JPH096366A - Fuzzy noise canceller - Google Patents

Abstract

(57) [Abstract] [Objective] A noise canceller that actively cancels noise, and a noise canceller that can perform active noise cancel flexibly with a simple configuration without using an adaptive filter such as a DSP. The purpose is to provide. [Structure] A transfer function of an unknown system is estimated by estimating a transfer function of an unknown system whose transfer function is unknown and actively canceling noise by applying a signal for canceling the noise. A fuzzy adaptive filter that generates a noise cancellation signal to be applied to the output, divides a plurality of time-series sampling data of the input signal into N equal parts, and determines a membership function for each N equal parts of the input signal. And the output error signal are used to actively update the membership function, and fuzzy inference is performed using the input signal and the membership function to actively generate the optimum noise cancellation signal.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a fuzzy noise canceling apparatus that actively cancels noise by performing fuzzy inference.

[0002] The noise of a signal processing system such as an active noise control device, a vibration suppression control system, an image processing system, a voice processing system is actively eliminated by applying a fuzzy theory.

[0003]

2. Description of the Related Art FIG. 5 shows a system of a conventional active noise canceller, in which an input signal X _j propagates (unknown system).
It represents a system that estimates the transfer function of and actively generates a noise canceling signal by an adaptive filter and applies it to the output signal of the unknown system to cancel the noise.

In FIG. 5, 110 is an unknown system,
It transmits the input signal X _j . Reference numeral 111 denotes an adder which adds noise N _j (disturbance) to the output g _j of the unknown system 110.
Is to be added.

Reference numeral 112 denotes an adder which adds the noise elimination signal G _j to a signal containing noise. 113
Is an adaptive filter for generating an optimum noise canceling signal based on the input X _j and the output E _j .

Reference numeral 114 denotes an optimum coefficient calculation unit of the adaptive filter, which calculates the optimum coefficient by the NLMS method. The operation of the configuration of FIG. 5 will be described.

The input signal X _j propagates through the unknown system 110 and outputs the output g _j . The output g _j of the unknown system 110 is the addition unit 1
The noise N _j is added at 11 and the noise elimination signal G _j is applied at the adder 112 to take the difference between g _j and G _j . The error E _j to which the noise canceling signal is applied is input to the optimum coefficient calculating unit 114, and the optimum coefficient calculating unit 114 calculates the coefficient of the adaptive filter optimum for noise canceling. The adaptive filter 113 generates the noise elimination signal G _j based on the input signal X _j by the optimum coefficient calculated by the optimum coefficient calculation unit 114. However, the acoustic feedback system is omitted in FIG.

[0008]

In the conventional active noise canceller, the adaptive filter is composed of an FIR filter,
Since the error signal (output) is controlled by a unique scalar amount, it is not possible to perform adaptive control with flexibility.
Further, since the circuit is constructed by using a DSP or the like, the circuit scale becomes large, and in particular, when the frequency response of the transmission system becomes long, the number of taps of the digital FIR filter becomes large and the processing speed is delayed.

It is an object of the present invention to provide a fuzzy noise canceller capable of actively canceling noise with a simple structure without using an adaptive filter such as a DSP.

[0010]

SUMMARY OF THE INVENTION The present invention is a noise canceller for estimating a transfer function of an unknown system whose transfer function is unknown, and applying a signal for canceling the noise to actively cancel noise. A fuzzy adaptive filter for estimating a transfer function of an unknown system and generating a noise cancellation signal to be applied to its output, divides a plurality of time-series sampling data of an input signal into N equal parts, and divides them into N equal parts. A membership function is defined for each, a discrete section is interpolated according to the shape of the membership function, the membership function is updated actively by the input signal and the error signal of the output, and fuzzy inference is performed by the input signal and the membership function. Then, the optimum noise canceling signal is actively generated.

FIG. 1 is a diagram showing the basic configuration of the present invention.
In FIG. 1, 1 is an unknown system.

Reference numeral 2 is a sampling unit. 3 is fuzzy
It is an adaptive filter. 4 is an adder, which is a fuzzy
Noise canceling signal G output from the adaptive filter 3 _jUnknown system
Output of unknown system in addition to the signal output of_jWhat is the difference between
It is.

In the fuzzy adaptive filter 3, 11 is a defuzzification section. 12 is a fuzzy inference unit.

Reference numeral 14 denotes a grade calculation unit which obtains a grade by a membership function with the input sampling value X _j as a support. A membership function holding unit 15 holds a membership function.

Reference numeral 16 is a membership function updating unit,
The membership function is updated. Membership
The update of the loop function is based on, for example, the method of least squares.
, Μ _{S, j + 1}= Μ_{S, j}+ KX_j(G_j-G_j) / ΣX_j
²Follow Where s is the n time-sampling of the input signal
The ring data is divided into N equal sections. Σ is in section s
It is the sum for each j.

Numeral 17 is an old grade holding part,
The grade calculated by the code calculation unit 14 is the old grade
(Μ_j) Is held as. 18 is an error calculator
It is. Enter the error and, for example, by the NLMS method
What finds the optimal value needed to update the ship function
Then, KX for each section_j(G_j-G_j) / ΣX_j ²Calculate
Is what you do. After that, KX_j(G_j-G_j) / ΣX _j
²= Β, and β in the section S is β_SRepresent.

[0017]

The operation of the basic configuration of the present invention shown in FIG. 1 will be described with reference to FIG. FIG. 2A shows time series data {X _j } of sampling X _j of the input signal, where S ₁ , S ₂ , ..., S
_N is a section obtained by dividing the input sampling sequence X _j at a predetermined cycle, and a membership function is assigned to each section. For example, S ₁ is the membership function M _1, S ₂ is the membership function M _2, S _N assigns the membership function M _N.

FIG. 2B shows the membership function (M₁,
M₂・ ・ ・ ・ ・ ・ M_N), And each input sample
Assigned to the segment. For example, MIN
-When fuzzy inference is performed by the MAX method, the sampling of each interval
The grade is obtained with the
Adopt for reasoning. In Fig. 2 (b), μ₁, Μ₂, ...
., Μ_NIs the minimum grade value thus obtained
You. That is, μ₁Is the section S₁For the sampling data of
Membership function M₁Of the grade when applying
It is a small price. In the case of FIG. 2 (b), S₁Of the section grade
Minimum value is support V ₂(Corresponding to the sampling value of j = 2
)) Grade. μ₂Is the section S₂Sampling
Membership function M₂If you apply
Is the minimum grade. μ_NIs the section S_NThe sampler
Membership function M_NIf you apply
It is the minimum value of the grade.

FIG. 2 (c) shows the inference result thus obtained. Figure 2 (d) is an integration of the inference results of Figure 2 (c). G _j is the center of gravity of the integration result.

Input sequence {X of sampling signal of FIG. 2 (a)
_j} Is time t_jIn. Grade calculation unit 1
4 is the membership function for each section as shown in Fig. 2 (b)
(M ₁, M₂・ ・ ・ ・ ・ ・ M_N) And input signal X_jBased on
X for each section_jDetermine the grade (j = 0-4).
The obtained grade is held in the old grade holding unit 17.

The fuzzy inference unit 12 makes a fuzzy inference for each section based on the grade of each section, for example, by its minimum value (see FIG. 2 (c)). The defuzzification unit 11 integrates the inference results of the respective sections, obtains G _j from the center of gravity, and outputs it.

On the other hand, the error calculation unit 18 uses the error E_jBased on β
(= KX_j(G_j-G_j) / ΣX_j ²). Σ is
It is the sum for j in each interval. That is, section S₁
Β is obtained for j = 0 to 4 in FIG.
₂, ..., S_NAbout β ₂, ..., β_NSeeking
You. The membership function updating unit 16 is an error calculating unit.
Β for each section obtained from 18_SAnd according to the old grade
, Μ_{S, j + 1}= Μ_{S, j}+ Β_SThe membership function is
It is held in the membership function holding unit 15.

At the next time t _{j + 1} , the input sampling sequence {X
_The above calculation is performed based on _{j + 1} } to obtain and output the noise elimination signal G _{j + 1} at time t _{j + 1} . According to the present invention, it is possible to provide a noise canceller capable of flexibly and actively canceling noise with a simple configuration without using an adaptive filter such as a DSP.

[0024]

EXAMPLE FIG. 3 shows an example of the present invention. In FIG. 3, reference numeral 21 is an unknown system for transmitting an input signal.

A sampling pulse generator 22 samples the input signal. 23
Is a multiplication unit, and is a sampling pulse generator 2
The two pulses are multiplied by the input signal to generate the sampling input signal X _j .

Reference numeral 24 denotes an adder which adds noise N _j to the output of the unknown system. Reference numeral 25 denotes an adder which adds the noise elimination signal G _j to the output of the unknown system including noise.

Reference numeral 26 is a fuzzy adaptive filter. 27
Is a white noise source (M-seq). In the fuzzy adaptive filter 26, 34 is a fuzzy inference unit.

Reference numeral 34 'is a minimum value determination unit for determining the minimum value of the grade in each section.
Reference numeral 35 is a defuzzification unit.

Reference numeral 35 'is a center of gravity computing section, which is a unit for calculating the center of gravity.
The result of fuzzy reasoning is integrated to find the center of gravity.
36 is a membership function updating unit,_{S, j + 1}= Μ
_{S, j}+ KX_j(G_j-G_j ) / ΣX_j ²According to
It updates the badship function. s represents an interval,
Σ represents taking the sum of j in each section, for example, j =
The arithmetic mean in each discrete section of J + 1 to (n + 1) J
Take.

Reference numeral 37 denotes an error calculation unit, which is β (= KX _j (g _j) necessary for obtaining the optimum filter coefficient by NLMS.
-G _j ) / ΣX _j ² ) is obtained. A K holding unit 38 holds a step gain K.

Reference numeral 39 is a grade calculator. 40 is a membership function holding unit 40. Reference numeral 41 is an old grade holding unit.

FIG. 4 is a diagram for explaining the operation of the embodiment of the present invention. FIG. 4A shows a time-series signal of the sampling data of the input signal, the sampling time interval is 250 μs, and the time is divided every 125 μs (every 5 samplings). t ₁ , t ₂ , ... T ₈ are time divisions, and the respective time divisions are S ₁ , S ₂ , S ₃ , ...
..., and S _8. FIG. 4 (a) represents sampling of a 100 Hz signal.

FIG. 4B shows the membership function at each time interval. M ₁ is the membership function in S ₁ , M ₂ is the membership function in S ₂ , and M ₃ is S
The membership function in ₃ and M ₈ are the membership functions in S ₈ . The horizontal axis (support) of each membership function is a sampling value in each section. In M ₁ , a ₀ , a ₁ , a ₂ , a ₃ , a ₄ are
These are sampling values at j = ₀ , ₁ , 2, 3, and 4, and grades are μ _1,0 , μ _1,1 , μ
_{They are 1,2} , μ _1,3 and μ _1,4 . Similarly, at M ₂ ,
b ₀ , b ₁ , b ₂ , b ₃ and b ₄ are j = 5, respectively.
Sampling values at 6, 7, 8, and 9, and grades are μ _2,0 , μ _2,1 , μ _2,2 , μ _2,3 , μ
_2,4 . At M ₃ , d ₀ , d ₁ , d ₂ , d ₃ ,
d ₄ is the sampling value at j = 10, 11, 12, 13, and 14, and the grade thereof is μ
_{They are 3,0} , μ _{3, 1} , μ _3,2 , μ _3,3 , and μ _3,4 . In M ₈ , h ₀ , h ₁ , h ₂ , h ₃ and h ₄ are respectively j
= 35, 36, 37, 38, 39 sampling values, and grades are μ _8,0 , μ _8,1 , μ
_8,2, μ _8,3, is μ _8,4.

In the fuzzy inference unit 34, the minimum value judging unit 34 'finds the minimum value of the grade in each section. Section _{S 1, S 2, S 3} , S 4, ···, the minimum grade in _{S 8 μ 1, μ 2,} μ 3, ···, and mu _8. Fig. 4 (c) shows the membership function and support (MA of the average value in the sampling data section) in Fig. 4 (b).
This is the result of fuzzy inference by taking X.

In the above, the membership function is updated by μ _{s, (j + 1)} = μ _{s, j} + β _S (β _S is KX _j (g _j −G _j ) / ΣX _j ² ) in the interval S. is there.

The center of gravity G _j is calculated by the following equation.

[0037]

[Equation 1]

The operation of the embodiment shown in FIG. 3 will be described with reference to FIG. The input signal is sampled according to the sampling timing of the sampling pulse generator 22 to generate time-series data {X _j } of the input signal. The grade calculator 39 determines the grade of each sampling X _j (see FIG. 4 (b)). The obtained grade is held in the old grade holding unit 41. Further, the fuzzy inference unit 34 finds the minimum value of the grade of each section and performs the fuzzy inference (FIG. 4 (c).
reference). In the defuzzification unit 35, the center of gravity calculation unit 35 '
Integrates each inference result, finds the center of gravity, and outputs it as G _j .

The error calculator 37 inputs the error E _j and obtains β _S in each section. The membership function updating unit 36 calculates an optimum membership function from the input sampling sequence {X _{J + 1} }, E _j and β _{S at the} next time. If the membership function of section S at time j is μ _{S, j , then} time j
+1 μ _{S, j + 1} = μ _{S, j} + β _S. However, β _S is K (g _j −G _j ) / ΣX _j ² in the section s.

In the configuration of FIG. 3, the white noise source 27 is used to add the white noise to the unknown system and obtain the frequency response characteristic when the system is started up.

[0041]

According to the present invention, it is possible to provide a noise canceller capable of flexibly and actively canceling noise with a simple structure without using an adaptive filter such as a DSP.

[Brief description of drawings]

FIG. 1 is a diagram showing a basic configuration of the present invention.

FIG. 2 is an explanatory diagram of a basic configuration of the present invention.

FIG. 3 is a diagram showing an embodiment of the present invention.

FIG. 4 is an explanatory diagram of an example of the present invention.

FIG. 5 is a diagram showing a conventional technique.

[Explanation of symbols]

 1: Unknown system 2: Sampling unit 3: Fuzzy adaptive filter 4: Addition unit 11: Defuzzification unit 12: Fuzzy inference unit 14: Grade calculation unit 15: Membership function holding unit 16: Membership function updating unit 17: Old Grade holding unit 18: Error calculation unit

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Internal reference number FI Technical display location H03H 21/00 8842-5J H03H 21/00 // G05B 13/02 0360-3H G05B 13/02 N

Claims (5)

[Claims] 1. A transfer function of an unknown system is estimated in a noise canceller that actively cancels noise by applying a signal for canceling the noise by estimating the transfer function of an unknown transfer function. And a fuzzy adaptive filter that generates a noise canceling signal to be applied to its output, and divides a plurality of time-series sampling data of the input signal into N equal parts,
A membership function is determined for each of the N equally divided sections, the membership function is actively updated by the input signal and the output error signal, and fuzzy inference is actively performed by the input signal and the membership function. A fuzzy noise canceller characterized by generating an optimum noise canceling signal. 2. The fuzzy system according to claim 1, further comprising a white noise source, wherein the white noise is input to an unknown system, and the frequency response characteristic of the system including the unknown system is obtained by fuzzy inference. Noise canceller. 3. The fuzzy inference based on the sampling data in the section and the membership function in the section, the fuzzy inference results of each section are integrated, and a noise cancellation signal is output based on the integrated result. 1. The fuzzy noise canceller according to 1 or 2. 4. When the grade of the sampling signal X _j in the section s is μ _{S, J} , the grade in X _{j + 1} is μ _{S, j + 1} = μ _{S, j} + KX _j E _j / ΣX _j ² ( However, K
Is a step gain, E _j is an error, and Σ is a sum for j in each section). The fuzzy noise canceller according to claim 1, 2 or 3, wherein 5. The Σ of μ _{S, j + 1} = μ _{S, j} + KX _j E _j / ΣX _j ^{2 in} the normalized least squares algorithm is j = n.
J + 1 to (n + 1) J (J is a number in each discrete section)
5. The fuzzy noise canceller according to claim 4, wherein the averaging is performed in each discrete section.