![](bg1.png)
Sensor Reduction on EMG-based Hand Gesture Classification
Giovanni Costantini
1
, Gianni Saggio
1
, Lucia Quitadamo
1
, Daniele Casali
1
,
Alberto Leggieri
1
and Emanuele Gruppioni
2
1
Departement of Electronic Engineering, University of Rome “Tor Vergata”, Rome, Italy
2
Centro protesi INAIL, Budrio, Bologna, Italy
Keywords: Neural Networks, EMG, Hand-Gesture, Classification, Feature Selection.
Abstract: This work concerns a system based on EMG sensors, signal conditioning circuitry, classification algorithm
based on Artificial Neural Network, and virtual avatar representation, useful to identify hand movements
within a set of five. This is to potentially make any trans-radial upper-limb amputee able to drive a virtual or
real limb prosthetic hand. When using six EMG sensors, the system is able to recognize with an accuracy of
88.8% the gestures performed by a subject, and replicated by an avatar. Here we focused on differences
resulting with the adoption of a different number of sensors and therefore, by means of a very simple
heuristic method, we compared different subsets of features, excluding the less significant sensors. We
found optimal subsets of one, two, three, four and five sensors, demonstrating a decrease of the performance
of only 0.8% when using five sensors, while with three sensors the accuracy can be as high as 81.7%.
1 INTRODUCTION
The electrical activity of a muscle can be detected by
sensors able to convert electro-myogram (EMG)
signals into electric ones. Surface and intramuscular
EMGs differ from invasiveness and feasibility, and
we deal with the surface one for practical reasons.
In the recent years, different systems were
proposed to use surface EMG (sEMG) signal
acquired on human forearms as input data to control
a real prosthesis (Matrone et al., 2010) or a virtual
device (Li et al., 2010), either for interactive or
clinical/rehabilitative (Scheme and Englehart, 2011)
purposes.
Most of the EMG-controlled device users are
radial upper-limb amputees, i.e. amputation occurred
below elbow. For these people, the replacement of
missing arm functionalities could be a significant
improvement to their quality of life. Moreover
research showed that the visual-sensorial feedback
provided by following the prosthetic or virtual hand
movements can be useful to alleviate the phantom
limb pain (Castellini et al., 2009, Alphonso et al.,
2012), an invalidating condition that affects between
50% and 80% of amputees (Flor H, 2002).
Standard EMG-controlled devices have usually
relied on the detection of weak/strong contractions
of just two forearm muscles to perform very simple
movements (e.g. hand opening and closing) and this
has restricted their usability by amputees (Zlotolow
and Kozin, 2012). To avoid these limitations, pattern
recognition on multiple forearm muscle signals has
been proposed to discriminate hand movements
(Chowdhury et al., 2013). Extracted patterns of
EMG activity, which are different for each hand
movement, allow to increase the amount of usable
information and to realize a more natural, and hence
satisfactory, reproduction of the gestures. A pattern
recognition-based system is tipically structured in
three main steps:
1. EMG signal acquisition and condition by means
of an array of sensors and electronic circuitry;
2. feature extraction, consisting in the calculation
of relevant characteristics from the signals, e.g.
mean, energy, waveform length, etc.
(Phinyomark et al. 2012)
3. feature translation, or classification, to assign
the extracted features to the class (gesture) they
most probably belong to.
Once the gesture attempted by the user of the
system is recognized, it can be mapped towards the
controlled device.
In order to develop a fully reliable system to
classify the intended hand gesture of the amputee, it
seems reasonable to utilize as many EMG sensors as
138
Costantini G., Saggio G., Quitadamo L., Casali D., Leggieri A. and Gruppioni E..
Sensor Reduction on EMG-based Hand Gesture Classification.
DOI: 10.5220/0005040501380143
In Proceedings of the International Conference on Neural Computation Theory and Applications (NCTA-2014), pages 138-143
ISBN: 978-989-758-054-3
Copyright
c
2014 SCITEPRESS (Science and Technology Publications, Lda.)