See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/350383774

ECG analysis, a hypothesis: are hidden lower harmonics a sign of chest pain to
come?

Method · March 2021

CITATIONS READS

0 700

1 author:

Christian Mira
Free researcher, after retirement, no affiliation
312 PUBLICATIONS 3,444 CITATIONS

SEE PROFILE

All content following this page was uploaded by Christian Mira on 25 March 2021.

The user has requested enhancement of the downloaded file.

1. Introduction.
An ECG (electrocardiogram) is a measurement device which records the signal generated by
heart beats during a small time interval, that of a cardiologist consultation. An ECG holter
records the same signal during a large time interval, generally one day. For healthy relaxed adult
patients, the ECG consists of a basic signal (heartbeat interval, Fig. 1), periodically repeated at a
"normal" (or "fundamental") frequency f f N of heart beats, 50 f N 90Hz. Some signs of heart
disease are: distorsion of this basic signal related to amplitude parameters (QRS, ST), duration
parameters (QRS, QT, and PR), combined parameters (Q/R ratio, S/R ratio), and frequency f out
of the above range. It is also the case of amplitude increase of the signal harmonics, called higher
harmonics (with respect to f N ), but knowing that the use of high-bandpass filters (for the higher
harmonics analysis) may lead to interpretation error of the ST segment [Buendía-Fuentes et al.,
2012]. Lower harmonics are either subharmonics (f f N /q, q integer), or fractional harmonics
(f pf N /q, p,q integers, p q). They are hidden by artifacts (such as muscular activities), and
different types of noises, located in the lower frequencies part of the heart bandwidth. Until now,
it seems that the negative role played by lower harmonics (as sign of health abnormality) has not
been considered while, in engineering, a resulting increase in the oscillation amplitude is source
of possible damages for the system working, by "suffering" of some of its components.
*

Fig. 1. ECG signal interval, periodically repeated at the fundamental frequency rate for a healthy patient.
**
Interpreting ECG abnormalities requires an offline analysis from the record data, analysis
made by a heart rhythm specialist (cardiac electrophysiology). Initially, by considering an ECG
as output of a nonlinear oscillating system, this interpretation has led to a closed co-operation
between cardiologists and signal processing specialists. Since then, the analysis of recorded data
has been the subject of many publications, related to different mathematical methods, with their
adaptation for understanding abnormal heart behaviors. Due to the author’s incompetency in the
field of cardiology, the purpose of this paper is not to review, or to define advantages of each
analysis method. This information is given in specialized publications which refer to signal
processing tools, met under the names: Fourier transform (FT, the better known for estimating
frequency spectra) and its related applications (role of higher harmonics in a diagnosis), fast
Fourier transform (FFT), wavelet transform (WT), artificial-intelligent (AI) approaches, wavelet
neural networks, artificial neural network (ANN), fuzzy hybrid neural networks, grey relational
analysis (GRA), heart rate variability (HRV), vector cardiography (VCG). The traditionally long
introductions of publications in cardiac electrophysiology, and their numerous quotations, deal
with this matter, and develop the matter related to the above abbreviations (for example cf.
[Chia-Hung Lin, 2008] [Chao Lin, 2012]). More elaborated methods are also described in such
publications. Among them: conjoint analysis of the ECG signal in six dimensions [Tereshchenko
& Josephson, 2015], dynamic approach to multifractal analysis [Shekatkar S.M., et al., 2017],
method using Support Vector Machine algorithm (SVM) [Fairooz T. & Khammari H.S., 2016]. It
is also worth to quote the proceedings of the ICASSP (IEEE internal Conference on Acoustic,
Speech and Signal Processing) which has published numerous articles devoted to the heart
dynamics (for example [Fangyou Li et al. 2019], [Dia N. et al. 2019]). In another setting,
[Vaschillo et al., 2011] is devoted to low frequency resonances in the cardiovascular system,
caused by rhythmical muscle tension, but without any relation with the paper hypothesis: lower
harmonics resonance as sign of an eventual chest pain to come.
With respect to these specialized contributions, the purpose of this paper is more modest.
Knowing the "negative role" played by lower harmonics (f f N ) in engineering dynamic systems,
it is about to set the problem of detecting their eventual presence, in a record of ECG holter, this
in the two forms of such harmonics:
(a) subharmonics (f f N /q, q integer), or fractional irreducible harmonics (f pf N /q,
p,q integers, p q, the fraction p/q being mathematically irreducible),
(b) the fraction p/q is mathematically reducible, but the fractional harmonics pf N /q is
"dynamically" irreducible, because the dynamic behaviors are different, before and after the
fraction reduction (cf. below sec. 2, Fig. 2).
In engineering and physics, when lower harmonics are located in the system bandwidth, they
give rise to maxima in the frequency spectrum. This situation leads to a significant increase in
the oscillation amplitude, with possible damages in the system working, due to a "suffering" of
some of its components. For mechanical systems, the amplitude increase can cause mechanical
breaks. In electrical systems, it may result to their destruction by over intensity in some of their
circuits, or to a working not in accordance with what is claimed by the system manufacturer.
From the point of view of dynamic behavior, the "heart system" (or cardiovascular system) is
a nonlinear oscillating system, with its bodily environment behaving like a low-pass filter.
Therefore, the "heart system" shares some common features with the oscillating systems studied
in engineering (or physics). It is in this setting that the paper hypothesis is formulated :
In ECG, an eventual presence of lower harmonics, might be sign of heart exhaustion, and
perhaps even to chest pains to come.
Nevertheless, from a practical point of view, detecting this presence must take into account
some measurement problems in the low frequencies range, owing to lower harmonics (cf. sec. 2)
coexisting with several types of artifacts, which implies particular filtering methods (cf. below
sec. 3).
Until now, the essentially negative role played by lower harmonics, as abnormal heart
behavior, seems to have not been taken into account. Nonetheless, it is worth to mention that, in
another context, subharmonics behaviors on steady states oscillatory potentials elicited in an
electroencephalogram (EEG) by flicker stimulation (so an external excitation) have been studied
[Fangyu Li et al., 2019].
Section 2 is devoted to a short reminder of signal processing approaches and, more
particularly, to the distinction between the two forms of lower harmonics (see above). Section 3
identifies some difficulties for detecting lower harmonics in ECGs, and shows how to overcome
them. Section 4 defines the setting of the paper hypothesis, this on the basis of a similarity of
dynamic behavior common to all oscillating systems. It also mentions some existing heart
mathematical models, and simulators, which might be used for checking the paper hypothesis.
Section 5 proposes a simple example of engineering system, that of a hybrid system (model
associating a discrete part with its switching conditions, and a continuous part). This example
generates two coexisting stable steady states, a "fundamental" oscillation (frequency f f N ), and a
subharmonics one (resonance when f f N /q), each state having its basin (domain of influence in
the phase space). Then, starting from the "fundamental" oscillation, a perturbation of initial
conditions, or a parameters perturbation, may lead to the stable subharmonics oscillation,
associated with a very important increase of the oscillation amplitude (sec. 5, Fig. 5, "image" of a
supposed heart exhaustion?). Section 5 concludes by mentioning the steps which might led to
assess the paper hypothesis.
*
2. Signal processing methods.
Concerning oscillations phenomena, the signal processing field is related to the nonlinear
systems theory (for example cf. [Hundewale, 2012]). This theory was initially based on models of
physical systems described by ordinary differential equations (ODE), and also in the form of
discrete time dynamic models (recurrence relationships). Except some very particular cases,
solutions of such equations are very complex transcendental functions not defined, and not
tabulated, in the classical mathematical analysis. In the general case, the problem is dealt through
three methods referring to three adjectives: analytical, qualitative, numerical.
- Analytical methods (very rarely used in the ECG case) consist in the representation of
nonlinear behaviors (nonclassical transcendental function) by convergent (or at least
asymptotically) convergent series expansions, or via an averaging approach. In the former Soviet
Union, from Poincaré’s results this method was essentially developed by the Kiev School
[Bogoliubov & Mitropolsky, 1958]). An improvement of this approach (by using the notion of
natural parametric dependence) is given in [Gumowski, 1989]. For its part, the book [Hayashi,
1964] associates the analytical approach with numerical methods.
- Qualitative methods, also based on Poincaré’s results, concern the identification of two
spaces associated with behaviors of dynamic systems. The first one, called phase space
represents the nonclassical transcendental functions by their singularities: equilibrium points,
periodic solutions and their stability, phase trajectories, boundary of the influence domain of a
stable stationary state (i.e. basin) etc .... The second one, deals with the evolution of these
singularities when the system parameters vary, in particular via the determination of bifurcation
sets (qualitative changes of a solution) in a parameter space. This type of study has attained its
higher development in the former Soviet Union with the Andronov’ School (or Gorki’ School,
now Nizhny Novgorod) [Boïko, 1983] [Shilnikov L. P., 1997] [Mira, 1996].
- Qualitative methods are associated with numerical ones via adapted algorithms to follow
the bifurcations in the system behavior. In this setting, from the years 1950, the Japanese School
has given a first rank contribution with [Hayashi, 1964] and his disciples: [Kawakami, 1984],
[Ueda, 1991]. For details on Kawakami’s researches, in particular with results on lower
harmonics, cf. the hypertext link
https://www. researchgate.net/scientific-contributions/H-Kawakami-62843113.
In this setting the notion of fractional harmonics f pf N /q, p,q integers, f f N ) must be better
defined by distinction of two cases: (a) the fractional harmonics p/q is mathematically
irreducible, (b) the fraction p/q is mathematically reducible but "dynamically" irreducible,
because the dynamic behaviors before, and after reduction, are different (cf. below Fig. 2). It
seems that a first complete qualitative study has been made in three papers [Kawakami & Sakaï,
1983, 1984, 1988] which give the bifurcation curves related to subharmonic synchronization 1/k,
k 2, . . . , 5. On this basis, the notions of reducible and irreducible fractional harmonics were
defined for the non autonomous Duffing-Rayleigh ordinary differential equation (ODE)
[Kawakami & Sakaï, 1988] [Qriouet & Mira, 1993] [Touzani & Mira, 2000]:
2
dx/dt y, dy/dt 1 dx/dt dx/dt x3 B cos t, 0
 with examples of irreducible fractional harmonics [leading to a stable period k cycle defined
by a Poincaré’ section T at t 2 ], and "dynamically" irreducible fractional harmonics r/ kr ,
k 2, r 1, 3, 4, 5, (same "mathematical" period k of the ODE solution, but differing from their
dynamic behaviors). This point appears in the continuous phase plane [x(t),y(t)]: the ODE
solution r/ kr corresponds to a close curve , distinct from that related to the subharmonics 1/k
(r 1). Fig. 2 shows the reducible harmonics r/ kr , k 2, r 1, 3, 4, 5, defined by the r
points resulting from a correctly chosen local section of .

Fig. 2. 0. 2. Fractional harmonics r/ kr , k 2 (k is the signal period), and closed curve , in the
continuous phase plane [x(t),y(t)]. is designated by Nu (rad/sec). A local section of gives r points. With
r 1, the fractional harmonics 1/2 is mathematically irreducible. When r 3, 4, 5, the ratios r/ kr are
mathematically reducible, but "dynamically" irreducible, because they correspond to different dynamic behaviors.
For each r, one of the point is given by its coordinates x, y .
*
In the continuous phase plane x, y with k 1, mathematically reducible fractional
harmonics, but "dynamically" irreducible lower harmonics r/r, are shown in Fig. 2 of [Mira &
Djellit, 1993]).
*
3. ECG: detection of lower harmonics in artifacts presence.
As common "engineering systems", the "heart system" behaves as a low pass filter which
attenuates harmonics whose frequencies are higher than f N (also called "over- harmonics"), out
of the system bandwidth. In the heart frequency response, this means that lower harmonics (when
they exist) take place in a low frequencies region without signal attenuation. Then, this may give
rise to spectrum local maxima at f f N , whose amplitude can be even greater than that related to
f N . Such a low frequency information is non perceptible with insufficiently large recording
windows, which is not the case with a holter ECG record. Nevertheless, ECG measurements
come up against a problem: the lower harmonics are perturbed by artifacts. The most important
are: baseline wander (caused by movements of the chest, or when an arm or leg is moved during
the ECG data acquisition), inherent noise in electronics equipments (or power line noise),
ambient noise (electromagnetic radiation), EMG (electromyography) noise, and electrode motion
artifacts (changes in the muscle due to relative motion) [Amrutha & Arul, 2011].
Then detecting lower harmonics, hidden by the above artifacts, implies particular filtering
methods (subject of many publications) among them: electronic filter of variable band width
[Koschita et al. 2011, 2013, 2019], [Amrutha & Arul, 2011], Least Mean Squares (LMS) filters.
Other methods are also proposed : photoplethysmograpghy [Fangyu Li, et al., 2019], adaptation
of the harmonic Power Spectral Density, which improves frequency resolution of lower
harmonics [Names et al. 2002], the subharmonics suppression method [Fangyu Li et al., 2019]
which might be inversely adapted as filter (???) for detecting subharmonics, etc.
4. Setting of the paper hypothesis and models of the heart system.
With respect to engineering dynamic systems, it is known that establishing a mathematical
model of concrete "living systems" is a more difficult task. In spite of this problem, the
consideration of the "heart system" as a nonlinear oscillating system (therefore sharing common
dynamic features with "engineering systems") has allowed the development of cardiac
electrophysiology, through the co-operation of specialists in signal processing. It is this context
that led to the paper hypothesis.
On these foundations, sec. 5 proposes an example of engineering system, which shows how
the presence of lower harmonics in the system bandwidth leads to a significant amplitude
increase in the system output, with negative effects on the system working. This example can be
considered as a very rough deductive model (behavioral similarity) of the heart system dynamics.
Before a wide clinical investigation, based on sufficiently numerous analyses of ECG holter
records by heart rhythm specialists (field out of my competency), a validation of the above
hypothesis implies a first step to check it by using more advanced "heart system" (or
cardiovascular system) models proposed by several authors in specialized journals. It’s about to
prompt these authors to observe eventual significant signal increases (cf. below Fig. 5 for having
an idea of the increase ratio) due to lower harmonics of the heart beats basic frequency f N .
About models design, two distinct, but essentially equivalent methods, are available. These
methods constitute merely a particular adaptation of the classical microscopic and macroscopic
viewpoint of physics, leading to what can be respectively called deductive and inductive models
[Gumowski & Mira, 1968] [Mira, 2016b].
- Deductive model. Starting from the observed properties of the complete system, variability
included, a global (phenomenological) model of its behavior is established. The variability of
individual components is expressed implicitly by the variability of the complete system. It is the
way which has led to a co-operation between cardiologists and signal processing specialists for
ECGs interpretation. The advantage of this approach consists in the fact that it is possible to start
from a very coarse mathematical representation, and then add features as required to match the
set of system properties chosen. Such a procedure avoids dealing with component properties,
known or unknown explicitly, which have a negligible effect on the behaviour of the complete
system. The agreement between the mathematical model, and the chosen set of observed system
properties, is established by a process of successive refinements, deduced from a
phenomenological interpretation of the system properties.
At first sight the determination of a deductive model appears to be quite artificial but, if
properly applied, it will generally lead to a simpler mathematical formulation than the inductive
method. Within the specified precision, the agreement between the mathematical model, and the
chosen set of observed system properties, is established by a process of successive refinements.
The refinements, deduced from a phenomenological interpretation of the system properties, are
stopped when the error of the mathematical representation has become sufficiently small.
- Inductive model. Starting from a detailed knowledge of component properties, elementary
interactions between them are determined, and from these interactions properties of the whole
system a mathematical model is deduced. The variability of the system is thus expressed
explicitly in terms of the variability of individual components. The determination of an inductive
model is quite straightforward, but it leads generally to a very unwieldy mathematical
formulation. This unwieldiness is due to the fact that all properties of each individual component
do not have an equally important effect on the properties of the complete system.
4.1. Deductive models of the heart system.
Based on a similarity of dynamic behavior common to all oscillating systems, as above said,
the roughest deductive model is that of this paper, illustrated by the two examples of sec. 5 which
show how lower harmonics, present in the system band width, lead to an amplitude increase in
the system output, with negative effects on the system working.
Deductive mathematical models, with higher degrees of improvement, have already been
proposed by several authors. A first step of the hypothesis validation would be an interest of
these authors in observing eventual significant signal increases due to lower harmonics
resonances. A next step would be a confirmation via cardiovascular system simulators.
Among the different refinements of the heart model, we can quote the use of ordinary
differential equations (ODEs) of the cardiovascular system in [Myers et al., 2017] from ether a
pair of first order ODEs (ref. [15] of [Myers et al., 2017]), to systems of more than 40 coupled
differential-delay equations (ref. 6, 9, 25, 26 of [Myers et al., 2017]). This last paper also devotes
a section to a dim4 ODE with sinusoidal excitation, with a Fig. 1 evoking a possible presence (?)
of a rank 6 subharmonics, and Fig. 2 another subharmonics, but without any mention of such
behaviors.
Another refinement step considers a coupled system of ODEs and quasilinear hyperbolic
Partial Differential Equations (PDEs) [Ruan, 2008] which takes into account the fluid dynamics
character of the blood flow in the vessels. This approach is also proposed in [Borsche, et al.
2020] which uses hyperbolic partial differential equations (PDEs) with boundary conditions
driven by switched differential algebraic equations (DAEs). As association of discrete (switching
conditions) and continuous parts (ODEs, or PDEs), this model is called hybrid system in control
engineering (a simple example is given in [Mira, 2017, 2018a, 2018b], or below in sec. 5).
4.2. Inductive models. Simulators.
With respect to engineering systems, a "heart system" mathematical model will always give
an approximate description of the real system behavior. The most sophisticated inductive models
are simulators, designed via an inductive way, often by integrating "technological" components.
A first model, based on the control engineering notion of transfer function, is proposed in
[Stauss et al., 2009] (frequency response of whole body autoregulation of blood flow in rats). A
refinement of the method is given from the notion of generalized transfer functions, available to
compute the more relevant central blood pressure waveform from a more easily measured radial
waveform, method here adapted to variations in pulse pressure (PP) amplification (ratio of radial
to central PP) [Gao et al., 2016]. In this setting a simple adaptive transfer function (ATF) was
developed. In a first step the transfer function is defined in terms of the wave travel time and
reflection coefficient parameters of an arterial model. Knowing that the transfer function notion
is based on the linear control theory for a given input amplitude, it seems that the ATF takes into
account this fact from adaptation to an amplitudes range for the freqency response of the system,
via a parameters estimation from measured radial blood pressure waveform.
About simulations in physiology, it is worth quoting:
- The Lab Simulations in Physiology (Indiana University Northwest), [Timothy, 1991] and
the [Mukkamala, 2000] thesis (computer model of the human cardiovascular system for the
purposes of analyzing practical system identification methods).
- CVSim [Heldt et al.,2010] which is a lumped-parameter model of the human cardiovascular
system, developed and used for research and for teaching quantitative physiology courses at MIT
and Harvard Medical School since 1984.
- Until now, it seems that the most advanced simulator is the SIMULIA Living Heart Human
Model of Dassault Systèmes
(https://www.3ds.com/products-services/simulia/solutions/life-sciences/living-heart-human-model
It is presented as a high-fidelity multiphysics model of a healthy, 4-chamber adult human heart
and proximal vasculature. The dynamic response of the Living Heart is governed by realistic
electrical, structural, and fluid (blood) flow physics. With this model, medical professionals,
researchers, and device manufacturers will be able to rapidly conduct virtual experiments in a
highly realistic 3D environment. The Living Heart can readily be used to study cardiac defects or
diseased states and explore treatment options by modifying its geometry, loading, or
electromechanical properties. In addition, medical devices can be inserted into the model to study
their influence on cardiac function, validate their efficacy, and predict their reliability under a
wide range of operating conditions.
5. A "coarse" deductive model illustrating subharmonics effects.
Based on a similarity of dynamic behavior, common to all oscillating systems, one of the
roughest and simplest deductive models is that of Dim2 non autonomous ordinary equation
(ODE) of Duffing type submitted to a sinusoidal excitation. The papers [Khammari H. et al.,
2004] [Khammari & Mira, 2009] show examples of frequency spectra with a sub-harmonic line
higher than that of the fundamental harmonics. On the same base, a "less coarse" model is that a
three-phases rectifier with current feedback which, as hybrid system the rectifier described in
[Mira, 2017, 2018] seems to have a behavioral vicinity (?) with [Borsche, et al. 2020] (cf. sec.
4.1).
In this setting, Fig. 3 represents the diagram of a rectifier which feeds a circuit made up of a
resistance R and an inductance L, the link BC corresponding to a current feedback, the primary
transformer being fed by v E sin t. The transformer secondary is made up of N-phases, each
one connected to a thyristor, acting as a bistable switch. The block K/p (p d/dt) is the transfert
function of an integrator with gain K, e is the constant input which controls the mean value of the
residual ripple I t the frequency of which is f N /2 .
Transformer
Secondary Circuit

Pulses
Generator

Fig. 3. The N-phases rectifier diagram. K/p (p d/dt) is the transfert function of an integrator with gain K;
e is the constant input which controls the mean value of the residual ripple I t .
*
The study of this system dynamics is made from a model in a map form (recurrence
relationship) T . This map is the result of a Poincaré’ section of the continuous system, at times
t n , t n 1 , whose general condensed form (itemized in [Mira, 2018b]) is T :
Xn 1 Fn tn, tn 1, Xn , Xn 0, g k X, t 0, (1)
where t n is a switching time, and X n X t n is the state vector, X x, y , x RI/M (M is
the transformer mutual inductance), y is the voltage at the integrator output (cf. fig. 5 of [Mira,
2018b]). X n n X n , t n , is the state just after the switching time (t t n ), used as initial state for
obtaining the ODE solution for t n t t n 1 . X n is the state just before the switching (t t n ).
The index k corresponds to two different possible forms of the function g k (g 0 is a jump
condition): k n 1 when t t n with X X n , and k n when t t n 1 with X X n 1 . The
time t n is the smallest root (t t n 1 ) of g n 1 X, t, X n 0. The time t n 1 is the smallest root
(t t n ) of g n X, t, X n 0. Between two consecutive switching times t n , t n 1 , the X t evolution
is given by the solution of the ODE describing the continuous part. The very complex form of
F n t n , t n 1 , X n , X n is given in sec. 3 of [Mira, 2018b]).
The study is made via the qualitative methods of the theory of nonlinear oscillations (cf. sec.
2), associated with numerical ones via adapted algorithms conceived by A. Giraud ([Mira, 2017,
2018a,b]). In this setting, Fig. 4 represents the system behavior in the phase plane (x,y), where
the abscissa x RI/M (M is the transformer mutual inductance), the ordinate y is the voltage at
the integrator output (cf. fig. 5 of [Mira, 2018a]). The phase plane (x,y) analysis of (1) shows the
coexistence of two attractors with their respective basins (domain of influence):
- a stable fixed point (black square, related to the residual ripple f N /2 , corresponding to
the accepted higher ripple amplitude specified by the user to the rectifier manufacturer), with its
basin (green domain of initial conditions giving a phase trajectory tending toward this point),
- stable period 4 cycle (4 red points with a blue basin: f N /8 related to a subharmonics
1/4). Under initial conditions perturbation, it immediately appears that a change of the system
qualitative behavior may occur with increase of the residual ripple amplitude, when the green
area boundary is crossed. Figure 5 shows this qualitative change.
*

*
Fig. 4. Rectifier with current feedback, K 5, T 10, e 0. 3, N 3. (a) Green colored area: basin of
the fixed point (black square related to the I t fundamental residual ripple frequency N /2 ). (a) Blue colored
area: basin of the subharmonics 1/4 (four red points related to the I t residual ripple having the frequency N /8 )

(a ) (b )
Fig. 5. Fig. 3 rectifier with current feedback, K 5, T 10, e 0.3, N 3. (a) I t fundamental residual ripple (red
curve with frequency f N /2 , black square fixed point in Fig. 4). (b) I t subharmonics 1/4 (red curve related
to the Fig.4 four red points of I t with f N /8 ) and the resulting amplitude increase. The green curve is that of
y t . The yellow curve defines the thyristors firing at commutation times (details in [Mira, 2018b]).
*
Qualitative methods of the nonlinear systems theory introduce another important
information, that of the system parameters space which defines parameter regions leading to well
identified domains, each one related to a given stable dynamic state. In the Dim2 case (parameter
plane T, K, for the Fig. 3 system). For the same parameters, the coexistence of two stable steady
states (Fig. 4) shows that this space is foliated, i.e. as made up of overlapping sheets, each one
related to a well defined subharmonics 1/k (about the parameter plane foliation and its
bifurcations cf. [Mira, 2016a]).
In the parameter plane T, K , Fig. 6 represents a large red colored sheet (k 1, the stable fixed
point of the map (1) i.e. the fundamental residual ripple specified by the rectifier user to the
rectifier manufacturer. This sheet is overlapped by a yellow one (k 4, that of the subhamonic
1/4). Other overlapping sheets appear. A brown sheet (k 8, subharmonics 1/8 coexisting with
the fundamental residual ripple) which results from the classical period doubling of the cycle
k 4. In the middle left part: two green areas (k 3, subharmonics, coexisting with the fundamental
residual ripple), and their resulting k 6 (light blue areas) by period doubling. In the left upper
corner the blue region (k 2) results from the period doubling of the fundamental residual ripple
(k 1), followed by its period doubling (yellow area k 4, subharmonics 1/4, without coexistence
with the fundamental residual ripple). Ditto for the uppest left green area (k 3, subharmonics
1/3, without coexistence with the fundamental residual ripple). Figure 6 is the result of a
parameter plane scanning with its inherent numerical jumps from a sheet to another one, leading
to an abrupt color changes, here useful for showing the plane foliation (kind of windows opened
toward the foliation strata).
The parameter point (T 10, K 5, of the yellow sheet, subharmonics 1/4, Fig. 5b), which
overlaps the same point of the red sheet (fundamental residual ripple, Fig. 5a), is not far from the
boundary of the yellow sheet. This proximity with the yellow boundary implies that a T, K
perturbation from the yellow sheet may lead to a situation with only one attractor, the
fundamental residual ripple. An ideal working situation occurs when the parameter point (T, K)
is fully located in the red region without any overlapping with other sheets, ie. perturbation of the
parameter point, or of intinial conditions, cannot lead to a subharmonic behavior, with a
significant increase of the output oscillation. It is what is specified by the user to a rectifier
manufacturer.
Remark. When N 2, T 15, K 8, e 0.3, 4 attractors coexist: k 1,3,9,12 (cf. Figs. 19, 20,
22 of [Mira, 2018b])

Fig. 6. Foliated parameter plane N 3 (abscissa T, ordinate K). The "yellow" black point (T 10, K 5, k 4,
subharmonics 1/4, Fig. 5b) overlaps the same point of the red area (fundamental residual ripple, k 1, Fig. 5a) .
"Chaos" (left part of the figure) means large period subharmonics (k 20)
6. Conclusion. Steps for a possible hypothesis assessing.
In spite of having no expertise in cardiology and in its different branches, including
rhythmology, the paper’s author has taken the risk of formulating a hypothesis concerning this
scientific field: in ECG subharmonics, or fractional harmonics might be sign of heart exhaustion,
and perhaps even to chest pains to come. It is on the basis of a dynamic behavior similarity,
common to all oscillating systems, that sec. 5 gives an engineering illustrative example showing
the negative role played by lower harmonics (significant amplitude increase in the system
output), situation which might occur in more refined models of the "heart system" (?). Then,
before a cardiologists valuer’s opinion, assessing the paper hypothesis requires two intermediate
steps. The first is a checking by experts in signal processing, who have already proposed
mathematical models of the heart dynamics. The second is the confirmation by a simulator, such
as the SIMULIA Living Heart Human Model of Dassault Systèmes. In particular, such
verifications would allowed an eventual cardiologic interpretation of the coexistence of two
stable steady states (blood pressure?): one corresponding to a satisfactory working of the heart
system (Fig. 5a, possible image of heart beats of healthy relaxed patient?), the other to a
subharmonics and its resulting amplitude increase in the system output (Fig. 5b, possible image
of heart exhaustion?).
*
REFERENCES
- Amrutha N. & Arul V.H. [2011]: A review of noises in EMG signals and its removal. Intern.
J. of Sci. and Res. Publications 7(5), May 2011, 22-27 2011, 4284-4287.
https://www.researchgate.net/publication/220736086
- Bogoliubov N.N. & Mitropolsky Iu. A. [1958]: Asymptotic methods in the theory of
nonlinear oscillations. Fizmatgiz, Moscow, 1958.
- Boïko E.S. [1983] : The Academician A.A. Andronov’ School (in Russian). Ed. Nauka,
Moscow 1983.
- Borsche R., Kocoglu D. & Trenn S. [2020]: A distributional solution framework for linear
hyperbolicPDEs coupled to switched DAEs. Mathematics of Control, Signals, and Systems
(2020) 32:455–487.
- Buendía-Fuentes F., Arnau-Vives M. A., Arnau-Vives A., Jiménez-Jiménez Y.,
Rueda-Soriano J., Zorio-Grima E., Osa-Sáez A., Martínez-Dolz L. V., Almenar-Bonet L., and
Palencia-Pérez M. A. [2012]: High-Bandpass Filters in Electrocardiography: Source of Error in
the Interpretation of the ST Segment. https://www.hindawi.com/journals/isrn/2012/706217/
- Chia-Hung Lin [2008] : Frequency domain features for ECG beat discrimination using grey
relational analysis-based classifier. Computers & Mathematics with Applications. Vol. 55, Issue
4, February 2008, 680-690.
- Chao Lin [2012] : P and T analysis in ECG signals using Bayesian methods. Thèse de
doctorat (Phd) en Signal, Image,et Optimization, soutenue le 02-07-2012 à l’Ecole doctorale
Mathématiques, Informatique et Télécommunications (INPT, Toulouse) https://oatao.
univ-toulouse;fr/8990/1/lin.pdf
- Das S. & Maharatna K. [2013]: Fractional dynamic model of ECG like signals from filtered
coupled Van-der-Pol oscillators. Publication of Scool of Electronics and Computers Science.
Compute methods programs Biomed, 112(3), 2013 Dec., 490-507.
- Dia N., Fontcave-Jalon J., Gumèri P., Rivet B. [2019]: Heart Rate Estimation from
Phoncardiogram SignalsUsing Non-Negative Matrix Factorization. Proc. of the IEEE Int. Conf.
on Acoustics, Speech, and Signal Processing, ICASSP 2019, 1293-1297.
- Fairooz T. & Khammari H.S. [2016] : SVM classificaion of CWT signal features for
predicting sudden cardiac death. Biomedical Physics & Engineering Express 2(2), 2016.
- Fangyu Li, WenZan Song , Changwei Li, Aiying Yang [2019]: Non-Harmonic Analysis
Based Instantaneous Heart Rate Estimation from Photoplethysmograpghy. Proceedings of
ICAASP 2O19, 1279-1283.
-Gao M., Rose W., Fetics B., Kass D. A., Chen C.-H., & Ramakrishna M. [2016] :A Simple
Adaptive Transfer Function for Deriving the Central Blood Pressure Waveform from a Radial
Blood Pressure Waveform. Scientific Reports 6(1): 33230.
- Gumowski I. & Mira C. [1968] : Optimization in control theory and and practice.
Cambridge at the University Press.
- Gumowski I. [1981]: Some relations between differential equations, recurrences and
functional iterates. 9th ICNO Kiev (1981), summary in Proc.9th ICNO, Kiev 1984, vol. 2,
98-103.
- Gumowski I. [1989]: Oscillatory evolution processes. Quantitative analyses arising from
applied sciences. Manchester University Press, Nonlinear Science, theory and applications, 1989.
- Hayashi Ch. [1964] : Nonlinear oscillations in physical systems.McGraw-Hill,
Electricaland Electronic Engineering Series.
- Hayashi Ch. & Kawakami H. [1984]: Behavior of solutions for Duffing-Van der Pol
equation. Proc. 10th Int. Conf. on Nonlinear oscillations, Varna, 1984, 27-40.
- Heldt T., Ramakrishna M., Moody G.B., Mark R.G. [2010]: CVSim: An Open-Source
Cardiovascular Simulator for Teaching and Research. Open Pacing Electrophysiol Ther J. 2010;
3: 45-54.
- Hundewale N. [2012] : The application of methods of nonlinear dynamics for ECG in
normal sinus rhythm. Intern. J. of Computer Science Issues, vol. 9, Issue 1, N°1, Jan. 2012,
458-468.
- Kawakami, H. & Sekaï, M. [1983] : Synchronization phenomena observed in a
Duffing-Raleigh equation. " Publication NLP84-7,pp. 25-33 (in Japanese).
- Kawakami, H. & Sekaï, M. [1984] : Synchronization phenomena and Hopf bifurcation."
Publication NLP84-7,pp. 25-33 (in Japanese).
- Kawakami H. [1984] : Bifurcation of periodic responses in forced dynamic nonlinear
circuits: computation of bifurcation values of the system parameters. IEE Trans.on Circuits and
Syst. 31(3), March 1984, 248-260.
- Kawakami, H. & Sekaï, M. [1988] : Synchronization and chaos in a Duffing-Rayleigh
oscillator. Asymptotic Methods of Mathematical Physics (Collection of scientific papers), Kiev,
Naukova Dumka, pp. 83-92.
- Khammari, H., Mira, C. & Carcasses, J. P. [2005]: Behavior of harmonics generated by a
Duffing type
equation with a nonlinear damping. Part I. Int. J. Bifurcation and Chaos 15, 3181–322.
- Khammari H. & Mira C. [2009] : Behavior of harmonics generated by a Duffing’s equation
with a nonlinear damping. Part II. International Journal of Bifurcation & Chaos, 19 (4)
1227-1254 (2009).
- Koshita, S., Miyoshi, K., Abe, M., & Kawamata, M. [2019]: A State-Space Approach to
High-order Adaptive Band-Pass Filtering. IEEE 23rd International Conference on Digital Signal
Processing, Vol. 2018-November). Institute
- Koschita S., Kumamoto Y., Abe M., Kawamata M. [2013] : Anedaptive IIR
Band-Pass/Band-Stop Filtering Using High-Order Transfer Function and Frequency
Transformation. Interdisciplinary Information Sciences Vol. 19, n°2 (2013), 163-172.
- Koschita S., Kumamoto Y., Abe M., Kawamata M. [2011]: High-order center-frequency
adaptive filter using block diagram-based frequency transformation. Proc. of the IEEE Int. Conf.
one Acoustics, Speech, and Signal Processing, ICASSP 2011, Prague May 21-27.
- Labecki M. et al. [2016]: Nonlinear origin of SSVEP Spectra - A combined experimental
and modeling study. Front Comput Neurosci. 2016; 10: 129. Published online 2016 Dec. 27.
- Lin Chao [2012] :
- Mcnames J., Crespo C., Aboy M., Bassale J., Jenkins L., Goldstein B. [2002]: Harmonic
spectrogram for the analysis of semi-periodic physiologic signals. Research Gate publications,
htpps://reserchgate.net/publication/3978230.
- Mira C., & Djellit I. [1993a] : Bifurcations structure in a model of frequency modulated
CO2 Laser. Int. J. Bifurcation and Chaos 3(1), 97-129.
- Mira C. , [1996] : Some historical aspects of nonlinear dynamics: possible trends for the
future. International Journal of Bifurcation & Chaos, 7 (9) 2145-2173 (1999).
- Mira C. [2016a] : Shrimp" fishing, or searching for foliation singularities in the parameter
plane? Part I: basic elements of the parameter plane. Research Gate article,
 htpps://reserchgate.net/publication/302960218.
- Mira C. [2016b] : "A priori" mathematical models and "a posteriori" models . Research
Gate publication, November 2016, https://www.researchgate.net/publication/310806972
- Mira C. , Touzani-Qriouet M. & Kawakami H. [1999] : Bifurcation structures generated by
the non autonomous Duffing equation, International Journal of Bifurcation & Chaos, 9 (7)
1363-1379 (1999).
- Mira C. [2017]: Hybrid systems. (I) A former step toward a theory. Research Gate
publications, htpps://researchgate.net/publication/321214134.
- Mira C. [2018a]: Hybrid system. Part (II), simple example:two-phases rectifier with current
feedback. The Giraud pioneering results. Research Gate publications,
htpps://researchgate.net/publication/329128946.
- Mira C. [2018b]: Hybrid system. Part (IIa). A first simple example:rectifiers with voltage
feedback. Research Gate publications, htpps://researchgate.net/publication/324068942.
- Mukkamala R. [2000] : A Forward Model-Based Analysis ofCardiovascular System
Identification Methods. MIT thesis, Departement of Electrical Engineering and computer
Science. https://core.ac.uk/download/pdf/4385519.pdf.
- Myers T. G., Ripoll V.R., Cuenca A.S.T.C., Sarah L. Mitchell S.L., & Mark J. McGuinness
M.J. [2017] : Modelling the cardiovascular system for assessing the blood pressure curve. Math.
Ind. Case Stud. 2017; 8(1): 2.
- Qriouet M. & Mira C. [1994] : Fractional harmonic synchronization in the
Duffing-Rayleigh differential equation, International Journal of Bifurcation & Chaos, 4 (2),
411-426 (1994).
- Ruan W. [2008] : A coupled system of ODEs and quasilinear hyperbolic PDEs arising in a
multiscale blood flow model. J. Math. Anal. Appl. 343 (2008) 778–798.
- Shekatkar S.M., Kotriwar Y., Harikrishnan K.P., Ambika G. [2017]. Detecting
abnormalities in heart dynamics from multifractal analysis of ECG signals.
www.nature.com/scientificreports, 7, n°15127 (2017).
- Shilnikov L. P. [1997] ”Mathematical problem of nonlinear dynamics: a tutorial”. Int. J.
Bifurcation and Chaos, 7(9), 1953-2001.
- Stauss H. M. , Rarick K. R., Deklotz R. J. , and Sheriff D. D. [2009]: Frequency response
characteristics of whole body autoregulation of blood flow in rats. American Journal of
Physiology-Heart and Circulatory Physiology. Vol. 296, No. 5.
- Tereshchenko L.G. & Josephson M.E. [2015]: Frequency Content and Characteristics of
Ventricular Conduction. J. Elecrocardiol. 48(6), 2015 Nov-Dec, 933-937.
- Timothy L. D. [1991] : Teaching physiology through interactive simulation of
hemo-dynamics. Master’s thesis, Massachusetts Institute of Technology, Cambridge,MA 02139,
February 1991.
- Touzani-Qriouet M. & Mira C. : Reducible fractional harmonics generated by the non
autonomous Duffing-Rayleigh equation. Pockets of reducible harmonics and Arnold’s tongues,
International Journal of Bifurcation & Chaos, 10 (6), 1345-1366 (2000).
- Ueda Y. [1991]: Survey of regular and chaotic phenomena in the forced Duffing oscillator.
Chaos, Soliton & Fractals. Vol. 1, n°3, 1999-231.
- Vaschillo E.G., Vaschillo B., Pandina R.J., and Bates M.E. [2011]: Resonances in the
Cardiovascular System Caused by Rhythmical Muscle Tension. Psychophysiology. 2011 Jul;
48(7): 927–936.

View publication stats