EP0814635B1 - Hearing aid - Google Patents

Abstract

The hearing aid comprises an input converter (12), an amplifier - and transmission arrangement (10), an output converter (14) and a calculation arrangement (20), which realises fuzzy-logic functions, and which responds to a output signal (22) of the amplifier- and transmission arrangement, providing a result signal (26). The result signal is conveyed to the amplifier- and transmission arrangement and influences their output signal (28). At least the calculation arrangement is provided as a digital circuit.

Description

The invention relates to a hearing aid according to the preamble of Claim 1. Under a "signal" here is the course one or more physical quantities on one or several measuring points can be understood over time; each Signal can therefore consist of a bundle of individual signals.

Such a hearing device is known from EP-A-0 674 464, in which a fuzzy logic controller is provided to either the signal transmission characteristic of an amplifier and Change transmission device or set of parameters influencing the signal transmission characteristic to be selected automatically from a parameter memory.

EP-A-0 674 463 discloses a similar hearing aid an automatic gain control circuit (automatic gain control - AGC) assigned a fuzzy logic controller is.

In the hearing aids described in these published documents is however only an implementation of fuzzy logic functions provided in analog circuit technology. It follows the problem of high circuit complexity, the particular because of the necessary with hearing aids Miniaturization is a disadvantage.

The invention accordingly has the object of the problem mentioned to solve. In particular, the invention is intended to be a hearing aid be provided, which is with little developmental and Circuit effort can be made and an optimal Adaptation to the specific requirements of the hearing aid wearer allows.

According to the invention, this object is achieved by a hearing aid with the features according to claim 1.

A digital structure of a computing device, the fuzzy logic functions realized, offers a high degree of compatibility with digital signal processing: one additional implementation (analog / digital or digital / analog) is not required and the calculation device can completely or partially realized with the same components become like the rest of the processing of the signals. It follows the calculation device is easy to combine with conventional digital data and signal processing functions, such as in microprocessors or signal processors are common. In addition, digital technology offers Benefits such as increased immunity to interference and insensitivity against manufacturing tolerances.

The calculation device is preferably with conventional digital ones Components such as gates, flip-flops, memories, etc. educated; more generally with switching networks and switching mechanisms. she can be used in particular as an ASIC (application specific integrated circuit - application-specific integrated circuit) his. Alternatively, it is possible to use the calculation device as a microprocessor or microcontroller with one train associated program in a read-only memory (ROM; especially mask-programmed ROM, PROM, EPROM or EEPROM) or a read-write memory (RAM) is saved. Mixed forms are also possible; for example can use specific hardwired modules with a programmed control. This is particularly so useful for functions that are performed frequently and can be realized digitally relatively easily, for example for functions for calculating the maximum or minimums of several binary numbers.

The computing device is preferred in the hearing device according to the invention for direct signal processing and / or for the control of signal processing functions and / or for the automatic selection of hearing programs in the Hearing aid used.

Furthermore, the computing device of the hearing aid realizes the fuzzy logic functions are preferred by executing the Sub-steps fuzzification of sharp input variables, Evaluation of premises, evaluation of partial conclusions, Accumulation of initial terms and defuzzification. The calculations required for this are preferably based on several Calculation modules distributed, the local or shared storage can have.

Configuration parameters of the calculation device are preferred in a memory, for example a RAM or EEPROM, filed, so that reprogramming of the calculation device by the hearing care professional and / or even an adaptation of the function of the computing device during operation of the hearing aid is possible.

Other preferred embodiments are in the rest Subclaims defined.

Fig. 1 ein Blockschaltbild eines erfindungsgemäßen Hörgerätes,

Fig. 2 eine konzeptionelle Darstellung einer beispielhaften Verarbeitungsstruktur,

Fig. 3 Graphen von Zugehörigkeitsfunktionen zur Veranschaulichung der Fuzzyfizierung,

Fig. 4a bis Fig. 4c und Fig. 5a bis Fig. 5e Graphen von beispielhaften Zugehörigkeitsfunktionen,

Fig. 6 eine die Auswertung von Prämissen veranschaulichende Darstellung,

Fig. 7 eine Darstellung zweier Möglichkeiten zur Bestimmung des Aktivierungsgrades einer Teilkonklusion,

Fig. 8 eine Darstellung zweier Möglichkeiten zur Bestimmung der Aktivierung eines Terms,

Fig. 9 eine Darstellung zweier Möglichkeiten zur Akkumulation von Ausgangstermen,

Fig. 10 eine ein erstes Verfahren zur Defuzzyfizierung veranschaulichende Darstellung,

Fig. 11 eine ein zweites Verfahren zur Defuzzyfizierung sowie ein aufwandsreduziertes Verfahren veranschaulichende Darstellung,

Fig. 12 ein Blockschaltbild einer Berechnungseinrichtung eines erfindungsgemäßen Hörgerätes,

Fig. 13 ein Blockschaltbild einer ersten Ausführungsalternative der in Fig. 12 gezeigten Berechnungseinrichtung, und

Fig. 14 ein Blockschaltbild einer zweiten Ausführungsalternative der in Fig. 12 gezeigten Berechnungseinrichtung.

Embodiments of the invention will now be described in more detail with reference to the drawings. They represent:

1 is a block diagram of a hearing aid according to the invention,

2 shows a conceptual illustration of an exemplary processing structure,

3 graphs of membership functions to illustrate the fuzzification,

4a to 4c and 5a to 5e are graphs of exemplary membership functions,

6 is a representation illustrating the evaluation of premises,

7 shows a representation of two possibilities for determining the degree of activation of a partial conclusion,

8 shows two options for determining the activation of a term,

9 shows two possibilities for the accumulation of output terms,

10 shows a representation illustrating a first method for defuzzification,

11 shows a diagram illustrating a second method for defuzzification and a reduced-effort method,

12 is a block diagram of a computing device of a hearing aid according to the invention,

FIG. 13 shows a block diagram of a first alternative embodiment of the calculation device shown in FIG. 12, and

FIG. 14 shows a block diagram of a second alternative embodiment of the calculation device shown in FIG. 12.

In the hearing aid shown schematically in Fig. 1 sets a microphone acting as an input converter 12 produces a sound signal into an electrical signal and guides it an amplifier and transmission circuit 10 further. The Amplifier and transmission circuit 10 amplifies the incoming Signal and processes it, for example through selective Raising or weakening certain frequency or Volume areas. The output signal 28 processed in this way is output by a handset serving as an output transducer 14.

At least one suitable point of the amplifier and Transmission circuit 10 is a tap signal 22 from the Tapped signal path of the hearing aid and a signal processing device 16 fed. The tap signal 22 can furthermore have individual signals which are transmitted by further input converters, of controls or sensors for Monitoring system properties (e.g. the Battery voltage).

The signal conditioning device 16 prepares the tap signal 22 suitable, for example by rectification, Averaging or derivative over time to make it one Computing device 20, which realizes fuzzy logic functions, to supply as input signal 24. With regard to the Design of the signal processing device 16 and with regard to the individual signals that make up the tap signal 22 is composed, the content of EP-A-0 674 464 hereby expressly included in the present description.

The calculation device 20 has a memory 18, the intermediate results and, if necessary, configuration parameters the calculation device 20 stores. The calculation device 20 processes the input signal supplied to it 24 in the manner described in more detail below the principles of fuzzy logic and gives the result as Result signal 26 to the amplifier and transmission device 10, their gain and transmission properties by the result signal acting as a control signal 26 can be changed within wide limits.

In one embodiment of the invention, only that Calculator 20 digitally executed while the other assemblies, except where necessary Analog-digital and digital-analog converters, as analog Circuits are formed. In an alternative implementation are, however, the amplifier and transmission device 10, the signal conditioning circuit 16 and the calculation device 20 executed essentially digitally, and that Tap signal 22, the input signal 24 and the result signal 26 are digital signals that are preferred as consecutive Binary numbers on several lines in parallel be transmitted. In this alternative embodiment points only the amplifier and transmission device 10 an analog-digital converter for that of the input converter 12 originating signal and a digital-to-analog converter on the the output signal 28 passed to the output converter 14 generated.

In the embodiment of the invention shown in FIG. 1 Hearing aid controls the result signal 26 the transmission characteristic the amplifier and transmission device 10 directly by the individual signals of the Result signal 26 individual parameters of the amplifier and Transmission device 10, for example the gain certain frequency bands or response and fall times an automatic gain control control - AGC).

In an alternative embodiment, the amplifier and Transfer device 10 has a memory that several contains preset or programmed parameter sets. A parameter set of this memory is based on the result signal 26, selected, for example, that the digital result signal 26 as a memory address serves.

In a further alternative embodiment, the amplifier and transmission device 10 no immediate Signal path from input converter 12 to output converter 14 on. Rather, the signal path runs from the input converter 12 via a first part of the amplifier and transmission device 10 to the signal conditioning device 16, from there to the calculation device 20, from there as a result signal 26 to a second part of the amplifier and transmission device 10 and from there as an output signal 28 to Output converter 14. In the second part of the amplifier and Transmission device 10 becomes the digital result signal 26 only converted into an analog signal and possibly filtered.

Regel 1:

WENN (A ist groß) UND (B ist groß)
DANN (X ist groß) UND (Y ist groß)

Regel 2:

WENN (A ist klein) ODER (B ist groß)
DANN (X ist klein)

Regel 3:

WENN (A ist klein) UND (B ist klein)
DANN (X ist klein) UND (Y ist klein)

The fuzzy logic used in the hearing aid according to the invention permits the processing of signals and information according to fuzzy regulations, a so-called rule set. This rule set can be, for example, as follows:

Rule 1:

IF (A is large) AND (B is large)
THEN (X is large) AND (Y is large)

Rule 2:

IF (A is small) OR (B is large)
THEN (X is small)

Rule 3:

IF (A is small) AND (B is small)
THEN (X is small) AND (Y is small)

The expression between IF and THEN is called the premise; the expression to the right of the THEN is called a conclusion designated. The sub-expressions in parentheses become corresponding referred to as partial premises and partial conclusions.

In the following, the individual are illustrated using this rule set as an example Sub-functions of the calculation device 20 explained.

1) Fuzzyfizierung 50 der scharfen Eingangsvariablen. Gleichzeitig wird dadurch der Erfüllungsgehalt der Teilprämissen ermittelt.

Auswertung 52 der Prämissen, d.h. Ermittlung des Erfüllungsgehaltes der Prämissen.

3) Auswertung der Teilkonklusionen, d.h. Ermittlung der Aktivierung der Teilkonklusionen. In der Darstellung von Fig. 2 ist dieser Schritt in die beiden Teilschritte Bestimmung 54 des Aktivierungsgrades der Teilkonklusionen und Bestimmung 56 der Aktivierung der Terme der Ausgangsvariablen unterteilt.

4) Akkumulation 58 der Ausgangsterme, d.h. Ermittlung der Aktivierung der Konklusionen.

5) Defuzzyfizierung 60 der aktivierten Konklusionen. Dadurch werden wieder scharfe Ausgangsvariablen bestimmt.

Fig. 2 shows the conceptual structure of the processing of the above rule set. It consists of the following essential sub-functions:

1) Fuzzification 50 of the sharp input variables. At the same time, the fulfillment content of the sub-premises is determined.

Evaluation 52 of the premises, ie determination of the fulfillment content of the premises.

3) Evaluation of the partial conclusions, ie determination of the activation of the partial conclusions. 2, this step is divided into the two sub-steps determining 54 the degree of activation of the partial conclusions and determining 56 the activation of the terms of the output variables.

4) Accumulation 58 of the starting terms, ie determining the activation of the conclusions.

5) Defuzzification 60 of the activated conclusions. This again determines sharp output variables.

In the digital implementation of the calculation device according to the invention 20 serves the structure shown in FIG. 2 only for the conceptual representation of a fuzzy logic calculation, because in actual implementation one arbitrary assignment of the partial functions shown in FIG. 2 to one or more modules of the calculation device 20 can be done.

Step 1) - Fuzzification of the input variables

Fuzzification determines the value of a membership function of every linguistic term of corresponding linguistic variables in the current Value of the input variable.

This is shown by way of example in FIG. 3. The example set of rules contains two linguistic variables A and B, each with two linguistic terms, namely (A is small), (A is large) and (B is small), (B is large). The graphs shown in FIG. 3 represent the membership functions of these terms: µ _small (A), µ _large (A) and µ _small (B), µ _large (B). The input values a and b are mapped to the corresponding values µ _small (A = a) µ _large (A = a) and µ _small (B = b), µ _large (B = b) of the membership functions.

• Völlig freier Verlauf der Zugehörigkeitsfunktion (Fig. 4a); eine Beschränkung ergibt sich - wie auch in den folgenden Klassen - lediglich durch die Quantisierung der Kurven. Jede Zugehörigkeitsfunktion muß - entsprechend der angewandten Quantisierung - in Form ihrer Einzelwerte abgespeichert werden. Dies ist relativ speicheraufwendig. Auch die Weiterverarbeitung ist rechenaufwendig.
• Linearer Verlauf des Funktionswertes zwischen beliebig angebbaren Eckwerten (Fig. 4b und Fig. 5a). Durch diese Einschränkung ergibt sich ein reduzierter Speicher- und Rechenaufwand. Jede Zugehörigkeitsfunktion kann, entsprechend der Anzahl M der Eckwerte, als eine Folge von x-y-Wertepaaren (x₁, y₁, x₂, y₂, ..., x_M, y_M) dargestellt werden.
• Linearer Verlauf des Funktionswertes zwischen maximal vier Eckwerten, als deren Ordinatenwerte nur 0 und 1 zulässig sind. Alle damit möglichen Kurvenverläufe sind in Fig. 4c dargestellt: linke Schulterfunktion 72, Trapezfunktion 70, Dreiecksfunktion 74, rechte Schulterfunktion 76. Diese Einschränkung ergibt eine optimale Reduzierung des Speicheraufwandes. Da maximal vier Eckwerte vorhanden sind und als y-Werte nur 0 und 1 verwendet werden, kann jede in diese Klasse fallende Zugehörigkeitsfunktion allein durch ihre vier x-Werte (x₁, x₂, x₃, x₄) eindeutig beschrieben werden. In Fig. 5b ist dies für die Trapezfunktion 70 gezeigt, in Fig. 5c für die linksseitige Schulterfunktion 72 (hierbei gilt x₁ = x₂), in Fig. 5d für die Dreiecksfunktion 74 (hierbei gilt x₂ = x₃) und in Fig. 5e für die rechtsseitige Schulterfunktion 76 (hierbei gilt x₃ = x₄).

The exemplary membership functions shown in FIGS. 4a to 4c and 5a to 5e can be divided into three classes:

• Completely free course of the membership function (Fig. 4a); the only limitation - as in the following classes - is the quantization of the curves. Each membership function must be saved in the form of its individual values, depending on the quantization used. This is relatively memory intensive. Further processing is also computationally complex.
• Linear course of the function value between arbitrarily definable basic values (Fig. 4b and Fig. 5a). This limitation results in a reduced storage and computing effort. Each membership function can be represented according to the number M of the basic values as a sequence of xy value pairs (x ₁ , y ₁ , x ₂ , y ₂ , ..., x _M , y _M ).
• Linear course of the function value between a maximum of four basic values, the ordinate values of which only 0 and 1 are permissible. All possible curves are shown in FIG. 4c: left shoulder function 72, trapezoidal function 70, triangular function 74, right shoulder function 76. This restriction results in an optimal reduction in the amount of memory. Since there are a maximum of four basic values and only 0 and 1 are used as y values, each membership function falling in this class can be clearly described by its four x values (x ₁ , x ₂ , x ₃ , x ₄ ). This is shown in FIG. 5b for the trapezoidal function 70, in FIG. 5c for the left shoulder function 72 (here x ₁ = x ₂ ), in FIG. 5d for the triangular function 74 (here x ₂ = x ₃ ) and in Fig. 5e for the right shoulder function 76 (here x ₃ = x ₄ ).

In order to determine the degree of fulfillment of the partial premises during the fuzzification to be able to calculate before the start of fuzzyfication each input value on the internally used abscissa normalized. It is further assumed that the input values are already standardized.

In the case of the free course shown in Fig. 4a Affiliation function is used to determine the degree of fulfillment by reading out the corresponding x value assigned y value from the memory.

1. Wenn V < x₁ ist, dann ist µ... (V=v) = 0.

2. Durchlaufe alle Werte x_m = x₂, ..., x_M, bis gilt x_m > V. Dann berechne den Erfüllungswert nach der Vorschrift µ... (V=v) = ym-1 + ym - ym-1 xm - xm-1 * (V - xm-1) und breche den Durchlauf ab.

3. Wenn bei obigem Durchlauf niemals die genannte Bedingung erfüllt worden ist, dann ist µ... (V=v) = 0.

In the linearized membership function shown in FIGS. 4b and 5a, the value of the membership function μ ... (V = v) is to be determined according to the following rule:

1. If V <x ₁ , then µ ... (V = v) = 0.

2. Run through all values x _m = x ₂ , ..., x _M , until x _m > V applies. Then calculate the compliance value according to the regulation µ ... (V = v) = y m-1 + y m - y m-1 x m - x m-1 * (V - x m-1 ) and cancel the run.

3. If the above condition has never been met in the above run, then µ ... (V = v) = 0.

If a negated variable occurs in the set of rules, the value of the inverse membership function must be determined. This is calculated from the value of the non-inverted membership function: inv [µ ... (V = v)] = 1 - µ ... (V = v).

1. Wenn V < x₁ ist, dann ist µ... (V=v) = 0 [1].

2. Wenn x₂ > V ist, dann berechnet sich der Erfüllungsgrad nach der Vorschrift

3. Wenn x₃ > V ist, dann ist µ... (V=v) = 1 [0].

4. Wenn x₄ > V ist, dann berechnet sich der Erfüllungsgrad nach der Vorschrift

5. Wenn keine der genannte Bedingung erfüllt ist, dann ist µ... (V=v) = 0 [1].

In the case of the greatest possible simplification of the membership functions shown in FIGS. 4c and 5b to 5e, the - now also simpler - calculation rule for the value of the membership function is µ ... (V = v)

1. If V <x ₁ , then µ ... (V = v) = 0 [1].

2. If x ₂ > V, the degree of fulfillment is calculated according to the regulation

3. If x ₃ > V, then µ ... (V = v) = 1 [0].

4. If x ₄ > V, the degree of fulfillment is calculated according to the regulation

5. If none of the above conditions is met, then µ ... (V = v) = 0 [1].

If a negated variable occurs in the set of rules, it is Value of the inverse membership function according to the above to determine the specified formula. You can choose at the Calculate the values given in square brackets above be used.

Step 2) - Evaluation of the premises

The values of the membership functions calculated in step 1), which is the degree of fulfillment of the partial premises (A large), (B is large) and so on, are in the example set of rules used here by linguistic ANDund OR operators on the premises of the individual rules connected.

The calculation of the AND and OR operations of the partial premises is preferably done by calculating the Minimum or maximum of the corresponding degree of fulfillment, as shown in FIG. 6. The result of this Operation is the degree of fulfillment of the respective premise [(A is large) AND (B is large)], [(A is large) OR (B is large)] and so on. This calculation is done for all the rules.

Step 3) - Evaluation of the partial conclusions

To evaluate a partial conclusion, the first Partial step determines the degree of activation of the partial conclusion. The principle applies that every partial conclusion in is activated to the extent assigned to it in the set of rules Premises are fulfilled.

• Bildung des Maximums der Erfüllungsgrade der Prämissen, oder
• Bildung der (auf Eins) beschränkten Summe der Erfüllungsgrade der Prämissen.

If a partial conclusion is only mentioned once in the set of rules, its degree of activation is equal to the degree of fulfillment of the corresponding premise. If a partial conclusion is mentioned in several rules, ie the degree of activation depends on several premises, the degrees of activation of the premises in question must be offset against each other in a suitable manner. There are in particular the two options shown in FIG. 7:

• Formation of the maximum of the degree of fulfillment of the premises, or
• Formation of the (to one) limited sum of the degrees of fulfillment of the premises.

The result of this operation is the degree of activation of the Sub-conclusion. This calculation is done for all partial conclusions.

• Beschränkung der maximalen Werte der Zugehörigkeitsfunktion auf den Wert des Aktivierungsgrades, oder
• Multiplikation des Verlaufes der Zugehörigkeitsfunktion mit dem Wert des Aktivierungsgrades.

In a second step of evaluating the partial conclusions, the activation of the terms of the output variables is determined. Each partial conclusion activates a corresponding term of an output variable. These terms are described by their membership functions. Their activation, ie the measure with which they are currently effective, corresponds to a partial area under this membership function. This partial area is in turn determined by the degree of activation of the partial conclusion (determined in the first partial step above). One of the two methods shown in FIG. 8 is preferably used to determine the activation of the corresponding term from the degree of activation of a partial conclusion:

• Limiting the maximum values of the membership function to the value of the degree of activation, or
• Multiplication of the course of the membership function by the value of the degree of activation.

This calculation is done for all terms of all output variables.

Step 4) - Accumulation of the starting terms

• Bildung des Maximums der die aktivierten Terme umgebenden Funktionsverläufe für jeden Abszissenwert, oder
• Addition der die aktivierten Terme umgebenden Funktionsverläufe für jeden Abszissenwert.

Each linguistic output variable usually consists of several terms. Its activation has now been determined for each of these terms. The individual activated terms of each output variable must now be appropriately superimposed (accumulated). For this purpose, the two methods shown in FIG. 9 are preferably provided:

• Formation of the maximum of the functional curves surrounding the activated terms for each abscissa value, or
• Addition of the functional curves surrounding the activated terms for each abscissa value.

This accumulation happens for each output variable.

Step 5) - Defuzzification

• Bestimmung des Mittelwerts der Maxima (Fig. 10), oder
• Schwerpunktsbestimmung (Fig. 11).

Defuzzification determines a sharp output value from the accumulated terms of each output variable. The defuzzification operation is thus applied to each output variable. The following two methods are possible:

• Determination of the mean of the maxima (Fig. 10), or
• Center of gravity (Fig. 11).

In the type of defuzzification shown in FIG. 10 by determining the mean value of the maxima, the sharp initial value x is calculated as the mean value of the positions of the maxima of f _active (X).

Dies entspricht der Berechnung der x-Komponente des Flächenschwerpunktes. In the center of gravity method illustrated in FIG. 11, the following calculation method is used to calculate the sharp output value x on the accumulated terms of each output variable:

This corresponds to the calculation of the x component of the centroid.

In the case of a digital implementation of the calculation, the integrations are to be replaced by totals. The following then applies:

In order to shorten the calculation, the area over which integration or summation is carried out is preferably limited to the interval between X _min and X _max ; the interval between the smallest and the largest X value, for which f _active (X)> 0 applies. This information arises when the starting terms are accumulated.

The method described below allows a reduced effort Calculation of the steps from activation the terms of the output variables until defuzzification.

If the activation of the associated starting term is determined from the degree of activation of the conclusion, then this operation can be described by two mapping functions: the degree of activation of the conclusion is mapped on the one hand to the activated area F _{n of} the starting term, and on the other hand it is centered on a position S _n this activated surface imaged. Both mapping rules do not have to be evaluated at runtime of the system, since they are only dependent on the starting terms and the method of converting the degree of activation of the conclusion into the activation of the terms (maximum formation or multiplication) shown in FIG. 8.

für jede Ausgangsvariable. Hierbei steht N für die Anzahl der Terme der Ausgangsvariablen. In dem in Fig. 11 gezeigten Beispiel ergibt sich damit ein Gesamtschwerpunkt x = (S1*F1 + S2*F2)/(F1 + F2). 11 illustrates the described transition to two separate mapping rules. The accumulation of the initial terms and the defuzzification now take place simultaneously by executing the calculation rule

for each output variable. Here N stands for the number of terms of the output variables. In the example shown in FIG. 11, this results in an overall center of gravity x = (p 1 * F 1 + S 2 * F 2 ) / (F 1 + F 2 ).

This calculation method implicitly includes accumulation the terms by the method of addition.

12 is a first embodiment of the invention Calculator 20 shown, which described Executes fuzzy logic functions. The calculation device 20 has six calculation modules 30, which are connected in series via five buffers 32 are. Each calculation module 30 is also one Memory module 34 each assigned a configuration input 36. A control module 40 is with all calculation modules 30 and connected to a working memory 42 on which can be accessed externally via a connection 44.

Each partial function type 50, 52, 54 shown in FIG. 56, 58 and 60 corresponds to one of the calculation modules 30 first calculation module 30 receives the sharp input values as input signal 24; the last calculation module 30 there the calculated sharp result values as result signal 26 out. The transfer of the intermediate results between the calculation modules 30 takes place via the buffer 32.

In the memory module assigned to each calculation module 30 34 internal interim results can be stored. each Memory module 34 may also contain configuration information for that executed by the respective calculation module 30 Partial function included. Such configuration information can, for example, in the first calculation module 30 receives the input signal 24, the membership functions of the Be input variables. To configure the fuzzy logic functions of the calculation device 20 are the memory modules 34 writable from the outside via the configuration inputs 36.

The control module 40 coordinates the overall process and the cooperation of the calculation modules 30. For example the processing time in the individual calculation modules 30 be different. It is the task of the control module 40 then to notify each calculation module 30 if the intermediate results of the previous calculation module 30 for Pending further processing.

Also in the working memory assigned to the control module 40 42 can intermediate results and configuration information be filed.

The implementation of the calculation modules 30 and others Components of the computing device 20 in digital Circuit technology results directly in a known manner from the description of the corresponding sub-functions. she can be done by switching networks, switching mechanisms or a combination happen from both. Their exact function can be determined by configuration information be determined.

The number of those provided in the calculation device 20 Computation modules 30 do not necessarily need six his. There may be more or fewer calculation modules 30 be to calculate the fuzzy logic functions to divide finer or coarser. For example five calculation modules 30 corresponding to those described above Steps 1) to 5) can be used, or only a single one Calculation module 30 'as shown in FIG. 14 is.

13 shows an embodiment variant of the calculation device 20. All buffers shown in Fig. 12 32 and memory modules 34 and the working memory 42 summarized here to the single memory 18. This allows a more rational use of storage space because it is partitioned as desired and according to the individual modules Can be assigned to demand. So information, which are required by different modules, only once are stored in the memory 18.

14 shows a further embodiment variant of the calculation device 20. All calculation modules 30 are closed here summarized in a single calculation module 30 '. Becomes this calculation module 30 'as far as possible designed as a programmable operations center, so its Computing partitioned arbitrarily and the individual Sub-functions can be assigned. This ensures one optimal data throughput through the entire system.

In a further preferred embodiment, the calculation modules 30 (or the calculation module 30 ') access on a preferably hard-wired module for determination the minimum minimum and / or the maximum of two or more Binary numbers. This is advantageous because the formation of the Minimum and maximum two in many fuzzy logic subfunctions occurring basic functions are.

Claims (9)

1. Hearing aid with an amplifier and transmission means (10), which is connected on one side to an input transducer (12) and on the other side sends an output signal (28) to an output transducer (14), and with a calculating means (20) which implements fuzzy logic functions, responds to a tapped signal (22) which is tapped on the amplifier and transmission means (10) and supplies a result signal (26) which is sent to the amplifier and transmission means (10) and acts on the latter's output signal (28), characterised in that

   the amplifier and transmission means (10) has a signal path between the input transducer (12) and the output transducer (14), whose amplification and transmission characteristic can be acted on by the result signal (26) of the calculating means (20),

   the amplifier and transmission means (10) has a memory in which a plurality of sets of amplification and transmission parameters are stored and the result signal (26) of the calculating means (20) is used to select one of these parameter sets,

   a signal editing means (16) is provided which edits the tapped signal (22) which is tapped by the amplifier and transmission means (10) and sends it as an input signal (24) to the calculating means (20), and

   at least the calculating means (20) and the signal editing means (16) are executed using digital circuit technology.
2. Hearing aid according to Claim 1, characterised in that a signal path of the hearing aid proceeds from the input transducer (12) via a first part of the amplifier and transmission means (10), the calculating means (20) and a second part of the amplifier and transmission means (10) to the output transducer (14).
3. Hearing aid according to Claim 1, characterised in that the amplifier and transmission means (10) has an analogue-to-digital converter and a digital-to-analogue converter.
4. Hearing aid according to one of Claims 1 to 3, characterised in that the calculating means (20) has a control module (40), at least one memory (18; 32, 34, 42) and at least one calculating module (30; 30').
5. Hearing aid according to Claim 4, characterised in that a separate calculating module (30) and/or a separate memory module (34) is provided in the calculating means (20) for each step of a procedure to implement the fuzzy logic functions.
6. Hearing aid according to Claim 4 or 5, characterised in that a plurality of calculating modules (30) connected in series and at least one intermediate memory (32) for linking series-connected calculating modules (30) is provided in the calculating means (20).
7. Hearing aid according to one of Claims 1 to 6, characterised in that the calculating means (20) is equipped to implement the fuzzy logic functions in the sub-steps

   fuzzification (50) of sharp input variables,

   evaluation (52) of premises,

   evaluation (54, 56) of sub-conclusions,

   accumulation (58) of output terms, and

   defuzzification (60).
8. Hearing aid according to Claim 7, characterised in that the calculating means (20) is equipped to employ membership functions during fuzzification (50), said membership functions having a function value which proceeds linearly between a maximum of four comer values in each case, whereby the ordinate values of the maximum of four corner values are either 0 or 1 in each case.
9. Hearing aid according to Claim 7 or 8, characterised in that the calculating means (20) is equipped to implement the accumulation (58) of the output terms and the defuzzification (60) for each output variable simultaneously by executing the calculation rule

   

   whereby for each output variable the number of output terms of these output variables is designated by N, the activated surface of the n-th output term by F_n and the centre of gravity position of this surface by S_n.

Images