Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Sumit A. Raurale
Department of Electronics & Telecommunication
Government College of Engineering,
Amravati, India-444604
Email: sumit.raurale@gmail.com
AbstractIn the field of Robotics, prosthesis hand amputees time duration. This information is used to specify the function
are highly benefited for various active hand movements based on to be performed for the prosthesis assistive systems. Once the
wrist-hand mobility. The development of an advanced human- function is determined, it can be driven at a constant speed,
machine interface has been an interesting research topic in or its speed may be controlled in a proportional way for
the field of rehabilitation, in which biomedical signals such the myoelectric activities. Although these systems have been
as electromyography (EMG) signals, plays a significant role.
Sensing of EMG signals concerns with the signal capturing,
very successful, the amplitude and rate coded schemes do not
conditioning, feature extraction and classification of different provide sufficient information to reliably control more than one
active hand movements for controlled human-assisting robots or device.
prosthetic applications. This paper concerns with the acquisition The raw EMG contains valuable information about the
and analysis of EMG signals for multiple active hand movements
movements still it cannot be directly useful before quantifica-
based on wrist-hand mobility for control of prosthesis robotic
hand. To recognize the effectiveness of hand prosthesis, Anterior tion. Numerous signal processing techniques are functioned on
and Posterior forearm muscles are being considered for better raw EMG to get precise EMG signal. Several researches have
exploitation of EMG signals. The Feature is extracted using been made in biomedical signal processing for emerging bet-
statistical analysis and pattern classification is done by linear ter algorithms, improvement in methodologies, improvement
discriminant analysis (LDA) with estimated classification rate of in noise less signal detection and to acquire precise EMG
about (80-86)%. signals. Since muscles are deep benith the skin surface, thus
KeywordsActive hand movements; EMG signals; Feature
the power spectrum of EMG is limited to 500Hz. Thus the
extraction; Linear discriminant analysis (LDA); Prosthesis hand. electrodes can be positioned on the lower forearm muscles
which are responsible for hand movements. For acquiring and
discriminate six movements, namely open and close hand,
I. I NTRODUCTION hand pitch up or down, move the thumb in abduction or
adduction for prosthesis hand assessment. Thus the resulting
Bioelectrical signals are generally considered as electric
prosthesis can give the user a more natural grasping movement,
currents produced by the sum of electrical potential differences
also allowed to move the wrist. Furthermore the enhancement
across a dedicated tissue which results in different electro-
can be carried out for different gripping movements and also
chemical proceedings happening in the body. EMG signal is
to discriminate more natural movements. In our work, the
one of the optimum known bioelectrical signals which can be
EMG signals are acquired from just two electrodes which
sensed over the skin surface or directly with contact to the
are optimally processed in real time with statistical feature
muscles which generates the electrical activity of the muscle
evaluation and classification is done by linear discriminant
fibres during contraction or relaxation. Each movement of mus-
analysis (LDA) for higher classification rate.
cles relates to a specific pattern of activation of several muscle
fibres through which acquiring multi channels EMG signals
can be used to identify the corresponding movement. As per II. ACQUISITION OF EMG SIGNALS
the complex nature of EMG signals, comprehensive analysis A. Prosthesis Field
for feature evaluation and classification is often a difficult
task, especially if the EMG relates to dedicated movement. The hand prosthesis are differentiated as active and passive
For this purpose, different pattern recognition analysis which hand prostheses. Active hand prosthesis can freely actuate
consisting of feature extraction and classification, have been the hand movements to some degree with ease in accessible
applied on the raw EMG signals to extract the specific feature. operation as controlled by the patient. Operation wise active
This concept has been used for the development of myoelectric hand prosthesis can be classified as [9]:
prosthesis control systems obtained by feature analysis with Prostheses moved by the patient
pattern classification of EMG signals.
Prostheses with external source of energy, either myo-
Many EMG control systems are currently available that are electric command or electronic command.
capable of controlling a precise robotic prosthetic limb such
as a hand, an elbow, or a wrist. These systems extract control The regulation of an actuated prosthesis is usually based on
information from the EMG signals based on an evaluation of EMG, which is the electric activity of activated muscles, mea-
the amplitude or the rate of change of the EMG accordance to sured from surface electrodes. There are two types of interfaces
between the patient and the prosthesis are observed: invasive of the wrist can be considered in prosthesis field. Since the
and non-invasive. The former gather control signals directly muscles that move the thumb are very deep, we can only
from the users nervous system, either via brain implants or consider thumb extension and leave the control of thumb flex-
surgical use of electrodes which delivers a high signal quality, ion to an empirical supervisor. Some active hand movements
since the signals can be gathered exactly from the right areas; discrimination is illustrated in Fig. 2.
but they involve surgery and all related psychological issues.
On the other hand, non-invasive interfaces are easier to handle
and maintain, but require a much better signal conditioning,
since they usually work with surface signals and observation
tracking where the muscle activation potentials are gathered by
electrodes placed on the patients forearm skin; these potentials
can be used to track which muscles the patient is willing to
activate, and with what force. Surface EMG is therefore, in
principle, a cheap and easy way of detecting what the patient
wants the prosthesis to do.
B. Muscles Acting on Hand Prosthesis Fig. 2. Various Active hand movements: a) wrist extension b)wrist flexion
c)hand opened d)hand closed e) thumb abduction
For signal acquisition and defining scheme, there is a need
to understand which muscles are relevant and how they are
connected to the movements. There are many muscles in the
forearm devoted to move the wrist and the hand [11]; we have D. Electrodes and Signal Acquiring
enlist some important muscles and their related movements. EMG signals are the appearance of impulses which are
The Abductor pollicis muscle is devoted for thumb action; initially generated in the central nervous system and then travel
Extensor digitorum communis muscle is devoted for Extension to their final destination, where they produce the desired action.
of fingers and wrist action; Extensor pollicis longus muscle There are two main kinds of electrodes that might be used
is devoted for extension of thumb and abduction of wrist; for measuring EMG signals: surface and inserted electrodes.
Extensor indicis muscle is devoted for extension of thumb and Inserted electrodes are very thin wires or needles that are
abduction of index; Extensor digiti quinti muscle is devoted inserted inside the muscles which could be painful, and are
for extension of the little finger; Flexor digitorum profundus not removable without surgery [8]. Due to this factors surface
muscle is devoted for flexion of interphalanxes articulations electrodes are mostly preferred which are made of metal and
and wrist; Flexor pollicis longus muscle is devoted for flexion covered with a thin coat of silver chloride (AgCl).
of the thumb as shown in Fig. 1.
The muscles which are responsible of hand motion are
deeply placed in the forearm, while measuring the myoelectric
signals with surface electrodes it is not possible to exclusively
acquire the signal of the involved muscle, but we obtain the
superimposition of every signal generated between the point of
interest and the detection surface. This phenomenon is known
as Cross-talk. By using bipolar electrodes, the common
information registered by both electrodes is mostly eliminated
by a differential amplifier. The acquired signal contains noise
and an offset signal; furthermore not all the sequences are
observed properly as the electrode is in practice a low-pass
filter and distorts some spectral components of the signal. The
cut frequency of 5mm diameter electrode is about 360Hz and
for a 20mm diameter is about 100Hz. Equation 1 shows a
simple model of the EMG signal:
N
X 1
x(n) = h(r)e(n r) + w(n) (1)
r=0
Fig. 1. Classification of different forearm muscles
where x(n), modelled EMG signal; e(n), point processed repre-
senting the firing impulses; h(r), represents the MUAP; w(n),
C. Kinds of movement for the prosthesis zero mean addictive white Gaussian noise and N is the number
of motor unit firings.
Since the hand has many degrees of freedom (20 only
for the fingers) [11] and it is impossible to replicate all of III. D EVELOPMENT OF EMG M ODULE
them in a simple way, thus we only consider the movements
which allows the patient to manipulate objects in a sufficient The different module that evaluates the acquiesced EMG
way. There are only two movements available in commercial signals are Preamplifier; Band pass filter; Amplifier and Ana-
prosthesis, namely opening and closing the hand. Some more log to digital convertion as shown in Fig. 3. In presence of
movements like Abduction of the thumb, extension or flexion muscular activity, the EMG signal is picked up by the two
Fig. 3. EMG signal development modules
Fig. 4. Acquiesced forearm EMG signal pattern of wrist-hand movement X i = X i+1 X i (3)
VI. C ONCLUSION
EMG signal is obtained via the differentiation of individual
action potentials generated by irregular discharges of active
motor units in muscle fibers. It contains rich information that
can make myoelectric control a pioneer solution for prosthesis
application and human-assisting robots. The level of activity
Fig. 8. EMG signals for various phase movements of muscles in contraction is the most important factor to
be recognized in EMG classification. Therefore, we applied
different statistical analysis to the processed EMG signal which
To acquire the EMG signal we used two surface electrodes extracts the feature for prosthesis hand movements; also to
with noise amputations which reduces the complexity of the estimate the signal pattern, variance, MAV and length of signal
circuitry with efficient signal feature and classification anal- results in significant performance followed by Linear analysis
ysis. The data of each subject is used to create different set for feature classification.
of threshold levels for classifying the appropriate pattern of The accomplishments research in this paper have led to
different user active hand movements. The procedure for signal the modified strategies for the improvement of EMG signal
acquisition and pattern classification is as following. acquisition and classification. Since hand motions originates
from the concurrent activation of several small and large
two surface electrodes where applied on the forearm
forearm muscles, thus collecting data from different locality on
and a third on the wrist as reference to the EMG
the skin surface which involves more muscles for classification.
capturing circuit;
Thus, improves the number of functions that a system can
each of the 5 movements are sequentially repeated 10 manipulate efficiently. It also improves accuracy, by providing
times and the EMG signal are recorded for first 200 more discriminative patterns for input signals for each hand
ms each; motion by various subjects examined threshold database. For
development of the simplex EMG signal acquisition system,
each set of 10 movements constitutes a sequence, we has considered minimum 2 numbers of forearm muscles
which is stored in a feature threshold database; in the anterior an posterior forearm from which the efficient
signal classification rate of about (80-86)% is evaluated.
so 50 data sets of EMG signals are obtained from
which a total of 500 data sets for overall database is
created; R EFERENCES
[1] Ali Salman, Javaid Iqbal, Umer Izhar, Optimized Circuit for EMG
data in each pattern are averaged and the 10 move- Signal Processing, IEEE Trans, Vol. 12, No. 7, 2012.
ments are separated and individually evaluated and [2] A. B. Ajiboye, R. F. Weir A Heuristic Fuzzy Logic Approch to EMG
stored accordingly; Pattern Recognition for Multifunctional Prosthesis Control, IEEE Trans
NSR 13, No. 3, Sep-2005.
as the pattern information contained in the first 200 [3] B. Crawford, K. Miller, P. Shenoy and R. Rao, Real-time classification
ms of a movement, so the filter is set accordingly so of electromyographic signals for robotic control, Proc 20th AAAI, pp.
as to sustain the pattern for 200ms time duration only; 523-528, 2005.
[4] Bitzer S, van der Smagt P, Learning EMG control of a robotic
the threshold for every peak value is set for classifi- hand: towards active prostheses, In Proceedings of ICRA, International
cation of different pattern with the varying range. Conference on Robotics and Automation, Orlando, pp. 2819- 2823, 2006.
[5] Claudio Castellini, Patrick van der Smagt, Surface EMG in advanced
Data have been acquired twice from each person with a hand prosthetics, Springer (Biological cybernetics), Vol. 100, pp. 35-47,
Sep-2009 .
different position of the electrodes: lower (CD) or higher (CP)
on the forearm so as to extract the perfect feature for different [6] Ekvall S, Kragi D Grasp recognition for programming by demonstra-
tion, In Proceedings of ICRA, International Conference on Robotics and
movements. The frequency of acquisition signal is examined Automation, Barcelona, 2008
about (20-500)Hz. Thus, the experimental results also prove [7] Giuseppina Ginia, Matteo Arvettia, Acquisition and analysis of EMG
that appropriate statistical features can be used in various signals to recognize multiple hand movements for prosthetic applica-
classifying muscle activities which can be easily evaluated and tions, ABB (IOS Press), Vol. 9, pp. 145-155, July-2012.
[8] J. Perry, C. Schmidt Easterday, D. J. Antonelli Surface versus intramus-
cular electrodes for electromyography of superficial and deep muscles,
Physics Theory, Vol. 61, pp 7-15, 1981.
[9] J. C. K. Lai, M. P. Schoen, A. Perez Gracia, D. S. Naidu, S. W. Leung,
Prosthetic devices: challenges and implications of robotic implants
and biological interfaces, Proc in Mechanical Part for Engineering in
Medicine, Vol. 2, pp. 173-183, 2007.
[10] M. B. I. Reaz, M. S. Hussain and F. Mohd-Yasin, Techniques of EMG
signal analysis: detection, processing, classification and applications,
Biomedical Proc Online, Vol. 8, No. 1, pp. 11-35, 2006.
[11] Henry Gray, Anatomy of the Human Body, Bartebly.com books on
line, 1918