research article a robust nonlinear observer for a class...
TRANSCRIPT
![Page 1: Research Article A Robust Nonlinear Observer for a Class ...downloads.hindawi.com/journals/tswj/2014/215943.pdf · of real neural networks. e use of neural mass models has beenthepreferredapproachsince](https://reader034.vdocuments.us/reader034/viewer/2022042210/5eae538af1d79d650c588f46/html5/thumbnails/1.jpg)
Research ArticleA Robust Nonlinear Observer for a Class of Neural Mass Models
Xian Liu Dongkai Miao and Qing Gao
Key Lab of Industrial Computer Control Engineering of Hebei Province Institute of Electrical Engineering Yanshan UniversityQinhuangdao 066004 China
Correspondence should be addressed to Xian Liu liuxianysueducn
Received 12 November 2013 Accepted 31 January 2014 Published 20 March 2014
Academic Editors G Cheron and X Fan
Copyright copy 2014 Xian Liu et alThis is an open access article distributed under the Creative Commons Attribution License whichpermits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
A new method of designing a robust nonlinear observer is presented for a class of neural mass models by using the Lurrsquoe systemtheory and the projection lemmaThe observer is robust towards input uncertainty andmeasurement noise It is applied to estimatethe unmeasured membrane potential of neural populations from the electroencephalogram (EEG) produced by the neural massmodels An illustrative example shows the effectiveness of the proposed method
1 Introduction
Mathematical modelling provides a powerful tool for study-ing mechanisms involved in the generation of different elec-troencephalogram (EEG) rhythms and neuronal processesof neurological disorders There are two types of approachesto model neural signals One is on the basis of networksbuilt with a large number of elementary cells to describe theactivity of a given system The other is a lumped-parameterapproach in which neural populations are modeled as non-linear oscillators Neural mass models are based on the latterapproach These models comprise macrocolumns or corticalareas and represent themean activity of the whole populationby using one or two state variables It is seldom tractable tomodel EEG signals at the neuronal level due to the complexityof real neural networks The use of neural mass models hasbeen the preferred approach since 1970s Neural mass modelsoriginated from the seminal work of Lopes da Silva et al foralpha rhythm generation [1] and redesigned by Jansen andRit to represent the generation of evoked potentials in thevisual cortex [2] The dynamical analysis [3ndash6] and control[7ndash9] of the neural mass models have been widely studiedover the years Despite the existence of these neural massmodels for simulating distinct rhythms in EEG signals neuralactivity is always measured through observing just a singlevariable such as voltage A combination of noise in neuronsand amplifiers aswell as uncertainties in recording equipmentleads to uncertainty of the measurement The observationof states therefore plays significant roles in neuroscientificstudies for better understanding of the human brain [10]
In general neural mass models can be expressed asnonlinear systems of Lurrsquoe type [11] Observer design fornonlinear systems of Lurrsquoe type [12ndash14] and the neural massmodels [15] has been widely investigated over the years Wehere introduce a newmethod of designing a robust nonlinearobserver for the neural mass modelsThe Lurrsquoe system theoryand new tools in linear matrix inequality (LMI) method[16] are used to obtain the new reformulation We shouldmention that this new reformulation takes input uncertaintyand measurement noise into account The superiority of theproposedmethod is demonstrated in the last section which isdevoted to numerical comparisonsNotation The identity matrix is denoted by 119868 The symmetricblock component of a symmetric matrix is denoted by ⋆ Thevector norm is denoted by | sdot | The 119871
2norm is denoted by
sdot 2 The set of positive real numbers is denoted by R
+
2 Problem Formulation
Let us consider a class of neural mass models that can beformulated as the following mathematical structure
= 119860119909 + 119861119891 (119867119909) + 1198611119906
119911 = 119862119909 + 119863119908
(1)
where 119909 isin R119899 is the state vector 119906 isin R119901 is the input 119911 isin R119902
is the measurement output 119908 isin R119904 is the measurementnoise 119860 isin R119899times119899 119861 isin R119899times119898 119861
1isin R119899times119901 119862 isin R119902times119899
119867 isin R119898times119899 and 119863 isin R119902times119904 are constant matrices and
Hindawi Publishing Corporatione Scientific World JournalVolume 2014 Article ID 215943 5 pageshttpdxdoiorg1011552014215943
2 The Scientific World Journal
119891(sdot) R119899 rarr R119898 is a memoryless nonlinear vector valuedfunction which is continuously differentiable on R119898 Eachentry of the state-dependent nonlinearity119891(119867119909) is a functionof a linear combination of the states
119891119897= 119891119897(
119899
sum
119895=1
ℎ119897119895119909119895) = 119891
119897(ℎ119897119909) 119897 = 1 2 119898 (2)
where ℎ119897= [ℎ1198971 sdot sdot sdot ℎ
119897119899] It satisfies certain slope-restrictedcondition
0 le 1198911015840
119897(120590) le 120575
119897 forall120590 isin R 119897 = 1 2 119898 (3)
where 120575119897
ge 0 The models in David and Friston [4]Goodfellow et al [6] Jansen and Rit [2] and Wendling et al[3] can all be expressed as the form of (1) Let us construct thefollowing observer for plant (1)
119909 = 119860119909 + 119861119891 [119867119909 + 119870 (119862119909 minus 119911)] + 119871 (119862119909 minus 119911) + 1198611(119906 + 119889)
(4)
where 119909 is the estimation of state 119889 isin R119901 is the disturbanceof input and 119870 isin R119898times119902 119871 isin R119899times119902 are the observer matricesto be designed Defining the observer error as 119890 = 119909 minus 119909 thedynamics of it are governed by
119890 = (119860 + 119871119862) 119890 + 1198611119889 minus 119871119863119908 + 119861120578 (119890 119905) (5)
where 120578(119890 119905) = 119891(119881) minus 119891(119880) 119881 = 119867119909 + 119870(119862119909 minus 119911) and119880 = 119867119909 Note from (3) that each entry of the nonlinearity120578(119890 119905) satisfies
0 le120578119897(119890 119905)
V119897minus 119906119897
=119891119897(V119897) minus 119891119897(119906119897)
V119897minus 119906119897
le 120575119897 forall119890 isin R
forall119905 isin R+ 119897 = 1 2 119898
(6)
The observer design for (1) consists in finding observermatrices 119870 and 119871 such that the observer error 119890 satisfies thefollowing property for all 119905 ge 0
1198902 le 12058110038161003816100381610038161198900
1003816100381610038161003816 + 1205881198891198892 + 120588
1199081199082 (7)
where scalars 120581 gt 0 120588119889ge 0 and 120588
119908ge 0 The disturbance
gains from 119889 and 119908 to 119890 are 120588119889and 120588119908
3 Main Results
Theorem 1 Consider plant (1) and observer (4) Under theslope restrictions (3) if there exist a matrix 119875 = 119875
119879gt 0 a
diagonal matrix 119872 = diag(1198981 119898
119898) ge 0 matrices 119870 and
119871 nonsingular matrices 119866 and 119865 with appropriate dimensionsand scalar constants 120576 gt 0 120583
119908ge 0 and 120583
119889ge 0 such that
[[[[[[[[[[[[
[
119866 + 119866119879
Γ1
119866119861 1198661198611
minus119866119871119863
⋆ Γ2
Γ3
1198651198611
minus119865119871119863
⋆ ⋆ minus119872 0 minus1
2Δ119872119870119863
⋆ ⋆ ⋆ minus120583119889119868 0
⋆ ⋆ ⋆ ⋆ minus120583119908119868
]]]]]]]]]]]]
]
le 0
Γ1= 119866 (119860 + 119871119862) + 119875 minus 119865
119879
Γ2= 119865 (119860 + 119871119862) + (119860 + 119871119862)
119879119865119879+ 120576119868
Γ3= 119865119861 +
1
2(119867 + 119870119862)
119879119872Δ
(8)
then the observer error 119890 satisfies (7) for all 119905 ge 0 where Δ =
diag(1205751 120575
119898) 120581 = radic120582max(119875)120576 120588119889 = radic120583
119889120576 and 120588
119908=
radic120583119908120576
Proof The inequality (8) can be written as
Ω + 1198811[119866
119865]1198812+ 119881119879
2[119866
119865]
119879
119881119879
1le 0 (9)
where
Ω =
[[[[[[[[[[[[[
[
0 119875 0 0 0
⋆ 1205761198681
2(119867 + 119870119862)
119879119872Δ 0 0
⋆ ⋆ minus119872 0 minus1
2Δ119872119870119863
⋆ ⋆ ⋆ minus120583119889119868 0
⋆ ⋆ ⋆ ⋆ minus120583119908119868
]]]]]]]]]]]]]
]
1198811= [
119868 0 0 0 0
0 119868 0 0 0]
119879
1198812= [minus119868 119860 + 119871119862 119861 119861
1minus119871119863]
(10)
By using the well-known projection lemma in LMI method[16] (9) can be transformed into
[[[[[[[[[
[
Π1
Π2
1198751198611
minus119875119871119863
⋆ minus119872 0 minus1
2Δ119872119870119863
⋆ ⋆ minus120583119889119868 0
⋆ ⋆ ⋆ minus120583119908119868
]]]]]]]]]
]
le 0 (11)
where
Π1= (119860 + 119871119862)
119879119875 + 119875 (119860 + 119871119862) + 120576119868
Π2= 119875119861 +
1
2(119867 + 119870119862)
119879119872Δ
(12)
The Scientific World Journal 3
The derivative of 119881(119890) = 119890119879119875119890 is given by
= 119890119879119875119890 + 119890
119879119875 119890 le 119890
119879119875119890 + 119890
119879119875 119890
minus
119898
sum
119897=1
119898119897120578119897(119890 119905) [120578
119897(119890 119905) minus 120575
119897(V119897minus 119906119897)]
(13)
Applying (11) we have
le minus120576119890119879119890 + 120583119889119889119879119889 + 120583119908119908119879119908 (14)
from which it follows that
1198902le radic
120582max(119875)
120576
100381610038161003816100381611989001003816100381610038161003816 +
radic120583119889
1205761198892 + radic
120583119908
1205761199082
(15)
Hence (7) results from 120581 = radic120582max(119875)120576 120588119889 = radic120583119889120576 and
120588119908= radic120583119908120576
Theorem 1 shows that the observer design for (1) consistsin finding observer matrices 119870 and 119871 to satisfy (8) witha symmetric matrix 119875 gt 0 a diagonal matrix 119872 ge 0nonsingularmatrices119866119865 and scalar constants 120576 gt 0 120583
119908ge 0
and 120583119889ge 0 The feasible solution of (8) can be obtained by
solving the following optimization problem
minmax 120583119908 120583119889
st (8) 119875 gt 0 119872 ge 0 120576 gt 0 120583119908ge 0 120583
119889ge 0
(16)
Efficient numerical tools such as YALMIP in MATLAB areavailable for this task Once the values of 120583
119908and 120583
119889are
computed the disturbance gains 120588119908
and 120588119889can also be
derived When no input uncertainty and measurement noiseare taken into account Theorem 1 is simplified as follows
Theorem 2 Consider plant (1) and observer (4) with 119889 =
0 and 119908 = 0 Under the slope restrictions (3) if thereexist a matrix 119875 = 119875
119879gt 0 a diagonal matrix 119872 =
diag(1198981 119898
119898) ge 0 matrices119870 119871 nonsingular matrices 119866
119865 and scalar constants 120576 gt 0 120583119908ge 0 such that
[[[[
[
119866 + 119866119879
Γ1
119866119861
⋆ Γ2
Γ3
⋆ ⋆ minus119872
]]]]
]
le 0 (17)
where Γ1 Γ2 and Γ
3are defined as Theorem 1 then the origin
of the observer error system (5) is globally exponentially stable
4 Simulations
Let us consider a neural mass model developed by Jansenand Rit [2] This type of single cortical column model withaltered parameters is able to generate realistic patterns suchas alpha rhythms and epileptiform spikes in EEG It can beformulated as the form of (1) with the state vector 119909 =
[11990911199092119909311990941199095119909611990971199098]119879 where 119909
119894(119894 = 1 3 5 7) are
the mean membrane postsynaptic potentials and 119909119895(119895 =
2 4 6 8) are their time derivatives The input 119906 is the afferentinfluence from neighbouring or more distant columns and ismodeled by a Gaussian white noise with mean value 90 andstandard deviation 30The output 119911 is the EEGmeasurementavailable to the observer The system matrices are as follows
119860 = diag (1198601 119860
4)
119860119894= [
[
0 1
minus1205812
119894minus2120581119894
]
]
1205811= 1205812= 119886 120581
3= 119887
1205814= 119886119889
1198611= [0 0 0 120579
119886119886 0 0 0 0]
119879
119862 = [0 0 1 0 minus1 0 0 0] 119863 = 1
119861 =
[[[[
[
0 120579119886119886 0 0 0 0 0 120579
119886119886119889
0 0 0 12057911988611988611986220 0 0 0
0 0 0 0 0 12057911988711988711986240 0
]]]]
]
119879
119867 =
[[[[
[
0 0 1 0 minus1 0 0 0
11986210 0 0 0 0 0 0
11986230 0 0 0 0 0 0
]]]]
]
119891 (119867119909) = [119878 (1199093 minus 1199095) 119878 (119862
11199091) 119878 (119862
31199091)]119879
119878 (V) =21198900
1 + 119890119903(V0minusV)
(18)
The function 119878(sdot) satisfies (3) with 120575119897
= (12)1198900119903 (119897 =
1 2 3) All values of the constants in the model are set on aphysiological interpretation basis which can be found in [2]The standard values of these constants are given anatomicallyas
120579119886= 325mV 120579
119887= 22mV 119886 = 100 sminus1 119887 = 50 sminus1
V0= 6mV 119890
0= 25 sminus1
119903 = 056mVminus1 119886119889= 33 sminus1 119862
1= 135
1198622= 108 119862
3= 3375 119862
4= 3375
(19)
We design the robust nonlinear observer (4) for the neuralmass model The performance of the observer obtained fromTheorem 1 is presented in what follows Input disturbance119889 sim 119873(0 01
2) and measurement noise 119908 sim 119873(0 09
2) are
introduced in the design of robust nonlinear observer For therobust nonlinear observer we solve the optimization problem(16) to obtain119870 and 119871 The computed disturbance gains 120588
119908=
596 and 120588119889= 353 are derived by using the YALMIP toolbox
in MATLAB They are much less than the values given in[15] In the following simulations the initial states of theneural mass model and the observer are chosen as 119909(0) =
[1 05 1 05 1 05 1 05]119879 and 119909(0) = [0 0 0 0 0 0 0 0]
119879respectively Figure 1 presents the time evolutions of the states
4 The Scientific World Journal
0 05 1 15 2minus05
0
05
1
15
18 19 2005
01
015
0 05 1 15 2minus40
minus30
minus20
minus10
0
10
18 19 2minus2
minus1
0
1
2
0 05 1 15 2minus20
minus10
0
10
20
30
18 19 222
23
24
25
26
0 05 1 15 2minus500
0
500
1000
18 19 2minus400minus200
0200400
0 05 1 15 2minus10
0
10
20
30
40
18 19 210
15
20
0 05 1 15 2minus1000
minus500
0
500
1000
1500
18 19 2minus200
0
200
x1
and
its es
timat
ion
x2
and
its es
timat
ion
x3
and
its es
timat
ion
x4
and
its es
timat
ion
x5
and
its es
timat
ion
x6
and
its es
timat
ion
t (s)
t (s)
t (s) t (s)
t (s)
t (s)
0 05 1 15 2minus05
0
05
1
15
18 19 203
032
034
036
0 05 1 15 2minus15
minus10
minus5
0
5
10
18 19 2minus2
0
2
x7
and
its es
timat
ion
x8
and
its es
timat
ion
t (s) t (s)
Figure 1 The time evolutions of the states 1199091ndash1199098and their estimations
The Scientific World Journal 5
1199091ndash1199098(black lines) and their estimations that is the states
of observer (4) proposed in this study (red lines) and in [15](blue line) Insets are given to show the zoom-in on dataFigure 1 shows that the states of observer (4) obtained fromTheorem 1 do converge to a neighbourhoodof the states of theneural mass model It also shows that the observer proposedin this study performs better than that proposed in [15]
5 Conclusions
We have designed a robust nonlinear observer for a classof neural mass models by using the Lurrsquoe system theoryand the projection lemma The resulting observer inhibitsinput uncertainty and measurement noise We apply thisobserver to the neural mass model that generates alpharhythms to estimate the mean membrane potential of neuralpopulations from the EEG measurement We show that theproposed observer performs better than some existing onesThe proposed method can also be applied to other types ofneural models that have the typical structure of Lurrsquoe systems
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
This research is supported by the National Natural ScienceFoundation of China (61004050 61172095)
References
[1] F H Lopes da Silva A Hoeks H Smits and L H ZetterbergldquoModel of brain rhythmic activityrdquo Biological Cybernetics vol15 no 1 pp 27ndash37 1974
[2] B H Jansen and V G Rit ldquoElectroencephalogram and visualevoked potential generation in a mathematical model of cou-pled cortical columnsrdquo Biological Cybernetics vol 73 no 4 pp357ndash366 1995
[3] F Wendling J J Bellanger F Bartolomei and P ChauvelldquoRelevance of nonlinear lumped-parameter models in theanalysis of depth-EEG epileptic signalsrdquo Biological Cyberneticsvol 83 no 4 pp 367ndash378 2000
[4] O David andK J Friston ldquoA neuralmassmodel forMEGEEGcoupling and neuronal dynamicsrdquo NeuroImage vol 20 no 3pp 1743ndash1755 2003
[5] A Babajani and H Soltanian-Zadeh ldquoIntegrated MEGEEGand fMRI model based on neural massesrdquo IEEE Transactionson Biomedical Engineering vol 53 no 9 pp 1794ndash1801 2006
[6] M Goodfellow K Schindler and G Baier ldquoSelf-organisedtransients in a neural mass model of epileptogenic tissuedynamicsrdquo NeuroImage vol 59 no 3 pp 2644ndash2660 2012
[7] X Liu H J Liu Y G Tang and Q Gao ldquoFuzzy PID controlof epileptiform spikes in a neural mass modelrdquo NonlinearDynamics vol 71 no 1-2 pp 13ndash23 2013
[8] X Liu Q Gao B W Ma J J Du and W J Ren ldquoAnalysis andcontrol of epileptiform spikes in a class of neural mass modelsrdquoJournal of Applied Mathematics vol 2013 Article ID 792507 11pages 2013
[9] X Liu and Q Gao ldquoParameter estimation and control fora neural mass model based on the unscented Kalman filterrdquoPhysical Review E vol 88 no 4 Article ID 042905 2013
[10] S J Schiff Computational Neuroscience Neural Control Engi-neering The Emerging Intersection Between Control Theory andNeuroscience The MIT Press London UK 2011
[11] G A Leonov D V Ponomarenko and V B Smirnova Fre-quency-Domain Methods For Nonlinear Analysis Theory andApplications World Scientific Singapore 1996
[12] M Arcak and P Kokotovic ldquoNonlinear observers a circlecriterion design and robustness analysisrdquo Automatica vol 37no 12 pp 1923ndash1930 2001
[13] X Fan and M Arcak ldquoObserver design for systems withmultivariable monotone nonlinearitiesrdquo Systems and ControlLetters vol 50 no 4 pp 319ndash330 2003
[14] A Zemouche and M Boutayeb ldquoA unified 119867infin
adaptiveobserver synthesis method for a class of systems with bothLipschitz and monotone nonlinearitiesrdquo Systems and ControlLetters vol 58 no 4 pp 282ndash288 2009
[15] M Chong R Postoyan D Nesic L Kuhlmann and AVarsavsky ldquoA robust circle criterion observer with applicationto neural mass modelsrdquo Automatica vol 48 no 11 pp 2986ndash2989 2012
[16] S Boyd L E Ghaoui E Feron and V Balakrishnan Lin-ear Matrix Inequalities in System and Control Theory SIAMPhiladelphia USA 1994
Submit your manuscripts athttpwwwhindawicom
Neurology Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Alzheimerrsquos DiseaseHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Research and TreatmentSchizophrenia
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Neural Plasticity
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Parkinsonrsquos Disease
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Research and TreatmentAutism
Sleep DisordersHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Neuroscience Journal
Epilepsy Research and TreatmentHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Psychiatry Journal
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational and Mathematical Methods in Medicine
Depression Research and TreatmentHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Brain ScienceInternational Journal of
StrokeResearch and TreatmentHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Neurodegenerative Diseases
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Cardiovascular Psychiatry and NeurologyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
![Page 2: Research Article A Robust Nonlinear Observer for a Class ...downloads.hindawi.com/journals/tswj/2014/215943.pdf · of real neural networks. e use of neural mass models has beenthepreferredapproachsince](https://reader034.vdocuments.us/reader034/viewer/2022042210/5eae538af1d79d650c588f46/html5/thumbnails/2.jpg)
2 The Scientific World Journal
119891(sdot) R119899 rarr R119898 is a memoryless nonlinear vector valuedfunction which is continuously differentiable on R119898 Eachentry of the state-dependent nonlinearity119891(119867119909) is a functionof a linear combination of the states
119891119897= 119891119897(
119899
sum
119895=1
ℎ119897119895119909119895) = 119891
119897(ℎ119897119909) 119897 = 1 2 119898 (2)
where ℎ119897= [ℎ1198971 sdot sdot sdot ℎ
119897119899] It satisfies certain slope-restrictedcondition
0 le 1198911015840
119897(120590) le 120575
119897 forall120590 isin R 119897 = 1 2 119898 (3)
where 120575119897
ge 0 The models in David and Friston [4]Goodfellow et al [6] Jansen and Rit [2] and Wendling et al[3] can all be expressed as the form of (1) Let us construct thefollowing observer for plant (1)
119909 = 119860119909 + 119861119891 [119867119909 + 119870 (119862119909 minus 119911)] + 119871 (119862119909 minus 119911) + 1198611(119906 + 119889)
(4)
where 119909 is the estimation of state 119889 isin R119901 is the disturbanceof input and 119870 isin R119898times119902 119871 isin R119899times119902 are the observer matricesto be designed Defining the observer error as 119890 = 119909 minus 119909 thedynamics of it are governed by
119890 = (119860 + 119871119862) 119890 + 1198611119889 minus 119871119863119908 + 119861120578 (119890 119905) (5)
where 120578(119890 119905) = 119891(119881) minus 119891(119880) 119881 = 119867119909 + 119870(119862119909 minus 119911) and119880 = 119867119909 Note from (3) that each entry of the nonlinearity120578(119890 119905) satisfies
0 le120578119897(119890 119905)
V119897minus 119906119897
=119891119897(V119897) minus 119891119897(119906119897)
V119897minus 119906119897
le 120575119897 forall119890 isin R
forall119905 isin R+ 119897 = 1 2 119898
(6)
The observer design for (1) consists in finding observermatrices 119870 and 119871 such that the observer error 119890 satisfies thefollowing property for all 119905 ge 0
1198902 le 12058110038161003816100381610038161198900
1003816100381610038161003816 + 1205881198891198892 + 120588
1199081199082 (7)
where scalars 120581 gt 0 120588119889ge 0 and 120588
119908ge 0 The disturbance
gains from 119889 and 119908 to 119890 are 120588119889and 120588119908
3 Main Results
Theorem 1 Consider plant (1) and observer (4) Under theslope restrictions (3) if there exist a matrix 119875 = 119875
119879gt 0 a
diagonal matrix 119872 = diag(1198981 119898
119898) ge 0 matrices 119870 and
119871 nonsingular matrices 119866 and 119865 with appropriate dimensionsand scalar constants 120576 gt 0 120583
119908ge 0 and 120583
119889ge 0 such that
[[[[[[[[[[[[
[
119866 + 119866119879
Γ1
119866119861 1198661198611
minus119866119871119863
⋆ Γ2
Γ3
1198651198611
minus119865119871119863
⋆ ⋆ minus119872 0 minus1
2Δ119872119870119863
⋆ ⋆ ⋆ minus120583119889119868 0
⋆ ⋆ ⋆ ⋆ minus120583119908119868
]]]]]]]]]]]]
]
le 0
Γ1= 119866 (119860 + 119871119862) + 119875 minus 119865
119879
Γ2= 119865 (119860 + 119871119862) + (119860 + 119871119862)
119879119865119879+ 120576119868
Γ3= 119865119861 +
1
2(119867 + 119870119862)
119879119872Δ
(8)
then the observer error 119890 satisfies (7) for all 119905 ge 0 where Δ =
diag(1205751 120575
119898) 120581 = radic120582max(119875)120576 120588119889 = radic120583
119889120576 and 120588
119908=
radic120583119908120576
Proof The inequality (8) can be written as
Ω + 1198811[119866
119865]1198812+ 119881119879
2[119866
119865]
119879
119881119879
1le 0 (9)
where
Ω =
[[[[[[[[[[[[[
[
0 119875 0 0 0
⋆ 1205761198681
2(119867 + 119870119862)
119879119872Δ 0 0
⋆ ⋆ minus119872 0 minus1
2Δ119872119870119863
⋆ ⋆ ⋆ minus120583119889119868 0
⋆ ⋆ ⋆ ⋆ minus120583119908119868
]]]]]]]]]]]]]
]
1198811= [
119868 0 0 0 0
0 119868 0 0 0]
119879
1198812= [minus119868 119860 + 119871119862 119861 119861
1minus119871119863]
(10)
By using the well-known projection lemma in LMI method[16] (9) can be transformed into
[[[[[[[[[
[
Π1
Π2
1198751198611
minus119875119871119863
⋆ minus119872 0 minus1
2Δ119872119870119863
⋆ ⋆ minus120583119889119868 0
⋆ ⋆ ⋆ minus120583119908119868
]]]]]]]]]
]
le 0 (11)
where
Π1= (119860 + 119871119862)
119879119875 + 119875 (119860 + 119871119862) + 120576119868
Π2= 119875119861 +
1
2(119867 + 119870119862)
119879119872Δ
(12)
The Scientific World Journal 3
The derivative of 119881(119890) = 119890119879119875119890 is given by
= 119890119879119875119890 + 119890
119879119875 119890 le 119890
119879119875119890 + 119890
119879119875 119890
minus
119898
sum
119897=1
119898119897120578119897(119890 119905) [120578
119897(119890 119905) minus 120575
119897(V119897minus 119906119897)]
(13)
Applying (11) we have
le minus120576119890119879119890 + 120583119889119889119879119889 + 120583119908119908119879119908 (14)
from which it follows that
1198902le radic
120582max(119875)
120576
100381610038161003816100381611989001003816100381610038161003816 +
radic120583119889
1205761198892 + radic
120583119908
1205761199082
(15)
Hence (7) results from 120581 = radic120582max(119875)120576 120588119889 = radic120583119889120576 and
120588119908= radic120583119908120576
Theorem 1 shows that the observer design for (1) consistsin finding observer matrices 119870 and 119871 to satisfy (8) witha symmetric matrix 119875 gt 0 a diagonal matrix 119872 ge 0nonsingularmatrices119866119865 and scalar constants 120576 gt 0 120583
119908ge 0
and 120583119889ge 0 The feasible solution of (8) can be obtained by
solving the following optimization problem
minmax 120583119908 120583119889
st (8) 119875 gt 0 119872 ge 0 120576 gt 0 120583119908ge 0 120583
119889ge 0
(16)
Efficient numerical tools such as YALMIP in MATLAB areavailable for this task Once the values of 120583
119908and 120583
119889are
computed the disturbance gains 120588119908
and 120588119889can also be
derived When no input uncertainty and measurement noiseare taken into account Theorem 1 is simplified as follows
Theorem 2 Consider plant (1) and observer (4) with 119889 =
0 and 119908 = 0 Under the slope restrictions (3) if thereexist a matrix 119875 = 119875
119879gt 0 a diagonal matrix 119872 =
diag(1198981 119898
119898) ge 0 matrices119870 119871 nonsingular matrices 119866
119865 and scalar constants 120576 gt 0 120583119908ge 0 such that
[[[[
[
119866 + 119866119879
Γ1
119866119861
⋆ Γ2
Γ3
⋆ ⋆ minus119872
]]]]
]
le 0 (17)
where Γ1 Γ2 and Γ
3are defined as Theorem 1 then the origin
of the observer error system (5) is globally exponentially stable
4 Simulations
Let us consider a neural mass model developed by Jansenand Rit [2] This type of single cortical column model withaltered parameters is able to generate realistic patterns suchas alpha rhythms and epileptiform spikes in EEG It can beformulated as the form of (1) with the state vector 119909 =
[11990911199092119909311990941199095119909611990971199098]119879 where 119909
119894(119894 = 1 3 5 7) are
the mean membrane postsynaptic potentials and 119909119895(119895 =
2 4 6 8) are their time derivatives The input 119906 is the afferentinfluence from neighbouring or more distant columns and ismodeled by a Gaussian white noise with mean value 90 andstandard deviation 30The output 119911 is the EEGmeasurementavailable to the observer The system matrices are as follows
119860 = diag (1198601 119860
4)
119860119894= [
[
0 1
minus1205812
119894minus2120581119894
]
]
1205811= 1205812= 119886 120581
3= 119887
1205814= 119886119889
1198611= [0 0 0 120579
119886119886 0 0 0 0]
119879
119862 = [0 0 1 0 minus1 0 0 0] 119863 = 1
119861 =
[[[[
[
0 120579119886119886 0 0 0 0 0 120579
119886119886119889
0 0 0 12057911988611988611986220 0 0 0
0 0 0 0 0 12057911988711988711986240 0
]]]]
]
119879
119867 =
[[[[
[
0 0 1 0 minus1 0 0 0
11986210 0 0 0 0 0 0
11986230 0 0 0 0 0 0
]]]]
]
119891 (119867119909) = [119878 (1199093 minus 1199095) 119878 (119862
11199091) 119878 (119862
31199091)]119879
119878 (V) =21198900
1 + 119890119903(V0minusV)
(18)
The function 119878(sdot) satisfies (3) with 120575119897
= (12)1198900119903 (119897 =
1 2 3) All values of the constants in the model are set on aphysiological interpretation basis which can be found in [2]The standard values of these constants are given anatomicallyas
120579119886= 325mV 120579
119887= 22mV 119886 = 100 sminus1 119887 = 50 sminus1
V0= 6mV 119890
0= 25 sminus1
119903 = 056mVminus1 119886119889= 33 sminus1 119862
1= 135
1198622= 108 119862
3= 3375 119862
4= 3375
(19)
We design the robust nonlinear observer (4) for the neuralmass model The performance of the observer obtained fromTheorem 1 is presented in what follows Input disturbance119889 sim 119873(0 01
2) and measurement noise 119908 sim 119873(0 09
2) are
introduced in the design of robust nonlinear observer For therobust nonlinear observer we solve the optimization problem(16) to obtain119870 and 119871 The computed disturbance gains 120588
119908=
596 and 120588119889= 353 are derived by using the YALMIP toolbox
in MATLAB They are much less than the values given in[15] In the following simulations the initial states of theneural mass model and the observer are chosen as 119909(0) =
[1 05 1 05 1 05 1 05]119879 and 119909(0) = [0 0 0 0 0 0 0 0]
119879respectively Figure 1 presents the time evolutions of the states
4 The Scientific World Journal
0 05 1 15 2minus05
0
05
1
15
18 19 2005
01
015
0 05 1 15 2minus40
minus30
minus20
minus10
0
10
18 19 2minus2
minus1
0
1
2
0 05 1 15 2minus20
minus10
0
10
20
30
18 19 222
23
24
25
26
0 05 1 15 2minus500
0
500
1000
18 19 2minus400minus200
0200400
0 05 1 15 2minus10
0
10
20
30
40
18 19 210
15
20
0 05 1 15 2minus1000
minus500
0
500
1000
1500
18 19 2minus200
0
200
x1
and
its es
timat
ion
x2
and
its es
timat
ion
x3
and
its es
timat
ion
x4
and
its es
timat
ion
x5
and
its es
timat
ion
x6
and
its es
timat
ion
t (s)
t (s)
t (s) t (s)
t (s)
t (s)
0 05 1 15 2minus05
0
05
1
15
18 19 203
032
034
036
0 05 1 15 2minus15
minus10
minus5
0
5
10
18 19 2minus2
0
2
x7
and
its es
timat
ion
x8
and
its es
timat
ion
t (s) t (s)
Figure 1 The time evolutions of the states 1199091ndash1199098and their estimations
The Scientific World Journal 5
1199091ndash1199098(black lines) and their estimations that is the states
of observer (4) proposed in this study (red lines) and in [15](blue line) Insets are given to show the zoom-in on dataFigure 1 shows that the states of observer (4) obtained fromTheorem 1 do converge to a neighbourhoodof the states of theneural mass model It also shows that the observer proposedin this study performs better than that proposed in [15]
5 Conclusions
We have designed a robust nonlinear observer for a classof neural mass models by using the Lurrsquoe system theoryand the projection lemma The resulting observer inhibitsinput uncertainty and measurement noise We apply thisobserver to the neural mass model that generates alpharhythms to estimate the mean membrane potential of neuralpopulations from the EEG measurement We show that theproposed observer performs better than some existing onesThe proposed method can also be applied to other types ofneural models that have the typical structure of Lurrsquoe systems
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
This research is supported by the National Natural ScienceFoundation of China (61004050 61172095)
References
[1] F H Lopes da Silva A Hoeks H Smits and L H ZetterbergldquoModel of brain rhythmic activityrdquo Biological Cybernetics vol15 no 1 pp 27ndash37 1974
[2] B H Jansen and V G Rit ldquoElectroencephalogram and visualevoked potential generation in a mathematical model of cou-pled cortical columnsrdquo Biological Cybernetics vol 73 no 4 pp357ndash366 1995
[3] F Wendling J J Bellanger F Bartolomei and P ChauvelldquoRelevance of nonlinear lumped-parameter models in theanalysis of depth-EEG epileptic signalsrdquo Biological Cyberneticsvol 83 no 4 pp 367ndash378 2000
[4] O David andK J Friston ldquoA neuralmassmodel forMEGEEGcoupling and neuronal dynamicsrdquo NeuroImage vol 20 no 3pp 1743ndash1755 2003
[5] A Babajani and H Soltanian-Zadeh ldquoIntegrated MEGEEGand fMRI model based on neural massesrdquo IEEE Transactionson Biomedical Engineering vol 53 no 9 pp 1794ndash1801 2006
[6] M Goodfellow K Schindler and G Baier ldquoSelf-organisedtransients in a neural mass model of epileptogenic tissuedynamicsrdquo NeuroImage vol 59 no 3 pp 2644ndash2660 2012
[7] X Liu H J Liu Y G Tang and Q Gao ldquoFuzzy PID controlof epileptiform spikes in a neural mass modelrdquo NonlinearDynamics vol 71 no 1-2 pp 13ndash23 2013
[8] X Liu Q Gao B W Ma J J Du and W J Ren ldquoAnalysis andcontrol of epileptiform spikes in a class of neural mass modelsrdquoJournal of Applied Mathematics vol 2013 Article ID 792507 11pages 2013
[9] X Liu and Q Gao ldquoParameter estimation and control fora neural mass model based on the unscented Kalman filterrdquoPhysical Review E vol 88 no 4 Article ID 042905 2013
[10] S J Schiff Computational Neuroscience Neural Control Engi-neering The Emerging Intersection Between Control Theory andNeuroscience The MIT Press London UK 2011
[11] G A Leonov D V Ponomarenko and V B Smirnova Fre-quency-Domain Methods For Nonlinear Analysis Theory andApplications World Scientific Singapore 1996
[12] M Arcak and P Kokotovic ldquoNonlinear observers a circlecriterion design and robustness analysisrdquo Automatica vol 37no 12 pp 1923ndash1930 2001
[13] X Fan and M Arcak ldquoObserver design for systems withmultivariable monotone nonlinearitiesrdquo Systems and ControlLetters vol 50 no 4 pp 319ndash330 2003
[14] A Zemouche and M Boutayeb ldquoA unified 119867infin
adaptiveobserver synthesis method for a class of systems with bothLipschitz and monotone nonlinearitiesrdquo Systems and ControlLetters vol 58 no 4 pp 282ndash288 2009
[15] M Chong R Postoyan D Nesic L Kuhlmann and AVarsavsky ldquoA robust circle criterion observer with applicationto neural mass modelsrdquo Automatica vol 48 no 11 pp 2986ndash2989 2012
[16] S Boyd L E Ghaoui E Feron and V Balakrishnan Lin-ear Matrix Inequalities in System and Control Theory SIAMPhiladelphia USA 1994
Submit your manuscripts athttpwwwhindawicom
Neurology Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Alzheimerrsquos DiseaseHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Research and TreatmentSchizophrenia
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Neural Plasticity
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Parkinsonrsquos Disease
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Research and TreatmentAutism
Sleep DisordersHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Neuroscience Journal
Epilepsy Research and TreatmentHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Psychiatry Journal
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational and Mathematical Methods in Medicine
Depression Research and TreatmentHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Brain ScienceInternational Journal of
StrokeResearch and TreatmentHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Neurodegenerative Diseases
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Cardiovascular Psychiatry and NeurologyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
![Page 3: Research Article A Robust Nonlinear Observer for a Class ...downloads.hindawi.com/journals/tswj/2014/215943.pdf · of real neural networks. e use of neural mass models has beenthepreferredapproachsince](https://reader034.vdocuments.us/reader034/viewer/2022042210/5eae538af1d79d650c588f46/html5/thumbnails/3.jpg)
The Scientific World Journal 3
The derivative of 119881(119890) = 119890119879119875119890 is given by
= 119890119879119875119890 + 119890
119879119875 119890 le 119890
119879119875119890 + 119890
119879119875 119890
minus
119898
sum
119897=1
119898119897120578119897(119890 119905) [120578
119897(119890 119905) minus 120575
119897(V119897minus 119906119897)]
(13)
Applying (11) we have
le minus120576119890119879119890 + 120583119889119889119879119889 + 120583119908119908119879119908 (14)
from which it follows that
1198902le radic
120582max(119875)
120576
100381610038161003816100381611989001003816100381610038161003816 +
radic120583119889
1205761198892 + radic
120583119908
1205761199082
(15)
Hence (7) results from 120581 = radic120582max(119875)120576 120588119889 = radic120583119889120576 and
120588119908= radic120583119908120576
Theorem 1 shows that the observer design for (1) consistsin finding observer matrices 119870 and 119871 to satisfy (8) witha symmetric matrix 119875 gt 0 a diagonal matrix 119872 ge 0nonsingularmatrices119866119865 and scalar constants 120576 gt 0 120583
119908ge 0
and 120583119889ge 0 The feasible solution of (8) can be obtained by
solving the following optimization problem
minmax 120583119908 120583119889
st (8) 119875 gt 0 119872 ge 0 120576 gt 0 120583119908ge 0 120583
119889ge 0
(16)
Efficient numerical tools such as YALMIP in MATLAB areavailable for this task Once the values of 120583
119908and 120583
119889are
computed the disturbance gains 120588119908
and 120588119889can also be
derived When no input uncertainty and measurement noiseare taken into account Theorem 1 is simplified as follows
Theorem 2 Consider plant (1) and observer (4) with 119889 =
0 and 119908 = 0 Under the slope restrictions (3) if thereexist a matrix 119875 = 119875
119879gt 0 a diagonal matrix 119872 =
diag(1198981 119898
119898) ge 0 matrices119870 119871 nonsingular matrices 119866
119865 and scalar constants 120576 gt 0 120583119908ge 0 such that
[[[[
[
119866 + 119866119879
Γ1
119866119861
⋆ Γ2
Γ3
⋆ ⋆ minus119872
]]]]
]
le 0 (17)
where Γ1 Γ2 and Γ
3are defined as Theorem 1 then the origin
of the observer error system (5) is globally exponentially stable
4 Simulations
Let us consider a neural mass model developed by Jansenand Rit [2] This type of single cortical column model withaltered parameters is able to generate realistic patterns suchas alpha rhythms and epileptiform spikes in EEG It can beformulated as the form of (1) with the state vector 119909 =
[11990911199092119909311990941199095119909611990971199098]119879 where 119909
119894(119894 = 1 3 5 7) are
the mean membrane postsynaptic potentials and 119909119895(119895 =
2 4 6 8) are their time derivatives The input 119906 is the afferentinfluence from neighbouring or more distant columns and ismodeled by a Gaussian white noise with mean value 90 andstandard deviation 30The output 119911 is the EEGmeasurementavailable to the observer The system matrices are as follows
119860 = diag (1198601 119860
4)
119860119894= [
[
0 1
minus1205812
119894minus2120581119894
]
]
1205811= 1205812= 119886 120581
3= 119887
1205814= 119886119889
1198611= [0 0 0 120579
119886119886 0 0 0 0]
119879
119862 = [0 0 1 0 minus1 0 0 0] 119863 = 1
119861 =
[[[[
[
0 120579119886119886 0 0 0 0 0 120579
119886119886119889
0 0 0 12057911988611988611986220 0 0 0
0 0 0 0 0 12057911988711988711986240 0
]]]]
]
119879
119867 =
[[[[
[
0 0 1 0 minus1 0 0 0
11986210 0 0 0 0 0 0
11986230 0 0 0 0 0 0
]]]]
]
119891 (119867119909) = [119878 (1199093 minus 1199095) 119878 (119862
11199091) 119878 (119862
31199091)]119879
119878 (V) =21198900
1 + 119890119903(V0minusV)
(18)
The function 119878(sdot) satisfies (3) with 120575119897
= (12)1198900119903 (119897 =
1 2 3) All values of the constants in the model are set on aphysiological interpretation basis which can be found in [2]The standard values of these constants are given anatomicallyas
120579119886= 325mV 120579
119887= 22mV 119886 = 100 sminus1 119887 = 50 sminus1
V0= 6mV 119890
0= 25 sminus1
119903 = 056mVminus1 119886119889= 33 sminus1 119862
1= 135
1198622= 108 119862
3= 3375 119862
4= 3375
(19)
We design the robust nonlinear observer (4) for the neuralmass model The performance of the observer obtained fromTheorem 1 is presented in what follows Input disturbance119889 sim 119873(0 01
2) and measurement noise 119908 sim 119873(0 09
2) are
introduced in the design of robust nonlinear observer For therobust nonlinear observer we solve the optimization problem(16) to obtain119870 and 119871 The computed disturbance gains 120588
119908=
596 and 120588119889= 353 are derived by using the YALMIP toolbox
in MATLAB They are much less than the values given in[15] In the following simulations the initial states of theneural mass model and the observer are chosen as 119909(0) =
[1 05 1 05 1 05 1 05]119879 and 119909(0) = [0 0 0 0 0 0 0 0]
119879respectively Figure 1 presents the time evolutions of the states
4 The Scientific World Journal
0 05 1 15 2minus05
0
05
1
15
18 19 2005
01
015
0 05 1 15 2minus40
minus30
minus20
minus10
0
10
18 19 2minus2
minus1
0
1
2
0 05 1 15 2minus20
minus10
0
10
20
30
18 19 222
23
24
25
26
0 05 1 15 2minus500
0
500
1000
18 19 2minus400minus200
0200400
0 05 1 15 2minus10
0
10
20
30
40
18 19 210
15
20
0 05 1 15 2minus1000
minus500
0
500
1000
1500
18 19 2minus200
0
200
x1
and
its es
timat
ion
x2
and
its es
timat
ion
x3
and
its es
timat
ion
x4
and
its es
timat
ion
x5
and
its es
timat
ion
x6
and
its es
timat
ion
t (s)
t (s)
t (s) t (s)
t (s)
t (s)
0 05 1 15 2minus05
0
05
1
15
18 19 203
032
034
036
0 05 1 15 2minus15
minus10
minus5
0
5
10
18 19 2minus2
0
2
x7
and
its es
timat
ion
x8
and
its es
timat
ion
t (s) t (s)
Figure 1 The time evolutions of the states 1199091ndash1199098and their estimations
The Scientific World Journal 5
1199091ndash1199098(black lines) and their estimations that is the states
of observer (4) proposed in this study (red lines) and in [15](blue line) Insets are given to show the zoom-in on dataFigure 1 shows that the states of observer (4) obtained fromTheorem 1 do converge to a neighbourhoodof the states of theneural mass model It also shows that the observer proposedin this study performs better than that proposed in [15]
5 Conclusions
We have designed a robust nonlinear observer for a classof neural mass models by using the Lurrsquoe system theoryand the projection lemma The resulting observer inhibitsinput uncertainty and measurement noise We apply thisobserver to the neural mass model that generates alpharhythms to estimate the mean membrane potential of neuralpopulations from the EEG measurement We show that theproposed observer performs better than some existing onesThe proposed method can also be applied to other types ofneural models that have the typical structure of Lurrsquoe systems
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
This research is supported by the National Natural ScienceFoundation of China (61004050 61172095)
References
[1] F H Lopes da Silva A Hoeks H Smits and L H ZetterbergldquoModel of brain rhythmic activityrdquo Biological Cybernetics vol15 no 1 pp 27ndash37 1974
[2] B H Jansen and V G Rit ldquoElectroencephalogram and visualevoked potential generation in a mathematical model of cou-pled cortical columnsrdquo Biological Cybernetics vol 73 no 4 pp357ndash366 1995
[3] F Wendling J J Bellanger F Bartolomei and P ChauvelldquoRelevance of nonlinear lumped-parameter models in theanalysis of depth-EEG epileptic signalsrdquo Biological Cyberneticsvol 83 no 4 pp 367ndash378 2000
[4] O David andK J Friston ldquoA neuralmassmodel forMEGEEGcoupling and neuronal dynamicsrdquo NeuroImage vol 20 no 3pp 1743ndash1755 2003
[5] A Babajani and H Soltanian-Zadeh ldquoIntegrated MEGEEGand fMRI model based on neural massesrdquo IEEE Transactionson Biomedical Engineering vol 53 no 9 pp 1794ndash1801 2006
[6] M Goodfellow K Schindler and G Baier ldquoSelf-organisedtransients in a neural mass model of epileptogenic tissuedynamicsrdquo NeuroImage vol 59 no 3 pp 2644ndash2660 2012
[7] X Liu H J Liu Y G Tang and Q Gao ldquoFuzzy PID controlof epileptiform spikes in a neural mass modelrdquo NonlinearDynamics vol 71 no 1-2 pp 13ndash23 2013
[8] X Liu Q Gao B W Ma J J Du and W J Ren ldquoAnalysis andcontrol of epileptiform spikes in a class of neural mass modelsrdquoJournal of Applied Mathematics vol 2013 Article ID 792507 11pages 2013
[9] X Liu and Q Gao ldquoParameter estimation and control fora neural mass model based on the unscented Kalman filterrdquoPhysical Review E vol 88 no 4 Article ID 042905 2013
[10] S J Schiff Computational Neuroscience Neural Control Engi-neering The Emerging Intersection Between Control Theory andNeuroscience The MIT Press London UK 2011
[11] G A Leonov D V Ponomarenko and V B Smirnova Fre-quency-Domain Methods For Nonlinear Analysis Theory andApplications World Scientific Singapore 1996
[12] M Arcak and P Kokotovic ldquoNonlinear observers a circlecriterion design and robustness analysisrdquo Automatica vol 37no 12 pp 1923ndash1930 2001
[13] X Fan and M Arcak ldquoObserver design for systems withmultivariable monotone nonlinearitiesrdquo Systems and ControlLetters vol 50 no 4 pp 319ndash330 2003
[14] A Zemouche and M Boutayeb ldquoA unified 119867infin
adaptiveobserver synthesis method for a class of systems with bothLipschitz and monotone nonlinearitiesrdquo Systems and ControlLetters vol 58 no 4 pp 282ndash288 2009
[15] M Chong R Postoyan D Nesic L Kuhlmann and AVarsavsky ldquoA robust circle criterion observer with applicationto neural mass modelsrdquo Automatica vol 48 no 11 pp 2986ndash2989 2012
[16] S Boyd L E Ghaoui E Feron and V Balakrishnan Lin-ear Matrix Inequalities in System and Control Theory SIAMPhiladelphia USA 1994
Submit your manuscripts athttpwwwhindawicom
Neurology Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Alzheimerrsquos DiseaseHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Research and TreatmentSchizophrenia
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Neural Plasticity
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Parkinsonrsquos Disease
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Research and TreatmentAutism
Sleep DisordersHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Neuroscience Journal
Epilepsy Research and TreatmentHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Psychiatry Journal
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational and Mathematical Methods in Medicine
Depression Research and TreatmentHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Brain ScienceInternational Journal of
StrokeResearch and TreatmentHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Neurodegenerative Diseases
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Cardiovascular Psychiatry and NeurologyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
![Page 4: Research Article A Robust Nonlinear Observer for a Class ...downloads.hindawi.com/journals/tswj/2014/215943.pdf · of real neural networks. e use of neural mass models has beenthepreferredapproachsince](https://reader034.vdocuments.us/reader034/viewer/2022042210/5eae538af1d79d650c588f46/html5/thumbnails/4.jpg)
4 The Scientific World Journal
0 05 1 15 2minus05
0
05
1
15
18 19 2005
01
015
0 05 1 15 2minus40
minus30
minus20
minus10
0
10
18 19 2minus2
minus1
0
1
2
0 05 1 15 2minus20
minus10
0
10
20
30
18 19 222
23
24
25
26
0 05 1 15 2minus500
0
500
1000
18 19 2minus400minus200
0200400
0 05 1 15 2minus10
0
10
20
30
40
18 19 210
15
20
0 05 1 15 2minus1000
minus500
0
500
1000
1500
18 19 2minus200
0
200
x1
and
its es
timat
ion
x2
and
its es
timat
ion
x3
and
its es
timat
ion
x4
and
its es
timat
ion
x5
and
its es
timat
ion
x6
and
its es
timat
ion
t (s)
t (s)
t (s) t (s)
t (s)
t (s)
0 05 1 15 2minus05
0
05
1
15
18 19 203
032
034
036
0 05 1 15 2minus15
minus10
minus5
0
5
10
18 19 2minus2
0
2
x7
and
its es
timat
ion
x8
and
its es
timat
ion
t (s) t (s)
Figure 1 The time evolutions of the states 1199091ndash1199098and their estimations
The Scientific World Journal 5
1199091ndash1199098(black lines) and their estimations that is the states
of observer (4) proposed in this study (red lines) and in [15](blue line) Insets are given to show the zoom-in on dataFigure 1 shows that the states of observer (4) obtained fromTheorem 1 do converge to a neighbourhoodof the states of theneural mass model It also shows that the observer proposedin this study performs better than that proposed in [15]
5 Conclusions
We have designed a robust nonlinear observer for a classof neural mass models by using the Lurrsquoe system theoryand the projection lemma The resulting observer inhibitsinput uncertainty and measurement noise We apply thisobserver to the neural mass model that generates alpharhythms to estimate the mean membrane potential of neuralpopulations from the EEG measurement We show that theproposed observer performs better than some existing onesThe proposed method can also be applied to other types ofneural models that have the typical structure of Lurrsquoe systems
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
This research is supported by the National Natural ScienceFoundation of China (61004050 61172095)
References
[1] F H Lopes da Silva A Hoeks H Smits and L H ZetterbergldquoModel of brain rhythmic activityrdquo Biological Cybernetics vol15 no 1 pp 27ndash37 1974
[2] B H Jansen and V G Rit ldquoElectroencephalogram and visualevoked potential generation in a mathematical model of cou-pled cortical columnsrdquo Biological Cybernetics vol 73 no 4 pp357ndash366 1995
[3] F Wendling J J Bellanger F Bartolomei and P ChauvelldquoRelevance of nonlinear lumped-parameter models in theanalysis of depth-EEG epileptic signalsrdquo Biological Cyberneticsvol 83 no 4 pp 367ndash378 2000
[4] O David andK J Friston ldquoA neuralmassmodel forMEGEEGcoupling and neuronal dynamicsrdquo NeuroImage vol 20 no 3pp 1743ndash1755 2003
[5] A Babajani and H Soltanian-Zadeh ldquoIntegrated MEGEEGand fMRI model based on neural massesrdquo IEEE Transactionson Biomedical Engineering vol 53 no 9 pp 1794ndash1801 2006
[6] M Goodfellow K Schindler and G Baier ldquoSelf-organisedtransients in a neural mass model of epileptogenic tissuedynamicsrdquo NeuroImage vol 59 no 3 pp 2644ndash2660 2012
[7] X Liu H J Liu Y G Tang and Q Gao ldquoFuzzy PID controlof epileptiform spikes in a neural mass modelrdquo NonlinearDynamics vol 71 no 1-2 pp 13ndash23 2013
[8] X Liu Q Gao B W Ma J J Du and W J Ren ldquoAnalysis andcontrol of epileptiform spikes in a class of neural mass modelsrdquoJournal of Applied Mathematics vol 2013 Article ID 792507 11pages 2013
[9] X Liu and Q Gao ldquoParameter estimation and control fora neural mass model based on the unscented Kalman filterrdquoPhysical Review E vol 88 no 4 Article ID 042905 2013
[10] S J Schiff Computational Neuroscience Neural Control Engi-neering The Emerging Intersection Between Control Theory andNeuroscience The MIT Press London UK 2011
[11] G A Leonov D V Ponomarenko and V B Smirnova Fre-quency-Domain Methods For Nonlinear Analysis Theory andApplications World Scientific Singapore 1996
[12] M Arcak and P Kokotovic ldquoNonlinear observers a circlecriterion design and robustness analysisrdquo Automatica vol 37no 12 pp 1923ndash1930 2001
[13] X Fan and M Arcak ldquoObserver design for systems withmultivariable monotone nonlinearitiesrdquo Systems and ControlLetters vol 50 no 4 pp 319ndash330 2003
[14] A Zemouche and M Boutayeb ldquoA unified 119867infin
adaptiveobserver synthesis method for a class of systems with bothLipschitz and monotone nonlinearitiesrdquo Systems and ControlLetters vol 58 no 4 pp 282ndash288 2009
[15] M Chong R Postoyan D Nesic L Kuhlmann and AVarsavsky ldquoA robust circle criterion observer with applicationto neural mass modelsrdquo Automatica vol 48 no 11 pp 2986ndash2989 2012
[16] S Boyd L E Ghaoui E Feron and V Balakrishnan Lin-ear Matrix Inequalities in System and Control Theory SIAMPhiladelphia USA 1994
Submit your manuscripts athttpwwwhindawicom
Neurology Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Alzheimerrsquos DiseaseHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Research and TreatmentSchizophrenia
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Neural Plasticity
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Parkinsonrsquos Disease
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Research and TreatmentAutism
Sleep DisordersHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Neuroscience Journal
Epilepsy Research and TreatmentHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Psychiatry Journal
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational and Mathematical Methods in Medicine
Depression Research and TreatmentHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Brain ScienceInternational Journal of
StrokeResearch and TreatmentHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Neurodegenerative Diseases
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Cardiovascular Psychiatry and NeurologyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
![Page 5: Research Article A Robust Nonlinear Observer for a Class ...downloads.hindawi.com/journals/tswj/2014/215943.pdf · of real neural networks. e use of neural mass models has beenthepreferredapproachsince](https://reader034.vdocuments.us/reader034/viewer/2022042210/5eae538af1d79d650c588f46/html5/thumbnails/5.jpg)
The Scientific World Journal 5
1199091ndash1199098(black lines) and their estimations that is the states
of observer (4) proposed in this study (red lines) and in [15](blue line) Insets are given to show the zoom-in on dataFigure 1 shows that the states of observer (4) obtained fromTheorem 1 do converge to a neighbourhoodof the states of theneural mass model It also shows that the observer proposedin this study performs better than that proposed in [15]
5 Conclusions
We have designed a robust nonlinear observer for a classof neural mass models by using the Lurrsquoe system theoryand the projection lemma The resulting observer inhibitsinput uncertainty and measurement noise We apply thisobserver to the neural mass model that generates alpharhythms to estimate the mean membrane potential of neuralpopulations from the EEG measurement We show that theproposed observer performs better than some existing onesThe proposed method can also be applied to other types ofneural models that have the typical structure of Lurrsquoe systems
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
This research is supported by the National Natural ScienceFoundation of China (61004050 61172095)
References
[1] F H Lopes da Silva A Hoeks H Smits and L H ZetterbergldquoModel of brain rhythmic activityrdquo Biological Cybernetics vol15 no 1 pp 27ndash37 1974
[2] B H Jansen and V G Rit ldquoElectroencephalogram and visualevoked potential generation in a mathematical model of cou-pled cortical columnsrdquo Biological Cybernetics vol 73 no 4 pp357ndash366 1995
[3] F Wendling J J Bellanger F Bartolomei and P ChauvelldquoRelevance of nonlinear lumped-parameter models in theanalysis of depth-EEG epileptic signalsrdquo Biological Cyberneticsvol 83 no 4 pp 367ndash378 2000
[4] O David andK J Friston ldquoA neuralmassmodel forMEGEEGcoupling and neuronal dynamicsrdquo NeuroImage vol 20 no 3pp 1743ndash1755 2003
[5] A Babajani and H Soltanian-Zadeh ldquoIntegrated MEGEEGand fMRI model based on neural massesrdquo IEEE Transactionson Biomedical Engineering vol 53 no 9 pp 1794ndash1801 2006
[6] M Goodfellow K Schindler and G Baier ldquoSelf-organisedtransients in a neural mass model of epileptogenic tissuedynamicsrdquo NeuroImage vol 59 no 3 pp 2644ndash2660 2012
[7] X Liu H J Liu Y G Tang and Q Gao ldquoFuzzy PID controlof epileptiform spikes in a neural mass modelrdquo NonlinearDynamics vol 71 no 1-2 pp 13ndash23 2013
[8] X Liu Q Gao B W Ma J J Du and W J Ren ldquoAnalysis andcontrol of epileptiform spikes in a class of neural mass modelsrdquoJournal of Applied Mathematics vol 2013 Article ID 792507 11pages 2013
[9] X Liu and Q Gao ldquoParameter estimation and control fora neural mass model based on the unscented Kalman filterrdquoPhysical Review E vol 88 no 4 Article ID 042905 2013
[10] S J Schiff Computational Neuroscience Neural Control Engi-neering The Emerging Intersection Between Control Theory andNeuroscience The MIT Press London UK 2011
[11] G A Leonov D V Ponomarenko and V B Smirnova Fre-quency-Domain Methods For Nonlinear Analysis Theory andApplications World Scientific Singapore 1996
[12] M Arcak and P Kokotovic ldquoNonlinear observers a circlecriterion design and robustness analysisrdquo Automatica vol 37no 12 pp 1923ndash1930 2001
[13] X Fan and M Arcak ldquoObserver design for systems withmultivariable monotone nonlinearitiesrdquo Systems and ControlLetters vol 50 no 4 pp 319ndash330 2003
[14] A Zemouche and M Boutayeb ldquoA unified 119867infin
adaptiveobserver synthesis method for a class of systems with bothLipschitz and monotone nonlinearitiesrdquo Systems and ControlLetters vol 58 no 4 pp 282ndash288 2009
[15] M Chong R Postoyan D Nesic L Kuhlmann and AVarsavsky ldquoA robust circle criterion observer with applicationto neural mass modelsrdquo Automatica vol 48 no 11 pp 2986ndash2989 2012
[16] S Boyd L E Ghaoui E Feron and V Balakrishnan Lin-ear Matrix Inequalities in System and Control Theory SIAMPhiladelphia USA 1994
Submit your manuscripts athttpwwwhindawicom
Neurology Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Alzheimerrsquos DiseaseHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Research and TreatmentSchizophrenia
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Neural Plasticity
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Parkinsonrsquos Disease
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Research and TreatmentAutism
Sleep DisordersHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Neuroscience Journal
Epilepsy Research and TreatmentHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Psychiatry Journal
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational and Mathematical Methods in Medicine
Depression Research and TreatmentHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Brain ScienceInternational Journal of
StrokeResearch and TreatmentHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Neurodegenerative Diseases
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Cardiovascular Psychiatry and NeurologyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
![Page 6: Research Article A Robust Nonlinear Observer for a Class ...downloads.hindawi.com/journals/tswj/2014/215943.pdf · of real neural networks. e use of neural mass models has beenthepreferredapproachsince](https://reader034.vdocuments.us/reader034/viewer/2022042210/5eae538af1d79d650c588f46/html5/thumbnails/6.jpg)
Submit your manuscripts athttpwwwhindawicom
Neurology Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Alzheimerrsquos DiseaseHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
BioMed Research International
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Research and TreatmentSchizophrenia
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Neural Plasticity
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Parkinsonrsquos Disease
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Research and TreatmentAutism
Sleep DisordersHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Neuroscience Journal
Epilepsy Research and TreatmentHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Psychiatry Journal
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational and Mathematical Methods in Medicine
Depression Research and TreatmentHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Brain ScienceInternational Journal of
StrokeResearch and TreatmentHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Neurodegenerative Diseases
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Cardiovascular Psychiatry and NeurologyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014