and distinctive in terms of character-

izing the gait of healthy adults and those affected by

amputation as well as showing the existence of sym-

metry. Whether these parameter are correct will be

evaluated in the following study with TF amputees

where the gait analysis based on inertial sensors is in-

tended to use as pre- and posttest.

In addition to the changed swing-stance-relation

an oscillating motion of the upper body is characteris-

tic for the amputees in comparison to the normal gait

(Goujou-Pillet et al., 2008). These motion can not

be registered with two sensors attached to the lower

shank, though. Consequently, our gait analysis has

to be expanded to the upper body by using additional

sensors. Furthermore it is necessary for the evaluation

in the pre- and posttest of the following study with TF

amputees to have more measurements of normal gait.

REFERENCES

Ambrosini, E., Ferrante, S., Pedrocchi, A., Ferrigno, G.,

Guanziroli, E., and Molteni, F. (2011). A novel

biofeedback cycling training to improve gait symme-

try in stroke patients: A case series study. IEEE

International Conference on Rehabilitation Robotics

(ICORR), pages 1–6.

otz-Neumann, K. (3. Auﬂage, 2011). Gehen verstehen

- Ganganalyse in der Physiotherapie. Georg Thiem

Verlag KG.

Goujou-Pillet, H., Sapin, E., Fod, P., and Lavaste, F. (2008).

Three-dimensional motion of trunk and pelvis during

transfemoral amputee gait. Arch Phys Med Rehabil,

89:87–94.

Gouwanda, D. (17-19 December 2012). Further valida-

tion of normal symmetry index and normal cross-

correlation in identifying gait asymmetry on restricted

knee and ankle movement. IEEE Conference on

Biomedical Engineering and Sciences, pages 423–

427.

Jiang, S., Zhang, B., Xu, W., and Wei, D. (2010). Gait

symmetry analysis based on afﬁnity propagation clus-

tering. IEEE 10th International Conference on Com-

puter and Information Technology (CIT), pages 1673–

1678.

Muro-de-la-Herran, A., Garcia-Zapirain, B., and Mendez-

Zorrilla, A. (2014). Gait analysis methods: An

overview of wearable and non-wearable systems,

highlighting clinical applications. Sensors, 14:3362–

3394.

Murray, M. P. (1980). Gait patterns of above-knee amputees

using constant -friction knee components. Bulletin of

Prothetics Reserach, 17 (No. 2):35–45.

Orlowski, K. and Loose, H. (2013). Evaluation of kinect

and shimmer sensors for detection of gait parame-

ters. Proceedings of BIOSIGNALS 2013, Int. Confer-

ence on Bio-Inspired Systems and Signal Processing,

Barcelona, Spain, 11-14 Feb. 2013, pages 157–162.

Orlowski, K. and Loose, H. (2014). Simple algorithms for

the determination or the walking distance based on

the acceleration sensor. Proceedings of BIOSIGNALS

2014, Int. Conference on Bio-Inspired Systems and

Signal Processing, Angers, France, 3-6 March 2014,

pages 264–269.

Perry, J. (2010). Gait Analysis - Normal and Pathological

Function. Slack Inc.

Petersen, A. O., Comins, J., and Alkjr, T. (03/2010). As-

sessment of gait symmetry in transfemoral amputees

using c-leg compared with 3r60 prosthetic knees. JPO

Journal of Prosthetics and Orthotics, 22 (2):106–112.

Robinson, R., Herzog, W., and Nigg, B. (1987). Use of

force platform variables to quantify the effects of chi-

BIOSIGNALS2015-InternationalConferenceonBio-inspiredSystemsandSignalProcessing

262

ropractic manipulation on gait symmetry. Journal

of Manipulative and Physiological Therapeutics, 10

(4):172–176.

Sant’Anna, A. P. and Wickstr

om, N. (2010). A symbol-

based approach to gait analysis from acceleration sig-

nals: Identiﬁcation and detection of gait events and a

new measure of gait symmetry. IEEE Transactions

on Information Technology in Biomedicine, 14 (No.

5):1180–1187.

Tura, A., Raggi, M., Rocchi, L., Cutti, A. G., and Chiari, L.

(2010). Gait symmetry and regularity in transfemoral

amputees assessed by trunk acceleration. Journal of

NeuroEngineering and Rehabilitation, 7:4.

Wafai, L., Zayegh, A., Begg, R., and Woulfe, J. (2013).

Asymmetry detection during pathological gait using a

plantar pressure sensing system. 7th IEEE Conference

and Exhibition (GCC), pages 182–187.

Wafai, L., Zayegh, A., Woulfe, J., and Begg, R. (2014).

Automated classiﬁcation of plantar pressure asymme-

try during pathological gait using artiﬁcial neural net-

work. Middle East Conference on Biomedical Engi-

neering (MECBME), pages 220–223.

Wang, P. and Low, K. (2009). Qualitative evaluations of

gait rehabilitation via emg muscle activation pattern:

Repetition, symmetry, and smoothness. IEEE Interna-

tional Conference on Robotics and Biomimetics (RO-

BIO), pages 215–220.

AnalyzingtheTransfemoralAmputeeGaitusingInertialSensors-IdentifyingGaitParametersforInvestigatingthe

SymmetryofGait-APilotStudy

263