JP2002271886A - Equalizer and its filter coefficient decision method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To generate an equalizing characteristic data stream at an equal frequency interval with smoothness equivalent to that of spline interpolation in spite of simple processing. SOLUTION: A sample-and-hold section 61 applies N-point sampling and holding to each of frequency characteristic data at a set logarithmic scale, a data output section 63 regards the frequency characteristic data stream with the logarithmic scale subjected to sample-and-holding as time series data and provides an output of the data, a digital low-pass filter applies filtering to the time series data to smooth the equalizing characteristic, a frequency characteristic data generating section 65 extracts data from the frequency characteristic data stream with the logarithmic scale outputted from the low-pass filter at an equal frequency interval, and an inverse FFT section 66 applies inverse FFT processing to the frequency characteristic data at the equal frequency interval to decide a filter coefficient of an FIR filter being a component of equalizing device 52.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an equalizer comprising a digital filter and a method for determining the filter coefficient, and more particularly to an equalizer for determining and setting a filter coefficient from frequency characteristic data discretely set on a logarithmic scale. And a filter coefficient determination method.

[0002]

2. Description of the Related Art An equalizer in an audio device controls a frequency characteristic of an audio signal so as to output a sound having a desired sound quality by a user. FIG. 7 is a configuration diagram of an audio device provided with an equalizer, 1 is an audio source such as a CD player, a mini-disc player, and an AM / FM tuner, 2 is an equalizer that corrects the frequency characteristics of an audio signal output from the audio source, 3
Is an amplifier, 4 is a speaker, 5 is an equalizing characteristic setting unit that sets equalizing characteristics, 6 is a spectrum analyzer that calculates and outputs the spectrum of each frequency band of the audio signal output from the equalizer, 7 is an operation unit,
Reference numeral 8 denotes a display unit for displaying a reception band / reception frequency display, a spectrum display, and other displays, and 9 denotes an audio control unit for controlling the entire audio apparatus.

[0003] The equalizer 2 corrects the frequency characteristics of the audio signal based on the equalizing characteristics set by the equalizing characteristics setting section 5 and outputs the corrected audio signal. The equalizing characteristic is calculated every octave or 1/3 on the frequency axis.
It is a gain characteristic that is set discretely for each octave,
For example, 125Hz, 250Hz, 500Hz, 1kHz, 2kHz, 4kHz, 8kHz,
Set by the gain at 16 kHz. That is, the equalizing characteristics are set on a logarithmic scale.

Recently, an equalizer 2 has been used for digitizing.
Filter, for example, an FIR filter.
Have been done. FIG. 8 shows the structure of an n-tap FIR filter.
The diagram is DL₁~ DLn-1 is one sampler of input signal
Delay part, MP that delays ₀~ MPn-1 is output of each delay section
And filter coefficient C₀Multiplying circuit for multiplying .about.Cn-1 by ADD
Is an adder for adding and outputting the outputs of the respective multiplier circuits. Person in charge
Number C₀Input audio by appropriately setting ~ Cn-1
Frequency correction can be applied to the signal. Such FIR files
The filter configuration equalizer has a set logarithmic scale.
It is necessary to determine the filter coefficient from the equalizing characteristics
is there. In general, filter coefficients are discrete
Inverse FFT processing (inversefa
St Fourier transform). But before
As described, the equalizing characteristics are set on a linear scale
Instead, they are set discretely on a logarithmic scale. That's it
To calculate the filter coefficients,
Linearize the given discrete equalizing characteristic data
Discrete equalizing characteristics at equal frequency intervals on the scale
Need to convert to data.

Heretofore, discrete equalizing characteristic data at equal frequency intervals on a linear scale has been obtained by interpolation from discrete equalizing characteristic data given on a logarithmic scale, and spline interpolation is employed as an interpolation method. . FIG. 9 is an explanatory diagram of a conventional filter coefficient determination process using an interpolation method, and the same parts as those in FIG. 7 are denoted by the same reference numerals. The equalizing characteristic setting unit 5 sets the equalizing characteristic on a logarithmic scale. For example, 125Hz, 250Hz, 500Hz, 1kHz, 2kHz, 4kHz, 8kHz, 16kHz
The logarithmic scale equalizing characteristic is set by setting the gain (dB) at.

A spline interpolation section 9a of the audio control section 9 performs spline interpolation between logarithmic scale discrete equalizing characteristic data to generate a linear scale equalizing characteristic data sequence shown in FIG. However, the spline interpolation unit 9a performs spline interpolation with the gain of the frequency of 0 Hz and the audible limit frequency of 22.05 kHz set to 0 (dB). Also, the gains indicated by * in the figure are frequency
125Hz, 250Hz, 500Hz, 1kHz, 2kHz, 4kHz, 8kHz, 16kHz
It is assumed that is set in.

Next, the spline interpolation section 9a extracts n / 2 (eg, n / 2 = 128) data at predetermined frequency intervals, for example, 250 Hz intervals, from the equalizing characteristic data sequence of the linear scale. Then, the extracted equal frequency intervals
The 128 data and the 128 data obtained by folding the 128 data to the high frequency band side are combined to generate a total of n (= 256) data at equal frequency intervals (see FIG. 11). , In parallel to the inverse FFT operation unit 9b. The reason for combining by folding back to the high frequency band side is as follows. That is, when n discrete digital signals of the sampling time Δt are subjected to the FFT processing, the amplitude spectrum is turned around the frequency n · Δf / 2 (however, Δf = 1 / n · Δt) (symmetry of spectrum). . Therefore, in order to obtain a discrete digital signal by performing an inverse FFT operation on the amplitude spectrum, it is necessary to fold the amplitude spectrum at the frequency n · Δf / 2.

The inverse FFT operation section 9b performs an inverse FFT process on n (= 256) data at equal frequency intervals, and outputs data at n predetermined sampling time intervals. The n (= 256) pieces of data output from the inverse FFT operation unit become the coefficients c _{0 to} c _{n−1 of the n} taps of the FIR filter (FIG. 8) constituting the equalizer 2. FIG. 12 shows the spline interpolation result in FIG.
In the case of 0, 128 frequency data is extracted at 250 Hz intervals and turned back to the high frequency band side.
Coefficients c _{0 to} c _n−1 (n = 25) of taps of the FIR filter obtained by inputting six data to the inverse FFT operation unit 9b
6) is an example.

[0009]

According to the above-mentioned spline interpolation method, while it is possible to generate an equalizing characteristic data sequence at equal frequency intervals on a linear scale smoothly, it is necessary to solve a three-dimensional simultaneous equation, which requires a matrix operation. The algorithm becomes complicated, and the amount of calculation increases. For this reason, there is a problem that it takes time until a required filter coefficient is obtained, and a high-speed arithmetic device is required, which makes the device expensive. From the above, it is an object of the present invention to enable a logarithmic scale data string to be converted to a linear scale equally spaced data string by simple processing. Another object of the present invention is to be able to convert a logarithmic scale equalizing characteristic data sequence into a linear scale data sequence at equal frequency intervals with the same level of smoothness as spline interpolation and with good precision. Another object of the present invention is to enable a filter coefficient of an FIR filter constituting an equalizer to be determined by a simple process from a logarithmic scale equalizing characteristic data sequence.

[0010]

The first aspect of the present invention is a data scale conversion method for converting a data sequence given on a logarithmic scale into a data sequence at an equal interval on a linear scale.
(1) Each point of the logarithmic scale is sampled and held at N points. (2) The sampled and held logarithmic scale data string is regarded as time-series data and input to a digital low-pass filter. From the logarithmic scale data string output from the low-pass filter, data at equally spaced linear scales are extracted, thereby generating data strings at equally spaced linear scales. According to this data scale conversion method, it is possible to convert a logarithmic scale data string into a linear scale equally spaced data string with the same level of smoothness as spline interpolation by simple processing.

A second aspect of the present invention is a method for determining a filter coefficient of a digital filter constituting an equalizer, wherein (1) frequency characteristic data is set discretely on a logarithmic scale, and (2)
Each frequency characteristic data on the logarithmic scale is sampled and held at N points, and (3) the sampled and held logarithmic scale frequency characteristic data string is regarded as time-series data and input to a digital low-pass filter. ) Data is extracted at equal frequency intervals from a logarithmic scale frequency characteristic data string output from the low-pass filter to generate a linear scale frequency characteristic data string, and (5) an inverse FFT is applied to the linear scale frequency characteristic data. (6) The reverse F
A filter coefficient of a digital filter constituting the equalizer is determined based on a result of the FT processing. According to this filter coefficient determination method, a logarithmic scale equalizing characteristic data string can be converted into a linear scale equal frequency characteristic equalizing characteristic data string by a simple process, and further, the audio signal can be converted according to the set equalizing characteristic. Can be determined.

In this case, the first and second frequency characteristic data of 0 (dB) are respectively adjacent to the lowest and highest bands where the frequency data of the lowest and highest frequencies on the logarithmic scale are valid.
A second band is provided, and a third band having the same frequency characteristic data as the lowest band is provided adjacent to the first band; and the highest band is provided adjacent to the second band. A fourth band having the same frequency characteristic data is provided, and in each of the first to fourth bands, each frequency characteristic data is sampled and held at N points, and the sample and hold data is added before and after the time series data. To the low-pass filter. By doing so, it is possible to convert the logarithmic scale equalizing characteristic data sequence into linear scale equal frequency characteristic equalizing characteristic data sequences with the same smoothness and accuracy as the spline interpolation,
It is possible to determine a filter coefficient capable of accurately correcting the audio signal according to the set equalizing characteristics.

A third aspect of the present invention is an equalizer composed of a digital filter, (1) a characteristic data setting unit for discretely setting frequency characteristic data on a logarithmic scale, and (2) each frequency characteristic on the logarithmic scale. A sample-and-hold unit for sample-holding data at N points; (3) means for outputting the sampled and held logarithmic scale frequency characteristic data sequence as time-series data; and (4) a digital low-level signal to which the time-series data is input. A band-pass filter, (5) a linear-scale frequency characteristic data generator for extracting data at equal frequency intervals from a logarithmic-scale frequency characteristic data string output from the low-pass filter to generate a linear-scale frequency characteristic data string (6) Inverse FFT processing is performed on the frequency characteristic data of the linear scale to obtain a filter of a digital filter constituting an equalizer. Inverse FFT to output the data coefficient
A processing unit. According to this equalizer, a logarithmic scale equalizing characteristic data string can be converted into a linear scale equal frequency characteristic equalizing characteristic data string by a simple process, and the audio signal can be corrected according to the set equalizing characteristic. Possible filter coefficients can be determined.

The equalizer includes (7) first and second frequency characteristic data of 0 (dB) adjacent to the lowest band and the highest band in which the frequency data of the lowest frequency and the highest frequency of the logarithmic scale are respectively effective. A third band adjacent to the first band and having the same frequency characteristic data as the lowest band; and a third band adjacent to the second band and having the same frequency characteristic data as the highest band. And a means for sampling and holding each point of the frequency characteristic data at N points in the first to fourth bands is added. The output means converts the sample and hold data of the first to fourth bands. The time-series data is added before and after the time-series data and input to the low-pass filter. In this way, the logarithmic scale equalizing characteristic data string can be converted into linear scale equal frequency characteristic equalizing characteristic data strings with the same smoothness and accuracy as spline interpolation, and audio can be converted according to the set equalizing characteristics. The signal can be corrected accurately.

[0015]

DESCRIPTION OF THE PREFERRED EMBODIMENTS (a) Filter Coefficient Determination Apparatus FIG. 1 is a block diagram of a filter coefficient determination apparatus according to the present invention.
Reference numeral 51 denotes a frequency characteristic setting unit; 52, an equalizer having an FIR configuration; and 53, a filter coefficient determination device. The filter coefficient determination device 53 is provided, for example, in an audio controller (not shown) that configures the audio system. The frequency characteristic setting unit 51 sets an equalizing characteristic on a logarithmic scale. For example, 125Hz, 250Hz, 500Hz, 1kHz, 2
The logarithmic scale equalizing characteristics are set by setting the gain (dB) at kHZ, 4 kHz, 8 kHz, and 16 kHz.

The N-point sample-and-hold unit 61 of the filter coefficient determination unit 53 operates at each frequency (125 Hz, 250 Hz, 500 Hz, 1
The frequency characteristic data set to kHz, 2 kHz, 4 kHz, 8 kHz, and 16 kHz are sampled and held by the required number of interpolation data points (N) as shown in FIG. That is, between the frequency fbi of the logarithmic scale and the next frequency fbi + 1, N gains (dB) set to the frequency fbi are inserted and held at equal intervals. N is, for example, 2 ⁷ (= 64). or,
As shown in FIGS. 2B and 2C, the N-point sample-and-hold unit 61 has a minimum band Bmin (125 to 250 Hz) in which a gain set to a minimum frequency of 125 Hz is valid and a gain set to a maximum frequency of 16 kHz. The first band B1 (0-125 Hz) and the second band B2, which are respectively adjacent to the highest band Bmax (16-32 kHz)
(32 to 64 kHz), N frequency characteristic data having a gain of 0 (dB) are inserted and held.

The symmetric characteristic adding section 62 includes a first band B1 (0 to
A third band B3 (-125 Hz to 0 Hz) adjacent to (125 Hz) is set, and the same N frequency characteristic data (gain) as the lowest band (125 to 250 Hz) are copied to the third band. Similarly, the symmetric characteristic adding unit 62 sets a fourth band B4 (64 to 128 kHz) adjacent to the second band B2 (32 to 64 kHz), and sets the fourth band to the same as the highest band (16 to 32 kHz). The N frequency characteristic data (gain) are copied. N point sample hold unit 61
And the first to fourth bands B1
The reason for adding N to B4 and inserting N data in each band will be described later. The data output unit 63 arranges N frequency data (gains) in each band from the lower band, and sequentially outputs the data as time-series data to a next-stage low-pass digital filter (FIR filter) 64 at predetermined time intervals.

The digital filter 64 performs a low-pass filtering process on the input time-series data and outputs it. That is, the digital filter 64 has a cutoff frequency of about 2π / 2N (rad) and a number of taps of (N + 1).
The filter coefficient is set, a low-pass filtering process is performed on an input data sequence, and a processing result is output. The filter coefficients can be calculated and set in advance using general-purpose signal processing simulation software. By the low-pass filtering process, the sampled and held frequency characteristic data is smoothly interpolated. FIG. 3 shows an equalizing curve when a logarithmic scale frequency characteristic data string output from the digital filter 64 is mapped and connected on a gain-frequency axis of a linear scale. FIG. 3 shows a case where the same equalizing characteristics as in the conventional example are set on a logarithmic scale for comparison. By comparing the equalizing characteristics shown in FIGS. 3 and 10, it is possible to obtain equalizing characteristics on a linear scale equivalent to that of the conventional spline interpolation according to the present invention.

The frequency characteristic data generator 65 extracts data at equal frequency intervals from the logarithmic scale frequency characteristic data string output from the digital filter 64 and creates a linear scale frequency characteristic data string. For example, N =
If 2 ⁷ (= 64), the frequency characteristic data creation unit 65
Samples 64 frequency data at 250 Hz intervals from the highest band (16 to 32 kHz) and 2 from the next band (8 to 16 kHz)
32 frequency data are sampled at 50 Hz intervals, frequency data is similarly sampled from each band in the same manner, and 25
A total of 128 frequency characteristic data strings are created at 0 Hz intervals. Next, the frequency characteristic data generation unit 65 generates a frequency characteristic data sequence of 256 data sequences at equal frequency intervals comprising 128 data sequences at 250 Hz intervals and a data sequence obtained by folding the data sequence toward the high frequency band side. Are generated and input to the inverse FFT operation unit 66 in parallel.

The inverse FFT operation section 66 performs an inverse FFT process on the 256 pieces of data at equal frequency intervals, and outputs 256 pieces of data at predetermined sampling time intervals. These 256 data become the coefficients c _{0 to} c ₂₅₅ of the 256 taps of the FIR filter (see FIG. 8) constituting the equalizer 2.
FIG. 4 shows the FI output from the inverse FFT operation unit 66 of the present invention.
This is an example of tap coefficients c _{0 to} c _n−1 (n = 256) of an R filter. By comparing the tap coefficients of FIG. 4 and FIG. 12, it is possible to obtain tap coefficients equivalent to those of the conventional spline interpolation according to the present invention.

(B) Reason for adding the first to fourth bands B1 to B4 The first to fourth bands B1 to B4 are used for the first to fourth bands B1 to B4.
The reason for adding the fourth bands B1 to B4 is to make the operation result (filter coefficient) of the inverse FFT a real value and a discrete value in time. Discrete Fourier transform of the signal is defined by the following equation _{X (ω) = Σ n x} (nT) exp (-jωnT) (n = -∞~ + ∞). If T = 1 second and normalized the _{X (ω) = Σ n x} (n) exp (-jωn) (n = -∞~ + ∞). From the above equation, the Fourier transform is a periodic function having a period of 2π. That, X (ω + 2π) = Σ n x (n) exp (-j (ω + 2π) n) = Σ n x (n) exp (-jωn) · exp (-j2πn) (n is an integer) = Σ _n x ( n) exp (-jωn) = X (ω) holds, and the Fourier transform is a periodic function with a period of 2π.

The Fourier transform of an arbitrary signal x (t) that is continuous in time is generally not a periodic function. However, by discretizing at a constant period, the Fourier transform X (ω) becomes a periodic function. Although the frequency spectrum increases periodically even though discrete values are obtained from the continuous-time waveform, it may seem strange, but it can be considered as follows. That is, as shown in FIG. 5A, the discrete-time waveform certainly takes only discrete values, but when viewed from the opposite side, the values of many parts of the continuous-time waveform are suppressed to zero. . The signal component added to suppress this interval is the repetition part of the frequency spectrum shown in FIG. As described above, discretization of the time waveform causes repetition of the frequency spectrum from angular frequencies −π to π. As described above, in order to make the inverse FFT operation result a discrete value (discrete time waveform), the frequency spectrum must have periodicity.

Next, the conditions of the frequency spectrum necessary for the time waveform obtained by the inverse FFT operation to become a real value will be described. The time signal x (n) is first assumed to be a complex number, and its real part is x _r (n) and its imaginary part is x _i (n), and x (n) = x _r (n) + jx _i (n) (J is an imaginary unit). When this is the Fourier _{transform, X (ω) = Σ n} x (n) exp (-jωn) (n = -∞~ + ∞) = Σ n (x r (n) + jx i (n)) (cos ( ωn) -jsin (ωn)) = Σ n {(x r (n) cos (ωn) + x i (n) sin (ωn)) + j (x i (n) cos (ωn) -x r (n ) sin (ωn))｝, where X _r (ω) is the real part of the frequency spectrum X (ω) and X _i (ω) is the imaginary part, and X _r (ω) =) _n (x _r (n) cos _{(ωn) + x i (n} ) sin (ωn)) X i (ω) = Σ n (x i (n) becomes _{cos (ωn) -x r (n} ) sin (ωn)). Here, the time signal x (n) has only a real part, that is, if x _i (n) = 0, X _r (ω) = Σ _n x _r (n) cos (ωn) X _i (ω) = Σ _n ( _{-x r (n) sin (ωn} )) becomes.

The real part X _r (ω) of the frequency spectrum X (ω)
Is the superposition of the even function cos with respect to the angular frequency ω, so that it becomes line-symmetric about the angular frequency ω = 0 [red] as shown in FIG. Further, since the imaginary part X _i (ω) of the frequency spectrum X (ω) is a superposition of the odd function sin with respect to the angular frequency ω, the angular frequency ω = 0 as shown in FIG.
It becomes point symmetric about [rad]. From the above, in order for the time waveform x (n) obtained by the inverse FFT to be a real value, (1)
The real part X _r (ω) of the frequency spectrum X (ω) is line-symmetric about the angular frequency ω = 0 [rad], and (2) the imaginary part X _i (ω) of the frequency spectrum X (ω) is It is necessary to be line-symmetric about the angular frequency ω = 0 [rad].

Actually, since the frequency spectrum set by the frequency characteristic setting unit 51 does not include the imaginary part X _i (ω), in order for the time waveform x (n) to be a real value, the “frequency spectrum The real part X _r (ω) of X (ω) is represented by an angular frequency ω = 0 [ra
d]. In summary, in order for the filter coefficient obtained by the inverse FFT operation to be a real value and a discrete value in time, "the angular frequency -π ~
The frequency characteristic of 0 [rad] must be line-symmetric with the frequency characteristic of 0 to π [rad], and the frequency characteristic of -π to π [rad] must be repeated. .

Therefore, the N-point sample-and-hold unit 61 of the present invention inserts N sample data with a gain of 0 into the first band (0 Hz to 125 Hz) B1, and the symmetric characteristic adding unit 62 sets the first band B1 (0 Third band B3 (-125Hz to 0H) adjacent to
z), the same N frequency characteristic data (gain) as the lowest band (125 to 250 Hz) are copied, and the second band (32 to 64 k
Hz) N pieces of sample data with a gain of 0 are inserted into B2, and the symmetric characteristic adding unit 62 outputs the fourth band B adjacent to the second band B2.
It is assumed that the same N frequency characteristic data (gain) as the highest band (16 to 32 kHz) is copied to 4 (64 to 128 kHz) and the above condition A is satisfied. By adding this symmetrical portion, the frequency characteristics are accurately smoothed on a linear scale. As described above, the present invention has been described with reference to the embodiments. However, the present invention can be variously modified in accordance with the gist of the present invention described in the claims, and the present invention does not exclude these.

[0027]

As described above, according to the present invention, a logarithmic scale data string can be converted into a linear scale equally spaced data string with the same smoothness and accuracy as in spline interpolation by simple processing. Further, according to the present invention, it is possible to convert the logarithmic scale equalizing characteristic data sequence into a linear scale equal frequency characteristic equalizing characteristic data sequence by simple processing, and furthermore, to convert the audio signal according to the set equalizing characteristic. A filter coefficient that can be corrected can be determined. Also,
According to the present invention, a logarithmic scale equalizing characteristic data sequence can be converted into a linear scale equal frequency characteristic equalizing characteristic data sequence with the same smoothness and accuracy as spline interpolation, and the set equalizing characteristic , A filter coefficient capable of correcting the audio signal can be correctly determined. Further, according to the present invention, an audio signal can be accurately corrected according to the set equalizing characteristics.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of a filter coefficient determination device of the present invention.

FIG. 2 is an explanatory diagram of an operation of an N sample hold unit.

FIG. 3 is an equalizing characteristic diagram of the linear scale of the present invention.

FIG. 4 is an example of an FIR coefficient of the present invention.

FIG. 5 is a diagram illustrating the periodicity of the discrete Fourier transform.

FIG. 6 is an explanatory diagram of symmetry of a frequency spectrum.

FIG. 7 is a configuration diagram of an audio device.

FIG. 8 is a configuration diagram of an FIR filter included in the equalizer.

FIG. 9 is an explanatory diagram of a conventional filter coefficient determination process.

FIG. 10 is an equalizing characteristic diagram of a linear scale by spline interpolation.

FIG. 11 is an explanatory diagram of frequency characteristic data input to an inverse FFT processing unit.

FIG. 12 is an example of an FIR coefficient by a spline interpolation method.

[Explanation of symbols]

51: frequency characteristic setting section 52: equalizer having FIR filter configuration 53: filter coefficient determination device 61: N-point sample hold section 62: symmetry characteristic adding section 63: data output section 64: digital low band Pass filter (eg FIR
Filter) 65 frequency characteristic data creation unit 66 inverse FFT operation unit

Claims (12)

[Claims] 1. A data scale conversion method for converting a data sequence given on a logarithmic scale into a data sequence on a linear scale, wherein each data of the logarithmic scale is sampled and held at N points, and the sampled and held logarithmic data is stored. The sequence is regarded as time-series data, input to a digital low-pass filter, and data of a linear scale are extracted and output from a log-scale data sequence output from the low-pass filter. Data scale conversion method. 2. The data scale conversion method according to claim 1, wherein the logarithmic scale data sequence is a logarithmic scale equalizing characteristic data sequence, and generates a data sequence at equal frequency intervals. 3. A method for determining a filter coefficient of a digital filter constituting an equalizer, wherein frequency characteristic data is discretely set on a logarithmic scale, and each frequency characteristic data on the logarithmic scale is sampled and held at N points. The logarithmic scale frequency characteristic data sequence is regarded as time-series data and input to a digital low-pass filter, and data is extracted from the logarithmic scale frequency characteristic data sequence output from the low-pass filter at equal frequency intervals. Generating a frequency characteristic data sequence of a linear scale, performing inverse FFT processing on the frequency characteristic data of the linear scale, and determining a filter coefficient of a digital filter constituting the equalizer based on the result of the inverse FFT processing. Characteristic filter coefficient determination method. 4. A frequency characteristic data sequence on a linear scale is generated from the extracted data sequence at equal frequency intervals and a data sequence obtained by folding the data sequence toward a higher frequency band. 3. The filter coefficient determination method according to 3. 5. The first and second frequency characteristic data of 0 (dB) are respectively adjacent to the lowest band and the highest band where the frequency data of the lowest frequency and the highest frequency of the logarithmic scale are valid.
And a third band having the same frequency characteristic data as the lowest band is provided adjacent to the first band, and the same frequency as the highest band is provided adjacent to the second band. A fourth band having characteristic data is provided, and each of the frequency characteristic data is sampled and held at N points in the first to fourth bands, and these sample and hold data are added before and after the time series data to form the low frequency band. 4. The method according to claim 3, wherein the input is made to a pass filter. 6. N equal-frequency intervals are extracted from the highest frequency band, and N equal-frequency intervals are extracted from all frequency bands up to the lowest frequency band. And a 4N linear scale frequency characteristic data sequence is generated by combining the data sequence with the 2N data sequence obtained by folding the data sequence toward the high frequency band side.
The filter coefficient determination method according to claim 5, wherein an FT process is performed. 7. An equalizer composed of a digital filter, a characteristic data setting section for discretely setting frequency characteristic data on a logarithmic scale, a sample and hold section for sampling and holding each frequency characteristic data on the logarithmic scale at N points, Means for assuming and outputting the logarithmic-scale frequency characteristic data string sampled and held as time-series data, a digital low-pass filter to which the time-series data is input, and a logarithmic-scale frequency characteristic output from the low-pass filter A linear scale frequency characteristic data generator for extracting data from the data sequence at equal frequency intervals to generate a linear scale frequency characteristic data sequence, and performing an inverse FFT process on the linear scale frequency characteristic data to configure the equalizer Inverse for outputting filter coefficient of digital filter An equalizer, comprising: an FFT processing unit. 8. The frequency characteristic data creation unit extracts data at equal frequency intervals from a logarithmic scale frequency characteristic data sequence output from the low-pass filter, and extracts the extracted data sequence with equal frequency intervals, 8. The equalizer according to claim 7, wherein a frequency characteristic data sequence of a linear scale is generated by a data sequence obtained by folding the data sequence to a high frequency band side. 9. The equalizer further includes first and second frequency characteristic data of 0 (dB) adjacent to the lowest band and the highest band, respectively, where the frequency data of the lowest frequency and the highest frequency of the logarithmic scale are valid. Set up a band,
Further, a third band having the same frequency characteristic data as the lowest band is provided adjacent to the first band, and has the same frequency characteristic data as the highest band adjacent to the second band. Means for providing a fourth band, and sampling and holding each frequency characteristic data at N points in the first to fourth bands, wherein the output means converts the sample and hold data of the first to fourth bands to the The equalizer according to claim 7, wherein the equalizer is added before and after time-series data and input to the low-pass filter. 10. The linear scale frequency characteristic data generating section extracts data of N equal frequency intervals from the highest band and N equal frequency intervals of all frequency bands up to the lowest band. Data is extracted to create 2N data strings at equal frequency intervals, and 4N data strings of the 2N data strings at equal frequency intervals and 2N data strings obtained by folding the data strings toward the high frequency band side. The equalizer according to claim 9, wherein a frequency characteristic data string of the number of linear scales is generated and input to the inverse FFT processing unit. 11. 4N output from the inverse FFT processing unit
The equalizer according to claim 10, wherein the result of the inverse FFT processing is used as a filter coefficient of a digital filter having 4N taps constituting the equalizer. 12. The number of taps constituting the equalizer is four.
The equalizer according to claim 11, wherein the N digital filters are FIR digital filters.