JP2010028756A - Apparatus and method for designing finite impulse response filter - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an apparatus and method for designing an FIR (finite impulse response) filter capable of calculating filter coefficients of an FIR filter that copes with data structure of a unique user-defined type, without having to carry out weighting by adaptation of a window function. <P>SOLUTION: The FIR filter designing apparatus 10 includes: a white noise signal generation means 2 for generating white noise signal; a reference filter means 3 for converting the white noise signal to a signal of a predetermined frequency characteristic and outputting the converted signal as a target signal; an adaptation filter means 4, having an FIR filter section 41 that functions as an FIR filter that corresponds to the data structure of a user-defined type, making the filter coefficients variable, and outputting the white noise signal as a response signal; a subtraction means 5 for subtracting the response signal from the target signal to derive an error signal; and a convergence monitoring means 6 for detecting convergence of the error signal. The convergence monitoring means 6 outputs variable filter coefficients of the FIR filter section 41, when the convergence of the error signal is detected. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention provides a filter coefficient or transfer function of a finite impulse response (hereinafter referred to as FIR) filter that outputs a signal having a desired frequency characteristic by extracting or removing a specific component from an input signal. The present invention relates to an FIR filter design apparatus and an FIR filter design method that can calculate the above.

  The FIR filter is used for development of a simulation application, a modem, an echo canceller, or an LSI (Large-scale integration), and designing the FIR filter is calculating a filter coefficient. The design of this FIR filter includes a specific data type according to the programming language specification, usually an integer (Integer: INT) type in which the decimal point is declared as 16 bits as a fixed-point number, or a real number type. It is used in combination with a standard data type such as a float type floating point number.

  On the other hand, the design of FIR filters has been simplified in recent years due to high integration in specifications required by customers, and the data structure of standard data types has been simplified, and FIR filters have been developed with unique user-defined data structures. It is requested to do. Here, the unique user-defined data structure refers to a data structure other than a C data type that is one of programming languages. C data types include variable types such as char, int, float, and double, modifiers such as signed, unsigned, long, and short that change the meaning of the type, and structures and memory equivalent to Pascal records. There are also unions that share each other. In addition, there are access modifiers such as const and volatile. There are also register types and enumeration types.

  For this reason, all FIR filters constructed with a unique user-defined data structure need to obtain a transfer function with a unique user-defined data structure in order to calculate filter coefficients.

  Therefore, in the design of the conventional FIR filter, when the transfer function is determined by a user-defined type that defines a unique data structure (for example, an integer type in which a decimal point is declared as 5 bits), the transfer function is determined. Is simulated (for example, an analog quantity is defined as an integer type of 5 bits) and weighted by adaptation of a window function (for example, Hanning window, Hamming window, Blackman window).

  On the other hand, the conventional audio apparatus sets a predetermined transfer characteristic by determining the coefficient of the filter means based on the layout of the front speaker. The filter means receives an audio signal through a band-pass filter, convolves the audio signal with a transfer characteristic, and outputs a desired target signal. The adaptive signal processing apparatus receives an error signal and an audio signal that are the difference between the target signal and the music signal at the observation point, and performs adaptive signal processing so that the power of the error signal is minimized (see, for example, Patent Document 1). .

  In addition, the conventional active noise control device is provided with an entire system filter that simulates the characteristics of the entire system from the noise detection microphone to the error detection microphone, and the coefficient update of the noise control filter is stopped and fixed at intervals. A differential total system filter providing a characteristic corresponding to a difference between two transfer functions provided by the total system filter obtained for the first and second coefficients, and a transmission corresponding to the first and second coefficients A differential noise control filter having a function difference as a characteristic, and an adaptive filter cascaded to the differential noise control filter, and when white noise is applied to the entire system filter and the differential noise control filter, The coefficient of the adaptive filter is updated so that the difference between the output of the system filter and the output of the adaptive filter is minimized, and the adaptive filter obtained when the difference is minimized From the coefficient for calculating the coefficients of the error path filter (e.g., see Patent Document 2).

Further, the conventional direct spread spectrum digital filter includes a plurality of shift registers that sequentially delay the input baseband signal at a predetermined timing, and a plurality of weights that respectively weight the outputs of the baseband signal and the shift register. A coefficient unit and an adder for adding outputs from the coefficient units are provided. Each tap coefficient of the coefficient unit is a value determined by convolution of a spreading code for despreading and a coefficient for noise removal and multiplying this by a window function (for example, Patent Documents). 3).
JP-A-9-22293 JP 11-22293 A JP 2000-68894 A

  In conventional FIR filter design, transfer function rounding and window function adaptation are repeated, and an FIR filter corresponding to a data structure of a standard data type such as a 12-bit integer type or a floating-point float type is used. There are software and verification tools to design. However, in the design of a conventional FIR filter, there is no software or verification tool for designing an FIR filter corresponding to a unique user-defined data structure. For this reason, in order to design an FIR filter corresponding to a unique user-defined data structure, it is necessary to construct a new development environment and to ensure the reliability of the newly constructed development environment itself. There was a problem.

  Further, in designing a conventional FIR filter, the determination of the optimum value of the filter coefficient by quantization or sampling involves repeated work. In particular, the designer of the FIR filter needs to perform trial and error so as to be close to an ideal transfer function, and it is necessary to repeat rounding of the transfer function and adaptation of the window function, thereby increasing the number of iteration steps and determining the filter coefficient. Therefore, there is a problem that it takes a long time.

  Further, in the design of the conventional FIR filter, rounding of the transfer function and adaptation of the window function cause a rounding error and an error due to information loss or digit loss caused by the expansion of the window function. In other words, there is a problem that it is difficult to determine the coefficient of the transfer function considering the configuration of the FIR filter to the maximum.

  The conventional audio apparatus is already equipped with a digital filter (FIR type filter), and the coefficient (target transfer characteristic) of the filter means is determined based on the layout of the front speakers and observation points in the vehicle interior acoustic space. By determining, it is possible to easily perform correction outside the processing band and to listen to a desired excellent sound at the observation point. For this reason, the present invention is different from the present invention which aims to calculate the filter coefficient of the FIR filter in the design and development stage.

  Similarly, the conventional direct spread spectrum digital filter and the active noise control device are existing actual products, and the present invention aimed at calculating the filter coefficient of the FIR filter at the stage of design development is the invention. It is different.

  The present invention has been made to solve the above-described problem, and calculates filter coefficients of an FIR filter corresponding to a unique user-defined data structure without performing weighting by adaptation of a window function. An FIR filter design apparatus and an FIR filter design method capable of performing the above are provided.

  The FIR filter design apparatus according to the present invention includes a white noise signal generating means for generating a white noise signal that simulates the frequency characteristic of white noise, and a reference for converting the white noise signal into a predetermined frequency characteristic and outputting the target signal as a target signal. As a finite impulse response filter corresponding to a filter means and a user-defined data structure, the filter means has an FIR filter unit that functions so that the filter coefficient can be varied, and the white noise signal is used as a response signal via the FIR filter unit. Adaptive filter means for outputting; subtracting means for calculating an error signal by subtracting the response signal from the target signal; and convergence monitoring means for detecting convergence of the error signal based on a transition of the error signal. When the convergence monitoring means detects the convergence of the error signal, the filter of the variable FIR filter unit is changed. Number, or it outputs a transfer function of the FIR filter section for the filter coefficients and coefficient.

  The disclosed FIR filter design apparatus reduces the process of repeatedly rounding the transfer function and adapting the window function without performing weighting in the adaptation of the window function. There is an effect that the filter coefficient of the FIR filter corresponding to the data structure can be calculated.

(First embodiment of the present invention)
FIG. 1 is a block diagram showing a schematic configuration of the FIR filter design apparatus according to the first embodiment, FIG. 2 is an explanatory diagram for explaining an example of the internal configuration of the adaptive filter means shown in FIG. 1, and FIG. FIG. 3 is an explanatory diagram for explaining an example of a transfer function set in the reference filter means shown in FIG. 1, and FIG. 3B is a block diagram of the FIR filter section in the adaptive filter means shown in FIG.

  The white noise signal generation means 2 generates a white noise signal x (n) that simulates the frequency characteristics of white noise based on the activation signal from the input means 1 such as a mouse or a keyboard. A white noise signal x (n) is output to the means 4. White noise is a wave that oscillates up and down irregularly, and is subjected to Fourier transform to form a power spectrum, which has the same intensity at all frequencies.

  The reference filter means 3 converts the input white noise signal x (n) into a predetermined frequency characteristic targeted by an operator of the FIR filter design apparatus 10 (hereinafter referred to as a filter designer), It outputs to the subtraction means 5 mentioned later as target signal d (n).

  Note that the reference filter means 3 has a transfer characteristic for conversion to a predetermined frequency characteristic, and is not a unique user-defined data structure (a standard data type data structure), but a transfer with very little calculation error. A function (hereinafter referred to as an ideal model) is set in advance by the input means 1. That is, the reference filter means 3 converts the white noise signal x (n) into a target signal d (n) having a predetermined frequency characteristic by using this ideal model.

  For example, as shown in FIG. 3A, this ideal model is obtained from a white noise signal x (n) from a low frequency band (about 0.0167 × fs Hz to 0.2004 × fs Hz (fs = sampling frequency 8000 Hz). )) Components are taken out, and a transfer function for removing the other frequency band components is set. The reference filter means 3 may input a frequency characteristic diagram as shown in FIG. 3A as an image by the input means 1, but is constructed with a data structure of a standard data type, and the filter You may input the transfer function of the linear filter of the fixed magnification by which the coefficient is specified.

  The adaptive filter unit 4 includes an FIR filter unit 41 that functions as an FIR filter corresponding to a unique user-defined data structure, and the white noise signal x (n) is transmitted through the FIR filter unit 41 to a response signal y ( n). The adaptive filter means 4 combines the FIR filter unit 41 and the calculation unit 42 to have a function as an adaptive filter, and the filter of the FIR filter unit 41 so as to minimize an error signal e (n) described later. Update the coefficient.

  In the present embodiment, as a unique user-defined data structure in the FIR filter unit 41, an integer data structure with a bit width of 5 bits is taken as an example, and a storage unit 41a described later corresponds to 5 bits. It is illustrated as a storage unit.

  The reason why the data structure of the integer type with a bit width of 5 bits is as follows. As the data type, since the signal-to-noise ratio (S / N) allowed in the telecommunications industry is 30 dB (6.02 dB × 5 = 30 dB), this signal-to-noise ratio (S / N) can be realized. This is because an integer type is required. In addition, as a bit width, there is a customer who needs to realize a signal-to-noise ratio (S / N) of about 10 dB. This is because a bit width of 5 bits can sufficiently cope with this customer. For this reason, in the present embodiment, as a unique user-defined data structure in the FIR filter unit 41, an integer data structure having a bit width of 5 bits is taken as an example.

  As shown in FIG. 3B, the FIR filter unit 41 is composed of a plurality of taps, and a storage unit 41a that stores a filter coefficient that is an arbitrary constant for each tap, and is stored in the storage unit 41a. A multiplier 41b for multiplying the white noise signal x (n) by a constant by the filter coefficient and outputting the result.

  The FIR filter unit 41 includes a delay unit 41c that delays the white noise signal x (n) by one sampling period (one clock) between adjacent taps, and outputs an output signal from the multiplication unit 41b of each tap. An adder 41d that adds and outputs a response signal y (n) is provided.

  Although the delay unit 41c is illustrated as a delay unit corresponding to 5 bits, it may be a delay unit corresponding to a bit width of the data structure in the FIR filter unit 41 or more. However, since there are calculation errors due to digit loss and information loss, for example, when the data width of the data structure in the FIR filter unit 41 is set to 5 bits, the number of digits exceeding 5 bits is not used, and the data width There is no particular effect of making the value larger than 5 bits.

  The number of taps of the FIR filter unit 41 is determined by the filter designer based on the sampling frequency and the target frequency band. However, the accuracy as a filter can be increased as the number of taps is increased. On the other hand, as the number of taps increases, a burden is imposed on a central processing unit (CPU) (not shown), and it takes time to converge an error signal e (n) described later. For this reason, the number of taps of the FIR filter unit 41 is appropriately set by the filter designer according to the specifications of the FIR filter that is the design target. It is known from the Nyquist theorem that the sampling frequency fs is a frequency fs higher than 2f, where f is the maximum frequency component included in the original signal.

  As shown in FIG. 2, the arithmetic unit 42 multiplies the white noise signal x (n) and an error signal e (n) from the subtracting means 5 described later by multiplying one tap of the FIR filter unit 41 by the first. A multiplier 42a that calculates a signal; and a subtractor 42b that calculates a third signal by subtracting the second signal and the first signal that are output via the storage unit 41a that stores the filter coefficient. . Note that the white noise signal x (n) is input to the multiplying unit 42a corresponding to the second and subsequent taps by being delayed by a delay unit 41c disposed between adjacent taps.

  In the present embodiment, an example is shown in which an error signal e (n) from a subtracting unit 5 described later is input to each multiplier 42a of the adaptive filter unit 4 without changing the value. However, a multiplication unit having a constant of 0 or more and less than 1 may be newly provided on a branched line between the subtracting unit 5 and the convergence monitoring unit 6 described later, and before the first-stage tap. This multiplication unit has an effect of mitigating a sudden change in the error signal e (n) and promotes suppression that the adaptive filter does not diverge.

  The subtracting means 5 calculates the error signal e (n) by subtracting the response signal y (n) from the adaptive filter means 4 from the target signal d (n) from the reference filter means 3. This error signal e (n) is input to each multiplier 42a of the adaptive filter means 4 and the convergence monitoring means 6 described later.

  The convergence monitoring means 6 monitors the convergence of the error signal e (n) based on the transition of the input error signal e (n). When the convergence monitoring unit 6 detects the convergence of the error signal e (n), the convergence monitoring unit 6 extracts filter coefficients stored in the storage units 41 a of the FIR filter unit 41 of the adaptive filter unit 4. Further, the convergence monitoring unit 6 displays the extracted filter coefficients of the FIR filter unit 41 by the output unit 7 such as a printer or a display.

The convergence monitoring unit 6 does not display the extracted filter coefficient of the FIR filter unit 41 by the output unit 7 but outputs the transfer function of the FIR filter unit 41 using the extracted filter coefficient of the FIR filter unit 41 as a coefficient. You may display by the means 7. In this case, since the filter designer can obtain a polynomial whose coefficient in the transfer function of the FIR filter unit 41 is undetermined based on the number of taps set in the adaptive filter means 4 as an initial value, the convergence monitoring means 6 The transfer function of the FIR filter unit 41 can be acquired by applying the extracted filter coefficients of the FIR filter unit 41 to the uncertain coefficients of the polynomial.
Next, the processing operation by the FIR filter design apparatus 10 according to the present embodiment will be described.

First, the filter designer operates the input unit 1 to set an ideal model for the reference filter unit 3, specifies the number of taps of the FIR filter unit 41 for the adaptive filter unit 4, and sets the filter coefficient. An arbitrary value (for example, coefficient “0” is set in all storage units 41a) is set as an initial value. In particular, the ideal model corresponds to a standard data type data structure, and the FIR filter unit 41 is a transfer function corresponding to a unique user-defined data structure.
Then, the filter designer operates the input unit 1 to activate the white noise signal generation unit 2.

  The white noise signal generating means 2 generates a white noise signal x (n) that simulates the frequency characteristics of white noise based on the activation signal, and sends the white noise signal x (() to the reference filter means 3 and the adaptive filter means 4. n) is output.

The reference filter unit 3 converts the white noise signal x (n) input from the white noise signal generation unit 2 into a predetermined frequency characteristic, and outputs it to the subtraction unit 5 as a target signal d (n).
The adaptive filter means 4 performs signal processing on the white noise signal x (n) input from the white noise signal generating means 2 by an adaptive filter (FIR filter section 41, calculation section 42) as a response signal y (n). Output to the subtracting means 5.

  Then, the subtracting means 5 calculates an error signal e (n) that is a difference between the target signal d (n) from the reference filter means 3 and the response signal y (n) from the adaptive filter means 4, and the error signal e (N) is output to the convergence monitoring means 6 and the adaptive filter means 4.

  Further, the adaptive filter means 4 updates the adaptive filter so that the error calculation signal e (n) is minimized by a least-mean-square (hereinafter referred to as LMS) algorithm. By repeating this updating process until the filter coefficient of the FIR filter unit 41 becomes substantially constant, the FIR filter unit 41 has substantially the same transfer characteristic as that of the reference filter unit 3. Note that the algorithm for identifying the filter coefficient is not limited to the LMS method, and for example, a learning identification method may be used.

Further, the FIR filter unit 41 does not change the number of taps even when the filter coefficient is updated. For this reason, the FIR filter unit 41 whose filter coefficients are substantially constant has substantially the same transfer characteristics as the reference filter means 3, and has the same number of taps as a preset initial value.
The convergence monitoring means 6 stores the error signal e (n) from the subtracting means 5 and monitors the convergence of the error signal e (n) based on the transition of the error signal e (n).

  When the convergence monitoring unit 6 detects the convergence of the error signal e (n), the convergence monitoring unit 6 uses the filter coefficient stored in each storage unit 41a of the FIR filter unit 41 of the adaptive filter unit 4 or the filter coefficient as a coefficient. The transfer function of the FIR filter unit 41 is extracted from the FIR filter unit 41 of the adaptive filter unit 4 and output to the output unit 7.

  Then, the output means 7 displays the filter coefficient or transfer function data from the convergence monitoring means 6 so that the filter designer can select the optimum filter coefficient or FIR filter corresponding to the unique user-defined data structure. A transfer function can be obtained.

  As described above, in the FIR filter design apparatus 10 according to the present embodiment, the development environment construction and the actual function are integrated and evaluated by computer simulation, and has substantially the same transfer characteristics as the ideal model. The filter coefficient or transfer function of the FIR filter corresponding to the unique user-defined data structure can be calculated.

  In the present embodiment, the unique user-defined data structure in the FIR filter unit 41 is illustrated as an integer data structure with a bit width of 5 bits. Is not limited to 5 bits.

  In particular, in the case of an integer type having a bit width of less than 16 bits, if it is existing software, it is necessary to correct the data type itself, but if it is the FIR filter design device 10 according to the present embodiment, There is an effect that an integer type having a bit width of less than 16 bits can be handled without correcting the data type.

  Further, when error accuracy such as quantization error is not required from the customer, an excellent bit width such as 16 bits is not necessary, and the FIR filter unit 41 is required to increase the processing speed of the FIR filter design apparatus 10. The smaller the size (bit width) of the storage unit 41a, the better.

(Second embodiment of the present invention)
FIG. 4 is an explanatory diagram for explaining an example of a test environment using the FIR filter design apparatus according to the second embodiment. 4, the same reference numerals as those in FIGS. 1 to 3 denote the same or corresponding parts, and the description thereof is omitted.

  The FIR filter 141 is realized by implementing the FIR filter unit 41 in the first embodiment as hardware, and is applied to a predetermined product 100 (for example, car audio) and disposed in a sound environment (for example, in an automobile). It is what is done.

  In particular, the FIR filter 141 is preferably applied to a product 100 corresponding to a JTAG (Joint Test Action Group) used as a means for accessing a CPU or a field programmable gate array (FPGA) for the purpose of debugging embedded system software. .

  As shown in FIG. 4, the FIR filter 141 is composed of a plurality of taps, and for each tap, a memory 141a that stores a filter coefficient that is an arbitrary constant, and a white noise signal by a filter coefficient stored in the memory 141a. a multiplier 141b for multiplying x (n) by a constant. The memory 141a may be a non-volatile memory such as a flash memory that can change the filter coefficient flexibly.

  The FIR filter 141 includes a delay device 141c that delays the white noise signal x (n) by one sampling period (one clock) between adjacent taps, and adds the output signal from the multiplier 141b of each tap. And an adder 141d for outputting a response signal y (n).

  Note that the FIR filter design apparatus 10 according to the second embodiment differs from the first embodiment only in that the FIR filter unit 41 is realized by hardware as the FIR filter 141, and hardware (sound environment). Except for the operational effects of the combination of the software and the software (test environment 200), the same operational effects as the first embodiment are achieved.

  The FIR filter design apparatus 10 in the first embodiment is a computer simulator that performs a simulation experiment on a computer, and has an FIR filter characteristic that is optimal for the sound environment (local conditions) surrounding the product 100 to which the FIR filter 141 is applied. Filters cannot always be designed. For this reason, in the present embodiment, the FIR filter 141 separated from the calculation unit 42 of the adaptive filter unit 4 is incorporated in the product 100, and a more optimal filter coefficient or transfer function of the FIR filter 141 suitable for the sound environment is calculated. There is an effect of being able to.

  In particular, the FIR filter design apparatus 10 according to the present embodiment operates the adaptive filter means 4 (FIR filter 141, calculation unit 42) for the initial lot of the product 100, and optimizes the FIR filter 141 suitable for the sound environment. Calculate the correct filter coefficient. For a mass production lot, by mounting only the FIR filter 141 with the calculated filter coefficient fixed on the product 100, there is an effect that it is possible to reduce unnecessary circuits of the adaptive filter.

[Appendix] The following appendices are further disclosed with respect to the embodiment.
(Supplementary note 1) White noise signal generating means for generating a white noise signal simulating the frequency characteristic of white noise, reference filter means for converting the white noise signal into a predetermined frequency characteristic and outputting it as a target signal, and user-defined As a finite impulse response filter corresponding to a data structure of a type, an adaptive filter means having an FIR filter section that functions so that a filter coefficient can be varied, and outputting the white noise signal as a response signal via the FIR filter section; Subtracting means for calculating an error signal by subtracting the response signal from the target signal; and convergence monitoring means for detecting convergence of the error signal based on a transition of the error signal, the convergence monitoring means When the convergence of the error signal is detected, the variable filter coefficient of the FIR filter unit or the filter coefficient is related. An apparatus for designing a finite impulse response filter, which outputs a transfer function of the FIR filter unit as a number.

(Supplementary Note 2) The finite impulse response filter design apparatus, wherein the adaptive filter means includes an arithmetic unit that updates a filter coefficient of the FIR filter unit so as to minimize the error signal.

(Supplementary Note 3) A multiplication unit that calculates the first signal by multiplying the error signal and the white noise signal for each tap of the FIR filter unit, and a storage unit that stores the filter coefficient. And a subtracting unit that subtracts the first signal and the first signal to calculate a third signal, and the third signal is updated as the filter coefficient in the storage unit An apparatus for designing a finite impulse response filter.

(Supplementary Note 4) The finite impulse response filter design apparatus, wherein the user-defined data structure is an integer type having a bit width of less than 16 bits.

(Supplementary Note 5) Corresponding to a step of generating a white noise signal that simulates the frequency characteristics of white noise, a step of converting the white noise signal into a predetermined frequency characteristic and outputting it as a target signal, and a user-defined data structure Outputting the white noise signal as a response signal via an FIR filter unit functioning as a finite impulse response filter, calculating an error signal by subtracting the response signal from the target signal, and the error signal Updating the filter coefficient of the FIR filter unit to minimize the error signal, detecting the convergence of the error signal based on the transition of the error signal, and updating when detecting the convergence of the error signal The filter coefficient of the FIR filter unit, or the transmission of the FIR filter unit using the filter coefficient as a coefficient. Finite impulse response filter design method characterized by comprising the step of outputting function, the.

It is a block diagram which shows schematic structure of the FIR filter design apparatus which concerns on 1st Embodiment. It is explanatory drawing for demonstrating an example of the internal structure of the adaptive filter means shown in FIG. (A) is explanatory drawing for demonstrating an example of the transfer function set to the reference | standard filter means shown in FIG. 1, (b) is a block diagram of the FIR filter part in the adaptive filter means shown in FIG. It is explanatory drawing for demonstrating an example of the test environment using the FIR filter design apparatus which concerns on 2nd Embodiment.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Input means 2 White noise signal generation means 3 Reference filter means 4 Adaptive filter means 5 Subtraction means 6 Convergence monitoring means 7 Output means 10 FIR filter design apparatus 41 FIR filter part 41a Memory | storage part 41b Multiplication part 41c Delay part 41d Adder 42 Operation Part 42a Multiplying part 42b Subtracting part 100 Product 141 FIR filter 141a Memory 141b Multiplier 141c Delayer 141d Adder 200 Test environment

Claims (5)

White noise signal generating means for generating a white noise signal that simulates the frequency characteristics of white noise;
Reference filter means for converting the white noise signal into a predetermined frequency characteristic and outputting it as a target signal;
As a finite impulse response filter corresponding to a user-defined data structure, an adaptive filter having an FIR filter unit that functions to make a filter coefficient variable and outputting the white noise signal as a response signal via the FIR filter unit Means,
Subtracting means for calculating an error signal by subtracting the response signal from the target signal;
Convergence monitoring means for detecting the convergence of the error signal based on the transition of the error signal;
With
The convergence monitoring means outputs a variable filter coefficient of the FIR filter section when the convergence of the error signal is detected, or a transfer function of the FIR filter section using the filter coefficient as a coefficient. A finite impulse response filter design device. In the finite impulse response filter design apparatus according to claim 1,
The finite impulse response filter design apparatus, wherein the adaptive filter means includes an arithmetic unit that updates a filter coefficient of the FIR filter unit so as to minimize the error signal. In the finite impulse response filter design apparatus according to claim 2,
The arithmetic unit is output via a multiplication unit that calculates the first signal by multiplying the error signal and the white noise signal for one tap of the FIR filter unit, and a storage unit that stores the filter coefficient. A subtracting unit that subtracts the second signal and the first signal to calculate a third signal,
The apparatus for designing a finite impulse response filter, wherein the third signal is updated in the storage unit as the filter coefficient. In the finite impulse response filter design apparatus according to claim 1,
The apparatus for designing a finite impulse response filter, wherein the user-defined data structure is an integer type having a bit width of less than 16 bits. Generating a white noise signal that mimics the frequency characteristics of the white noise;
Converting the white noise signal into a predetermined frequency characteristic and outputting it as a target signal;
Outputting the white noise signal as a response signal via an FIR filter unit functioning as a finite impulse response filter corresponding to a user-defined data structure;
Subtracting the response signal from the target signal to calculate an error signal;
Updating a filter coefficient of the FIR filter unit to minimize the error signal;
Detecting a convergence of the error signal based on a transition of the error signal;
Outputting the updated filter coefficient of the FIR filter unit or the transfer function of the FIR filter unit using the filter coefficient as a coefficient when the convergence of the error signal is detected;
A method for designing a finite impulse response filter.

Images