WO2003047097A1

WO2003047097A1 - Digital filter designing method, designing apparatus, digital filter designing program, digital filter

Info

Publication number: WO2003047097A1
Application number: PCT/JP2002/011897
Authority: WO
Inventors: Yukio Koyanagi
Original assignee: Neuro Solution Corp.
Priority date: 2001-11-29
Filing date: 2002-11-14
Publication date: 2003-06-05
Also published as: TW200304075A; JP2003168958A

Abstract

It is possible to easily design an FIR filter having an arbitrary frequency characteristic only by inputting a waveform of a desired frequency characteristic as an image without having any special knowledge, i.e., by inputting a waveform of a desired frequency characteristic as a numerical value series and performing reverse FFT of this to obtain a filter coefficient group. Moreover, by performing a special rounding calculation to the numerical value series obtained by the reverse FFT, it is possible to simplify the filter coefficient value without lowering the filter characteristic accuracy and significantly reduce the number of times a multiplier as a filter constituting element is used. Furthermore, by performing window multiplication to the result of the reverse FFT, it is possible to increase the length of the numerical value series input firstly so as to minimize the frequency error and minimize the number of filter coefficients, thereby simplifying the configuration of the digital filter to be designed.

Description

Description Digital filter design method and design device, digital filter design program, digital filter technical field

The present invention relates to a digital filter design method and a digital filter design program, a digital filter design program, and a digital filter. In particular, the present invention includes a tapped delay line composed of a plurality of delay units, and multiplies the signal of each tap by several times. Later, it relates to the design method of the FIR filter that adds and outputs. Background art

In various electronic devices provided in various fields such as communication, measurement, voice / image signal processing, medical care, seismology, and the like, it is usual to perform some kind of digital signal processing internally. One of the most important basic operations in digital signal processing is filtering, which extracts only signals in the required frequency band from input signals containing various signals and noise. For this reason, electronic devices that perform digital signal processing often use digital filters.

As digital filters, IIR (Infinite Impulse Response) filters and FIR (Finite Impulse Response) filters are often used. Among these, the FIR filter has the following advantages. First, the circuit is always stable because the pole of the transfer function of the FIR filter is only at the origin of the z-plane. Second, perfectly accurate linear phase characteristics can be achieved.

When the filters are classified based on the arrangement of the passband and stopband, Filter, high-pass filter, band-pass filter, and band-stop filter. In the IIR filter, the fundamental is the low-pass filter, and other high-pass filters, band-pass filters, and band-reject filters perform frequency conversion and other processing from the one-pass filter. Is guided. Even at FIR filters, high-pass filters are derived from low-pass filters.

For example, when designing a high-pass filter, first design a basic low-pass filter and convert it to a frequency. Furthermore, a high-pass filter having a desired frequency characteristic is designed by repeating the design of the mouth-pass filter and the frequency conversion as needed. In the frequency conversion and processing here, convolution operation using a window function, Chebyshev approximation, or the like is performed based on the ratio between the sampling frequency and the cut-off frequency, and so on. The transfer function is obtained, and the process is further replaced with frequency components.

However, the conventional filter design method described above requires a high degree of expertise such as frequency conversion, and has a problem that the filter cannot be easily designed. In addition, although it is possible to design a typical filter such as a high-pass filter, a band-pass filter, or a band-stop filter, it is necessary to design a filter that has a complex analog-like waveform as its frequency characteristics. It was extremely difficult to do so. Also, frequency conversion using window function or Chebyshev approximation is very complicated. Therefore, if this is realized by software, the processing load becomes heavy, and if realized by hardware, the circuit scale becomes large.

The present invention has been made to solve such a problem, and it is an object of the present invention to be able to easily design an FIR digital filter having an arbitrary frequency characteristic. Another object of the present invention is to enable a simple design of an FIR digital filter capable of realizing a desired frequency characteristic with high accuracy on a small circuit scale. Disclosure of the invention

In order to solve the above problem, in the present invention, a numerical sequence or function representing a desired frequency characteristic is input, and the input numerical sequence or function is subjected to inverse Fourier transform, and a real number term of the result is extracted. The process of rearranging the first half and the second half of the numerical sequence consisting of the extracted real number terms, and multiplying the numerical sequence consisting of the above real number terms by 2 ⁿ times (n is a natural number) and rounding the decimal point After that, the result is reduced by 1 to 2 times, and the numerical sequence obtained as a result is determined as a filter coefficient group.

In another aspect of the present invention, a numerical sequence or function representing a desired frequency characteristic, which has a number of data points larger than the number of taps of the digital filter, is input, and the input numerical sequence or function is input. Inverse Fourier transform of the function to extract the real term of the result, rearrangement of the former half and the latter half of the numerical sequence consisting of the extracted real number term, and the numerical sequence consisting of the above real number term Is multiplied by a predetermined window function, and the numerical sequence obtained as a result is determined as a filter coefficient group. BRIEF DESCRIPTION OF THE FIGURES

FIG. 1 is a flowchart showing a processing procedure of a digital file design method according to the present embodiment.

FIG. 2 is a diagram showing an example of a desired frequency characteristic input in step S1 of FIG.

FIG. 3 is a diagram illustrating the relationship between the input data length and the maximum frequency error. FIG. 4 is a diagram for explaining the rearrangement process in step S3 of FIG.

FIG. 5 is a diagram showing the relationship between the number of taps limited by the width of the window function used in step S4 of FIG. 1 and the cutoff characteristic.

FIG. 6 is a diagram showing the function value of the Hanning window used in step S4 of FIG.

FIG. 7 is a diagram showing a filter coefficient group obtained by applying the filter design method of the present embodiment from the numerical sequence of the desired frequency characteristics shown in FIG. FIG. 8 is a diagram showing a frequency-gain characteristic (logarithmic scale) and a frequency-phase characteristic obtained by performing an FFT on the numerical sequence of the filter coefficient group shown in FIG. 7 obtained by the filter design method of the present embodiment. .

FIG. 9 is a diagram showing a frequency-gain characteristic (linear scale) obtained by performing FFT on the numerical sequence of the filter coefficient group shown in FIG. 7 obtained by the filter design method of the present embodiment.

FIG. 10 is a diagram showing the z-plane as a result of FFT of the numerical sequence of the filter coefficient group shown in FIG. 7 obtained by the filter design method of the present embodiment. FIG. 11 is a diagram illustrating a configuration example of a digital FIR filter configured using the filter coefficient group obtained by the filter design method of the present embodiment.

FIG. 12 is a diagram showing another example of the desired frequency characteristic input in step S1 of FIG.

FIG. 13 is a diagram showing a filter coefficient group obtained by applying the filter design method of the present embodiment from the numerical sequence of the desired frequency characteristics shown in FIG. 12. FIG. Frequency-gain characteristics obtained by FFT of the numerical sequence of the filter coefficient group shown in Fig. 13 obtained by the design method ( It is a figure which shows a logarithmic scale) and a frequency-phase characteristic.

FIG. 15 is a diagram showing a frequency-gain characteristic (linear scale) as a result of FFT of the numerical sequence of the filter coefficient group shown in FIG. 13 obtained by the filter design method of the present embodiment.

FIG. 16 is a diagram showing the z-plane of the result of FFT of the numerical sequence of the filter coefficient group shown in FIG. 13 obtained by the filter design method of the present embodiment. FIG. 17 shows step S in FIG. FIG. 6 is a diagram showing another example of a desired frequency characteristic input at 1.

FIG. 18 is a diagram illustrating a frequency characteristic as a result of FFT of a filter coefficient group obtained by applying the filter design method of the present embodiment from the numerical sequence of the desired frequency characteristic illustrated in FIG.

FIG. 19 is a diagram showing another example of the desired frequency characteristic input in step S1 of FIG.

FIG. 20 is a diagram illustrating a frequency characteristic as a result of FFT of a filter coefficient group obtained by applying the filter design method of the present embodiment from the numerical value sequence of the desired frequency characteristic illustrated in FIG. BEST MODE FOR CARRYING OUT THE INVENTION

Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a flowchart showing a processing procedure of a digital file design method according to the present embodiment. The digital filter designed here has a delay line with taps consisting of a plurality of delay units, and a type of FIR filter that multiplies the signal of each tap by a given filter coefficient group, then adds and outputs the result. It is.

The FIR filter has the impulse response represented by the finite time length as it is. It is the coefficient of Phil Yu. Therefore, designing an FIR filter means determining the filter coefficient group so that the desired frequency characteristics are obtained. Therefore, the flowchart of FIG. 1 shows a method of determining a filter coefficient group in such an FIR filter.

As shown in FIG. 1, first, a numerical sequence representing a waveform having a desired frequency characteristic is input (step S 1). At this time, it is preferable that the numerical sequence to be input has as many data numbers as possible. Originally, to construct an ideal filter, an infinite number of filter coefficients and an infinite number of filter taps were necessary. Therefore, in order to reduce the error with the desired frequency characteristic, it is preferable to increase the number of input data corresponding to the number of filter coefficients to such an extent that the frequency error falls within a required range. At least, enter a numerical sequence so that the number of filter coefficients is greater than the number of filter coefficients to be obtained (the number of taps of the digital filter).

In this data input, individual numerical values may be directly input, or a waveform of a desired frequency characteristic is drawn on a two-dimensional input coordinate for representing a frequency-gain characteristic, and the drawn waveform corresponds to the waveform. It may be replaced with a numeric string to be input. If the latter input method is used, the input can be performed overnight while confirming the desired frequency characteristics as an image, so that the input of the data representing the desired frequency characteristics can be easily performed intuitively. There are several means for realizing the above input method. For example, a two-dimensional plane representing frequency-gain characteristics is displayed on the display screen of the combination display, and a waveform of a desired frequency characteristic is drawn on the two-dimensional plane using a GUI (Graphical User Interface) or the like. Then, it can be converted into numerical data. Also, instead of the GUI on the computer screen, A pointing device such as an imager or a plotter may be used. The method described here is merely an example, and a numerical sequence may be input by other methods. Although the desired frequency characteristic is input as a numerical sequence here, it may be input as a function representing the waveform of the frequency characteristic.

Next, the input frequency characteristic is subjected to an inverse Fourier transform (inverse FFT) as a transfer function, and a real term of the result is extracted (step S2). As is well known, when Fourier transform (FFT) processing is performed on a certain numerical sequence, a waveform having a frequency-gain characteristic corresponding to the numerical sequence is obtained. Therefore, by inputting a numerical sequence or a function representing the waveform of the desired frequency-gain characteristic, inverse FFT it, and extracting its real term, the original necessary to realize the frequency-gain characteristic can be obtained. You get a sequence of numbers. This sequence of numerical values corresponds to the filter coefficient group to be obtained.

However, the numerical sequence itself obtained by inverse FFT is not arranged in an order that can be used as it is as a filter coefficient group. In other words, in any type of digital filter, the numerical sequence of filter coefficients has the largest median value, and the value gradually decreases as the distance from the center increases, repeating the amplitude. On the other hand, the numerical sequence obtained by inverse FFT has the smallest median value and the largest value at both ends. Therefore, by rearranging the first half and the second half such that the median of the numerical sequence obtained by the inverse FFT is located at both ends, the median becomes the maximum value and is symmetrical in front and rear (step S 3 ).

Although the numerical sequence obtained in this manner can be determined as a filter coefficient group as it is, in the present embodiment, a windowing operation is further performed (step S4). As described above, in the data input stage of step S1, the data is input to an extent that an error from a desired frequency characteristic falls within a required range. The number of force data is increased. This number of input data corresponds to the number of filter coefficients. Therefore, if the sequence of numerical values obtained by processing such as inverse FFT from this input data is used as it is as the filter coefficient group, the number of taps of the digital filter becomes very large, and the circuit scale becomes large. Therefore, the number of taps is reduced to the required number by performing windowing operation.

The window function used at this time includes various functions such as a rectangular window, a Hamming window, a Hanning window, and a heartlet window. Although any window function may be applied, it is particularly preferable to use a Hanning window. This is because the Hanning window is a function in which the values at both ends of the window are 0, and the values gradually decrease from the median toward both ends. For example, when a square window is used, the number of taps is forcibly cut off to a finite number, but this causes ringing (undulation phenomenon) in the filter characteristics. On the other hand, if the filter coefficient does not stop at a finite value but transitions smoothly to 0, the occurrence of ringing can be suppressed.

The numerical sequence obtained in this way can be used as it is as a filter coefficient group. However, the filter coefficient group obtained by inverse FFT and windowing has a very large number of digits below the decimal point, and is a complex and random set of values. Therefore, if this numerical sequence is used as it is as a filter coefficient group, the number of multipliers required for the digital filter becomes enormous, which is not practical.

Therefore, it is necessary to round the filter coefficients by truncating the decimal digits of the numerical sequence. However, with rounding by simple truncation, the resulting sequence of numbers is still a complex and random value, with only a reduced number of digits, and still requires many multipliers. In addition, with simple truncation, the accuracy of the obtained filter coefficient group is poor, and an error with the desired frequency The difference increases. Therefore, in the present embodiment, a rounding operation process as described below is performed (step S5). That is, the numerical sequence after windowing in step S4 is multiplied by 2 "(n is a natural number), rounded below the decimal point (converted to an integer), and the result is multiplied by 1Z2".

According to such a rounding operation, all filter coefficients have a value that is an integral multiple of 1 2 ". Therefore, the signal from each tap of the digital filter is an integral multiple of the signal. It is possible to configure a digital filter that multiplies the parts individually, adds all the multiplied outputs, and then multiplies them together by 12 ". In addition, the integer multiple can be represented by binary addition, such as 2 ¹ + 2 j + · · · (where i and j are arbitrary integers). As a result, the number of multipliers used in the entire digital filter can be greatly reduced, and the configuration can be simplified. In addition, since the numerical sequence obtained by the inverse FFT is rounded after being multiplied by 2 ", the rounding error can be reduced as compared to the case where the decimal portion of the numerical sequence is simply rounded. Thus, the filter coefficient group can be simplified without reducing the accuracy of the filter characteristics.

In the present embodiment, the numerical sequence obtained by such a rounding operation is finally determined as a filter coefficient group. It is to be noted that the processing in steps S3 to S5 described above does not necessarily have to be performed in this order, but it is sufficient if the rounding operation is performed at least after the windowing operation. For example, a windowing operation may be performed before sorting. In this case, multiply the Hanning window so that the coefficient value at both ends of the window is "1" and the coefficient value at the center of the window is "0". By performing the windowing operation at an early stage in a series of procedures in this manner, the number of data used for subsequent operations can be reduced, and the processing load on the operation can be reduced.

Hereinafter, the procedure of the filter design method according to the present embodiment described above will be described in detail along with specific examples. As shown in FIG. 2, in step S 1 Draws the frequency-gain characteristics of the filter standardized by "1" and converts it into numerical data. The input data is symmetrical about the center of the sampling frequency. At this time, the input data length (the length of the graph, that is, the number of numerical sequences) m is a value that falls within the required range of the frequency error, and is 2 to simplify the inverse FFT processing in step S2. ^k .

For example, when designing a FIR filter for a speech signal with a sampling frequency of 44.1 K I-Iz, the relationship between the input data length m and the maximum frequency error is as shown in FIG. The maximum frequency error referred to here corresponds to the frequency of one graduation of the daraf, and is obtained by the calculation of 44. l KHz z Zm. In the case of voice processing, if it is about 10 Hz, it will be within the permissible error range, so use 406 as the input data length m.

In the example of the graph shown in Figure 2, the sampling frequency is 44.1 KHz, the input data length m is 496, the cutoff frequency is 8 KHz, and the gain drops at the cutoff frequency. It shows frequency characteristics equivalent to a low-pass filter with an amount of —60 dB. In this case, the horizontal axis of the graph is equally divided into 496 scales (clocks). Assuming that the number of clocks is CK, the frequency f at the number of clocks CK is

f = C K X (44.1 / 4096) (KHz)

Becomes Therefore, the number of clocks C K1 corresponding to 8 KHz is

C K l = f X (4096 / 44.1 K) = 7 43.04

Becomes

In step S2, an inverse FFT process is performed using the input low-pass filter frequency characteristics as a transfer function as shown in FIG. 2, and the resulting real term is extracted. In the next step S3, in order to convert the numerical sequence obtained by the inverse FFT into an order that can be used as a filter coefficient group, as shown in FIG. Thus, the numerical sequence is divided into the first half and the second half, and they are sorted. In other words, the value of the 0th clock is replaced by the value of the 2048th clock (hereinafter, referred to as 0 → 2048), 1 → 2049, 2 → 2505, , 20447 → 495, 048 → 0, 209 → 1 · · · 495 → 20047.

Further, in step S4, a windowing operation is performed to reduce the number of taps. As described above, window functions include a square window, a Hamming window, a Hanning window, and a heartlet window. Here, a Hanning window whose both ends converge smoothly to 0 is used. Here, Fig. 5 shows the relationship between the number of taps limited by the width of the window function and the cutoff characteristics. As can be seen, the slope of the characteristic at the Katoff frequency becomes steeper as the number of sunsets increases.

Here, an example is shown in which the width of the window function is set so that the number of taps of the digital filter is 127. FIG. 6 is a diagram showing the function value of the Hanning window in this case. The central part of the numerical sequence (4096 data sequences) obtained by rearranging is multiplied by the changing window (127 data sequences) shown in Fig. 6. At this time, all coefficients outside the range of the Hanning window are calculated as 0. Then, in the last step S5, the numerical sequence after the windowing operation is multiplied by 2 ", the decimal part is rounded, and the result is multiplied by 12" (for example, 2 ¹ = 2 48).

Figure 7 shows the filter coefficient group (127 filter coefficients) obtained by the above calculation. FIG. 8 is a diagram showing frequency-gain characteristics and frequency-phase characteristics obtained by FFT of a numerical sequence of the filter coefficient group obtained as shown in FIG. 7 above. It is shown on a scale. FIG. 9 is a diagram showing gain on a linear scale for the same frequency-gain characteristic, and FIG. 10 is a z-plane view. These Figures 8 to 1 As can be seen from 0, the filter coefficient group obtained by the filter design method of the present embodiment realizes a low-pass filter characteristic with a cut-off frequency of 8 KHz almost accurately. Moreover, the amount of deduction at the cutoff frequency is more than 40 dB, and the phase characteristics are linear and stable. FIG. 11 is a diagram illustrating a configuration example of a low-pass filter configured using a filter coefficient group obtained by the filter design method according to the present embodiment. In this filter, connected in cascade 1 2 seven D-type flip-flop 1 - sequentially delaying by one clock CK input signals by ~ 1 _27. Then, against the signal extracted from the output tap of the D-type flip-flop 1 to 1 _27, the integer value of the filter coefficients 2 0 4 8 times result 1 2 seven coefficient unit 2 _, ~ 2 _ _{1 27} , And all the multiplied results are added by 127 adders 3- _| to 3 _ _{| 27} and output.

Then, in the multiplier 4 provided at the output stage of the adder 3_ _{| 27 at} the final stage, the added output is multiplied by 1/2048 to return the amplitude to the original value, and the result is supplied to the D-type flip-flop 5. After holding once, output. Here, 127 coefficient units and 127 adders are provided, respectively, but it is possible to omit the coefficient unit and the adder in the portion where the filter coefficient value is 0. Therefore, in practice, it is possible to configure a digital filter with fewer multipliers and adders than in Fig. 11. As described above, in the present embodiment, a special rounding operation is performed when obtaining a filter coefficient, so that the configuration of a digital filter to be designed can be simplified.

In the above, an example of designing a low-pass filter has been described. However, other digital filters can be similarly designed. For example, an example in which a band pass filter is set will be described below. Here, it is assumed that a numerical sequence of frequency characteristics as shown in FIG. 12 is input as a desired band-pass filter frequency characteristic. The desired frequency characteristic shown in Fig. 12 is In this case, only signals in the frequency band of 5 to 8 KHz are passed. In this case, the sampling frequency is assumed to be 44.1 入力 and the input data length is assumed to be 410.

For the input data shown in Fig. 12, the inverse FFT → rearrangement → windowing operation (the window is a Hanning window and the width is 1 2 7) in the same way as the low-pass filter described above. A filter coefficient group as shown in Fig. 13 is obtained.

FIG. 14 is a diagram showing frequency-gain characteristics and frequency-phase characteristics as a result of FFT of the numerical sequence of the filter coefficient group obtained as shown in FIG. Are shown on a logarithmic scale. FIG. 15 is a diagram showing gain on a linear scale for the same frequency-gain characteristics, and FIG. 16 is a z-plane view. As can be seen from FIGS. 14 to 16, the filter coefficient group obtained by the filter design method of the present embodiment has a bandpass filter characteristic in a pass frequency band of 5 to 8 kHz. Almost exactly. Moreover, the amount of deduction at the cut-off frequency is 40 dB or more, and the phase characteristics are linear and stable.

FIG. 17 is a diagram showing data input as desired frequency characteristics of a low-pass filter for sound quality adjustment used for a hearing aid, various acoustic devices, and the like. The frequency characteristics of this sound quality adjusting mouth-and-pass filter are continuously changed in an analog manner. Similarly, for the input data shown in Fig. 17, inverse FFT → rearrangement → windowing operation-rounding operation is performed, and the FIR coefficient group obtained by this is FFT. Such a frequency characteristic is obtained. As can be seen from the above, the filter coefficient group obtained by the filter design method of the present embodiment realizes almost exactly the desired frequency characteristic of the sound quality adjustment port-pass filter. In addition, although not shown in the figure, the phase characteristics are linear and stable. FIG. 19 is a diagram illustrating data input as desired frequency characteristics of a high-pass filter for sound quality adjustment used for a hearing aid, various acoustic devices, and the like. This high-pass filter for sound quality adjustment also has a continuously changing analog frequency characteristic. Similarly, for the input data shown in Fig. 19, inverse FFT—rearrangement—windowing operation—rounding operation is performed, and the FIR coefficient group obtained by this is subjected to FFT. A frequency characteristic like 0 is obtained. As can be seen from the above, the filter coefficient group obtained by the filter design method of the present embodiment realizes the desired frequency characteristic of the high-pass filter for sound quality adjustment almost exactly. In addition, although not shown in the figure, the phase characteristics are linear and stable.

An apparatus for realizing the digital filter design method according to the present embodiment described above can be realized by any of a hardware configuration, a DSP, and software. For example, when realized by software, the file design apparatus of the present embodiment is actually configured by a computer CPU or MPU, RAM, ROM, or the like, and stored in RAM, ROM, a hard disk, or the like. It can be realized by running the programmed program.

Therefore, the present invention can be realized by recording a program that causes a computer to perform the functions of the above-described embodiment on a recording medium such as a CD-R〇M and reading the program into the computer. As a recording medium for recording the above programs, in addition to a CD-ROM, a flexible disk, a hard disk, a magnetic tape, an optical disk, a magneto-optical disk, a DVD, a nonvolatile memory card, and the like. Can be used. In addition, the above program can be realized by downloading the program over a network such as the Internet.

In addition, the computer executes the supplied program to increase In addition to realizing the functions of the above-described embodiment, the above-described embodiment may be implemented in cooperation with an OS (operating system) or another application software in which the program is running on the console. When the functions of the above-described embodiment are realized, or all or a part of the processing of the supplied program is performed by the function expansion board or the function expansion unit of the computer, the functions of the above-described embodiment are realized. Such a program is also included in the embodiment of the present invention.

As described in detail above, in the present embodiment, a numerical sequence representing a waveform of a desired frequency characteristic is input as an image, and a filter coefficient group is obtained by performing an inverse Fourier transform on the image. Therefore, the coefficients of the FIR digital filter that achieve the desired frequency characteristics can be easily determined without any special mathematical knowledge or electrical engineering knowledge. Of particular note are not only single-pass filters, but also eight-pass filters, non-pass filters, band emiline filters, com filters, and analog filters with arbitrary waveforms. It can be easily designed using the same method.

Also, in the present embodiment, a special rounding operation is performed on the numerical sequence obtained by the inverse Fourier transform, thereby simplifying the filter coefficient group without reducing the filtering accuracy. This can greatly reduce the number of multipliers (dividers) used as filter components. Further, in the present embodiment, the result of the inverse Fourier transform is multiplied by a window function of a required length, so that the input data length is increased to suppress the frequency error to a small value. In addition, the number of filter coefficients (the number of taps of a digital filter) can be reduced. As a result, the configuration of the digital filter to be designed can be simplified, and the desired frequency characteristic can be realized with high accuracy. Note that each of the above embodiments is embodied in practicing the present invention. This is merely an example, and should not be construed as limiting the technical scope of the present invention. That is, the present invention can be embodied in various forms without departing from the spirit or main features thereof.

As described above, according to the present invention, a waveform of a desired frequency characteristic is input as a numerical sequence or a function, and a filter coefficient group is obtained by performing an inverse Fourier transform on the waveform. Even without expert knowledge, FIR with any frequency characteristics, such as mouth-to-mouth, noise-free, evening, and emi-line-short Digital filters can be easily designed.

Further, according to the present invention, a special rounding operation is performed on the numerical sequence obtained by the inverse Fourier transform, so that the filter coefficient to be obtained can be obtained without deteriorating the accuracy of the filter characteristic. The group can be simplified and the number of filter component multipliers used can be significantly reduced. This makes it possible to easily design a FIR digital filter capable of realizing a desired frequency characteristic with high accuracy on a small circuit scale.

Further, according to the present invention, since the windowing operation is performed on the result of the inverse Fourier transform, the first input numerical value sequence is lengthened to suppress the frequency error small, and at the same time, the filter coefficient The number of digital filters (the number of taps in the digital filter) can be reduced, and the configuration of the digital filter to be designed can be simplified. This makes it possible to easily design a FIR digital filter capable of realizing desired frequency characteristics with high accuracy on a small circuit scale. Industrial applicability

The present invention makes it possible to simplify FIR digital filters with arbitrary frequency characteristics. It is useful for making design possible. Further, the present invention is useful for easily designing an FIR digital filter capable of realizing desired frequency characteristics with high accuracy on a small circuit scale.

Claims

The scope of the claims

1. A method of designing a digital filter that multiplies a signal of each tap in a delay line with taps composed of a plurality of delay units by a given filter coefficient group, and adds and outputs the result.

A numerical sequence or function representing a desired frequency characteristic is input, and the input numerical sequence or function is subjected to inverse Fourier transform to extract a real number term as a result. The process of rearranging the first half and the second half, the process of multiplying the numerical sequence composed of the real terms by 2 ⁿ (n is a natural number), rounding the decimal part, and then multiplying the result by 1 2 "are performed. A digital filter design method, wherein the numerical sequence obtained by the above is determined as the filter coefficient group.

2. A method of designing a digital filter that multiplies a signal of each tap in a tapped delay line composed of a plurality of delay units by a given filter coefficient group, and adds and outputs the result.

A numerical sequence or function representing a desired frequency characteristic and having a number of data points or more than the number of taps of the digital filter is input, and the input numerical sequence or function is subjected to inverse Fourier transform. Extracting the real number term of the result, rearranging the former half and the latter half of the numerical sequence composed of the extracted real number term, and applying a predetermined window function to the numerical sequence composed of the real number term. A digital filter design method, characterized in that a numerical sequence obtained by the above is determined as the above-described filter coefficient group.

3. The numerical sequence consisting of real terms of the result of the inverse Fourier transform before or after sorting, or the numerical sequence after the above window function is multiplied by 2 "times (n Is a natural number). 3. The method for designing a digital filter according to claim 2, wherein a process of multiplying the result by 1/2 "is performed.

4. A digital filter device which multiplies a signal of each tap in a tapped delay line composed of a plurality of delay units by a given filter coefficient group, and adds and outputs the multiplied signal.

An input means for inputting a numerical sequence or a function representing a waveform having a desired frequency characteristic;

An inverse Fourier transform unit for performing an inverse Fourier transform on the numerical sequence or the function input by the input unit, and extracting a real term of the result;

Reordering means for rearranging the first half and the second half of the numerical sequence obtained by the inverse Fourier transform;

The process of multiplying the number sequence of the real term before or after sorting by the above sorting means by 2 ⁿ (where n is a natural number), rounding the decimal point, and multiplying the result by 1/2 " And a rounding means for performing

A digital filter design apparatus, wherein a numerical sequence that is rearranged by the rearranging unit and rounded by the rounding unit is determined as the filter coefficient group.

5. A digital filter design apparatus that multiplies a signal of each tap in a tapped delay line composed of a plurality of delay units by a given filter coefficient group, and then adds and outputs the multiplied signal.

Input means for inputting a numerical sequence or a function representing a waveform of a desired frequency characteristic, the numerical sequence or the function having more data points than the number of taps of the digital filter;

The first half and the second half of the numerical sequence obtained by the inverse Fourier transform Sorting means for changing the order;

Window processing means for applying a predetermined window function to the numerical sequence before or after the sorting by the sorting means,

A digital filter design device, wherein the numerical sequence that is rearranged by the rearranging unit and windowed by the window processing unit is determined as the filter coefficient group.

6. The numerical sequence before or after sorting by the sorting means, or the numeric sequence after windowing by the window processing means is 2. 6. The digital filter designing apparatus according to claim 5, further comprising a rounding means for performing a process of multiplying by 2 (n is a natural number), rounding the decimal portion, and multiplying the result by 1 Z 2 ⁿ .

7. The input means is means for drawing a waveform of the desired frequency characteristic on two-dimensional input coordinates for representing frequency-gain characteristics, and for inputting the drawn waveform as the numerical sequence or function. 5. The design device for a digital filter according to claim 4, comprising:

8. The input means includes means for drawing a waveform of the desired frequency characteristic on two-dimensional input coordinates for representing frequency-gain characteristics, and inputting the drawn waveform as the numerical sequence or function. 6. The design of the digital filter according to claim 5, characterized in that

9. A digital filter design program for causing a computer to function as each means described in claim 4.

10. A digital filter design program for causing a computer to function as each means described in claim 5.

1 1. Digital designed using the design method described in claim 1. filter.

1 2. A digital filter designed using the design method described in claim 2.

13. A digital filter designed using the design apparatus according to claim 4.

14 4. A digital filer designed using the design apparatus according to claim 5.