FR2790620A1 - ELECTRONIC DEVICE, PARTICULARLY PROGRAMMABLE, WITH CHARACTERISTIC OF NON-LINEAR TRANSFER, AND APPLICATION TO A CONTROL DEVICE, IN PARTICULAR A CARDIAC STIMULATOR - Google Patents

Abstract

The invention relates to a non-linear transfer device consisting of or approached by a bit linear function, whereby the bits are defined according to intervals of variation in an input signal pertaining to a device comprising the following: first means (20, 22, 24) for determining, in the case of each Q input signal value, the interval to which said value belongs and the degree of belonging for said interval; second means (26, 28, 30) for producing a Z output signal of the device as a function of the information on the belonging interval and the degree of belonging of the value of the input signal that were previously determined, in addition to that of the bit linear function.

Description

TECHNICAL FIELD AND PRIOR ART

  The invention relates first of all to the field of electronic circuits for

  nonlinear transfer function.

  In the field of electronic systems, there is the problem of producing circuits implementing non-linear transfer functions or characteristics. The function, or transfer characteristic of a circuit, is

  the ratio between the output signal and its input signal.

  More specifically, there is the problem of producing circuits that can be easily implemented so as to implement a non-transfer function.

linear given.

  The invention also relates to the field of methods and control devices. In particular, it finds application to methods and control devices implementing fuzzy logic rules. It finds particular application to the control of devices implanted in the human body, such as for example pacemakers with frequency

enslaved to the effort.

  Fuzzy logic-based control is applied to many technical fields. The implementation of control algorithms using fuzzy logic rules is possible, either using software implemented on

  standard or dedicated microprocessors, or using analog circuits.

  Each technique has its own advantages and disadvantages.

  The implementation using software allows a great flexibility and a

  great flexibility of programming. Moreover, it allows a certain precision.

  However the cost for all these benefits lies in a power consumed

  high and a relatively low speed.

  Analog designs provide higher speeds and lower power consumption, but are less flexible or

  allow to obtain only a small precision.

  Pacemaker technology, meanwhile, is evolving quite rapidly. New versions of stimulators are offered almost every year by the manufacturers, who thus regularly improve the

properties.

2/2790620

  Users, however, do not have a truly satisfactory device at their disposal, especially because of the control algorithms, which

are not quite suitable.

  Moreover, in the opinion of the users the programming of the control device is too complicated to allow an optimal personalization or

satisfactory stimulator.

  There is therefore the problem of finding a new method and a new control device, applicable in particular to the control of a pacemaker, and which is preferably easily programmable and cost-effective.

energy consumption.

  There is also the problem, in the case of application to a pacemaker, optimizing the rate of cardiac activity from control signals, the optimization criteria selected being: the adaptation of the heart rate, at rest and during an effort, at the

metabolic demand.

  - the appropriate increase in the heart rate at the beginning of the effort - the decrease in the heart rate at the end of an effort There is also the problem of simplifying and optimizing the adaptation procedures of a control device with each variation in the design of a system to be regulated, in particular in the case of a pacemaker. Finally, it is important to have control systems that can be adapted to any type of pacemaker, whatever the type of pacemaker.

  signal or control signals used.

SUMMARY OF THE INVENTION

  The invention firstly relates to an electronic device with non-linear transfer characteristic, constituted by, or approximated by, a piecewise linear function, the pieces being defined over intervals of variation of an input signal. the device, comprising:

3 2790620

  first means for determining, for each value of an input signal, the interval at which this value belongs, and the degree of membership for this interval; second means for producing an output signal of the device, function of the membership interval information and the degree of membership of the previously determined input signal value, as well as the function

linear in pieces.

  According to another definition, the subject of the invention is also an electronic device with a non-linear transfer characteristic constituted by or approximated by a piecewise linear function, the pieces being defined over intervals of variation of a signal of input of the device, comprising: - a first and a second current source respectively for shifting a value of an input signal by means of a first offset value, and multiplying the signal thus obtained by a first multiplicative coefficient, - third and fourth current sources, respectively making it possible to multiply the second value obtained by a second multiplicative coefficient, and then to shift the third value thus obtained by a certain intensity using a second value. offset - means for producing a digital signal identifying a range of variation in which the input signal is located and p to vary or determine the multiplicative coefficients and offset values as a function of the interval

  in which the input signal is located.

  A device according to the invention makes it possible to easily implement a nonlinear transfer function, constituted by or approximated by a piecewise linear function, the pieces being defined over intervals of variation.

an input signal.

  Such a device may advantageously furthermore comprise digital storage means, for storing data defining or characterizing the piecewise linear function on each of the intervals on

  which pieces of the linear function are defined.

4 2790620

  The first means can also produce a first output signal representative of the interval at which the value Q belongs and a second

  signal representative of the degree of membership in this interval.

  In particular, the first means comprise: means for producing an analog signal defining the degree

  where the value of the input signal belongs to the interval to which it belongs.

  means for producing a digital signal defining the interval

  the value of the input signal belongs to.

  Such an embodiment makes it possible to obtain a device which, moreover, has a

low energy consumption.

  Advantageously, a digital feedback loop makes it possible to modify the determination of the degree of membership of a value of the input signal.

  depending on the interval to which it belongs.

  For example, means can be provided for comparing the second output signal, representative of the degree of membership, with first and second limit values, respectively lower and upper, and for modifying the degree of membership, as well as the first signal. representative of the degree of membership, when the second signal falls below the first limit value or when

  the second signal becomes greater than the second limit value.

  A device according to the invention can be applied to a device for controlling a physical quantity Z of a system. Such a control device then comprises a device as above, the input signal comprising at least one measurement signal, and the output signal being a control signal of the physical magnitude Z. An example of a system physical that can be so regulated is that

pacemaker.

  An example of a non-linear transfer function that can be implemented with the invention consists of a piecewise linear transfer function, in particular of the type derived from a fuzzy logic system. Thus, the invention also relates to a fuzzy logic control system,

  implement a device as explained above.

2790620

  Such a regulating device may also advantageously comprise means for limiting the variations of the output signal. In particular, this device may comprise at least two circuits in parallel, as already described above, one for calculating the values of the physical quantity Z, at least one of the others making it possible to calculate a maximum variation of this

  magnitude from a measured value of that same magnitude.

  According to a preferred embodiment, the device comprises three circuits in parallel, as already described above, a first circuit for calculating the values of the physical quantity Z, a second circuit whose input signal is the actual value of Z, measured on or in the system and for output signal the maximum possible increase of Z, the third also having for its input signal the actual value of Z, and for output signal the maximum possible decrease of Z. comparator circuit may furthermore make it possible to limit the correction of Z, calculated using the first circuit, to a value taking into account the values of maximum possible increase and maximum decrease.

calculated.

  Consequently, the circuit according to the invention comprises: means for determining, for each measured value Q, the interval at which this value belongs, and the degree of membership for this interval; means for calculating a value of a signal output or a value of a control signal, from the information on the interval at which

  the Q value and membership degree for that interval.

  In particular, this device may comprise: means for producing a digital signal defining the membership of an input value for an interval, and for producing a signal

  analog degree of membership of said input value at said interval.

  digital means for determining coefficients (c, d) depending on the interval to which the input value belongs, for calculating a value of the output or control signal Z. analog means for calculating said control value by function of the degree of belonging to the subset, and according to the coefficients

(c, d).

6 2790620

  The means for producing an analog signal and a digital signal may comprise an analog part and a digital part, mounted in

closed loop.

  For example, a control signal of the digital part is established by means of a comparator, which compares the output of the analog part with

predetermined values.

  The circuit as defined above makes it possible to produce a control circuit with low energy consumption, which is extremely important and critical in the case of certain applications, particularly for the realization of

pacemakers.

BRIEF DESCRIPTION OF THE FIGURES

  The features and advantages of the invention will become more apparent

  light of the description that follows. This description focuses on the examples of

  embodiment, given for explanatory and nonlimiting purposes, with reference to the appended drawings in which: - Figure 1 schematically shows a device according to the invention - Figure 2 is an example of nonlinear transfer characteristic, and its approximation 3 shows in more detail a device according to the invention - FIGS. 4 and 5 show detailed embodiments of certain components of the device according to the invention - FIGS. 6 to 9 are transfer characteristics of FIG. certain internal variables and the output variable, with respect to the input variable, of a

device according to the invention.

  FIG. 10 represents a regulation system implementing a

device according to the invention.

  FIG. 11 is a diagram representing, for a parameter constituted by a given input or control signal, a membership function for a fuzzy logic control algorithm; FIG. 12 is a diagram representing a linear transfer function; in pieces, for a fuzzy logic control algorithm

7 2790620

  FIG. 13 schematically represents a regulation system of a pacemaker, using a device according to the invention; FIG. 14 represents a comparison between a simulation and a

circuit according to the invention.

  DETAILED DESCRIPTION OF EMBODIMENTS OF

THE INVENTION

  An example of a circuit according to the invention is given in FIG.

  The circuit comprises an input 21, for an input signal Q, and an output 29 for outputting an output signal Z. Each of these signals can be represented

  for example by an input current I ,, and an output current Iout.

  A nonlinear transfer characteristic to be implemented is illustrated by the curve C of FIG. 2. The variable or input signal Q (or Tin) varies over a range of input signal values between a minimum value Iin, min = Io = (or Qmin = Qo) and a maximum value I3 = Im, mnx (or Q3 = Qmax). According to one example, and to simplify the explanations, we can take

Qmin = 0.

  As can be seen in Figure 2, the curve C can be approximated or approximated by pieces of linear functions on the intervals [Qo, QI], [Qi, Q2] and [Q2, Q3]. The book by RWHamming entitled "Numerical Methods for Scientists and Engineers", Mac Graw Hill, New York, 1962, Part II deals with the problems of approximation by first degree polynomials and linear interpolation, questions which are directly related to that of the approximation of a nonlinear function by a piecewise linear function. The transfer function could also be a function

linear in pieces.

  The points (Xi, Yi) indicate points of concordance between the transfer function C and the linear pieces or segments. Each interval, and each piece, can be identified by a number N: N = 0, N = 1, N = 2. To each value No can be associated data or a set of data making it possible to identify the corresponding linear function, for example, the coordinates of the points of concordance (Xi, Yi): we then associate with the piece or the interval N =

8 2790620

  0 coordinates (X0, Y0) and (X1, Y1), to the piece N = 1 the coordinates (X1, Yi) and

  (X2, Y2) and at the piece N = 2 the coordinates (X2, Y2) and (X3, Y3).

  According to the invention, a value Q (or Ini) of the input variable, or of the input signal, is assigned the corresponding piece or linear segment, therefore the value of N identifying this segment: according to the example of Figure 2, N = 1

for Q '(or ln).

  Q is also assigned a p value to identify or characterize its position in the interval [QI, Q2]. For example, we can assign Q the value t = 0 for Q = QI and the value pi = 1 for Q = Q2. In other words, the current value Q has been assigned the value: it = (QQI) / (Q2-QI) (1)

  (or at the current value I of the input current the value it = (I - IJ) / (I2 - I,) (1 ')).

  This value (on function) it is the value (or function) of membership of

  the input variable at the interval [QI, Q2].

  In other words, each value of the input signal Q is associated, on the one hand, with a discrete quantity N, and on the other hand with a continuous quantity gi identifying the

  position of Q in the interval [QI, Q2] of the input values.

  When the values N and pi are determined, it is possible to determine the corresponding output quantity Z: Z = Pi (Y2-Y)) + Yi or, more generally: Z = 1t (Yi + 1-Yi) + Yi ( 2), with i = 0, 1 or 2 depending on whether Q (value of the input signal) belongs to [QoQi] or

[QQ2] or [Q2Q3].

  The method according to the invention has been described with the example of a membership function i = (Q - Qi) / (Qi + - Qi), but other functions could be formed or chosen, with a corresponding adaptation of the function making it possible to calculate the output quantity Z. In other words, according to the invention, a nonlinear transfer function can be defined beforehand. This function is approximated by a piecewise linear function on intervals of the variation domain of the variable or

  of the input signal, for example on a succession of such intervals.

  Each value of the variable, or of the input signal, is assigned a discrete quantity N identifying the interval at which the input value belongs, and a value of a membership function gi of this value for this interval. From N and g, the value of the output signal can be calculated according to parameters

9 2790620

  identifying or characterizing or defining, in the interval N, the linear function

  who approaches the transfer function.

  An example of a circuit developed to implement the method according to the invention is shown in FIG. 1. This circuit comprises: means 20, 22, 24 for determining, for each value of the input signal Q, the interval N to which this value belongs, and the degree of membership kt of this value for this interval - means 26, 28, 30 for calculating the value of the output signal Z, from the information on the interval N to which the value belongs Q and its degree

  of membership p for this interval.

  The means 20 make it possible to assign to each measured input value Q two output signals: a first signal, digital, N, and a second signal, analog pt, where N represents the subset QiQ to which the measured value Q

  belongs and represents the degree of belonging of Q to this subset.

  These means 20 comprise, for example, means 22, which are purely analog or else made in analogue-digital technology, which make it possible to transform the quantity or variable Q (or the corresponding input current Iin) into a second output signal p, or signal of belonging. For example, kt is calculated according to a formula of type: p = a. (Iin - Ib) (3) where a and b are coefficients whose values are programmable, depending on

  N, and o Ib is a current that depends on b.

  According to the example given above: - for N = 0: Ib = X0 and a = (XI-X0) 1 - for N = 1: Ib = XI and a = (X2-Xl) - '- for N = 2: Ib = X2 and a = (X3-X2) 'The means 20 may also comprise means 24 which make it possible to obtain the number N, as well as the corresponding coefficients a and b for the means 22. The coefficients a and b can be represented in analog or digital form. In fact, these coefficients can be stored in a

  storage device 32 (digital memory).

2790620

  The means 26 make it possible to calculate the output quantities Z as a function of t, of N and coefficients c and d which can be stored in memory

  in storage means, for example the means 32.

  In the case of several quantities or input variables Q. Q ', Q "..., the means 26 can be connected to several blocks or means 20, each block

  corresponding to an input variable ..

  These means 26 comprise, for example, means 28, which are purely analog or else made in analogue-digital technology, which make it possible to transform the signal pt into an output signal Z. For example, Z is calculated according to a formula of the following type: Z = c . (1i + Id) (4) where c and d are coefficients whose values are programmable according to

  N, and o Id is a current that depends on d.

  According to the example given above: - for N: 0: c = Y, -Y0 and Id = Y0 - for N = 1: c = Y2 -Y, and Id = Y1 - for N = 2: c = Y3 - Y2 and Id = Y2 The means 26 may also comprise means 30 which make it possible to obtain the values or the coefficients c and d for the means 28. c and d may be represented in analog or digital form. The input values of block 30, or means 30, are on the one hand the value of N and the other

  the values or coefficients stored in memory 32.

  The memory 32 thus makes it possible to store coefficients which describe an appropriate transfer characteristic, linear in parts or in sections, which approaches the real nonlinear characteristic C. Each value of N is associated with a set of coefficients, stored in memory at the same time. determined addresses. For each new value of N, the corresponding areas of the memory 32 are addressed, and the appropriate values of the coefficients are transmitted from the

memory at circuits 24, 30.

  Thus, in the example already given above, four values of coefficients a, b, c, d are assigned to each value of N. Therefore, for each value of N, four addresses of the memory 32 are generated, which determine the appropriate values of a, b, c, d. The addresses for a and b are generated by means 24 and addresses for c and d by means 30. When N increases or decreases by one, the appropriate addresses are generated by means 24 and 30 and sent to the memory, and the modified values of a, b, c, d are transmitted from the memory

  to the means 24 and 30, for controlling the current sources.

  The operation of a control device as illustrated above can be explained with the example of the transfer characteristic C of FIG. 2, which can be piece-wise approximated by linear functions. Data defining or defining the linear pieces or segments are stored in the memory area 32. The actual transfer function is approximated

  by Nmax + I linear sections (Nmax = 2 in the case of the example of Figure 2).

  Preferably, in a first step, the input signal Q is normalized using the means 22. The normalization makes it possible to facilitate the processing in the block 26. The signal is shifted (which can be done using of a current source, as will be explained in more detail below), then its slope is normalized (this can be done using a multiplier circuit, as will be explained in more detail below ), these steps corresponding to equation (1)

above.

  The normalized signal is input into the circuit 24 which determines N. The circuit 24 increments N by one unit when it grows beyond an upper limit value (equivalent to the value "1" of the membership function) and decrements N by one unit when It decreases below a lower limit value (equivalent to the value "0" of the membership function). Such a modification of the value of N causes a corresponding modification of the coefficients a, b, c, d generated by the means or circuits 24 and 30, from the

values stored in memory 32.

  A normalized value p is supplied to the circuit 26, 28. The slope of the signal is then modified to obtain the slope of the output signal in the interval QiQi + that the input value Q belongs to (this can also be realized at using a current source) and then the signal is shifted to obtain the output signal defined by the transfer characteristic in the interval QiQi + 1, these steps

  corresponding to equation (4) above.

  Figure 3 shows in more detail an example of the

  structure of a device according to the invention.

  The analog input signal Q is represented by a current Iin. At each value of this current, the block or the means 20 correspond to an analog signal, or a current, 12 and a digital signal N. The signal I2 corresponds to the

signal pI of FIG.

12 2790620

  The means 22 make it possible to transform Iin, (Q) into I2 (pi) according to the equation

(1) or (1 ') above.

  The means 24 consist respectively of means 34 and 36. The means 34 make it possible to generate the coefficients a and b, as a function of the values stored in the digital memory 32 and the number N generated by the means 36. The range on which the number N varies depends on the number of linear sections used to approximate the transfer characteristic. Nnax is equal to

number of sections minus 1.

  The output current 12, generated by the means 22, is controlled by the means 36. When, as already explained above, this current exceeds a limit value, the value of N is increased by one unit. When this current reaches

  a lower limit value, the value of N is reduced by one unit.

  The means 26 make it possible to calculate the output value I or (or the output signal Z), as a function of the inputs constituted by the analog value I 2, the number N, and the coefficients c and d stored in the memory 32. The function of

  transformation from I2 to Iou is determined by means 28.

  out is calculated according to 12 according to formula (2) above.

  The means 30 generate the coefficients c and d to be supplied to the means 28, taking as input parameters the coefficients stored in the

  digital memory 32 and the number N generated by the means 36.

  The means 22 and 28 can be made using mirrors of

  current and a digitally controlled current gain.

  FIG. 4 represents an embodiment of the multiplier circuit which

  multiplies the input current by a certain factor (a digital value).

  This circuit can be used for the realization of the means 38 and 44 of FIG. 3. They implement a MOS technology, and in particular the

NMOS transistors M 12-M22.

  FIG. 6 represents an embodiment of the controlled current source of

  digitally, which can be used in blocks 40 and 42 of FIG.

  Here again, the technology used uses NMOS transistors MlMl I. In these two figures 4 and 5, aO, a1, a2, b0, b1, and b2 represent individual bits (generated by means 30 and 34) of words which control the current sources 38, 40, 42, 44. Each bit switches a

13 2790620

  MOS transistor (M20, M21, M22 in Fig. 4, M9, M10, Mi1 in Fig. 5).

  Thus currents I, I2, 13, 14 are digitally controlled.

  The circuits of FIGS. 4 and 5 are current mirrors in CMOS technologies, which generate digitally controlled output currents. The currents in the different branches are controlled by the input bits

bO, bI, b2, aO, a I, a2.

  For example, for the device of FIG. 5, if 10 is the current flowing through M9, and which is switched by the bit b0, then the current flowing in M10, switched by the bit b1, has the value 2.10, and the current flowing in MlI 1, switched by the bit b2, has the value 4.10. Thus, the total output current I 2 is proportional to the value of the coefficient b represented digitally by b 0, b 1, b 2. In the same way, the current Ic is generated from the

  digital representation d0, dl, d2 (not shown in FIG. 5) of the coefficient c. In Figure 4, the same principles of operation apply. A

  Ia current (or Id) is generated, which depends on the value of a (or d), represented digitally by aO, al, a2 (respectively dO, dl, d2, not shown in Figure 4). The current I = const. is a bias current. Its value is chosen by the user of the circuit and will be the result of a compromise: if it is too high, the circuit will consume more energy than necessary; if it is too weak the circuit

will have less accuracy.

  In the analog part of the device, the means 36 can be

  made using a simple comparator.

  The complexity of the digital part of the system (means 24, 30, 34, 36 of FIG. 4) depends on the environment in which the controller can be used (data buses 27, 31 and address 25, 29, the organization of digital memory, etc.) and the need for speed and accuracy. Typically, this digital part is a simple sequential circuit that can be automatically realized at

  from a functional description.

  FIGS. 6-9 are examples of the functions describing the transfer characteristics of the systems of FIGS. 1 and 3, and in particular: the dependence of the current II as a function of the input current Ij, (FIG. of the current I2 as a function of the input current Iin (FIG. 7) - the dependence of the current 13 as a function of the input current Iin (FIG.

  the dependence of the current I4 as a function of the input current I1, (FIG 9).

For current Ii (fig.6):

14 2790620

  N = 0 => Il = In, - X0 N = 1 => I, = In- X1 N = 2 Z Il = Iin- X2 For the dependence of I2 as a function of In (fig.7): N = 0> I2 = (Ii, - Xo) / (X1-Xo) N = 1 => 12 = (Iin-X1) / (X2-X1) N = 2 = I2 = (Iin-X2) / (X3-X2) For the dependence of I3 as a function of Ii (fig.8): N = 0 => 13 = (Iin-Xo) (Y1Yo) / (Xi-Xo) N = 1 => I3 = (In-X1) (Y2- Y1) / (X2-XI) N = 2 => 13 = (Iin-X2) (Y3Y2) / (X3-X2) For the dependence of I4 as a function of Ii. (fig.9): N = 0 => 14 = [(IiJ-Xo) (Y1-Y0) / (X1-X0)] + Y0 N = I> I4 = [(Iin-X1) (Y2-Y1) / (X2X1)] + YI N = 2 =:> I4 = [(Iin-X2) (Y3-Y2) / (X3-X2)] + Y2 The transfer characteristic of this device (FIG. 9) has the same speed than that shown in FIG. 2, the axis of the abscissae corresponding to the current

  Im input and ordinate axis at Iout output current.

  In the circuit according to the invention, the input analog signal is represented, internally, by the analog signal I2 and by the digital signal corresponding to the number N. A first input signal conversion in this double representation

  analog-digital is performed using the means 22.

  The analog signal I2 and the digital signal N are then separately processed and combined by the means 26, 28 to provide the output signal.

analog lout.

  This structure makes it possible to produce a universal controller for processing an analog signal, but which is programmed from a digital point of view in a rather simple manner. After determining the piecewise linear function, the programming comprises the following steps: - determination of the coefficients on each piece of the linear function - storage of these coefficients or these data in the zone

memory 32.

2790620

  The circuit according to the invention makes it possible to produce a nonlinear transfer characteristic, which can notably be approximated by a piecewise linear function. This circuit represents the input signal, during the processing of said signal, by a pair of signals: - the digital signal, which determines the interval [Qi, Qi + 1] at which the input value belongs: thus is determined the part or the linear section of the approximate characteristic - the analog signal whose value represents a particular point in

  this section (membership function).

  The circuit according to the invention makes it possible to program, as described above, a piecewise linear piece transfer characteristic. It is also possible to modify previously programmed values of Nmax and / or a, b, c, d. The circuit according to the invention makes it possible to achieve a nonlinear transfer characteristic, which can notably be approximated by or consisting of a piecewise linear function combining digital technology and analog technology. Programming is done using the part

  digital, the analog part carries out the signal processing.

  The digital part is not very complex, and does not require a high clock frequency. It can also be realized in NMOS or CMOS technology, which makes it possible to further improve the consumption

  for special applications (pacemakers).

  The analog part can operate with very low currents. In addition, the entire circuit does not require to consume power continuously. When new data is input, the circuit processes it as described above, and then stops. In this way, most circuit components are generally in the off state and do not consume

no energy.

  The invention can be applied to a physical input signal Q, the device 2 delivering a physical signal Z output. The invention and the device 2 (FIG. 10) can thus also be applied to a physical signal Z to be regulated, for example a physical system 50 to be regulated, a physical signal Q being measured and making it possible to deduce a value of the signal Z according to a characteristic of

  given transfer approximated by a piecewise linear function.

16; 2790620

  According to the invention, therefore, at least one physical signal Q of the

system 50 to be regulated.

  For example, in the case of a pacemaker, a motion signal or breath signal or the QT interval, which reflects the activity of the pacemaker, is measured. Exemplary embodiments of a device

  regulation and a pacemaker are given later.

  For each measured value Q of the signal, the interval

  this value belongs, as well as the degree of membership for this interval.

  In other words, for a given signal level Q (t), at a given instant t, the degree of membership of Q, for the corresponding interval, is

  determined. We thus obtain a membership function pt (Q (t)).

  There is therefore two pieces of information from which the value of Z is determined: - the interval at which the measured value belongs

  the degree of membership for said interval.

  Moreover, the invention can be applied not only to a single input signal Q but also to several input signals Q (1, Q (2), .... In this case, the system of FIGS. 3 is realized, with as many blocks 20 as input signal Q (i) The invention therefore applies to the realization of a system implementing a nonlinear and multidimensional transfer characteristic., The characteristic of transfer, for a specific Q (i) variable, being approximated by

  a piecewise linear function.

  A set of fuzzy logic rules, or a control algorithm based on fuzzy logic rules, have been developed to realize a fuzzy logic control method. This method can be applied in particular to

  control of cardiac activity using a cardiac simulator.

  As illustrated in FIG. 11, each signal varies in a field of

  value that is split into NQ intervals with: Qo <QI <Q2 <Q3 <Q, ...

  For each measured value Q of signal, it is determined at which interval

  this value belongs, and the degrees of membership for that interval.

  In other words, for a value of the given signal Q, belonging to an interval QiQi + 1, it is possible to determine the degrees of membership for this interval, and for the neighboring intervals, Qi [Q, and Qi + 1Q1 + 2 . So, for the value Q =

  (QI + Q2) / 2 (Fig. 11): H0 = 0.5, l = 0.5 and J, 2 = 0.

17 2790620

  The control algorithm implements in particular the following parameters, for each control signal: a maximum signal level, and a minimum signal level: Qmin, Qnmax a given number of steps n (for example n = 4, 8, 16 ...): Qmin = Ql <Q2 = <... <Qn = Qmax The control value Z is imposed for each step: Z, for QI, Z2 for Q2,

  ., Z, for Qn. In other words, it is possible to determine beforehand, for each of the Qi, a desired value Zi of the control or regulation variable Z. In..DTD: the pacemaker case is a desired cardiac activity rate H ,.

  The control variable Zc (Q), calculated from a measured value Q (t) of a signal Q, is given by: 1 5 z (Q) = E 11, (OZi / .. Pt i For the case shown in Figure 1 l, only two neighboring intervals Qi Qi + l and Q, + I Qi + 2 can have a non-empty intersection (and so that for all i, only pi and pi + 1 are different from zero) The sum of Pi is equal to 1, for each value of Q. So: z = - 1 (n -1) (Q-Q, + Qz + (6) Qmax Z Qin Qm (- Q mm1 Indeed, we then: pi bQ, +, - Q (7) Q-QI

, - - QQ, (8)

  It is therefore possible to calculate the desired value Zc from pi and Zi or from Qi and Zi. In the case of two control signals Q (<) Q (2), weights W1, W2 positive for each signal can be specified, in order to take into account

18/2790620

  each signal according to its greater or lesser importance. Wl and W2 can be programmed and stored. In this case, equation (6) above becomes:

Z (W Z (Q () + W2 Z (Q (2)

  z (Pm, Q (2) -wi + 1WZwil + W, + w2 9 The equation (6) above can be transformed into the following equation (10):

Z, (Q) Z, + Q I (Z '+, -Z,) (10)

Q Q +, - Q I

  Graphically (Figure 12) we see that this function is a piecewise linear function, which can be implemented by a device according to

  the invention as already explained above in the general case.

  In some systems, the control variable can not be changed instantly arbitrarily. It is then necessary to impose a maximum instantaneous variation of the control variable. This is particularly the case for a pacemaker, the rate of cardiac activity can not be changed, for security reasons, both instantaneously and arbitrarily. In this case, a number of steps m is defined between a minimum value of the control variable Z, n and a maximum value Zmx:

  Zmin = Z (1) <Z (2) <... z (m-1) <z (m): Zmax.

  For each Z (i), a maximum increase and a decrease, Zi + and Zi

* are specified. The maximum increase Z, 4 for a given real Zr (Zi) <Zr <z (i + l)) is calculated as follows: _ 1 (n1) (Z, -Z + '+ ((-l ) (Zr-Z ')) z + () max mi max min ax In Similarly, the maximum decrease Z- is calculated in the following way: (((n-l1) (Zr -Z <A _____________) Z (Zr ) = Zx Z + Z Z (Zmax - Zin mi - ax - Z m - n)

19; 2790620

  Finally, the control value provided by the control system is a combination of the calculated Z value (by one of equations (7) or (8)) and maximum increase and decrease values. For a given real value Zr and a signal level Q, the control variable in the next step is given by (13): Z (Z ,, Q) = Zr-Z '(Zr) if Zc (Q) < ZrZ '(Zr) Z (Zr, Q) = Zc (Q) if Zr-Z (Zr) <Zc (Q) <Zr + Z + (Zr) Z (Zr, Q) = Zr + Z + (Zr) if Zc (Q)> Zr + Z + (Zr) The method and the device according to the invention then make it possible to avoid abrupt changes in the value of the control variable due to sudden increases or sudden decreases in the Q signal level. This method of calculating the control variable can be carried out using: a processor of the type already described above for calculating Zó (Q), the input variable being Q, a processor of the type already described above (with fuzzy logic) for the calculation of Z + (Zr) - of a processor of the type already described above (with fuzzy logic) for the calculation of Z '(Zr) A comparator allows to perform the step corresponding to (13) and

  delivers the appropriate signal for regulation.

  In the language of the fuzzy sets, the means 20 of FIGS. 1 and 3 are also called "fuzzifier", while the means 26 constitute an engine

  of inference, or a "defuzzifier".

  FIG. 13 schematically represents a system for monitoring the cardiac activity of a patient provided with a pacemaker 56. An electrode 58, implanted on the heart 54 of the patient, makes it possible to transmit to this organ the electrical pulses necessary for its operation. activation. The system 2 according to the invention

is implemented.

  The control algorithm makes it possible, for example, to take into account a

  or two control signals measured with one or more sensors 52.

  Existing stimulators use: patient activity, measured using an accelerometer or piezoelectric transducer, and / or minute ventilation

2790620

  2o determined from a measurement of thoracic impedance, or a time interval between two waves of an electrocardiogram (QT interval). other

  control signals could be taken into account.

  These signals are used as input signals Q in the device of FIGS. 1 and 3. FIG. 14 represents a comparison between calculation results obtained with equation (13) above (thick line curve), and an output signal of a control system according to the invention, for a pacemaker (fine line curve). The scale on the y-axis is in beat per minute (bpm), and the time axis is in seconds. We see that the simulation and the circuit give very similar results. Values of Z + = I and Z = 0.5 were chosen for

this example.

  The example has been given of an application to the realization of a control system of a pacemaker. Other possible applications concern all fields of electronic and / or automatic control implementing a nonlinear transfer function, which can be approximated by a piecewise linear function. Examples of applications are: - the realization of non-linear response functions of integrated semiconductor sensors, for the measurement of physical and / or chemical quantities - the realization of programmable non-linear control characteristics, in automatic control devices - the realization Programmable nonlinearities, used in circuits

  based on chaos theory.

21 2790620

Claims (15)

  An electronic device with a nonlinear transfer characteristic, constituted by or approximated by a piecewise linear function, the pieces being defined over intervals of variation of an input signal of the device, comprising: first means (20,22, 24,34,36,38,40) for determining, for each value of an input signal Q, the interval at which this value belongs, and the degree of membership for that interval, - second means (26, 28, 30, 42, 44) for producing an output signal Z of the device, based on information on the membership interval and membership degree of the previously determined input signal value , as well as   of the piecewise linear function.   2. Electronic device with a nonlinear transfer characteristic, constituted by, or approximated by, a piecewise linear function, the pieces being defined over intervals of variation of an input signal of the device, comprising: a first and a second current source (38, 40) for respectively shifting a value of an input signal (Q, Iin) with a first offset value (Ia), and multiplying the signal as obtained by a first multiplicative coefficient (b), to obtain a second signal (I2), - a third and a fourth current source (42, 44), respectively making it possible to multiply the second signal by a second multiplication coefficient c to obtain a third signal (I3), and shifting the third signal thus obtained by means of a second offset value (Id) to obtain a fourth signal (I4, Iout), - means for producing a digital signal (N) identifying an i the range of variation in which the input signal is located and to vary or determine the multiplicative coefficients (b, c) and the offset values (Ia, Id) as a function of   the interval in which the input signal is located.   22; The device of claim 1 or 2, further comprising digital storage means (32), for storing data (a, b, c, d) defining or characterizing the piecewise linear function on each of the   intervals on which the pieces of the linear function are defined.   4. Device according to one of claims I to 3, the first means producing   a first output signal (N) representative of the interval at which the value Q belongs and a second signal (I2) representative of the degree of membership for that interval 5. Apparatus according to claim 4, the first means including means (40) ) to produce an analog signal (I2) defining the degree   where the value of the input signal belongs to the interval to which it belongs.   6. Device according to one of claims 4 or 5, the first means   having means (36) for producing a digital signal (N) defining   the interval at which the value of the input signal belongs.   7. Device according to one of claims I to 6, the first means determining   the interval at which the value of the input signal belongs according to the degree   membership of this value for this interval.   8. Device according to one of claims 1 to 7, the first means detrminant   the degree of membership of the value of the input signal as a function of the interval   the value of the input signal belongs to.   9. Device according to one of claims I to 8, the degree of membership of the   value of the input signal being determined according to coefficients (a, b)   depending on the interval at which this value belongs. 23 2790620   10. Device according to claim 9, the degree of membership Y being dependent on the value Q of the input signal, according to the formula: Y = b (Iin la), where Ia is a signal or current depending on a ..   He. Device according to one of claims I to 10, comprising a loop of   digital feedback (36, 34) for changing the determination of the degree of membership of a value of the input signal according to the interval at which she belongs to.   12. Device according to one of claims I to 11, further comprising   means (36) for comparing the second output signal, representative of the degree of membership for an interval, with first and second limit values, respectively lower and upper, and for changing the degree of membership, as well as the first representative signal degree of membership, when the second signal falls below the first limit value or when the   second signal becomes greater than the second limit value.   13. Device according to one of claims I to 12, the second means   (26,28,30,42,44) capable of producing an output signal, depending on the degree of membership of the input signal value and on the coefficients (c, d) depending on the membership interval.   Device according to Claim 13, the second means (26, 28, 30, 42, 44) being able to produce an output signal Z, as a function of the second output signal Y of the first means (20, 22, 24, 34, 36,38,40), according to the formula: Z = cY + Id, o Id   is a signal or current depending on d.   An electronic device with a non-linear transfer characteristic, constituted by, or approximated by, a piecewise linear function, the pieces being defined over intervals of variation of an input signal of the device, comprising at least a first and a second device according to one of 24 2790620   1 to 14, one of these devices having for input signal Q and for   output signal a calculated value of Z, while one of the other circuits has as input signal a measured value of Z and output a calculated maximum variation of Z. 16. Device according to claim 15, comprising three devices according to Moon   Claims 1 to 14, a first circuit for calculating the values of   the physical quantity Z, a second circuit whose input signal is the actual value of Z, measured on or in the system and for output signal the maximum possible increase of Z, the third also having as its input signal the actual value of Z, and for output signal the maximum possible decrease of Z.   17. Device according to one of claims 15 or 16, a comparator circuit   furthermore making it possible to limit the correction of Z, calculated using the first circuit, to a value taking into account the calculated maximum possible variation value, or the values of maximum possible increase and decrease. maximum possible calculated.   18. Device for controlling a physical quantity Z of a system (50),   comprising a device according to one of claims 1 to 17, the input signal   having at least one measuring signal measured in the system, and the output signal being a control signal of the physical variable Z. 19. Device for controlling a pacemaker having a device   according to one of claims 1 to 18.   20. A method for controlling a fuzzy logic system, comprising: - defining a set of fuzzy logic rules between given values Qi of an input variable Q and values Zi of a variable of output Z: if Q = Qi, then Z = Zi 2790620   the determination of a transfer function, piecewise linear   - The implementation of a device according to one of claims 1 to 19.