JP5388542B2 - Audio signal processing apparatus and method - Google Patents

Description

  The present invention relates to an audio signal processing apparatus and method, and more particularly to a technique for improving quantization distortion of a digital audio signal. More specifically, the present invention relates to a technique for generating data in which the number of bits of a digital audio signal is expanded.

In recent years, digitalization of audio signal processing in audio equipment has become common. For example, 16-bit quantization for a CD or the like, and a sampling frequency of 44.1 kHz (a digital audio tape standard is a sampling frequency of 48 kHz) are defined as standards.
On the other hand, a natural sound that is generated in an analog manner is generally used as the source of the audio signal, and the dynamic range of the input level is very wide. Therefore, waveform distortion occurs due to quantization in an audio signal that changes at a minute level.

In Patent Document 1 below, in order to cope with such signal distortion due to the insufficient number of bits of quantization, (a) among the data changes for each sample data of the input digital data, Using at least one of the low-pass filters having different cutoff frequencies as a criterion for the change time length, a signal output is obtained, and (b) the level change rate is determined from the LSB change time interval. In addition, it is described that after a data string lower than the LSB that smoothes the waveform change in the determined period is generated, data less than the LSB is generated before and after the LSB change point to perform bit extension.

Japanese Patent Publication No. 7-73186 Supervised by Takao Hinamoto, Mitsuji Muneyasu, Ryo Otaguchi, “Nonlinear Digital Signal Processing”, Asakura Shoten, March 20, 1999, p. 72-74, p. 106-108

  The above Non-Patent Document 1 will be mentioned later.

  In the method according to Patent Document 1, the cut-off frequency is selected only by the time length of the LSB change. That is, the cut-off frequency changes every time there is an LSB change. In addition, it is difficult to accurately select the cut-off frequency only by the time length of the LSB change. For example, even with a sine wave having a constant frequency, the time length of the LSB change is not constant. For this reason, the process joints do not change smoothly over a long time range, and the distortion is not sufficiently corrected.

  The present invention has been made to solve the above problems, and an object thereof is to provide an audio signal processing apparatus and method capable of generating audio having a waveform closer to the audio signal waveform before quantization. Another object of the present invention is to make it possible to output an audio signal without losing the specific frequency component when the audio signal includes a specific frequency component.

The audio signal processing device of the present invention is
an original data bit shift unit that shifts an n-bit (n is a positive integer) audio signal by α bits (α is a positive integer) to generate an n + α-bit audio signal;
A frequency amplitude estimation unit that detects an inflection point of the n-bit audio signal and estimates the frequency and amplitude of the audio signal from the inflection point;
A specific frequency detector that determines whether or not the intensity of a spectrum of a predetermined specific frequency of the audio signal is greater than a predetermined threshold;
When the intensity of the spectrum of the specific frequency is determined to be equal to or less than the predetermined threshold by the specific frequency detection unit, the frequency amplitude estimation unit estimates the audio signal output from the original data bit shift unit. Output an audio signal obtained by performing edge preserving smoothing processing using a low-pass filter coefficient generated based on the frequency and amplitude,
A selective filter processing unit that directly outputs the audio signal output from the original data bit shift unit when the specific frequency detection unit determines that the spectrum intensity of the specific frequency is greater than the threshold value. It is characterized by that.

  According to the present invention, harmonics generated by quantization are estimated by estimating the frequency and amplitude from an n-bit speech signal and smoothing using a low-pass filter coefficient determined based on the frequency and amplitude. The components can be removed, and waveform distortion of the audio signal due to quantization can be corrected. Further, by performing the edge preserving smoothing filter process, it is possible to prevent the reproducibility of the audio signal having a region where the signal amplitude changes sharply and greatly. Further, when the audio signal includes a component of a specific frequency, the edge preserving smoothing filter is not applied, so that it can be output as it is without losing the component of the specific frequency.

Embodiment 1 FIG.
FIG. 1 is a diagram showing a configuration of an audio signal processing device according to Embodiment 1 of the present invention. The audio signal processing apparatus according to Embodiment 1 includes an input terminal 1, a frequency amplitude estimation unit 2, a specific frequency detection unit 6, a filter coefficient generation unit 3, an original data bit shift unit 4, and a filter unit 5.

  In the audio signal processing apparatus shown in FIG. 1, an n-bit (n is a positive integer) audio signal (input audio signal) X is converted from an input terminal 1 to an original data bit shift unit 4, a frequency amplitude estimation unit 2, and a specific frequency. Input to the detection unit 6.

  The original data bit shift unit 4 filters the n + α-bit audio signal (bit-shifted audio signal) X ′ obtained by bit-shifting (bit extension) the n-bit audio signal X by α bits (α is a positive integer). Output to.

  The frequency amplitude estimator 2 detects an inflection point of the input audio signal X, estimates the frequency of the audio signal X from the interval of the inflection points that follow each other, and calculates the signal of the inflection point that follows. The amplitude of the audio signal is estimated from the difference in value, and the frequency F and the amplitude A are output to the filter coefficient generation unit 3.

  The specific frequency detection unit 6 calculates a spectrum intensity R of a predetermined specific frequency Fd of the audio signal X, and a signal (intensity determination result signal) DR indicating whether or not the intensity R is greater than a predetermined threshold value. Is output to the filter coefficient generation unit 3. The intensity determination result signal DR takes a first value, for example, “1” when the intensity R is greater than a predetermined threshold value RTH, and a second value, for example, “0” when the intensity is equal to or less than the predetermined threshold value RTH. "I take the.

The filter coefficient generation unit 3 generates a filter coefficient C based on the strength determination result signal DR, the frequency F, and the amplitude A, and outputs the filter coefficient C to the filter unit 5.
This filter coefficient C is generated based on the frequency F and the amplitude A when the intensity determination result signal DR is the second value (when the spectrum intensity R is equal to or less than the threshold value RTH). When the intensity determination result signal DR is the first value (when the spectrum intensity R is larger than the threshold value RTH), the filter unit 5 is operated as a smoothing filter (low-pass filter coefficient). This is for outputting the input signal as it is.

  The filter unit 5 filters the n + α-bit audio signal X ′ output from the original data bit shift unit 4 using the filter coefficient C, and outputs an n + α-bit audio signal Y. As described above, when the spectrum intensity R is relatively small, an edge-preserving smoothing process is performed, and when the spectrum intensity R is relatively large, a process of outputting the input as it is is performed. It is.

  When the filter coefficient generation unit 3 and the filter unit 5 have the spectrum intensity R of the specific frequency equal to or lower than a predetermined threshold, the low-frequency signal generated on the basis of the frequency R and the amplitude A is reduced with respect to the audio signal X ′ after the bit shift. The speech signal Y obtained by performing the edge preserving smoothing process using the pass-pass filter coefficient is output, and when the spectrum intensity R of the specific frequency is larger than the threshold, the speech signal X ′ after the bit shift is A selective filter processing unit 10 is configured to output as it is.

2A to 2C are diagrams illustrating an example of an audio signal processed by the audio signal processing device of FIG. FIG. 2A shows a sinusoidal analog audio signal having a frequency f1. 2A and 2B, the horizontal axis indicates time t, and the vertical axis in FIG. 2A indicates the signal level. The analog signal is sampled and quantized to generate a digital audio signal having an n-bit frequency f 1 as shown in FIG. 2B and input to the input terminal 1. The vertical axis in FIG. 2B indicates the level of the signal X (i). The digital audio signal X (i) shown in FIG. 2 (b) is a sequence of digital signals at successive sampling points (represented by black circles), and the time (representing numerical value) i is the sampling point. Increases by 1 each time.
The solid line in FIG. 2C indicates the frequency spectrum of the audio signal X (i) in FIG. 2B, and the dotted line indicates the frequency characteristics of a low-pass filter (described later). In FIG. 2C, the horizontal axis indicates frequency f, and the vertical axis indicates power and gain. As shown in the frequency spectrum of FIG. 2 (c), the analog signal before quantization is a sine wave of frequency f1, but the digital audio signal X (i) obtained by quantization has second harmonics f2 and 3 Harmonics such as the second harmonic f3 are included.

  In the audio signal processing device of FIG. 1, for example, the frequency f1 as shown by the solid line in FIG. 2C is estimated by the frequency amplitude estimation unit 2, and the strength of the specific frequency Fd is determined by the specific frequency detection unit 6 as a predetermined threshold value. When it is determined that it is below, a low-pass filter coefficient for realizing the cut-off frequency characteristic as shown by the dotted line in FIG. 2C is generated. At this time, the cutoff frequency fc1 is set between f1 and f2. If the frequency f1 can be estimated, f2 is obvious. By performing smoothing using the generated filter coefficient having the characteristic of the cut-off frequency fc1, harmonics such as f2 and f3 can be removed from the audio signal.

Hereinafter, components of the audio signal processing apparatus of FIG. 1 will be described in order.
FIGS. 3A and 3B are diagrams for explaining the operation of the original data bit shift unit 4. The horizontal axis indicates time i, and the vertical axis indicates the signal level. 3A shows an n-bit audio signal X (same as the signal shown in FIG. 2B), and FIG. 3B shows an n + α-bit audio signal X ′. The original data bit shift unit 4 bit-shifts the n-bit audio signal as shown in FIG. 3A by α bits, and the n + α-bit audio signal as shown in FIG. Output.

  FIG. 4 is a diagram showing a detailed configuration of the frequency amplitude estimation unit 2. The frequency amplitude estimation unit 2 includes an inflection point detection unit 7, a frequency estimation unit 8, and an amplitude estimation unit 9, as shown in FIG.

The inflection point detection unit 7 detects a point at which the n-bit audio signal starts or ends a series of monotone increases and a point at which a series of monotone decreases starts or ends as inflection points. Here, “a series of monotonous increases” means a continuous increase in which no decrease occurs in the middle (the same value may continue). Within this series of monotonically increasing intervals,
X (i) ≦ X (i + 1)
Are continuously satisfied (at all sampling points i).
Similarly, “a series of monotonous decreases” means a series of decreases in which no increase occurs in the middle (the same value may continue). Within this series of monotonically decreasing intervals,
X (i) ≧ X (i + 1)
Are continuously satisfied (at all sampling points i).

The illustrated inflection point detection unit 7 includes a first derivative calculation unit 11 and a sign change point detection unit 12.
The primary differential calculation unit 11 calculates and outputs primary differential data D (i) from the input n-bit audio signal X (i). For example, the primary derivative calculation unit 11
The n-bit audio signal at each sampling point is represented by X (i),
When the n-bit audio signal at the next sampling point is represented by X (i + 1),
D (i) = X (i + 1) −X (i)
D (i) obtained in step 1 is output as first-order differential data.

The sign change point detection unit 12 detects a point at which the sign of the primary differential data D has changed positively and a point at which the sign has changed negatively as an inflection point. More specifically, the sign change point detection unit 12 changes the primary differential data D from a state where the sign is negative or a state where the value is zero to a state where the sign is positive (hereinafter referred to as “positive”). ), And a point where the sign is positive or a state where the value is zero is changed to a state where the sign is negative (hereinafter referred to as “change point to negative”). This is detected as a point where the sign has changed. The output of the sign change point detection unit 12 is binary data PM, and takes a first value (for example, “1”) from “change point to positive” to “change point to negative”. The second value (eg, “0”) is taken from “change point to” to “change point to positive”.
Such an operation of the sign change point detection unit 12 is an operation of converting the primary differential data D (i) into binary data PM having only a sign (however, in the case of “0”, the sign of the previous data is used). You can also.
The inflection point detection unit 7 including the first-order differential calculation unit 11 and the sign change point detection unit 12 has a point where an n-bit audio signal starts a series of monotone increases and a point where a series of monotone decreases starts. Is detected as an inflection point, and a signal SL that has a first value, for example, “1” at the position of the detected inflection point, and a second value, for example, “0”, is generated at other positions.

The frequency estimator 8 calculates the section length of the inflection point of the audio signal and the previous (immediately preceding) inflection point and outputs it as the frequency F.
The amplitude estimation unit 9 outputs the level difference between the inflection point of the audio signal and the audio signal at the previous (immediately) inflection point as the amplitude A.

FIGS. 5A to 5D are diagrams for explaining the operation of the frequency amplitude estimation unit 2. The horizontal axis indicates time i.
FIG. 5A shows an n-bit audio signal X (i), FIG. 5B shows primary differential data D (i) of the audio signal, FIG. 5C shows binary data PM, FIG. 5D shows a signal SL indicating the detection result of the inflection point position of the audio signal.
When the n-bit audio signal shown in FIG. 5A is input, the primary differential calculation unit 11 calculates primary differential data D (i) as shown in FIG.
The sign change point detector 12 converts the primary differential data D into binary data PM having only a sign as shown in FIG. 5C, and the binary data PM changes as shown in FIG. 5D. The positions ic = i1, i2, and i3 (the positions on the time axis are indicated by dotted lines extending in the vertical direction) are detected, and the value of the signal SL is set to “1” at the detection position.
Therefore, the frequency estimation unit 8 and the amplitude estimation unit 9 obtain F = 1 / (i2−i1) as the frequency and A = | X (i2) −X (i1) | In the interval ˜i3, F = 1 / (i3−i2) as the frequency and A = | X (i3) −X (i2) | as the amplitude, respectively. The obtained frequency F and amplitude A are supplied to the filter coefficient generation unit 3.

  As described with reference to FIGS. 5A to 5D, the position where the sign of the first derivative data of the audio signal calculated by the frequency amplitude estimation unit 2 shown in FIG. 4 changes (“change point to positive”). ”And“ change point to negative ”) are treated as inflection points of the audio signal, and it can be estimated that the reciprocal of the time interval of the inflection points that follow each other is the frequency of the audio signal. In addition, since the inflection point is the start point of a series of monotone increases or the start point of a series of monotone decreases, it can be estimated that the difference between the signal values of successive inflection points is the amplitude.

  The frequency amplitude estimation unit 2 shown in FIG. 4 can calculate the frequency by estimating the half cycle of the highest frequency component included in the audio signal, and simultaneously calculate the amplitude from the difference in signal value at the inflection point. can do.

  Then, the filter coefficient determined based on the frequency and the amplitude estimated in this way is used for filtering on the signal value of each section (intersections between successive inflection points). For example, using the filter coefficient determined by the frequency and amplitude determined by the inflection points i1 and i2 shown in FIG. 5A to FIG. 5D, the sample point next to the inflection point i1 is changed. Filtering is performed on the data up to the music point i2.

The inflection point detection unit 7 detects a point at which an n-bit audio signal starts a series of monotone increases and a point at which a series of monotone decreases starts as an inflection point. The inflection point detection unit may be configured to detect, as an inflection point, a point at which a bit sound signal ends a series of monotone increases and a point at which a series of monotone decreases ends. In that case, as the first derivative calculation unit,
D (i) = X (i) -X (i-1)
D (i) obtained in (1) is output as primary differential data, and the sign change point detection unit is the closest positive point before the point where the sign of the primary differential data D changes to negative , And a point that detects the closest negative point before the point where the sign of the primary differential data D changes to positive may be used.

  The specific frequency detector 6 calculates the intensity of the spectrum only for a predetermined specific frequency. Based on the result of setting the frequency (specific frequency) Fd to be detected, generating a cosine wave (COS wave) and a sine wave (SIN wave) of that frequency, and superimposing the audio signal X on the basic pattern The intensity R of the spectrum of the specific frequency Fd is calculated.

Specifically, the intensity R (i) of the spectrum is expressed by equation (1).
Here, X (i) is the input level, Fd is the specific frequency, Fs is the sampling frequency, k is the relative position from the point of interest (defined by the distance and direction expressed by the number of samples), and m is preset. The order (number of samples used in the calculation), R (i), is the intensity of the spectrum of the frequency Fd in the m sample section centered on time i.

  FIGS. 6A and 6B, FIGS. 7A to 7D, and FIGS. 8A to 8D are diagrams for explaining the operation of the specific frequency detection unit 6. FIG. The operation of the specific frequency detector 6 when the frequency to be detected (specific frequency) is Fd = Fs / 4 will be described as an example. FIGS. 6A and 6B show basic patterns used for spectrum intensity calculation. FIG. 6A shows a basic pattern of COS waves (represented by cos {(2πFd / Fs) k}) when Fd = Fs / 4, and FIG. 6B shows a case where Fd = Fs / 4. The SIN wave basic pattern (represented by sin {(2πFd / Fs) k}) is shown.

7A to 7D are diagrams for explaining the operation of the specific frequency detector 6 when the input signal X (i) includes a specific frequency. 7A to 7D, the horizontal axis represents time i. In FIG. 7A, the vertical axis indicates the level of the signal X (i), and in FIGS. 7B to 7D, the vertical axis indicates the power.
FIG. 7A shows a sine wave audio signal X (i) having the same frequency as the specific frequency Fd. FIG. 7B shows Rc (i) that is a superposition result of the audio signal of FIG. 7A and the basic pattern of FIG. FIG. 7C shows Rs (i) that is a result of superimposing the audio signal of FIG. 7A and the basic pattern of FIG. 6B. FIG. 7D shows the sum of squares of Rc (i) in FIG. 7B and Rs (i) in FIG. 7C (R (i) = {Rs (i)} ² + {Rc (i) } ² ), which shows the intensity R (i) of the spectrum of the specific frequency Fd.

FIGS. 8A to 8D are diagrams for explaining the operation of the specific frequency detector 6 when the input signal X (i) does not include a specific frequency. 8A to 8D, the horizontal axis indicates time i. In FIG. 8A, the vertical axis indicates the signal level X (i), and in FIGS. 8B to 8D, the vertical axis indicates the power.
FIG. 8A shows a sine wave audio signal having a frequency (Fd / 2) that is a half of the specific frequency Fd as an example of a frequency different from the specific frequency Fd. FIG. 8B shows Rc (i) that is a result of superimposing the audio signal of FIG. 8A and the basic pattern of FIG. 6A. FIG. 8C shows Rs (i) that is a result of superimposing the audio signal of FIG. 8A and the basic pattern of FIG. 6B. FIG. 8D is a sum of squares of Rc (i) in FIG. 8B and Rs (i) in FIG. 8C, and shows the intensity R (i) of the spectrum of the specific frequency Fd.

  As shown in FIGS. 7D and 8D, when the input signal includes the specific frequency Fd, the spectrum intensity R (i) is a large value, and when the input signal does not include the specific frequency Fd, The spectrum intensity R (i) is a small value.

FIGS. 9A and 9B are diagrams for explaining the operation of the specific frequency detector 6 when the sine wave audio signal X having the frequency f1 is input. 9A and 9B, the horizontal axis represents time i, the vertical axis in FIG. 9A represents the signal level, and the vertical axis in FIG. 9B represents the spectrum intensity.
Here, it is assumed that the specific frequency Fd is set in the vicinity of Fs / 2. Since the audio signal X as shown in FIG. 9A is a sine wave having a frequency f1 different from the specific frequency Fd (illustrated as f1 = 2Fd / 9 in the illustrated example), the spectrum intensity R is As shown in 9 (b), the value is very small.

  As described above, the specific frequency detection unit 6 calculates the intensity R of the spectrum having a large value if the specific frequency is included in the audio signal, and a small value if the specific frequency is not included, and determines whether the intensity R is greater than the threshold value RTH. A signal DR can be output.

The filter coefficient generation unit 3 outputs the input as it is when the intensity determination result signal DR output from the specific frequency detection unit 6 takes the first value (indicating that the spectrum intensity R is greater than the threshold value RTH). When the intensity determination result signal DR output from the specific frequency detector 6 takes the second value (indicating that the spectrum intensity R is equal to or less than the threshold value RTH), the frequency F is generated. The low pass filter coefficient C is selected based on the amplitude A and output to the filter unit 5.
Here, the threshold value RTH is a reference value for determining whether or not a specific frequency is included in the audio signal, and is set in advance.

  In order to cause the filter unit 5 to output the input as it is, the filter coefficient generation unit 3 sets “1” as the coefficient for the sample value of the target point and “0” as the coefficient for the other sample values as the filter coefficient. To supply.

When generating the low-pass filter coefficient C, the cutoff frequency characteristic of the low-pass filter is determined based on the frequency of the audio signal, and the order of the low-pass filter is determined based on the frequency and amplitude.
For example, in the case of an audio signal having a frequency characteristic as shown by a solid line in FIG. 10, a low-pass filter coefficient for realizing a cutoff frequency characteristic as shown by a dotted line in FIG. 10 is generated. In this case, if the frequency f1 can be estimated, f2 is self-explanatory, and therefore a cutoff frequency fc1 is set between f1 and f2.

  Also, the order is determined by obtaining the slope from the frequency and amplitude. For example, if the frequency is the same, the order is reduced as the amplitude A increases.

  As the filter coefficient generation unit 3, a filter coefficient table (an example of which is shown in FIG. 11) that stores a plurality of sets of low-pass filter coefficients having different cut-off frequencies and orders that are generated based on the above consideration as candidates. May be selected based on the estimated frequency F and amplitude A and output as a low-pass filter coefficient C. The filter coefficients stored in the filter coefficient table shown in FIG. 11 have different cutoff frequencies and orders.

  As described above, the filter coefficient generation unit 3 generates a filter coefficient that outputs an input as it is when the audio signal includes a specific frequency, and cuts off the higher harmonic component of the estimated frequency F when it does not include it. As the estimated amplitude A increases, a filter coefficient having a lower order of the low-pass filter is generated.

When the low-pass filter coefficient is supplied, the filter unit 5 operates as an edge-preserving smoothing filter.
Here, the edge preserving smoothing filter will be described. An edge-preserving smoothing filter is a filter that smoothes only small changes while preserving the sharpness of a portion where there is a steep and large change. An ε filter, a trimmed average value filter (DW-MTM filter), There are bilateral filters. The ε filter and the trimmed average value filter (DW-MTM filter) are described in Non-Patent Document 1 described above. Hereinafter, an edge preserving smoothing filter will be described as an ε filter as an example.

One-dimensional processing by the ε filter is expressed by Expression (2).
Here, x (i) is an input level, y (i) is an output level, _ak is a low-pass filter coefficient, k is a relative position from a point of interest, m is an order, and ε is a threshold value.

  When the absolute value of u is less than or equal to the threshold ε, f (u) = u, and when the absolute value of u is greater than the threshold ε, f (u) = 0. In Expression (2), u is x (ik) -x (i).

Difference x (ik) −x (i) between the signal level x (i) of the point of interest i and the signal level x (ik) of the position i−k around the point of interest i
Is small,
f (u) = f {x (ik) -x (i)}
Is approximately linear, and assuming that the sum of the coefficients _ak is 1, Equation (2) is rewritten as Equation (3), which is equal to the weighted average filter.

On the other hand, the difference x (ik) −x (i) between the signal level x (i) at the point of interest i and the signal level x (ik) at the position i−k around the point of interest i.
Is large,
f (u) = f {x (ik) -x (i)}
Is 0, and Expression (2) is rewritten as Expression (4), and the value of the point of interest is output as it is.
y (i) = x (i) (4)
Therefore, the ε filter smooths only small changes while preserving a portion where there is a steep and large change.

FIG. 12 is a diagram for explaining the operation of the ε filter. 12A shows the input signal x (i) of the ε filter, and FIG. 12B shows f {x (i1-k) −x (when the position i1 in FIG. i1)}
Is shown.

  When the signal having a gentle slope as shown in FIG. 12A is an input signal of the ε filter, if the order is too large, the difference between the point of interest and the surrounding points exceeds the threshold value, as shown in FIG. Since the area is set to 0 and the weighted average is taken, the smoothing effect is weakened.

  Many edge-preserving smoothing filters such as the ε filter have the advantage of being able to preserve steep and large changes compared to low-pass filters. However, when the signal has a gentle slope, the smoothing performance is low. Therefore, in the audio signal processing apparatus of FIG. 1, the inclination of the signal is taken into account by estimating the frequency and amplitude of the audio signal. When the frequency is the same and the amplitude is large, the slope becomes steep, so the order of the filter is reduced and the smoothing performance of the ε filter is improved.

  In the following description, it is assumed that the filter unit 5 functions as an ε filter and performs a smoothing process when a low-pass filter coefficient is supplied. The ε filter handles the number of bits increased by the bit shift in the original data bit shift unit 4 as a small amplitude component. That is, when bit shifting from n bits to n + α bits, the threshold ε of the ε filter is set to 2 to the power of α. Thereby, a step of 1 LSB or less in n bits is smoothed.

  FIGS. 13A and 13B are diagrams for explaining the operation of the filter unit 5 when operating as an ε filter. The horizontal axis indicates time i, and the vertical axis indicates the signal level. 13A shows an n + α-bit audio signal X ′ (i), and FIG. 13B shows an n + α-bit audio signal Y (i) subjected to the ε filter process. The filter unit 5 performs smoothing processing using a filter coefficient C that realizes a cut-off frequency characteristic that removes harmonics from an n + α-bit audio signal, whereby an n + α-bit audio as shown in FIG. Waveform distortion due to quantization is corrected from the signal, and an n + α-bit audio signal as shown in FIG. 13B is output.

  In this way, by operating the filter unit 5 as an edge-preserving smoothing filter, for example, an ε filter, harmonics are removed from an n-bit audio signal, and an n + α-bit audio signal with corrected waveform distortion is output. it can.

  Next, the operation of the audio signal processing apparatus in FIG. 1 when an n-bit audio signal including the specific frequency Fd is input will be described. Here, the specific frequency Fd is a high frequency near half the sampling frequency.

FIGS. 14A to 14C are diagrams showing an example of an audio signal including a high frequency having a frequency near half the sampling frequency input to the audio signal processing apparatus of FIG.
FIG. 14A shows a sinusoidal analog audio signal having a frequency f8 near the half of the sampling frequency. In FIG. 14A, the horizontal axis indicates time t, and the vertical axis indicates the signal level. The analog signal is sampled and quantized to generate a sine wave digital audio signal with an n-bit frequency f8 as shown in FIG. In the figure, the vertical dotted line indicates the sampling timing. In the illustrated example, it is assumed that 1/2 of the amplitude of the audio signal (1/2 of the difference between the peak and the bottom) corresponds to the LSB step width of the n-bit signal.
In FIG. 14B, the horizontal axis indicates time i, and the vertical axis indicates the signal level. The digital audio signal shown in FIG. 14B is a sequence of digital signals for each successive sampling point, and the time (representing numerical value) i increases by 1 for each sampling point.
Since the sample frequency is slightly different from half the frequency of the audio signal, as shown in FIG. 14B, the sample phase is gradually shifted from the phase of the audio signal, and the sample value levels are continuously the same. An interval (from ia to ib) occurs (a signal collapse occurs). The solid line in FIG. 14C indicates the frequency spectrum of the audio signal in FIG. 14B, and the dotted line indicates the frequency characteristics of a low-pass filter (described later). In FIG. 14C, the horizontal axis indicates the frequency f, and the vertical axis indicates the power and gain. As shown in the frequency spectrum of FIG. 14C, since the frequency f8 is near half of the sampling frequency, there are not many harmonics having a frequency higher than f8.

  Considering a single sine wave, the lower the amplitude and the higher the frequency, the more the signal is crushed by quantization, and the same level tends to continue. This phenomenon is particularly noticeable near half the sampling frequency. Further, as shown in FIG. 14C, when the fundamental frequency is near half the sampling frequency, there are not many harmonics, and the effect of removing the harmonics is small.

  The operation of the audio signal processing apparatus of FIG. 1 when a sine wave audio signal X of frequency f8 quantized with n bits as shown in FIG. 14B is input will be described. The audio signal X is input to the frequency amplitude estimation unit 2, the specific frequency detection unit 6, and the original data bit shift unit 4.

FIGS. 15A to 15D are diagrams for explaining the operation of the frequency amplitude estimator 2 when a sine wave audio signal X having a frequency f8 is input. The horizontal axis indicates time i.
15 (a) shows an n-bit audio signal X (i), FIG. 15 (b) shows primary differential data D (i) of the audio signal, FIG. 15 (c) shows binary data PM, FIG. 15D shows a signal SL indicating the detection result of the inflection point position of the audio signal.
When the n-bit audio signal shown in FIG. 15A is input, the primary differential calculation unit 11 calculates primary differential data (Di) as shown in FIG.
The sign change point detector 12 converts the primary differential data D into binary data PM having only a sign as shown in FIG. 15C, and the binary data PM changes as shown in FIG. Positions ic = i4 to i19 (the positions on the time axis are indicated by dotted lines extending in the vertical direction) are detected, and the value of the signal SL is set to “1” at the detection position.
Therefore, the frequency estimation unit 8 and the amplitude estimation unit 9 obtain and obtain F = 1 / (i5-i4) as the frequency and A = | X (i5) −X (i4) | as the amplitude in the i4 to i5 section, respectively. The obtained frequency F and amplitude A are supplied to the filter coefficient generation unit 3. At this time, the obtained F is substantially the same as the frequency f8 of the input signal. Similarly, the frequency F can be accurately obtained in the i5 to i11 interval and the i12 to i19 interval.

  On the other hand, in the section i11 to i12, F = 1 / (i12−i11) as the frequency and A = | X (i12) −X (i11) | as the amplitude, and the obtained frequency F and amplitude A are the filters. It is supplied to the coefficient generator 3. The obtained F is greatly different from the frequency f8 of the input signal.

  FIGS. 16A and 16B are diagrams for explaining the operation of the specific frequency detection unit 6 when the sine wave audio signal X having the frequency f8 is input. The horizontal axis indicates time i. Here, it is assumed that the specific frequency Fd is set in the vicinity of Fs / 2. Since the audio signal X as shown in FIG. 16A is a sine wave of the frequency f8 near the specific frequency Fd, the spectrum intensity R having a large value as shown in FIG. .

  The filter coefficient generation unit 3 receives the spectrum intensity R having a large value, and outputs a coefficient for outputting the input as it is to the filter unit 5.

  FIGS. 17A and 17B are diagrams for explaining the operation of the original data bit shift unit 4 when a sine wave audio signal X having a frequency f8 is input. The horizontal axis indicates time i, and the vertical axis indicates the signal level. FIG. 17A shows an n-bit audio signal X, and FIG. 17B shows an n + α-bit audio signal X ′. The original data bit shift unit 4 bit-shifts the n-bit audio signal as shown in FIG. 17A by α bits, and the n + α-bit audio signal as shown in FIG. Output.

18A and 18B are diagrams for explaining the operation of the filter unit 5 when the sine wave audio signal X having the frequency f8 is input. The horizontal axis indicates time i, and the vertical axis indicates the signal level. 18A shows an n + α-bit audio signal X ′ (i), and FIG. 18B shows a filtered n + α-bit audio signal Y (i).
Since the filter coefficient for outputting the input as it is is input to the filter unit 5, an n + α-bit audio signal as shown in FIG. 18A is output as it is (FIG. 18B).

  As described above, the audio signal processing apparatus of FIG. 1 can output an n + α-bit audio signal obtained by performing only bit shift when a specific frequency is included in the n-bit audio signal.

  Here, the operation when the specific frequency detection unit 6 is not provided will be described with reference to 19 (a) and (b). In FIGS. 19A and 19B, the horizontal axis indicates the frequency f. In the sections i4 to i11 and i12 to i19 in FIGS. 14 (a) to 14 (c) and FIGS. 15 (a) to 15 (d), the frequency is accurately obtained as described above. A low-pass filter coefficient that realizes a cutoff frequency characteristic as shown is generated. On the other hand, since the obtained frequency is incorrect in the sections i11 to i12, a low-pass filter coefficient having a cutoff frequency characteristic is generated as indicated by a dotted line in FIG. Therefore, the frequency f8 is also removed in the i11 to i12 section in the smoothing process.

  On the other hand, when the specific frequency detection unit 6 is provided, in the audio signal processing device of FIG. 1, when the input audio signal includes a high frequency that is crushed by quantization, Although the amplitude estimation unit 2 erroneously detects the frequency, the specific frequency detection unit 6 can accurately detect the frequency, so that the fundamental frequency is not removed.

  As described above, the audio signal processing apparatus of FIG. 1 estimates the frequency and amplitude of the audio signal, generates a low-pass filter coefficient having a cutoff frequency characteristic and an order based on the frequency and amplitude, and smoothes it. Therefore, the waveform distortion due to the quantization of the audio signal can be accurately corrected. On the other hand, since the edge-preserving smoothing filter is applied at the time of smoothing, the reproducibility of the audio signal input waveform having a region where the signal amplitude changes sharply and greatly is not impaired.

  In addition, when the specific frequency is included in the n-bit audio signal, it is possible to output the n + α-bit audio signal that is only bit-shifted.

  The filter unit 5 has been described so far as operating as an ε filter, but the audio signal processing apparatus of FIG. 1 is not limited to this configuration. The same effect can be expected even when the filter operates as another kind of edge-preserving filter such as a bilateral filter or a trimmed average value filter (DW-MTM filter).

  In addition, the specific frequency has been described as being approximately half the sampling frequency so far, but the audio signal processing apparatus in FIG. 1 is not limited to this frequency. For example, if the specific frequency is set as an arbitrary frequency, for example, 10 kHz, a signal including the arbitrary frequency is output without being smoothed.

FIG. 20 is a flowchart showing the processing steps of the audio signal processing apparatus of FIG. 1 described above.
First, an n-bit audio signal is input from the input terminal 1 to the frequency amplitude estimation unit 2, the original data bit shift unit 4, and the specific frequency detection unit 6. The original data bit shift unit 4 outputs an n + α bit audio signal obtained by shifting the n bit audio signal by α bits to the filter unit 5 (ST11). The frequency amplitude estimator 2 estimates the frequency F and the amplitude A from the n-bit audio signal and outputs them to the filter coefficient generator 3 (ST12). The specific frequency detection unit 6 calculates the intensity R of the spectrum of the specific frequency Fd and outputs a signal DR indicating whether or not the intensity R is greater than the threshold value RTH to the filter generation unit 3 (ST13). If intensity R is equal to or less than threshold value RTH (YES in step ST14), the process proceeds to step ST15, and if intensity R is greater than threshold value RTH (NO in step ST14), the process proceeds to step ST17.

In step ST15, a low-pass filter coefficient is generated based on the frequency F and the amplitude A, and is output to the filter unit 5.
In this case, since the low-pass filter coefficient is supplied, the filter unit 5 operates as an edge-preserving smoothing filter, performs low-pass filter processing on the n + α-bit audio signal, and performs n + α-bit processing. An audio signal is output (ST16).
In step ST17, the filter coefficient generation unit 3 generates a filter coefficient that outputs the input as it is.
In this case, the filter unit 5 multiplies the sample value of the target point of the input n + α-bit audio signal by a coefficient “1” and multiplies the sample values before and after that by “0”. The existing audio signal is output as it is (ST18).
The audio signal output from the filter unit 5 in step ST16 or the audio signal output from the filter unit 5 in step ST18 becomes the output of the audio signal processing device.

  The audio signal processing apparatus shown in FIG. 1 can also be realized by software, that is, by a programmed computer. In this case, the processing procedure by the computer is the same as that described above with reference to FIG.

Embodiment 2. FIG.
FIG. 21 shows an audio signal processing apparatus according to the second embodiment. The audio signal processing apparatus shown in FIG. 21 is generally the same as the audio signal processing apparatus according to the first embodiment described with reference to FIG. 1, but the configuration of the selective filter processing unit 10 is different. The selective filter processing unit 10 shown in the figure receives a signal indicating the frequency F and the amplitude A estimated by the frequency amplitude estimation unit 2 and generates a low-pass filter coefficient C, and a bit-shifted speech. Edge-preserving smoothing filter unit 15 that receives signal X ′ as input and outputs an audio signal after filtering using low-pass filter coefficient C; and output signal of edge-preserving smoothing filter unit 15 and bit-shifted And a selection unit 16 that selects and outputs one of the audio signals X ′.

The filter coefficient generation unit 13 causes the edge preserving smoothing filter unit 15 to perform edge preserving smoothing on the bit-shifted speech signal X ′ based on the frequency F and the amplitude A estimated by the frequency amplitude estimating unit 2. A low-pass filter coefficient C for processing is generated.
Such processing of the filter coefficient generation unit 13 is performed when the filter coefficient generation unit 3 shown in FIG. 1 performs the low-pass filter coefficient C, which is performed when the spectrum intensity R of the specific frequency is equal to or less than the threshold value RTH. The filter coefficients can be generated in the same manner as in the first embodiment.

The edge preserving smoothing filter unit 15 outputs the result of performing the filter processing on the bit-shifted audio signal X ′ using the filter coefficient C generated by the filter coefficient generation unit 13.
The selection unit 16 selects the output signal of the edge preserving smoothing filter unit 15 when the intensity R of the spectrum of the specific frequency detected by the specific frequency detection unit 6 is equal to or less than a predetermined threshold value RTH, and the specific frequency detection unit When the intensity R of the spectrum of the specific frequency detected by 6 is larger than the threshold, the bit-shifted audio signal X ′ is selected.
The selective filter processing unit 10 having such a configuration as a whole has the same function as the selective filter processing unit 10 of the audio signal processing apparatus of the first embodiment shown in FIG. This is the same as the first embodiment. Therefore, the audio signal processing device of the second embodiment also has the same effect as the audio signal processing device of the first embodiment.
As the edge preserving smoothing filter unit 15, an ε filter can be used in the same manner as described for the filter unit 5 in the first embodiment.

The operation of the audio signal processing apparatus according to the second embodiment can also be represented by the flowchart of FIG.
First, an n-bit audio signal is input from the input terminal 1 to the frequency amplitude estimation unit 2, the original data bit shift unit 4, and the specific frequency detection unit 6. The original data bit shift unit 4 outputs an n + α bit audio signal obtained by shifting the n bit audio signal by α bits to the edge preserving smoothing filter unit 15 (ST11). The frequency amplitude estimator 2 estimates the frequency F and the amplitude A from the n-bit audio signal and outputs them to the filter coefficient generator 3 (ST12). The specific frequency detection unit 6 calculates the intensity R of the spectrum of the specific frequency Fd and outputs a signal DR indicating whether or not the intensity R is greater than the threshold value RTH to the filter generation unit 3 (ST13). If the intensity R is less than or equal to the threshold value RTH (YES in ST14), the process proceeds to step ST15, and if the intensity R is greater than the threshold value RTH (NO in ST14), the process proceeds to step ST22.

In step ST15, the filter coefficient generation unit 3 outputs the low-pass filter coefficient to the edge preserving smoothing filter unit 15 based on the frequency F and the amplitude A.
In this case, the edge-preserving smoothing filter unit 15 performs edge preserving smoothing filter processing on the n + α-bit audio signal using the supplied filter coefficient, and outputs an n + α-bit audio signal (ST16). ), The selection unit 16 selects the output of the edge-preserving smoothing filter unit 15 (ST21).
In step ST22, the selection unit 16 selects and outputs the output of the original data bit shift unit 4.
As a result of such processing, the audio signal output from the edge preserving smoothing filter unit 15 in step ST16 or the audio signal output from the original data bit shift unit 4 in step ST11 becomes the output of the audio signal processing device. .

  The audio signal processing apparatus shown in FIG. 21 can also be realized by software, that is, by a programmed computer. In this case, the processing procedure by the computer is the same as that described above with reference to FIG.

  The audio signal processing apparatus of the present invention can be applied to an audio signal processing apparatus such as a car audio.

It is a block diagram which shows the structure of the audio | voice signal processing apparatus which concerns on Embodiment 1 of this invention. (A)-(c) is a figure which shows an example of the audio | voice signal processed in Embodiment 1. FIG. (A) And (b) is a figure which shows operation | movement of the original data bit shift part 4. FIG. 4 is a block diagram illustrating an example of a frequency amplitude estimation unit 2. FIG. (A)-(d) is a figure which shows operation | movement of the frequency amplitude estimation part 2 of FIG. (A) And (b) is a figure which shows operation | movement of the specific frequency detection part 6. FIG. (A)-(d) is a figure which shows operation | movement of the specific frequency detection part 6. FIG. (A)-(d) is a figure which shows operation | movement of the specific frequency detection part 6. FIG. (A) And (b) is a figure which shows operation | movement of the specific frequency detection part 6. FIG. These are figures which show the frequency of an audio | voice signal, and the low-pass characteristic suitable for it. 6 is a diagram illustrating an example of a coefficient table stored in a filter coefficient generation unit 3. FIG. (A) And (b) is a figure for demonstrating the operation | movement of an epsilon filter. (A) And (b) is a figure which shows operation | movement of the filter part 5 when it operate | moves as an epsilon filter. (A)-(c) is a figure which shows an example of the n-bit audio | voice signal of Embodiment 1. FIG. (A)-(d) is a figure which shows the operation | movement of the frequency amplitude estimation part 2. FIG. (A) And (b) is a figure which shows operation | movement of the specific frequency detection part 6. FIG. (A) And (b) is a figure which shows operation | movement of the original data bit shift part 4. FIG. (A) And (b) is a figure which shows operation | movement of the filter part 5. FIG. (A) And (b) is a figure for demonstrating operation | movement when the specific frequency detection part 6 is not provided. 3 is a flowchart showing processing steps of the audio signal processing device according to the first embodiment. It is a block diagram which shows the structure of the audio | voice signal processing apparatus which concerns on Embodiment 2 of this invention. 10 is a flowchart showing processing steps of the audio signal processing device according to the second embodiment.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 Input terminal, 2 Frequency amplitude estimation part, 3 Filter coefficient production | generation part, 4 Original data bit shift part, 5 Filter part, 6 Specific frequency detection part, 7 Inflection point detection part, 8 Frequency estimation part, 9 Amplitude estimation part, 11 primary differential calculation unit, 12 sign change point detection unit, 13 filter coefficient generation unit, 15 edge preserving smoothing filter unit, 16 selection unit.

Claims (30)

an original data bit shift unit that shifts an n-bit (n is a positive integer) audio signal by α bits (α is a positive integer) to generate an n + α-bit audio signal;
A frequency amplitude estimation unit that detects an inflection point of the n-bit audio signal and estimates the frequency and amplitude of the audio signal from the inflection point;
A specific frequency detector that determines whether or not the intensity of a spectrum of a predetermined specific frequency of the audio signal is greater than a predetermined threshold;
When the intensity of the spectrum of the specific frequency is determined to be equal to or less than the predetermined threshold by the specific frequency detection unit, the frequency amplitude estimation unit estimates the audio signal output from the original data bit shift unit. Output an audio signal obtained by performing edge preserving smoothing processing using a low-pass filter coefficient generated based on the frequency and amplitude,
A selective filter processing unit that directly outputs the audio signal output from the original data bit shift unit when the specific frequency detection unit determines that the spectrum intensity of the specific frequency is greater than the threshold value. Audio signal processing device. The selective filter processing unit includes:
Based on a determination result signal indicating whether or not the spectrum of the specific frequency is greater than the predetermined threshold by the specific frequency detection unit, and a signal indicating the frequency and amplitude estimated by the frequency amplitude estimation unit. A filter coefficient generation unit for generating a filter coefficient;
A filter unit that performs a filtering process on the audio signal output from the original data bit shift unit and outputs the audio signal after the filtering process;
The filter coefficient generation unit
When the determination result signal indicates that the spectrum intensity of the specific frequency is determined to be less than or equal to the predetermined threshold, edge-preserving smoothing is performed on the audio signal output from the original data bit shift unit. A low-pass filter coefficient for performing the conversion processing,
When the determination result signal indicates that the spectrum intensity of the specific frequency is determined to be greater than the predetermined threshold, the audio signal output from the original data bit shift unit is directly input to the filter unit. Generate filter coefficients for output,
The said filter part performs a filter process with respect to the audio | voice signal output from the said original data bit shift part using the filter coefficient produced | generated by the said filter coefficient production | generation part. Audio signal processing device. The filter coefficient generation unit
When the intensity of the spectrum of the specific frequency is below the threshold,
The audio signal processing apparatus according to claim 2, wherein the order of the low-pass filter coefficient is decreased as the estimated amplitude is increased. The filter coefficient generation unit
As candidates for the low-pass filter coefficient,
The audio signal processing apparatus according to claim 3, further comprising a filter coefficient table that stores a plurality of low-pass filter coefficients having orders different from the cutoff frequency.   The audio signal processing apparatus according to claim 2, wherein the filter unit operates as an ε filter when the low-pass filter coefficient is given. The selective filter processing unit includes:
A filter coefficient generation unit that generates a low-pass filter coefficient based on a signal indicating the frequency and amplitude estimated by the frequency amplitude estimation unit;
An edge-preserving smoothing filter that receives the sound signal output from the original data bit shift unit, performs edge preserving smoothing filter processing using the low-pass filter coefficient, and outputs the sound signal after filtering And
A selection unit that selects and outputs either the output signal of the edge-preserving smoothing filter unit and the audio signal output from the original data bit shift unit;
The selection unit includes:
When the specific frequency detection unit determines that the spectrum intensity of the specific frequency is equal to or less than the predetermined threshold, the output signal of the edge preserving smoothing filter unit is selected,
The audio signal output from the original data bit shift unit is selected when the specific frequency detection unit determines that the spectrum intensity of the specific frequency is greater than the threshold value. The audio signal processing device described. The filter coefficient generation unit
The audio signal processing apparatus according to claim 6, wherein the order of the low-pass filter is decreased as the estimated amplitude is increased. The filter coefficient generation unit
As candidates for the low-pass filter coefficient,
The audio signal processing apparatus according to claim 7, further comprising: a filter coefficient table that stores a plurality of low-pass filter coefficients having different orders from the cutoff frequency.   The audio signal processing apparatus according to claim 6, wherein the edge preserving smoothing filter unit is an ε filter. The frequency amplitude estimator is
An inflection point detection unit that detects a point at which the n-bit audio signal starts or ends a series of monotone increases and a point at which a series of monotonic decreases starts or ends as the inflection points;
A frequency estimation unit that estimates a frequency from the section length of the position of the detected inflection point;
The audio signal processing apparatus according to claim 1, further comprising: an amplitude estimation unit that estimates a level difference of the detected inflection point as an amplitude. The inflection point detector is
A first derivative calculating unit for calculating first derivative data of the n-bit audio signal;
The audio signal processing device according to claim 10, further comprising: a sign change point detection unit that detects, as the inflection point, a point at which a sign of the primary differential data has changed positively and a point at which the sign has changed negatively. .   The sign change point detection unit is configured such that the primary differential data changes from a state where the sign is negative or a state where the value is zero to a state where the sign is positive, and a state or value where the sign is positive. The audio signal processing apparatus according to claim 11, wherein a point where the sign has changed to a negative state is detected as a point where the sign has changed.   The audio signal processing apparatus according to claim 1, wherein the specific frequency for calculating the intensity of the spectrum by the specific frequency detection unit is about half of the sampling frequency. The specific frequency detector is
14. The intensity of the spectrum of the specific frequency is calculated based on a result of superimposing the basic pattern of the cosine wave and sine wave of the specific frequency and the input audio signal. The audio signal processing apparatus according to 1. The specific frequency detector is
The square of the value of the signal obtained as a result of superimposing the basic pattern of the cosine wave of the specific frequency and the input audio signal;
The sum of squares of signal values obtained as a result of superimposing the basic pattern of the sine wave of the specific frequency and the input audio signal is calculated as the intensity of the spectrum of the specific frequency. 14. The audio signal processing device according to 14. an original data bit shift step of shifting an n-bit (n is a positive integer) audio signal by α bits (α is a positive integer) to generate an n + α-bit audio signal;
A frequency amplitude estimating step of detecting an inflection point of the n-bit audio signal and estimating a frequency and an amplitude of the audio signal from the inflection point;
A specific frequency detecting step of determining whether or not the intensity of a spectrum of a predetermined specific frequency of the audio signal is greater than a predetermined threshold;
When it is determined in the specific frequency detection step that the spectrum intensity of the specific frequency is equal to or less than the predetermined threshold, the frequency amplitude estimation step estimates the speech signal bit-shifted in the original data bit shift step. Output an audio signal obtained by performing an edge-preserving smoothing process using a low-pass filter coefficient generated based on the frequency and amplitude,
An audio signal processing method comprising: a selective filter processing step that outputs the n + α-bit audio signal as it is when the intensity of the spectrum of the specific frequency is determined to be greater than the threshold value by the specific frequency detection step. The selective filtering step includes:
Based on the determination result signal indicating whether or not the intensity of the spectrum of the specific frequency is greater than the predetermined threshold by the specific frequency detection step, and the signal indicating the frequency and amplitude estimated by the frequency amplitude estimation step. A filter coefficient generation step for generating a filter coefficient;
A filter step of performing a filtering process on the n + α-bit audio signal and outputting the filtered audio signal;
The filter coefficient generation step includes:
When the determination result signal indicates that the intensity of the spectrum of the specific frequency is determined to be equal to or less than the predetermined threshold value, the edge-preserving smoothing process is performed on the n + α-bit audio signal. Generate the low-pass filter coefficients of
When the determination result signal indicates that the intensity of the spectrum of the specific frequency is determined to be greater than the predetermined threshold, a filter coefficient for causing the filter step to output the n + α-bit audio signal as it is Produces
The audio signal processing method according to claim 16, wherein the filter step performs a filter process on the n + α-bit audio signal using the filter coefficient generated in the filter coefficient generation step. The filter coefficient generation step includes:
When the intensity of the spectrum of the specific frequency is below the threshold,
The audio signal processing method according to claim 17, wherein the order of the low-pass filter coefficient is decreased as the estimated amplitude is increased. The filter coefficient generation step includes:
As candidates for the low-pass filter coefficient,
The audio signal processing method according to claim 18, further comprising: a filter coefficient table that stores a plurality of low-pass filter coefficients having different orders from the cutoff frequency.   18. The audio signal processing method according to claim 17, wherein the filter step performs ε filter processing when the low-pass filter coefficient is given. The selective filtering step includes:
A filter coefficient generation step for generating a low-pass filter coefficient based on the signal indicating the frequency and amplitude estimated by the frequency amplitude estimation step;
An edge-preserving smoothing filter step for performing an edge-preserving smoothing filter process using the low-pass filter coefficient by using the n + α-bit sound signal as an input, and outputting the sound signal after the filtering process;
A selection step of selecting and outputting either the audio signal after the filtering process or the n + α-bit audio signal;
The selection step includes
When it is determined by the specific frequency detection step that the intensity of the spectrum of the specific frequency is equal to or less than the predetermined threshold, the filtered audio signal is selected,
17. The audio signal processing method according to claim 16, wherein the n + α-bit audio signal is selected when it is determined in the specific frequency detection step that the intensity of the spectrum of the specific frequency is greater than the threshold value. . The filter coefficient generation step includes:
The audio signal processing method according to claim 21, wherein the order of the low-pass filter is reduced as the estimated amplitude increases. The filter coefficient generation step includes:
As candidates for the low-pass filter coefficient,
The audio signal processing method according to claim 22, further comprising a filter coefficient table that stores a plurality of low-pass filter coefficients having orders different from the cutoff frequency.   The audio signal processing method according to claim 21, wherein the edge preserving smoothing filter step performs ε filter processing. The frequency amplitude estimating step includes:
An inflection point detecting step of detecting, as the inflection point, a point at which the n-bit audio signal starts or ends a series of monotonic increases and a point at which a series of monotonic decreases starts or ends;
A frequency estimation step of estimating the frequency from the section length of the position of the detected inflection point;
An audio signal processing method according to any one of claims 16 to 24, further comprising: an amplitude estimating step of estimating a level difference of the detected inflection point as an amplitude. The inflection point detecting step includes
A first derivative calculating step for calculating first derivative data of the n-bit audio signal;
The audio signal processing method according to claim 25, further comprising a sign change point detecting step of detecting, as the inflection point, a point at which a sign of the primary differential data has changed positively and a point at which the sign has changed negatively. .   In the sign change point detecting step, the first derivative data is changed from a state where the sign is negative or a value is zero to a state where the sign is positive, and a state or value where the sign is positive is zero. 27. The audio signal processing method according to claim 26, wherein a point at which the sign has changed to a negative state is detected as a point at which the sign has changed.   The audio signal processing method according to any one of claims 16 to 27, wherein the specific frequency for calculating the intensity of the spectrum in the specific frequency detecting step is about half of the sampling frequency. The specific frequency detecting step includes
29. The intensity of the spectrum of the specific frequency is calculated based on a result of superimposing the basic pattern of the cosine wave and sine wave of the specific frequency and the input audio signal. The audio signal processing method according to claim 1. The specific frequency detecting step includes
The square of the value of the signal obtained as a result of superimposing the basic pattern of the cosine wave of the specific frequency and the input audio signal;
The sum of squares of signal values obtained as a result of superimposing the basic pattern of the sine wave of the specific frequency and the input audio signal is calculated as the intensity of the spectrum of the specific frequency. 29. The audio signal processing method according to 29.

Images