Smart insole: A wearable system for gait analysis

Abstract

Gait analysis is an important medical diagnostic process and has many applications in rehabilitation, therapy and exercise training. However, standard human gait analysis has to be performed in a specific gait lab and operated by a medical professional. This traditional method increases the examination cost and decreases the accuracy of the natural gait model. In this paper, we present a novel portable system, called Smart Insole, to address the current issues. Smart Insole integrates low cost sensors and computes important gait features. In this way, patients or users can wear Smart Insole for gait analysis in daily life instead of participating in gait lab experiments for hours. With our proposed portable sensing system and effective feature extraction algorithm, the Smart Insole system enables precise gait analysis. Furthermore, taking advantage of the affordability and mobility of Smart Insole, pervasive gait analysis can be extended to many potential applications such as fall prevention, life behavior analysis and networked wireless health systems.

Full text

Smart Insole: A Wearable System for Gait Analysis
Wenyao Xu
Electrical Engineering
Department
University of California, Los
Angeles
Los Angeles, CA 90095
wxu@ee.ucla.edu
Ming-Chun Huang
Computer Science
Department
University of California, Los
Angeles
Los Angeles, CA 90095
mingchuh@cs.ucla.edu
Navid Amini
Computer Science
Department
University of California, Los
Angeles
Los Angeles, CA 90095
amini@cs.ucla.edu
Jason J Liu
Computer Science
Department
University of California, Los
Angeles
Los Angeles, CA 90095
jasonliu@cs.ucla.edu
Lei He
Electrical Engineering
Department
University of California, Los
Angeles
Los Angeles, CA 90095
lhe@ee.ucla.edu
Majid Sarrafzadeh
Computer Science
Department
University of California, Los
Angeles
Los Angeles, CA 90095
majid@cs.ucla.edu
ABSTRACT
Gait analysis is an important medical diagnostic process and
has many applications in rehabilitation, therapy and exer-
cise training. However, standard human gait analysis has to
be performed in a specific gait lab and operated by a medical
professional. This traditional method increases the exami-
nation cost and decreases the accuracy of the natural gait
model. In this paper, we present a novel portable system,
called Smart Insole, to address the current issues. Smart
Insole integrates low cost sensors and computes important
gait features. In this way, patients or users can wear Smart
Insole for gait analysis in daily life instead of participating in
gait lab experiments for hours. With our proposed portable
sensing system and effective feature extraction algorithm,
the Smart Insole system enables precise gait analysis. Fur-
thermore, taking advantage of the affordability and mobility
of Smart Insole, pervasive gait analysis can be extended to
many potential applications such as fall prevention, life be-
havior analysis and networked wireless health systems.
Categories and Subject Descriptors
H.4 [Information Systems Applications]: Miscellaneous
General Terms
Design
Keywords
Gait Analysis, Wireless Health, Wearable Device
1. INTRODUCTION
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies are
not made or distributed for profit or commercial advantage and that copies
bear this notice and the full citation on the first page. To copy otherwise, to
republish, to post on servers or to redistribute to lists, requires prior specific
permission and/or a fee.
PETRA’12, June 6 - 8, 2012, Crete Island, Greece.
Copyright c
2012 ACM ISBN 978-1-4503-1300-1/12/06... $10.00
Gait analysis is an important human locomotion study to
recognize normal or pathological patterns of walking, and its
results have plenty of applications in medical programs [1],
physical therapy [2], and sports training [3]. For example,
with detailed gait feature analysis, therapists can quantify
the rehabilitation progress of the patients after surgery, and
the corresponding treatment and training can be customized
according to an individual’s status [4].
Normally, the quantitative gait examination is operated at a
professional place, called the Gait Lab. A standard gait lab is
roughly a 1200 square foot facility. In order to achieve com-
prehensive sensing and observation of gait features, tradi-
tional equipment include eight high-speed surrounding cam-
eras, one big pressure sensitive carpet and a number of on-
body assistive devices such as accelerometers, gyroscopes
and color markers. When the patient walks in the gait lab,
his motion pattern can be captured by these devices. The
gait expert will analyze and recognize the walking patterns
with the combined quantified data. Although gait analy-
sis is important, it is difficult to accomplish accurately and
fairly because of two reasons. First, building a gait lab is
expensive. Without considering the cost and compensation
of medical professionals and related expenses, it takes more
than 300K dollars to setup the necessary facilities and e-
quipment. Second, due to the burdensome equipment and
precisely controlled environment, the patients always feel
uncomfortable during the data collection in a gait lab. In
most cases, they can not walk in their own regular style,
and sensing data of gait features might always be biased,
and even incorrect.
There are several prior research work on wearable devices for
human gait analysis in the past. Tao Liu et al. [5] proposed
a mobile gait analysis system with 3-axis accelerometer and
gyroscope. Bae et al. [6] presented a Force Sensing Resistive
(FSR) sensor array based system for gait analysis. Recently,
Xu et al. [7] developed a Compressed Sensing based algo-
rithm with a single acceleromater for accurate human activ-
ity recognition. All these works are helpful to understand

the human locomotion and walking status. However, they
cannot comprehensively and accurately address all gait fea-
tures used in actual medical applications. There is indeed
a demand of a portable system for professional level gait
analysis.
In this paper, we present a wearable and affordable technol-
ogy, called Smart Insole, which addresses the current issues
in gait analysis. Different from gait lab analysis, the smart
insole system integrates motion sensing components with-
in shoe insoles. With the intelligent analysis algorithm, all
important human gait features can be retrieved from the
sensor data. Therefore, the Smart Insole system can mon-
itor all types of activities in free-living without disturbing
the normal life of the subject. Contrary to the cost of set-
ting up a gait lab, the cost for the Smart Insole system can
be under 200 dollars for mass production.
The remainder of the paper is structured as following. Sec-
tion 2 describes the design and implementation of the Smart
Insole system. In Section 3, we discuss the preliminary data
from the Smart Insole System. In Section 4, we conclude
and outline the future research direction and related clinical
applications.
2. SYSTEM
In this section, we discuss the design concept, architecture
and implementation of the Smart Insole System.
2.1 Design Requirement
Gait analysis is a complicated process, and there are hun-
dreds of important features for identifying the normal and
pathological walking patterns [8]. Some features are general
and important for any kind of applications, such as cadence,
stride length, stride height and speed. Some features are
applicable for specific domains. For example, the pressure
balance of locomotion between two feet is important for an-
alyzing fall prevention [9], and pressure hotspot mobility is
critical for diabetes foot protection and ulcer prevention [10].
According to medical literatures, there are 12 critical gener-
al gait features for miscellaneous applications. We list them
and their corresponding definitions in Table 1. From an engi-
neering point of view, it requires precise motion sensing and
high-resolution under-foot pressure sensing to capture these
12 features. The detailed system architecture and hardware
design will be discussed in Section 2.2.
2.2 Architecture and Hardware
The Smart Insole system architecture is shown in Fig. 1.
There are three important subsystems. The first subsystem
is low cost sensors for gait characterization, including 48
pressure sensors, 3-axis accelerometer, 3-axis gyroscope and
3-axis compass. The pressure sensor array is used to obtain
the high-solution pressure map under foot. It is based on
advanced fabric sensor technique [11] and can be efficient-
ly integrated in the Smart Insole system. The accelerome-
ter and gyroscope are inertial sensors, and can measure the
movement information of the subject. The compass is used
as the baseline when the inertial sensors (accelerometer and
gyroscope) are calibrated. With 9 degrees of freedom mo-
tion sensing, the precision of motion sensing can be safely
achieved.
Figure 1: Sensors used in the Smart Insole system
The second subsystem is the signal acquisition and transmis-
sion module. The sensor data are quantized by 8-channel,
12-bit, 0-3.3 volt analogue-to-digital module. The sample
rate can be adapted to the specific applications, up to 100
samples per second (Hz). After that, the quantified sensor
data are streamed in real-time to a data aggregator. For the
sake of generality and low-power issues, we use Bluetooth to
transmit the data. With one 1200-mAh Li-battery, the sys-
tem can continuously work for over 24 hours. Therefore, the
Smart Insole system can be used daily without interruption
and without charging the battery.
The third subsystem is the sensor aggregation and process-
ing module. We developed a software on a smart phone to
store and analyze the raw data it receives. For the analysis,
the smart phone can calibrate the raw sensor data, fuse and
remove the noise, and compute the important gait features.
According to the computation load, the smart phone also
could switch into router mode and upload to a data center
with more powerful computing resources. The computed re-
sults can be visualized on the screen. Moreover, an effective
cloud service is implemented to coordinate the remote data
center. The software design will be elaborated in Section
2.3.
2.3 Software
The software system on the smart phone (see Fig. 2) consists
of three stacked layers. They are the physical driver inter-
face, data preprocessing module and data postprocessing.
For the sake of real-time computing, the software is imple-
mented with multithreading. In general, there are six main
threads in the software program. A device driver thread
handles asynchronous communication to Smart Insole over
the Bluetooth Serial Port Profile (BSPP). The device driv-
er synchronizes the incoming sensor data before forwarding
to the client programs over interconnect sockets. Another
thread performs data preprocessing, including de-noising for
pressure sensor data [11], calibrating inertial sensor values
with filtering, and initializing the baseline with compass da-
ta. After the above steps, the data is clean, compressive and
informative, and the following processing will be dispatched

Table 1: Important Features in Gait Analysis
No. Gait Feature Definition
1 Walking Speed The average speed while the subject is walking
2 Stride Height The maximum height from foot to ground during one step
3 Stride Length The displacement of one foot from lift to landing
4 Landing Position Position of foot when it lands, such as heel strikes first, foot
lands flat, toe hits first
5 Under-foot Pressure The fine-grained pressure distribution under foot
6 Path Distance The total length while the subject is walking
7 Path Variation The total length variation while the subject repeats experiments
8 Speed of Leg Movement The speed of each leg while the subject is walking
9Speed Difference The speed variation between left leg and right leg while walking
between L&R legs
10 Foot Orientation The foot orientation while the subject is walking
11 Cadence Average steps per minute while walking
12 Swing Time The transition time of the foot from lift to landing
to the corresponding services on the next layer. As shown in
Fig. 2, the remaining four threads receive the preprocessed
data and perform two local services and two remote services,
respectively.
Bluetooth Serial Port Profile (BSPP)
Data Preprocessing
Lightweight
Data Mining
Data
Visualization
(GUI)
Data
Uploading Feedback
Receiving
3G/WiFi Interface
Smart Insole
Data Center
Figure 2: Stacked Software Structure
The local services includes light weight data mining and
graphical user interface (GUI). Fig. 3 gives an example of
local processing results. For data mining, eight computed
results are achievable, including 1) total pressure profile, 2)
number of steps, 3) cadence, 4) step time, 5) swing time, 6)
stance time, 7) stance to swing ratio and 8) dual limb sup-
port time. In the GUI, pressure maps under two feet are
visualized (see the up left side in Fig. 3). At the same time,
the foot orientation is also displayed with the calibration
motion sensor information (see the up right side in Fig. 3).
The remote services are composed of data uploading for
heavy computing, and feedback receiving from data center.
The remote computing and inference engine can be used to
detect and classify the patient’s usage context of the Smart
Insole system. This accurate analytic information about the
insole permits care givers to monitor the patient’s walking
pattern and create the feedback model to train the proper
gait. The feedback also can be visualized on the screen for
SpatialUnder‐foot
Pressure Feature
Foot Orientation Feature
Pressure

Feature
Foot

Orientation

Feature
TemporalUnder‐footPressureFeature
Figure 3: Software and User Interface Design
the users.
3. EVALUATION
The preliminary results from the Smart Insole system are
shown in Fig. 4. The raw data streaming from the insole
system was saved on the smart phone. For simplicity of the
presentation, we visualize the data for 5 steps of walking
from the left side insole. The computing results are based
on 48 pressure sensors, 3-axis accelerometer, 3-axis gyro-
scope and 3-axis compass. To state the data intuitively, we
selected six results from local data mining threads. Fig.
4(a) shows the summation profile of five steps. Fig. 4(b)
shows the roll angular velocity. Fig. 4(c) shows the pitch
angular velocity. Fig. 4(d) shows the yaw angular velocity.
Fig. 4(e) shows the normalized walking speed. Fig. 4(f)
shows the swing speed during walking. Based on the ob-
servation and analysis, it indicates that pressure summation
profile and roll angular velocity are significantly related to
step numbers. Each step is prominent in the data visual-
ization. The general gait feature can be characterized by
walking speed and swing rate. When the subject is walking
constantly, the signals from these two metrics are constant
as well. The individual’s gait features are related to pitch
angular velocity and yaw angular velocity, which should be

analyzed in depth.
0 20 40 60 80 100 120 140 160
−10
0
10
[b]
roll
0 20 40 60 80 100 120 140 160
−20
0
20
[e]
speed
0 20 40 60 80 100 120 140 160
0
20
40 Preliminary Data
[a]
Pressure
0 20 40 60 80 100 120 140 160
−50
0
50
[f]
swing rate
time
0 20 40 60 80 100 120 140 160
−10
0
10
[c]
pitch
0 20 40 60 80 100 120 140 160
−50
0
50
[d]
yaw
Figure 4: Data from the Smart Insole system (left
foot) during subject walking activities (five steps).
(a) Pressure summation of 48 sensors; (b)Roll an-
gular velocity; (c) Pitch angular velocity; (d) Yaw
angular velocity; (e) Walking speed (normalized);
(f) Swing rate.
4. CONCLUSION
In this paper, we presented the design and development of
the Smart Insole system that utilizes the capabilities provid-
ed by the Telehealth architecture, which may be based on
commercially available microsensor, computing, and wire-
less technologies. We also presented preliminary data from
a patient using the Smart Insole system and showed that
the data can be used to analyze the patient’s usage of the
insole. Taking advantage of the ubiquitous internet access,
the Smart Insole system can transfer the data from the pa-
tient side to the centralized servers in the medical enterprise,
permitting caregivers to monitor the user gait in real-time
and over long periods. Over the cloud service, the feedback
could be offered by either medical professionals or statisti-
cal intelligent database. This also will enable a numerous of
future applications, such as foot ulcer prevention, running
pattern recognition and imbalance analysis.
Our future work to further develop and evaluate the Smart
Insole system includes several objectives. First, we can offer
the training to guide safe walking behavior for preventing
falls or reducing the risk of falls. Second, we plan to charac-
terize the walking patterns across a large group of patients
and develop statistical models that can identify and detect
the improper gait behavior leading to instability, falls and
exercise, in the elderly. We plan to use these statistical
models to guide and train these elderly patients to properly
use a cane. Furthermore, combining Smart Insole, Smart
Cane [12] and Smart Cushion [11] we developed, a complete
in-home monitoring health system, including infrastructure
and user end devices, could be built. Each technology is
complementary to each other and jointly they supply ubiq-
uitous sensing for complete activity monitoring.
Acknowledgement
This work is partially supported by the National Science
Foundation under Grant IIS-1238660.
5. REFERENCES
[1] M. Whittle and B. Heinemann, Gait Analysis: An
Introduction. New York: John Wiley and Sons, 2005.
[2] M. eastlack and et al., “Interrater reliability of
videotaped observational gait analysis assessments,”
Physical Therapy, vol. 105, pp. 465–472, December
1991.
[3] H. Stolze and et al., “Gait analysis during treadmill
and overground locomotion in children and adults,”
Electroencephalography and Clinical Neurophysiology,
Electromyography and Motor Control, vol. 105,
pp. 490–497, December 1997.
[4] R. El-Hawary, L. Karol, K. Jeans, and S. Richards,
“Gait analysis of children treated for clubfoot with
physical therapy or the ponseti cast technique,”
Journal of Bone and Joint Surgery, vol. 90,
pp. 468–477, July 2008.
[5] T. Liu, Y. Inoue, and K. Shibata, “Development of a
wearable sensor system for quantitative gait analysis,”
Measurement, vol. 42, pp. 978–988, February 2009.
[6] J. Bae, K. Kong, N. Byl, and M. Tomizuka, “A mobile
gait monitoring system for gait analysis,” in IEEE
International Conference on Rehabilitation Robotics,
(Kyoto, Japan), pp. 50–57, June 2009.
[7] W. Xu, M. Zhang, A. A. Sawchuk, and
M. Sarrafzadeh, “Co-recognition of human activity and
sensor location via compressed sensing in wearable
body sensor networks,” in IEEE International
Conference on Wearable and Implantable Body Sensor
Networks (BSN), (London, UK), pp. 1–6, May 2012.
[8] P. Jacquelin and et al., “Gait analysis: Normal and
pathological function,” Journal of Pediatric
Orthopaedics, vol. 12, pp. 815–822, November 1992.
[9] J. Verghese, R. Holtzer, R. Lipton, and C. Wang,
“Quantitative gait markers and incident fall risk in
older adults,” Journal of Gerontology, vol. 64,
pp. 896–901, January 2009.
[10] U. Uccioli and et al., “Manufactured shoes in the
prevention of diabetic foot ulcers,” Diabetes Care,
vol. 18, pp. 1376–1378, October 1995.
[11] W. Xu, Z. Li, M.-C. Huang, N. Amini, and
M. Sarrafzadeh, “ecushion: An etextile device for
sitting posture monitoring,” in BSN, pp. 194–199,
2011.
[12] W. Wu and et al., “The smartcane system: An
assistive device for geriatrics,” in IEEE International
Conference on Rehabilitation Robotics, (Tempe,
Arizona, USA), pp. 1–4, Sept. 2008.