Traitement du Signal

Vol. 36, No. 5, October, 2019, pp. 419-424

Journal homepage: http://iieta.org/journals/ts

A Footprint Extraction and Recognition Algorithm Based on Plantar Pressure

Chao Zhang1, Su Pan2, Yaowu Qi3, Yudan Yang4*
1
Changchun University, Changchun 130000, China
2
Department of Orthopedics, the Second Hospital of Jilin University, Changchun 130041, China
3
China railway Changchun railway bus co., LTD, Changchun 130062, China
4
Scientific Research Center, China-Japan Union Hospital of Jilin University, Changchun 130033, China

Corresponding Author Email: [email protected]

https://doi.org/10.18280/ts.360506 ABSTRACT

Received: 28 April 2019 Plantar pressure, an important index in gait analysis, has been widely adopted for clinical
Accepted: 2 August 2019 diagnosis and sports science. It is of great significance to accurately measure the plantar
pressure. In this paper, the data of plantar pressure are collected with the flexible force sensor,
Keywords: and then filtered by a self-designed filtering algorithm with time window. Then, an 8-
footprint recognition, plantar pressure, neighborhood connected-component labeling algorithm was proposed to segment and cluster
clustering, image segmentation the plantar pressure images. Finally, the footprints were recognized based on the plantar
pressure and shape of footprints. The experimental results show that the proposed footprint
extraction method extracted 99% of plantar pressure data accurately under normal walking
conditions, and the proposed footprint recognition method recognized more than 97% of
footprints in a correct manner.

1. INTRODUCTION flexible force sensor, clustered the data on plantar pressure by

k-means clustering (KMC), and then differentiated between
Plantar pressure, an important index in gait analysis, has the two feet based on the angle of the heel-toe vector. Redd
been widely adopted for clinical diagnosis and sports science. and Bamberg [2] marked the hardware of pressure insoles
Based on plantar pressure, it is possible to identify abnormal before collecting the plantar pressure of each foot, eliminating
gaits, assess the seriousness of foot disease and evaluate the the need to extract or recognize footprints.
rehabilitation condition. Thus, it is of great significance to With the aid of plantar pressure switches, Daliri [3]
accurately measure the plantar pressure. collected the single support time (left stance), single support
Traditionally, plantar pressure is measured by methods like time (right stance) and double support time, extracted the
pedography, sole barograph, pressure insole and force minimum, maximum, mean and standard deviation from the
measuring platform. With the development of computer three time series, and distinguished healthy people from
technology, pressure sensors have been applied to collect patients using the support vector machine (SVM) algorithm.
plantar pressure. Flexible force sensor is the most convenient Takayanagi et al. [4] classified normal gait from abnormal gait
and accurate pressure sensor, capable of capturing plantar based on four parameters, namely, walking speed, double
pressure in real time and displaying it in an intuitive manner. support time, gait balance ability, and static phase time. [5]
The application of large-area flexible force sensor can Gopalakrishna et al. [5] adopted fuzzy logic algorithm to
acquire both static and dynamic plantar pressures. The data calculate various gait parameters of plantar pressure.
thus collected accurately mirror the gait features of the subject. Bacarin et al. [6] obtained the static and dynamic features
The main problem with the gait analysis using large-area of plantar pressure with a single flexible force sensor, and
flexible force sensor lies in the processing of the collected recognized different people based on the two features. Before
plantar pressure. collecting the dynamic plantar pressure, the measuring points
To solve the problem, this paper designs a plantar pressure were determined through walking training. Then, the dynamic
platform based on flexible force sensor to collect plantar plantar pressure of each foot was measured by the sensor
pressure data, puts forward an 8-neighborhood connected- according to the expected walking speed. Jang et al. [7]
component labeling algorithm to segment and cluster the segmented the plantar pressure of each step by the KMC,
plantar pressure images, and proposes a dynamic footprint rotated the footprint image to determine six feature points by
recognition algorithm based on plantar pressure. The proposed scan lines, and recognized footprints based on the coordinates
platform and methods were verified through experiments. of feature points.
Powell et al. [8] predicted the possible trajectories of
footprints based on the feature parameters of one foot and the
2. LITERATURE REVIEW process parameters between two feet in the walking process.
The feature parameters were identified based on the travel
Flexible force sensor has been widely used in pressure direction and position. Ferber et al. [9] studied human behavior
insoles, plantar pressure switches, etc. Chu et al. [1] obtained perception with large-area flexible force sensor, and
the walking trajectories of left and right foot with a large-area recognized footsteps based on walking direction and

419
coordinates of key positions, revealing that footstep 3. ALGORITHM DESIGN
recognition based on plantar pressure is suitable for normal
walking with a small direction angle. 3.1 Data acquisition and outlier removal
Pataky and Goulermas [10] statistically segmented and
analyzed images on plantar pressure. Oliveira et al. [11] Currently, the plantar pressure collected by flexible force
presented a computational framework for automatic foot sensor has been widely applied in health care and
classification, using digital images on plantar pressure. Some rehabilitation, as well as gait analysis. The application effect
scholars meshed plantar pressure images through hinges on the authenticity and reliability of the collected data.
pedobarographic analysis, and applied it successfully to Since the walking habits are different from person to person
monitor the plantar pressure of social network users [12]. Crea and prone to external influence, it is necessary to standardize
et al. [13] analyzed the standing plantar pressure images in the the collection procedure of plantar pressure. In this paper, a
frequency domain. plantar pressure platform (Figure 1) is designed based on
flexible force sensor.

Figure 1. The proposed plantar pressure platform

In our platform, plantar pressure can be collected while the Despite its similarity to digital image, the collected data
previously collected data are being processed. The collected have time features in each frame. When the flexible force
data often contain outliers, i.e. the noises from system sensor collects plantar pressure, not all sensing points feel the
hardware and the flexible force sensor. The noises fall into pressure at the same time. Thus, there might be only a few
three categories: Gaussian noise, impact noise and jitter noise. nonempty pressure points in a frame of collected data. Before
Before using the plantar pressure, the outliers must be filtered filtering out the outliers, each pressure point should be judged
out on the platform from the collected data [14]. whether it is an outlier, based on the previous and subsequent
In data acquisition, a pressure value is collected at each frames.
sensing point of the flexible force sensor. The collected Drawing on hybrid image filters, this paper designs a new
pressure is divided to 256 levels, and displayed as the filtering algorithm with time window to remove the outliers.
corresponding RGB value. The greater the pressure, the darker This algorithm firstly identifies the outliers based on the
the displayed color. The inverse is also true. The plantar previous and subsequent frames, and then eliminates the
pressure 𝑃(t) collected by the flexible force sensor at time t outliers with a hybrid image filter. The specific steps of this
can be expressed as: algorithm are as follows:
Step 1. Judge whether each pressure point is an outlier. If
𝑝11 (𝑡) ⋯ 𝑝1𝑁 (𝑡) yes, go to Step 2 to determine if it is a noise. Otherwise, go on
𝑃(t) = [ ⋮ ⋱ ⋮ ] (1) to judge the next pressure point. Below is the criterion of
𝑝𝑀1 (𝑡) ⋯ 𝑝𝑀𝑁 (𝑡) outlier judgement: with the current pixel as the center,
generate a 3×3 window and count the number of valid pressure
where, 𝑝𝑖𝑗 (𝑡) is the pressure collected at sensing point (i, j) at points in the window; if the number is greater than the preset
time t; M and N are the row number and column number of threshold (here, the threshold is set to 4), then the current pixel
sensing point on the flexible force sensor, respectively. is a valid point; otherwise, it is an outlier.
According to the features of the flexible force sensor, the Step 2. Select the previous and subsequent N frames (here,
plantar pressure collected by the sensor can be regarded as a N is set to 5), i.e. a total of 2N frames. With the outlier
dynamically changing digital image, with 𝑃(t) being the pixel identified in Step 1 as the center, generate a 3×3 window in
value. Hence, the collected data can be filtered by digital each frame, forming a 3D filter window. Then, count the
image filtering algorithms. number of valid pressure points in each window; if the number

420
is greater than the present threshold (here, the threshold is set are compared in Table 1 below.
to 4), count that number in the window of the next frame. In
this way, the total number of valid pressure points of the 2N Table 1. Comparison of the three methods in denoising
frames can be obtained. If the total number is greater than a effects
preset threshold (here, the threshold is set to 5), then the
current pixel is a valid point. After that, adopt the hybrid image Algorithm
SNR (dB) SNR (dB)
filter to screen the 3×3 window of the current frame. If the total p=4, 𝝈=5 p=5, 𝝈=8
number is not greater than the threshold, set the pressure value MTM -8.43 -10.61
at the current point to zero, and return to Step 1 to filter the IMF -9.17 -11.18
next pressure point. Our algorithm -11.89 -12.50
Step 3. Repeat the above steps until all the frames have been
filtered. 3.2 Footprint extraction
Figure 2 illustrates the workflow of the proposed filtering
algorithm. In image processing, the purpose of pixel classification is to
identify the region in which the adjacent pixels are of the same
value. In this paper, the pixel classification is performed to
find out the pressure concentration region (PCR) of each frame,
based on the adjacent pressure points from the binary pressure
matrix. Then, the PCRs of different frames were matched. If
the preset requirement is met, then the PCRs are ascribed to
the same class, i.e. to the same footprint. The data of plantar
pressure can be displayed as images (Figure 3).

Figure 3. An example of plantar pressure image

Because the plantar pressure is dynamically changing in the

walking process, it is not suitable to use common clustering
Figure 2. The workflow of the proposed filtering algorithm algorithms to extract the footprint in each step and segment the
plantar pressure image. Under normal conditions, walking is
The filtering performance of our algorithm was evaluated movement with high regularity, which is reflected in the pace,
by signal-to-noise ratio (SNR). The SNR of an image refers to step length and step width. Based on the regularity, the plantar
the ratio of the power spectrum of the signal to the noise. Since pressure image was segmented under three empirical
the power spectrum is difficult to calculate, the ratio of the conditions:
variance of the signal to the noise is usually taken as an Condition 1: In normal walking, the step length and step
equivalent to the SNR. Therefore, the first step to compute the width both change in a reasonable range.
SNR is to calculate the local variance of each pixel in the Condition 2: The length and width of feet of a healthy
image. Then, the maximum and minimum of all local person are in a reasonable range (Figure 4).
variances are considered as the signal variance and the noise Condition 3: In normal walking, the interval between
variance, respectively. Then, the variance ratio between the frames of the same footstep falls within a reasonable range.
two could be computed and converted to decibels. Finally, the
variance ratio can be modified as:

∑𝑡 ∑𝑟𝑜𝑤 ∑𝑐𝑜𝑙(𝑂𝑢𝑡𝑝𝑢𝑡−𝑂𝑟𝑖𝑔𝑖𝑛)2
𝑆𝑁𝑅 = 10 log ∑𝑡 ∑𝑟𝑜𝑤 ∑𝑐𝑜𝑙(𝑁𝑜𝑖𝑠𝑒−𝑂𝑟𝑖𝑔𝑖𝑛)2
(2)

where, 𝑂𝑢𝑡𝑝𝑢𝑡 is the filtered output image; 𝑂𝑟𝑖𝑔𝑖𝑛 is the

original noise-free image; 𝑁𝑜𝑖𝑠𝑒 is the image with additive
noises.
Each image was added with a pulsating noise at the
percentage of p and a Gaussian noise with the standard
deviation of 𝜎 . For comparison, each noisy image was
processed by the modified trimmed mean (MTM) filter, the
intrinsic mode function (IMF) filter and out algorithm, Figure 4. An example of the length and width of feet of a
respectively [15]. The denoising results of the three methods healthy person

421
 In this paper, the plantar pressure image is segmented in The dynamic footprint recognition algorithm is
three steps: binarize the plantar pressure data of each frame; implemented in the following steps:
classify the binary data with the 8-neighborhood connected- Step 1: Extract the plantar pressure data of each footstep
component labeling algorithm [16]; allocate the classification based on the corresponding start timestamp, end timestamp
results to different classes under the three empirical conditions. and the coordinates interval in the data field of the flexible
The specific steps are as follows: force sensor. Binarize the data and fill them into the footstep
Step 1: Scan the plantar pressure data of each frame in turn, image matrix, and convert the matrix to a binary image.
map the pressure data to the image matrix, and binarize the Step 2: Magnify the small binary image through bilinear
image matrix. interpolation, and filter the magnified image with median filter,
Step 2: Classify the binary matrix with the 8-neighborhood creating a binary magnified image.
connected-component labeling algorithm, divide the plantar Step 3: Find the MBR of the binary magnified image by the
pressure data of each frame into different classes, compute the convex hull algorithm, and identify the angle between the
geometric center of each class, and record the start timestamp horizontal and vertical directions of the footprint.
of each class. Step 4: Rotate the binary magnified image by the angle
Step 3: Cluster each class according to its geometric center obtained in Step 3, and perform the nearest neighbor
and the preset thresholds (e.g. foot length, foot width, and interpolation.
interval of start timestamp), and iteratively separate each Step 5: Perform morphological image processing on the
footprint. rotated image, and smooth and filter the processed image.
Under the three empirical conditions, the interval between Step 6: Extract the outer edge of the processed image by the
the PCRs of the same footstep in normal walking should be edge detection algorithm, and compute the MBR by the
smaller than the length of a normal footstep, the footprint area convex hull algorithm.
fused by these PCRs should fall in the area range of normal Step 7: Scan the long-axis direction of the MBR, record the
footprints, and the footprint width fused by these PCRs should footprint width in the short-axis direction, and the coordinates
also fall in the width range of normal footprints. In each frame, of the upper and lower boundary points. Take the coordinates
there should be not spatial overlap between footsteps. In as the vertical values, the upper boundary of the MBR as the
addition, the interval of start timestamps between the PCRs of benchmark, and the distance between the upper and lower
the same footstep should fall within the normal range. boundary points as the pixel distance of the upper boundary of
In practical operations, the length, area, and width of a the MBR.
normal step are 26cm, 185cm2 and 10cm, respectively; the Step 8: Compress the range of the horizontal axis and the
interval of start timestamps between the PCRs of the same vertical axis to the range of the footprint length and the
footstep is 126ms. footprint width, highlighting the peaks and throughs of the
In this paper, the minimum bounding rectangle (MBR) is fitted curve.
used to cluster the PCRs and fully utilize the shape features of Step 9: Judge the footprints of left and right feet according
the MBR. The information of each footstep can be obtained to extreme values of fitting curve.
once all the frames are iterated.

3.3 Dynamic recognition of footprint 4. EXPERIMENTS AND RESULTS ANALYSIS

Both gait information and kinematic information are 4.1 Verification of footprint extraction accuracy
collected by the large-area flexible force sensor. To compute
the two types of information accurately and in real time, this The accuracy of footprint extraction was mainly measured
paper puts forward a dynamic footprint recognition algorithm by the start timestamp, end timestamp and the range of
based on plantar pressure. coordinates of the footprints in each footstep. These
The center of pressure (COP) trajectory reflects the integrity parameters were selected because they are the only metrics of
of plantar pressure. Let (𝑋𝑙1 , 𝑌𝑙1 ), (𝑋𝑙2 , 𝑌𝑙2 ), … , (𝑋𝑙𝑛 , 𝑌𝑙𝑛 ) be the clustering and segmentation effects of the plantar pressure
the coordinates of all the valid pressure points on the first data for each footprint.
footprint of the left foot, and 𝐹𝑙𝑛 be the force corresponding to A total of 67 subjects were selected for the experiment. All
each coordinates. The coordinates (𝑋𝑙𝑝 , 𝑌𝑙𝑝 ) at the COP of a their feet are normal. The test site was divided evenly into rows
frame of the left foot can be expressed as: and columns. A random row (column) was selected from the
site, and some rectangular boxes were marked, slightly larger
∑ 𝐹𝑙𝑛 ∗𝑋𝑙𝑛 than the footprints, on that row (column). The distance, step
𝑋𝑙𝑝 = (3)
∑ 𝐹𝑙𝑛 length and step width between the rectangular boxes were
similar.
∑ 𝐹𝑙𝑛 ∗𝑌𝑙𝑛
𝑌𝑙𝑝 = (4) Each subject was asked to step on the rectangular boxes in
∑ 𝐹𝑙𝑛
turn. The experimental process was fully monitored by a
camera. After the experiment, the landing and departure time
Then, the start frame of the first footprint of the left foot was of each step was extracted from the video shot by the camera.
extracted, and the all the coordinate points of the COP of that Then, the start and end timestamps of the footprints in each
foot of frame t were found. Those of the COP of the first step were compared, using the 8-neighborhood connected-
footprint of the right foot were obtained in a similar manner. component labeling algorithm. On this basis, the authors
Here, the direction of each footprint is determined based on evaluated whether the footprint range is consistent with the
the position relationship between the forefoot and arch, and the rectangular boxes marked in advance. The comparison of
left and right feet are differentiated based on the position footprint segmentation is shown in Table 2.
relationship between the inner arch and the outer arch.

422
 A long-time walking test was carried out to further verify connected-component labeling algorithm. Finally, the
the accuracy of the start and end timestamps of the footprints computed results were compared with those in the video. The
after segmentation. Each subject was asked to work normally number of correct and erroneous frames were counted to
in the test site for 3~5min. Then, the start and end timestamps evaluate the clustering accuracy (Table 3).
of all footprints were computed by the 8-neighborhood

Table 2. The comparison of footprint segmentation

Parameters
Step number
Footprint position Calculated position Recorded start and end timestamps Calculated start and end timestamps
1 (10, 10) (25, 30) (11, 9) (26, 29) 1000, 2100 1000, 2100
2 (15, 20) (20, 35) (14, 21) (20, 34) 2100, 3000 2100, 3000
3 (65, 35) (40, 25) (65, 36) (41, 26) 3200, 4000 3200, 4000
4 (10, 80) (40, 75) (11, 81) (41, 76) 4100, 4800 4100, 4800
5 (85, 60) (30, 35) (86, 60) (30, 36) 4900, 5250 4900, 5250
6 (30, 35) (50, 80) (31, 36) (51, 79) 5350, 5700 5350, 5700
7 (105, 70) (30, 65) (106, 79) (30, 66) 5860, 6050 5860, 6050
8 (110, 90) (70, 95) (111, 91) (71, 96) 6200, 6600 6200, 6600
9 (70, 30) (140, 65) (71, 31) (141, 66) 7200, 7600 7200, 7600
10 (100, 35) (50, 95) (101, 36) (51, 96) 8600, 9200 8600, 9200

Table 3. The comparison of footprint segmentation for the a self-designed filtering algorithm with time window. Inspired
long-time walking test by the theory of plantar dynamics, an 8-neighborhood
connected-component labeling algorithm was proposed to
Walking Sampling Number of Accuracy segment and cluster the plantar pressure images. Finally, the
time frames erroneous frames rate footprints were recognized based on the plantar pressure and
3 min 19000 112 99.41% shape of footprints. The experimental results show that the
4 min 26000 137 99.47% proposed footprint extraction method and footprint
5 min 32000 151 99.53% recognition method achieved desired segmentation and
recognition rates.
As shown in Table 3, the segmentation accuracy of our
algorithm was 99%, indicating that our algorithm can
accurately segment the start and end timestamps and ACKNOWLEDGEMENT
coordinates of each footstep in the normal walking process.
This paper was supported by the Jilin Provincial
4.2 Verification of footprint recognition accuracy Development and Reform Commission Project “Research on
the Health Assessment System of Human Biological Lines
A total of 67 subjects were selected and asked to walk Based on Dynamic Gait” (Project Number: 2019C048-6); the
normally, slowly, rapidly or at a random speed on the plantar Changchun University Fund Project “Research on Human
pressure platform, according to the prior process. The plantar Health System Based on the Plantar Micrology” (Project
pressure of each subject was measured, and used to recognize Number: ZK201806); and the Jilin Provincial Science and
the left and right feet. The recognition accuracy is listed in Technology Development Project “The Role of Bacterial
Table 4. DNA and Self-Mitochondrial DNA in Severe Multiple
Fractures with Acute Lung Injury by Regulating Caveolin-1
Table 4. The accuracy of footprint recognition and Neutrophil Membrane Surface TLR9” (Project Number:
20180414060GH).
Right foot Left foot Average
Walking speed recognition recognition recognition
rate rate rate
Normally 0.986 0.988 0.987 REFERENCES
Slowly 0.987 0.989 0.988
Rapidly 0.976 0.976 0.976 [1] Chu, Y.H., Cheng, C.H., Tang, P.F., Lin, K.H. (2012).
Random speed 0.968 0.966 0.967 Effects of age and step direction on behavioral
performances and center-of-pressure characteristics of
As shown in Table 4, the recognition rate of left foot, that volitional stepping in older and young adults. Biomedical
of right foot and the average recognition rate were higher in Engineering: Applications, Basis and Communications,
the three normal gait situations than the random speed 24(3): 207-216.
situation. Through analysis, it is learned that some subjects http://dx.doi.org/10.4015/S1016237212500032
failed to land their heels first in the random speed test, which [2] Redd, C.B., Bamberg, S.J.M. (2012). A wireless sensory
pushes up the recognition error. feedback device for real-time gait feedback and training.
IEEE/ASME Transactions on Mechatronics, 17(3): 425-
433. http://dx.doi.org/10.1109/TMECH.2012.2189014
5. CONCLUSIONS [3] Daliri, M.R. (2012). Automatic diagnosis of neuro-
degenerative diseases using gait dynamics. Measurement,
This paper collects the data of plantar pressure with the 45(7): 1729-1734.
flexible force sensor. Then, the collected data were filtered by http://dx.doi.org/10.1016/j.measurement.2012.04.013

423
[4] Takayanagi, N., Sudo, M., Fujii, M., Sakai, H., Morimoto, statistical parametric mapping (pSPM): A pixel-level
K., Tomisaki, M., Niki, Y., Tokimitsu, I. (2018). Foot approach to foot pressure image analysis. Journal of
pressure analysis of gait pattern in older Japanese Biomechanics, 41(10): 2136-2143.
females requiring different personal care support levels. http://dx.doi.org/10.1016/j.jbiomech.2008.04.034
Journal of Physical Therapy Science, 30(3): 461-466. [11] Oliveira, F.P., Sousa, A., Santos, R., Tavares, J.M.R.
http://dx.doi.org/10.1589/jpts.30.461 (2012). Towards an efficient and robust foot
[5] Gopalakrishna, K.P., Mothiram, K.P., Srinivasan, S. classification from pedobarographic images. Computer
(2001). A new method of analysis of standing foot methods in Biomechanics and Biomedical Engineering,
pressure images for detection of the plantar ulcers in 15(11): 1181-1188.
early-stage diabetic neuropathy. Frontiers of Medical and https://doi.org/10.1080/10255842.2011.581239
Biological Engineering: The International Journal of the [12] Li, B., Guo, C., Ning, T. (2018). An improved bacterial
Japan Society of Medical Electronics and Biological foraging optimization for multi-objective flexible job-
Engineering, 11(1): 31-43. shop scheduling problem. Journal Européen des
http://doi.org/10.1163/156855701750383178 Systèmes Automatisés, 51(4-6): 323-332.
[6] Bacarin, T.A., Sacco, I.C., Hennig, E.M. (2009). Plantar https://doi.org/10.3166/JESA.51.323-332
pressure distribution patterns during gait in diabetic [13] Crea, S., De Rossi, S.M.M., Donati, M., Reberšek, P.,
neuropathy patients with a history of foot ulcers. Clinics, Novak, D., Vitiello, N., Lenzi, T., Podobnik, J., Munih,
64(2): 113-120. http://dx.doi.org/10.1590/S1807- M., Carrozza, M.C. (2012). Development of gait
59322009000200008 segmentation methods for wearable foot pressure sensors.
[7] Jang, W.Y., Lee, D.Y., Jung, H.W., Lee, D.J., Yoo, W.J., In 2012 Annual International Conference of the IEEE
Choi, I.H. (2018). Inter-segment foot motion in girls Engineering in Medicine and Biology Society, pp. 5018-
using a three-dimensional multi-segment foot model. 5021. http://dx.doi.org/10.1109/EMBC.2012.6347120
Gait & Posture, 63: 184-188. [14] Bakker, M., Wicherts, J.M. (2014). Outlier removal and
http://dx.doi.org/10.1016/j.gaitpost.2018.05.007 the relation with reporting errors and quality of
[8] Powell, D.W., Long, B., Milner, C.E., Zhang, S. (2012). psychological research. PLoS One, 9(7): e103360.
Effects of vertical loading on arch characteristics and http://dx.doi.org/10.1371/journal.pone.0103360
intersegmental foot motions. Journal of Applied [15] Wang, J., Liu, Y., Zhang, C., Jiao, Y. (2009). Model for
Biomechanics, 28(2): 165-173. router-level Internet topology based on attribute
http://dx.doi.org/10.1123/jab.28.2.165 evolution. IEEE Communications Letters, 13(6): 447-
[9] Ferber, R., Hettinga, B.A. (2016). A comparison of 449. http://dx.doi.org/10.1109/LCOMM.2009.090437
different over-the-counter foot orthotic devices on multi- [16] Lang, F., Yang, J., Li, D., Zhao, L., Shi, L. (2013).
segment foot biomechanics. Prosthetics and Orthotics Polarimetric SAR image segmentation using statistical
International, 40(6): 675-681. region merging. IEEE Geoscience and Remote Sensing
http://dx.doi.org/10.1177/0309364615584660 Letters, 11(2): 509-513.
[10] Pataky, T.C., Goulermas, J.Y. (2008). Pedobarographic http://dx.doi.org/10.1109/LGRS.2013.2271040

424