HAL Id: hal-01174114
https://hal.inria.fr/hal-01174114
Submitted on 8 Jul 2015
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of sci-
entic research documents, whether they are pub-
lished or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diusion de documents
scientiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
Identifying representative muscle synergies in overhead
football throws
Ana Lucia Cruz Ruiz, Charles Pontonnier, Anthony Sorel, Georges Dumont
To cite this version:
Ana Lucia Cruz Ruiz, Charles Pontonnier, Anthony Sorel, Georges Dumont. Identifying representative
muscle synergies in overhead football throws. Computer Methods in Biomechanics and Biomedical
Engineering, Taylor & Francis, 2015, pp.2. �10.1080/10255842.2015.1070581�. �hal-01174114�
Identifying representative muscle synergies in overhead football throws
A.L. Cruz Ruiz
a,b
*, C. Pontonnier
a,b,c
, A. Sorel
a
, G. Dumont
a,b
a
IRISA/INRIA MimeTIC, Rennes, France;
b
ENS Rennes, Bruz, France;
c
Ecoles de Saint-Cyr Coëtquidan, Guer, France
Keywords: motor control; synchronous synergies; EMG; matrix factorization
1. Introduction
Human motion control requires the command of a
highly redundant and nonlinear musculoskeletal
system. This complex motor control problem has
been simplified through the theory of muscle
synergies. According to this theory, the control of
motion by the CNS (central nervous system) is made
by translating task-level commands into a reduced
number of modules or synergies (Muceli et al. 2010).
In the current abstract we show the existence of
such synergies for a group of dynamic motions of the
upper-body: overhead football throws. The purpose of
this study was to find a compact control
representation of throwing for the future design of
efficient motion controllers for avatar animation. The
controllers will adapt this representation allowing
muscle-actuated avatars to reproduce the recorded
and new throwing motions.
2. Methods
2.1 Data collection and processing
The motion consisted in a right-hand throw with an
American football to a 4m target (a description of this
motion is featured in Fig.1(top)). The muscle activity
of 16 right arm and trunk muscles was collected from
a healthy 32-year old male (stature, 1.86m, weight 72
kg), using surface electrodes (Cometa Waveplus
EMG system). The EMGs were amplified (gain
1000), digitized (1kHz), band-pass filtered (10-
450Hz), rectified, and low-pass filtered (6Hz). ECG
artifacts were removed using an ICA-based filtering
procedure. Motion was captured with a Vicon system
(15 cameras, 100Hz sampling rate).
2.2 Synchronous synergy model
A synchronous synergy
is defined as a -
dimensional vector of coefficients, capturing a
specific relationship in the strength of activation of -
muscles. Each synergy is paired with a time-varying
combination coefficient vector
, which determines
its temporal evolution. A set of -synergies can be
linearly combined to generate -muscle
patterns
:
(1)
Where, is the
matrix containing
the recorded muscle patterns, is the muscle
synergy matrix, and is the
combination coefficient matrix.
In our case matrix was created by concatenating
the muscle data for 8 throws.
2.3 Synergy extraction
A NMF (Non-negative matrix factorization)
algorithm (Kim & Park 2008) was used to extract a
set of muscle synergies and their corresponding
combination coefficients from the recorded EMG
patterns. Essentially, NMF decomposes a non-
negative matrix into a non-negative linear
combination of basis vectors, by solving the
following optimization problem,
W ,C
min
, subject to (2)
2.4 Synergy model order
The criterion for determining the synergy model order
was based on the average coefficient of determination
between the original and reconstructed muscle
patterns (d'Avella, et al. 2003; Muceli et al. 2010).
The chosen number of synergies corresponded to the
sharpest change in the slope of the
curve. For
throwing, the sharpest change occurred at 3 synergies
(
= 0.7352).
3. Results and discussion
The extracted synergies were used to reconstruct the
16 original muscle patterns of the right arm and trunk,
during an overhead football throw using Equation (1).
The performed motion and example reconstructions
are featured in Fig.1.
*Corresponding author. ana-lucia.cruz-ruiz@inria.fr
The reconstruction performance increases with the
number of synergies, however after 3 synergies there
is no abrupt change in performance.
The extracted synergies and their combination
coefficients are featured in Fig.2. Each synergy
recruits groups of muscles with similar biomechanical
action, and corresponds to specific motion phases.
Figure 1 Motion phases & wrist extensor group
excitation (blue-original, dark blue-reconstructed)
Figure 2 Combination coefficients and synergies
Synergy
, contains the biomechanical actions
needed during the late cocking phase. Therefore, the
most active muscles correspond to actions such as:
shoulder abduction, external rotation and scapular
retraction. This synergy was also active during the
follow through phase, as a decelerator. Synergy
,
comprises actions necessary during the acceleration
phase, such as: trunk extension, rotation and elbow
extension. Finally, synergy
includes actions that
contribute both to the early cocking phase and the
acceleration phase, such as shoulder internal rotation,
elbow and wrist flexion.
We ranked these 3 synergies according to their
degree of contribution or power contained in the
functions
(in the time domain). The synergy
displaying the highest power was synergy
,
followed by synergies
, and
. These results are
consistent with the expectation that power is highest
during the acceleration phase. We also observed that
the synergies themselves, their triggering order, and
degree of contribution were preserved in higher order
synergy models and in models derived from single
throws.
4. Conclusions
We have evidenced the presence of synchronous
muscle synergies in a throwing motion and analysed
their relationship with biomechanical outputs. The
results also evidence the robustness and repeatability
of this components across different throws.
Our future work, includes designing a more generic
synergy model for throwing by including throws to
2m and 7m in the NMF decomposition. We also plan
to validate further the optimal synergy model order
through a more robust measure of reconstruction
performance (ex. coefficient of determination within
a cross-validation procedure). Our final goal will be
to use the extracted components for the construction
of muscle-based controllers for avatar animation.
Acknowledgements
This work is supported by the French ANR project
ENTRACTE (Grant agreement: ANR 13-CORD-
002-01).
References
D'Avella A, Saltiel P, Bizzi E. 2003. Combinations of
muscle synergies in the construction of a natural
motor behavior. Nat. Neuroscience. 6(3): 300–
308.
Kim H, Park H. 2008. Nonnegative matrix
factorization based on alternating nonnegativity
constrained least squares and active set method.
SIAM J. Matrix Anal. Appl. 30(2): 713–730.
Muceli S, Boye AT, d'Avella A, Farina D. 2010.
Identifying representative synergy matrices for
describing muscular activation patterns during
multidirectional reaching in the horizontal
plane. J. Neurophysiol. 103(3): 1532–1542.
0 5 10 15
0
1
2
Time(s)
c2
0 5 10 15
0
1
2
Time(s)
c3
0 5 10 15
0
1
2
Time(s)
c1
Throw2
Throw1
Throw4
Throw5 Throw6
Throw7
Throw3 Throw8
0
0.5
1
w1
ErL
ErIlio
Lasti
TrapU
TrapM
TrapL
DeltP
DeltA
DeltM
PecC
PecS
Bic
TrpLg
TrpLt
WrstE
WrstF
0
0.5
1
w2
ErL
ErIlio
Lasti
TrapU
TrapM
TrapL
DeltP
DeltA
DeltM
PecC
PecS
Bic
TrpLg
TrpLt
WrstE
WrstF
0
0.5
1
w3
ErL
ErIlio
Lasti
TrapU
TrapM
TrapL
DeltP
DeltA
DeltM
PecC
PecS
Bic
TrpLg
TrpLt
WrstE
WrstF