Algorithm implementation for the analysis of gait and posture André Mendes Marques de Carvalho Mestrado Integrado em Bioengenharia Porto, Janeiro 2013
Algorithm implementation for the analysis of gait and
posture
André Mendes Marques de Carvalho
Mestrado Integrado em Bioengenharia
Porto, Janeiro 2013
André Mendes Marques de Carvalho
Algorithm implementation for the analysis of gait and
posture
Monografia do Mestrado Integrado em Bioengenharia
Faculdade de engenharia da Universidade do Porto
Porto, 2013
iii
Abstract
Some of the most common motor action are the postural control and gait
control. This is of critical importance as gait is the main way of human locomotion.
Both gait and the ability to remain in a standing position are highly affected by factors
such as somatosensorial inputs, ground inclination and health problems. The goal of
this study is to implement an algorithm capable of analyzing the EMG and force signal
recorded during standing and gait and associate the specific muscles activations during
body movements and standing maintenance. To accomplish that Matlab will be used
to operate previously collected data from both gait and standing movements. The
EMG data will be treated in the time domain and patterns of muscle activation are
expected to be found during cyclic movements. Also the force data will be examined as
to the magnitude, peak interval and event duration.
iv
Index
Abstract…………………………………………………………………………………iii
Introduction……………………………………………………………………………...1
Motor control…………………………………………………………………………….2
Musculoskeletal system………………………………………………………….3
Postural control…………………………………………………………………..6
Gait control………………………………………………………………………8
Instrumentation…………………………………………………………………………10
Force Plate………………………………...……………………………………11
Electromyography………………………………………………………………13
Work plan………………………………………………………………………………15
References………..…………………………………………………………………….16
1
Introduction
The complexity of the motor system is reflected on the complexity of the
actions humans can achieve (Enoka 2008). Nowadays gait and postural analysis are
mainly done for clinical research. This means that the purpose is not to study individual
problems but to study the specific condition of a group of patients. (Baker, A., 2006)
The objective of such analysis depends on the investigation. Some are interested in
gait patterns, speed, the work made by witch muscles during specific activities,
metabolic cost, etc. (Andreia, 2009). The goal is clear, to understand the mechanisms
associated with gait and posture control.
In order to study this mechanisms force plate, EMG, visual tracking, kinematics
as well as many other techniques can be applied. The main problem is to understand
and interpret correctly the signal of such measurements. To do that time or frequency
domain analysis of the collected signals are made. These signals can determine the
body center of pressure and the force exerted on the ground at all times and the rate
of muscular activation for a given task.
From this data a better understanding of the human gait and posture processes
can be achieved and proper medical evaluations can be made. It is known that gait and
postural pattern can be influenced by health state, age, shoes, body alignment and
many other factors. The main goal of this study is to implement an algorithm capable
of analyzing the EMG and force signal recorded during standing and gait and associate
the specific muscles activations during body movements and standing maintenance.
2
Motor Control
The complexity of the motor system is reflected on the complexity of the
actions humans can achieve (Enoka 2008). These actions are a result of the
interactions between the musculoskeletal and nervous systems, the first acting as a
support and the second as the coordinator (Ghez and Krakauer 2000). Being so, all
movement is achieved by the coordinated response of the muscles to neural
commands (Bawa 2002).
Some of the most common motor action are the postural control and gait
control. The postural control is a rather complex skill based on the interactions of
dynamic sensorimotor processes and can be separated into two main functional goals:
the postural orientation and postural equilibrium (Horak 2006). As for the Gait control
this is a complex action in which the body center of mass (COM) is continuously
unbalanced so the body weight is transferred from one leg to another with the
minimum energy consumption (Alonso et al. 2002).
For a motor action to take place it is needed a proper alignment of all body
sections so it can maintain its stability despite all the unbalancing forces (Mackey and
Robinovitch 2006). This way motor control can be defined as all the processes from
which the nervous system, based on visual, vestibular and somatosensory inputs,
controls the musculoskeletal system to create coordinated movements and skilled
actions while maintaining body posture (Junior and Barela 1996).
3
Musculoskeletal system
The musculoskeletal system is composed by over 200 bones and 600 muscles
that allows around 224 degrees of freedom and is the effector component of all
biomechanical movement (Duarte 2001). The bones act as the support and protective
system for various organs and tissues while the muscles can contract to realize
movement.
The muscle is a hierarchical structure that contains fascicles, muscle fibers,
myofibril and actin and myosin filaments as shown in figure 1.
Figure 1- Schematic of muscle components [from Medicalook 2013]
For a contraction to take place several structures of the nervous system must
be involved. At the lowest level of organization there’s the spinal cord which sends all
motoneurons and their associated muscle fibers form the neuromuscular system. This
system has been considered as the interface between muscles and nervous system
(kernel 2006).
The basic control unit is called motor unit. This term was introduced in 1925 by
Sherrington and Liddell, and it is from this unit that the nervous system controls the
movement and force of the muscle (Ghez and Krakauer 2000). The motor unit is
composed by one motor neuron and all the muscle fibers to which it is connected
(figure 2).
4
Figure 2 - Motor unit schematics [Baileybio 2012]
This is the smallest amount of muscle fibers the nervous system can activate to
perform a given task as once the motor neuron is activated all its associated muscle
fibers contract. Depending on the movement and the force necessary to perform it the
muscle fiber can be activated from only a few to all of them (Kernel 2003).
Once a motor neuron is activated it generates a potential difference in its
innervated muscle fiber resulting in an action potential (figure 3). This happens every
time an unbalanced distribution of ions inside and outside the cells is achieved (Loeb
and Ghez 2000). During cell depolarization there’s an influx of sodium (Na+) and an
efflux of potassium (K+).
5
Figure 3 - Membrane potential of a muscle cell [Wikipedia 2013]
During rest the membrane potential is usually around -70 mV. Once the
threshold potential is reached, approximately -55 mV, the cell depolarizes opening its
sodium channels. The influx of sodium channels raises the membrane potential up to
+40 mV. After this the sodium channels close and the potassium channels expel
potassium from inside the cell, lowering its membrane potential (Seeley 2003).
6
Postural control
Posture is referred to as the ability to maintain the body in a given position and
translates the result of a permanent muscular activity that opposes the instability
created by the gravitational force. As shown in figure 4 the postural control is the
result of an efficient coordination between the visual, vestibular, somatosensory and
musculoskeletal systems (Fernandes 1998)
Figure 4 - Postural control mechanisms [jneuroengrehab 2013]
The two main functional goals of postural control are postural orientation and
postural equilibrium. Postural orientation integrates sensory information to actively
align the head and body with respect to gravity visual surrounding, support surfaces
and internal references. This is achieved once all external and internal forces are
balanced. Postural equilibrium refers to the movements necessary to stabilize the
body center of mass during stability disturbances. It depends on individuals
experience, goals and postural displacement (Horak 2006).
The postural control is not an easy task, the surface area is quite small and its
body center of mass is located quite high making this a very instable position. To
maintain Postural control the nervous system is continuously making adjustments.
These adjustments are a result of the sensorial input and the appropriated muscular
responses (Pais 2005). The standing position is characterized by small and spontaneous
postural changes. There are two main contributions to the stability of the standing
position, the corporal alignment capable of reducing the gravitational forces effect as a
good alignment reduces the energy expenditure (figure 5) and the muscular tonus
(Shumway-Cook and Woollacott 2001).
7
Figure 5 - Body center of mass alignment and gravitational force [adapted from newbalancecalgary 2013]
As the standing position is unstable muscles from all over the body are
continuously contracting to maintain body balance. These muscular contractions give
rise to small movements which in turn act to stabilize the body position (Horak 2006).
8
Gait Control
The primary goal of walking is the locomotion of the body from one place to
another. To achieve this humans alternate their support from one leg to another,
always keeping one foot on the ground, in an action where the body center of mass is
displaced from its regular position allowing for the body to move forwards. The
movements that take place during walking are cyclic and optimized to decrease energy
expenditure (Silva 2009).
To perform a regular walking an individual needs to be able to maintain
movement patterns, be able to maintain a dynamic equilibrium between the body
center of mass and the moving support and to change his movement patter in
response to any external disturbance. The Walking cycle is every movement that
occurs between the initial contact of one foot in the ground and the instant that foot
touches the ground again (Monteiro 2004).
Each walking cycle has two steps, one for each leg, and it is divided by two main
actions, the step and the swing (figure 6). The step refers to the contact between the
foot and the ground and it correspond to 60% of the cycle time. The swing corresponds
to the other 40% and is the time in which a leg is progressing thru space (Gamble et al
1998).
Figure 6 - walking cycle [adapted from Perry 2005]
The two prevailing theories of human walking are the inverted pendulum and
the six determinants of gait as shown in figure 7. The inverted pendulum theory states
that it is beneficial for the stance leg to behave like a pendulum, where the kinetic
energy is converted into potential energy and vice versa, saving this way more than
9
50% of the required mechanical energy (Cavagna, 1996). The six determinant of gait
are kinematic features of gait proposed to minimize the energetic cost of locomotion
by reducing the vertical displacement (Kuo, 2007). Among those kinematic features we
have the pelvic rotation and the flexion of the knee as a way of reducing the body
center of mass vertical variation.
Figure 7 - Two major walking theories [kuo 2007]
Independently of the theory the muscular activity during the walking cycle can
be divided into the ground strike step in which the foot touches the ground with much
positive work from the quadriceps, the pre-load step in which the elastic energy is
stored into the Achilles tendon and the quadriceps continue doing positive work,
propulsion step with major positive work by the anterior tibialis and the swing step
where no force is being exerted to the ground. Despite the absence of ground contact
some muscular activity can still be found in anterior tibialis, hallucius and digitorum
longus extensor (Norkin, 1992).
10
Instrumentation
In order to study the gait and postural control mechanisms force plate, EMG,
visual tracking, kinematics as well as many other techniques can be applied. These
techniques work on the premise that the human body is not statically stable so a
control system is necessary to stabilize the body. Some of the techniques allow
monitoring the body center of pressure while others measure the small electric
potentials required to activate a muscle and other the position of the body segments.
It is known that ageing, lesion, footwear and many others factors contribute to
an abnormal postural and gate pattern. As the feet reaches the ground in different
angles or as the center of mass displaces from its regular position the stability
movements required to hold the body change.
Biomechanical instrumentations give us the ability to study gait and postural
patterns as well as all the variables and diseases that can change a normal pattern
increasing energy expenditure and increasing the probability of a fall. This is specially
important for elderly people as this is one of the major causes of dead and health
related problems (Stevens et al, 2006). It also allows quantifying motor impairments
and to keep tracking of diseases progressions. This kind of instrumentation is also used
in rehabilitation to monitor improvements of stroke patients.
11
Force plate
Force plates are used by many researchers to evaluate the postural stability by
analyzing the time-varying coordinates of the center of pressure. These measurements
have been shown to be highly correlated with the center of mass, this way the
horizontal or vertical ground reaction forces are proportional to its respective center of
mass acceleration (Hasan et al 1996).
There are several types of force platforms depending on its transducer being
the most common the strain gauge and the piezoelectric. These sensors are between
two rigid plates to minimize energy absorption in order to acquire the most precise
data possible. The strain gauge varies its resistance with an according deformation and
the piezoelectric generates a potential difference to an according deformation. The
most common force plates have four sensors located on the four corners of the
platform and each one measure the three force components and the three moments
of those forces as shown in figure 8 (Pais 2005).
Figure 8 - Force platform and measured force components [Pais 2005]
While using force plates it needs to taken into account the plate installation. A
proper installation can increase data quality. Force plates need to be screwed to a solid
and leveled plane. This way the plate is isolated from any vibrations produced in the
vicinity. Also the upper plate must be leveled with the ground so it doesn’t induce gait
alterations (AMTI 2013).
Force plates require auxiliary hardware to work as shown in figure 9. As the
plate output is an analogic signal it need to be converted to a digital signal so a
computer can read the data. This is done with an A/D converter. All the hardware
equipment is connected with shielded cables to avoid noise contamination of the
12
collected signal. There is also a piece of software that manages the data acquisition
and storage.
Figure 9 - Force plate components [AMTI 2013]
13
Electomiography
The activation of the muscle fibers generates an electrical field that can be
measured with electrodes, generating the electromyographic signal (Basmajian and
DeLuca, 1985). This way a direct measurement of the muscular activity can be made.
Also the EMG signal can provide information related to the degree of muscular
activation, the rate of force production, the number of motor units recruited, etc.
Theoretically the EMG signal would correspond to the sum of all motor units action
potentials (Basmajian and DeLuca, 1985).
The electric signals generated by muscular activity can be detected with two
different types of electrodes, intramuscular and surface electrodes (Farina et al, 2004).
The intramuscular electrodes are used to study individual motor units. They are
introduced with a needle and they get in direct contact with the muscles fibers. The
main problems regarding intramuscular electrodes are the limited amount of muscle
fibers that it can detect and the inconvenience related to invasive methods (Basmajian
and DeLuca, 1985). As for the surface electrodes they are painless and have a larger
area of activity can be pickup. The classic systems for surface detection are composed
of two large electrodes. This system delivers the sum of all the electric activity on the
detection area. It allows the study of the global activity of a muscle. Its main
disadvantages are its susceptibility to crosstalk; this means that a muscle activity can
be felt on the vicinity locations inducing to misleading readings (Rau et al. 1997).
Usually the frequency of an EMG signal can go from 25 to several kHz as for the
amplitude it usually has very small values ranging from 0.1 to 1 mV for surface
electrodes (DeLuca, 1993). Due to the small amplitude and the sensitivity of the
electrodes, the signal must be filtered, to avoid low frequencies noise contamination,
and amplified before it can be read by a computer. Figure 10 shows a typical EMG
signal. The main types of signal noise sources are the power supply, which introduces a
50-60Hz noise that can be eliminated with a high pass filter (Raez, 2006) and crosstalk.
Figure 10 - Typical EMG signal [adapted from Biopac Systems]
The processing of the EMG signal can be made either in the time or frequency
domain. The time domain signal analysis objective is to characterize the intensity of
14
the signal during muscle activation. As part of this analysis we have the rectification,
integration, and mean root square. The rectification is the elimination or
transformation of the negative values of the signal into positive ones, figure 11. The
mean root square is the technique that correlates in a more precise way the behavior
of the motor unit during activation. For this reason this is the most used procedure
(Basmajian and DeLuca, 1985). This value depends on the number of motor units
recruited, event duration, electric configuration and the characteristics of the
instruments. Despite the fact that the mean root square and the integral are both a
way of measure the area bellow the signal (figure 11), only the mean root square has a
physical meaning. Also as the mean root square is the mean square sum of the signal
values there’s no need for prior rectification (Oliveira, 2007).
Figure 11 - EMG signal processing example [Biopac Systems]
The Frequency domain signal characterization has been frequently used to
study muscle fatigue (Komi and Tesh, 1979). Stulen and DeLuca (1981) have shown a
linear relation between mean and median frequency and the conduction speed of
action potentials in muscle fibers.
15
Work plan
As the data for this analysis have already been collected from another
researcher there is no need to concern about an experimental procedure for EMG and
force plate. The data is going to be analyzed using Matlab. The first step is to acquire
all the necessary information about the data gathering procedures so frequency of
acquisition, layout and all other relevant experimental variables are checked.
After that a small algorithm for automatic reading and separation of the data is
going to be made. This allows storing each muscle EMG signal and force component on
separate arrays. The necessary normalization, filtering (if necessary), rectification, off-
set removal will then be made. Regarding the force data amplitude, peak intervals and
inflection points will be determined. As for the EMG signal this data will be studied as
for its frequency during activation, its mean root square, amplitude, duration and peak
interval. Both data must then be synchronized to study the influence of specific
muscular groups on the different gait cycle steps. Some modifications to the work plan
may occur if necessary. By the end of this work it is expected that a tool for
automatically analyses of EMG and force data from gait and posture control is created.
16
References
Alonso, V. K. et al. Análises cinemáticas da marcha em pacientes hemiparéticos.
Revista Fisio Brasil 55:16-23, 2002.
AMTI: www.amti.biz accessed in January 2013
Andreia, P. S. Análise da marcha baseada em correlação multifactorial. Dissertação de
mestrado, FEUP-UP, 2009
Basmajian, J. V. DeLuca, C. J. Muscles Alive. Their function revealed by
electromyography. Williams & Wilkins, 1985.
Bawa, P. Neural control of Motor Output: can training change it? Exerc Sport Sci Rev
30:59-63, 2002.
Cavagna, G.; Margaria, R. Mechanics of Walking. Journal of Applied Physiology.
21(1):271-278, 1966.
DeLuca, C. The use of surface electyromyography in biomechanics. Journal of applied
biomechanics, 13:135-163, 1993
Duarte, M. Modelagem do controlo postural humano. Anais IX Congresso Brasileiro da
Sociedade Brasileira de Buiomecânica 25-30, 2001.
Enoka, R. M. Neuromechanics of human Movement. Human Kinetics, 2008.
Farina, D. et al. The extraction of neural strategies from surface EMG. Journal of
Applied Physiology 96:1486-1495, 2004.
Fernades, J. Postural behavioral and visual impairment: the influence of high
performance sport upon the postural behavior of blind or low vision persons. Posture
et equilibre 105-113, 1998
Gamble J. G. et al. Marcha humana. Premier 2nd ed, 1998
Ghez, C.; Krakauer, J. The organization of movement. Principles of neural science,
McGraw-Hill, 2000.
Hasan, S. S. et al. Simultaneous measurement of body center of pressure and center of
gravity during upright stance. Gait and Posture 4:1-20, 1996.
Horak, F. B. Postural orientation and equilibrium: what do we need to know about
neural control of balance to prevent falls? Age and ageing, 2006.
17
Junior, P. F.; Barela, J. A. Alterações no funcionamento do sistema de controlo
postural de idosos. Uso de informação visual. Revista Portuguesa de Ciências do
desporto 6(1):94-105, 1996.
Kernell, D. Principles of force gradation in skeletal muscles. Neural Plast 10:69-76,
2003
Komi, P. V. Tesh, P. EMG frequency spectrum, muscle structure and fatigue during
dynamic contractions in man. European journal of applied physiology, 42:41-50, 1979.
Kuo, D. et al. The Six Determinants of gait in the Inverted Pendulum Analogy: A
Dynamic Walking Perspective. Human Movement Science. 26(4):617-656, 2007.
Loeb, G. E. Ghez, C. The motor unit and muscle action. Principles of neural science,
McGraw-Hill 674-693, 2000
Mackey, D. C.; Robinovitch, S. N. Mechanisms underlying age related differences in
ability to recover balance with the ankle strategy. Gait and Posture 23(1):59, 2006.
Monteiro, M. D. Comportamento postural dinâmico. Tese Doutoramento UTAD, 2004.
Norkin, C.; Levangie, K. Joint Structure and Function. A Comprehensive Analysis.
Library of Congress 2nd ed, 1992.
Oliveira, M. C. Análise electromiográfica da técnica de retropedalagem em polo
aquático. Dissertação de mestrado FADE-UP, 2007
Pais, M. R. Efeito de um programa de actividade física no equilíbrio estático e dinâmico
em idosos. Dissertação de mestrado FCDEF-UP, 2005
Raez, et al. Techniques of EMG signal analysis: Detecting, processing, classifications
and applications. Biology Proceedings, 8:8-35, 2006.
Rau G. et al. Noninvasive approach to motor unit characterization: muscle structure,
membrane dynamics and neuronal control. Journal of Biomechanics, 30:441-446,
1997.
Seeley R. R. Anatomy and Physiology. McGraw-Hill 6th ed, 2003
Shumway-Cook, A.; Woollacott, M. H. Motor control: Theory and practical
applications. Lippincott Williams & Wilkins, 2nd ed, 2001.
Silva, T. C. Análise de marcha em mulheres obesas e a sua relação com o índice de
massa corporal. Dissertação de metrado FCS-UB, 2009.
Stulen, F. B. DeLuca C. J. Frequency parameters of the myoelectric signal as a measure
of muscle conduction velocity. IEEE Trans Biomed Eng, 28:515-523, 1981.