JP2002015522A - Audio band extending device and audio band extension method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an audio band extending device, which expands the reproduction band of a digital audio signal whose upper limit of high-pass reproduction is determined by a sampling theorem, realizes improvement in the reproduced sound quality, and can reproduce sounds of satisfactory quality on audibility. SOLUTION: An oversampling type digital low-pass filter 102 performs band limiting of a digital signal inputted from an input terminal 101 and oversamples it. A nonlinear circuit 104, a dither-generating circuit 105, and filters 106 and 107 generate a high-pass signal which is equivalent to or higher than the band of an input signal. Level control circuits 108 and 109 control the level of the high-pass signal, on the basis of the spectral intensity or the like of a high-pass component of the input signal detected in a spectrum analysis circuit 103. Adder circuits 110 and 111 add the level-controlled high-pass signal to the input signal, and output them from an output terminal 111.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an audio band extending apparatus and an audio band extending device for improving the sound quality of an audio signal output from an audio device, particularly, a high sound range, and outputting a comfortable audio signal to a human ear. More particularly, the present invention relates to an audio band extending apparatus and an audio band extending method for processing a digital audio signal in a digital domain.

[0002]

2. Description of the Related Art Audio equipment, particularly equipment for performing digital processing (for example, analog / digital converter, digital / digital converter).
Analog converter, etc. or audio signal recording / reproducing device (eg, analog tape recorder, CD player, MD)
A recording / reproducing device or the like converts a frequency band of an audio signal (also referred to as an acoustic signal or an audio signal) to be processed or recorded / reproduced into a human audible band (for example, 20 Hz to 20 kHz).
Up to). However, conventionally, the audible band of humans was considered to be 20 kHz or less.
It has been found that humans can also sense signal components in the frequency band above 20 kHz. Since the output signal of the above-described device has only a component in a frequency band of 20 kHz or less, it is impossible to accurately reproduce the component originally having a frequency exceeding 20 kHz. However, by adding a pseudo high frequency component of 20 kHz or more to the output signal of the device, the output signal can be made closer to a more natural sound. Therefore, various voice band expansion devices and voice band expansion methods for adding a pseudo high-frequency component to an output signal have been proposed.

[0003] Japanese Patent Application Laid-Open No. 9-36685 discloses an audio signal in which a signal having a spectrum whose frequency exceeds the upper limit of the high frequency range of the reproduction frequency band or the upper limit of the high frequency range of the audible frequency band is added to the audio analog reproduction signal. A playback device and an audio signal playback method are disclosed. Hereinafter, Japanese Patent Application Laid-Open No. 9-3
A conventional audio signal reproducing device and an audio signal reproducing device described in Japanese Patent Publication No. 6685 will be described. FIG. 21 is a block diagram of a conventional audio signal reproducing device. In FIG. 21, 2
101 is an input terminal, 2102 is a buffer amplifier, 210
3 is a filter circuit, 2104 is an amplifier, 2105 is a detection circuit, 2106 is a time constant circuit, 2107 is a noise generator, 2108 is a filter circuit, 2109 is a multiplier, 21
Reference numeral 10 denotes an addition circuit, and 2111 denotes an output terminal.

An audio signal, which is an input signal, is input to an input terminal 210.
1 and input to the buffer amplifier 2102. The amplified output signal of the buffer amplifier 2102 is branched and one is input to the addition circuit 2110, and the other is input to the filter circuit 2103 which is a high-pass filter or a band-pass filter. The filter circuit 2103 controls a specific band (6 kHz to 20 kHz) of the input audio signal.
Hz). Amplifier 2104 amplifies the output signal of filter circuit 2103 to an appropriate level. Time constant circuit 2106 outputs an output signal specifying a time constant. The detection circuit 2105 detects the output signal of the amplifier 2104, smoothes the detected output signal with a time constant corresponding to the output signal of the time constant circuit 2106 (that is, delays the output signal with a constant time constant), and outputs it. The smoothed output signal is input to a multiplier 2109 as a level control signal,
The level of the noise component added to the original audio signal is adjusted.

[0005] The noise generator 2107 outputs a random noise signal. A filter circuit 2108, which is a high-pass filter or a band-pass filter, inputs a noise signal and passes a frequency band of 20 kHz or more. The multiplier 2109 receives the output signal of the filter circuit 2108, multiplies the output signal of the filter circuit 2108 by a level control signal, and outputs the result (that is, the amplification factor is proportional to the level control signal). Adder circuit 211
0 is the output signal of the buffer amplifier 2102 and the multiplier 21
09 and the added signal is output from an output terminal 2111. As described above, the high frequency range is expanded by adding random noise proportional to the output level of the high frequency sound of the original audio signal to the original audio signal.

[0006]

However, the above-mentioned conventional configuration has the following problems. A sound signal existing in the natural world (for example, the sound of a musical instrument) generally has a large signal level of a high-frequency component exceeding the audible band when the signal level of the audible band is large (at the same time). When it is small, the signal level of the high-frequency component exceeding the audible band is also small. Therefore, Japanese Patent Application Laid-Open
The output signal of the conventional audio signal reproducing device described in Japanese Patent Publication No. 36865, that is, an output signal to which a high-frequency component delayed by a certain time is added, lacks the naturalness of raw sound.

In the above-described conventional audio signal reproducing apparatus, a high-frequency component having a level dependent only on the output level of a component included in a certain band of the input signal is added to the output signal. The frequency distribution of the spectrum of the audio signal may be different from each other. Regardless of the sound properties of the input signal, it is not always necessary to add a certain level of high-frequency component to the input signal based solely on the level of the component of the input signal included in a certain band. There is a possibility that the sound quality may be degraded in some cases when the sound does not approach the sound. For example, even when a signal having a single spectrum such as a sine wave is input, the above-described conventional audio signal reproducing apparatus has a problem that signal degradation occurs because random noise is added. Was.

Further, since the above-mentioned conventional audio signal reproducing device is constituted by an analog circuit, there is a problem that its performance varies due to variations of components constituting the circuit or temperature characteristics. . Further, since the above-mentioned conventional audio signal reproducing apparatus is constituted by an analog circuit, there is a problem that the sound quality is deteriorated when the processing of the audio signal by the audio signal reproducing apparatus is repeated. In addition, the above-described conventional audio signal reproducing apparatus has a problem that, because it is constituted by an analog circuit, an attempt to improve the accuracy of a filter as a component increases the circuit scale and leads to an increase in cost. .

An object of the present invention is to provide a voice band extending apparatus and a voice band extending method which solve the above-mentioned conventional problems. SUMMARY OF THE INVENTION The present invention provides a sound band expansion for generating and outputting a more natural sound signal from an original sound signal for sound signals of various characteristics by adding a high frequency component according to the sound characteristics of an input signal (original sound signal). It is an object to provide an apparatus and a voice band extension method. For example, an object of the present invention is to provide a voice band extending apparatus and a voice band extending method in which signal deterioration does not occur even when a sine wave is input.

[0010] The present invention also provides a voice band extending apparatus and a voice signal, in which a circuit is made up of digital parts and digital processing is performed, so that performance variations do not occur due to variations in parts constituting the circuit or temperature characteristics. An object of the present invention is to provide a band extension method. It is another object of the present invention to provide an audio band extending apparatus and an audio band extending method which do not cause sound quality deterioration even if the audio signal processing by the audio signal reproducing apparatus is repeated. It is a further object of the present invention to provide an audio band extending apparatus and an audio band extending method which do not increase the circuit size and do not increase the cost as compared with the analog circuit structure even if the high-accuracy filter is employed. And It is another object of the present invention to provide a voice band extending apparatus and a voice band extending method which achieve the above object with a small circuit scale configuration.

[0011]

According to a first aspect of the present invention, an audio signal of an input band A (A is an arbitrary band) is oversampled at a sampling frequency of twice or more and aliasing noise is reduced. The output signal of the low-pass filter is separated from the low-pass filter to be removed and the output signal of the low-pass filter in units of time TB. A spectrum analyzing means for analyzing a spectrum and outputting an analysis result signal indicating a result of the analysis, a non-linear means for receiving an output signal of the low-pass filter and outputting an output signal subjected to a non-linear processing, a Gaussian distribution and a bell Dither generating means for generating an output signal having a probability density distribution of an arbitrary distribution between the band distribution and a band distribution which is a frequency band higher than the band A
(B is an arbitrary band higher than the band A), and a 1 / f filter according to the analysis result signal.
Characteristics (f is the frequency) and a second filter for changing the frequency characteristic in the region of between up to 1 / f ² characteristic, the output signal of the output signal and said dither generating means of said non-linear means the Filter means for passing the first filter and the second filter, level control means for controlling the output level of the output signal of the filter means in accordance with the analysis result signal, and the filter controlled by the level control means Means for adding an output signal of the means to an output signal of the low-pass filter, and outputting an addition signal.

According to the present invention, the output signal of the non-linear means and the output signal of the dither generation means are added to the input means at an appropriate level and an appropriate frequency characteristic based on the analysis result of the spectrum analysis means. Thereby, since the high-frequency components of the spectrum distribution according to the spectrum distribution of the input signal are added, it is possible to realize a voice band extending apparatus that generates a more natural sound.

The value of "Band A" is arbitrary. For example, the band A is a band of 20 kHz or less. The value of “time TB” is arbitrary. For example, the time TB is 2048 / (44.1k
Hz × 2) = 23 ms (sampling frequency fs
= 44.1 kHz, oversampling rate p = 2, and the number of sampling data for spectrum analysis is 2048). The value of “band B” is arbitrary. For example, band B is 2
It is a band from 0 kHz to 40 kHz. "Dither generation means" means means for generating a random or near random signal. “Probability density distribution” refers to a discrete value graph in which the horizontal axis represents the amplitude level and the vertical axis represents the frequency of occurrence of each amplitude level. The time TB is preferably a sufficiently long time so that the output signal of the dither generation means substantially approaches the probability density distribution with an infinite number of samples. "Distribution between Gaussian distribution and bell-shaped distribution" means, on a graph showing the probability density distribution,
A distribution in which the distribution curve is substantially included in a region between the Gaussian distribution and the bell-shaped distribution (including the Gaussian distribution and the bell-shaped distribution). Preferably, the distribution is a bell-shaped distribution close to a Gaussian distribution (bell-shaped distribution).

"Non-linear means" means means for passing an input signal through a certain non-linear circuit to generate a high-frequency component. The transfer function of the "non-linear means" is arbitrary. Preferably, the non-linear means is a full-wave rectifier circuit or a half-wave rectifier circuit. The output signal of the non-linear means which is a full-wave rectifier circuit to which the input signal sinx is input is as follows.

[0015]

[Formula 1]

Similarly, the output signal of the non-linear means which is a half-wave rectifier circuit to which the input signal sinx is input is as follows.

[0017]

[Formula 2]

From the 1 / f characteristic (f is frequency), 1 / f ²
The "in the region between the up characteristics, on coordinates showing a frequency characteristic, 1 / f characteristic (f is frequency) domain (1 / f characteristic and 1 / f ² characteristic between the 1 / f ² characteristic Is included). The configuration for “changing the frequency characteristic” may be, for example, a configuration in which a 1 / f characteristic filter and a 1 / f ² characteristic filter are switched alternatively. the addition ratio of the output signal of the filter output signal and the 1 / f ² characteristic may be configured to vary little by little.

The "filter means" may, for example, pass the output signal of the non-linear means and the output signal of the dither generation means through filters, and output the output signals of the respective filters separately. The output signal of the non-linear means and the output signal of the bell-shaped dither generating means may be added and then passed through a filter to output the added output signal. The former can determine the characteristics of the filter through which the output signal of the non-linear means passes and the characteristics of the filter through which the output signal of the dither generation means passes independently, so that a spectrum distribution more suitable for the spectrum distribution of the input signal is obtained. High frequency components can be added.
On the other hand, in the latter, since the output signal of the non-linear means and the output signal of the dither generation means are collectively passed through a filter, the circuit scale is reduced, and the frequency characteristic control and the level control can be further simplified.

The filter means may be configured to pass the input signal through the first filter and then through the second filter, or may be configured to pass the input signal through the second filter and then through the first filter. Further, another component (for example, a level control unit) may be interposed between the first filter and the second filter. Further, the first filter and the second filter may be integrated so that they cannot be separated.

The "level control means" may control the level of each output signal added by the addition means as a result. For example, it includes means for controlling the level of the input signal of the non-linear means, the output signal of the non-linear means or the input signal of the filter means, the output signal of the dither generation means or the input signal of the filter means, or the output signal of each filter means.

According to a second aspect of the present invention, the level control means includes a component based on an output signal of the non-linear means in an output signal of the filter means, and a component in the output signal of the filter means. 2. The voice band extending apparatus according to claim 1, wherein a level based on a component based on an output signal of said dither generating means is separately controlled.

According to the present invention, the output signal of the non-linear means and the output signal of the dither generating means are separately and appropriately controlled by the input means based on the analysis result of the spectrum analyzing means. Thereby, since the high-frequency components of the spectrum distribution according to the spectrum distribution of the input signal are added, it is possible to realize a voice band extending apparatus that generates a more natural sound.

"Independent level control" means that the level can be controlled separately. For example, an audio band extending device that allows the user to select that the level control is performed separately from the level control in conjunction, or that the level control is separately performed in conjunction with the level control in accordance with the input audio signal. Includes voice band extender that is automatically selected.

According to a third aspect of the present invention, the first filter is a band-pass filter or a high-pass filter that includes the band B in a pass band, and a cut-off frequency or cut-off on a lower side thereof. 3. The voice band extending apparatus according to claim 1, wherein at least one of frequency characteristics is switched according to the analysis result signal.

According to the present invention, by controlling at least one of the low-frequency cutoff frequency and the cutoff frequency characteristic of the first filter based on the analysis result of the spectrum analysis means, the input signal in the band A and the additional signal At the point where the high-frequency component of the band B contacts (the frequency at which the band A and the band B contact each other, for example, 20 kHz), and the respective frequency distributions (spectral analyzers) sandwiching the contact point. (A gradient on a graph with the frequency axis as the horizontal axis and the level of the frequency component as the vertical axis), which is the result of Fourier expansion, can be performed smoothly. As a result, it is possible to generate an output signal having a frequency characteristic having a spectrum distribution that naturally extends beyond, for example, 20 kHz, and thus has an effect of realizing a voice band extending apparatus that generates a more natural sound.

Preferably, both the cutoff frequency and the cutoff frequency characteristic on the lower frequency side are switched according to the analysis result signal. By switching between the two, disturbance of the frequency characteristics at the point where they are in contact (such as the spectrum (frequency component) taking a maximum value (or a minimum value) at the point) can be suppressed, and a more natural sound is generated. This has the effect of realizing a voice band extending device.

According to a fourth aspect of the present invention, the first filter includes the band A and the band B in a pass band, and a component of the band A included in an output signal of the first filter. The characteristic of the first filter is changed so that the ratio between the energy level of the first filter and the energy level of the component of the band B included in the output signal of the first filter approaches a constant value. Claims 1 to 3
A voice band extending apparatus according to any one of claims 1 to 4.

According to the present invention, the frequency characteristic of the first filter is controlled such that the ratio between the energy level of the component of the band A and the energy level of the component of the band B approaches a constant value. For example, level control is performed so that the level of each frequency component at the point where the input signal of the band A and the high frequency component of the band B added thereto are in contact with each other, and the energy of the component of the band A is controlled. The frequency characteristic of the first filter is changed so that the ratio between the level and the energy level of the component of the band B approaches a constant value as much as possible. As a result, the present invention has an effect that it is possible to realize a voice band extending device that generates a more natural sound.

According to a fifth aspect of the present invention, the non-linear means, the dither generation means, the filter means, the level control means or the addition means may be configured to output the sum signal based on an output signal of the non-linear means. A ratio of an energy level of a component and a component of the addition signal based on an output signal of the dither generation means in at least one of the band A and the band B is changed according to the analysis result signal. An audio band extending apparatus according to any one of claims 1 to 4.

According to the present invention, the band A between the component based on the output signal of the non-linear means and the component based on the dither generating means is used.
The energy level ratio in (or band B or bands A and B) is changed according to the analysis result signal. For example, the energy levels of the input signals in the low frequency band and the high frequency band, which are two frequency bands included in the band A, are analyzed (analysis by the spectrum analysis unit). Next, the ratio of the energy level of the signal obtained by adding the output signal of the non-linear means and the output signal of the dither generating means in two bands (low-frequency band and high-frequency band) is equal to the energy level ratio of the input signal. The addition ratio of the output signal of the non-linear means and the output signal of the dither generation means is changed so as to match.

The frequency component of the band B of the output signal of the non-linear means and the output signal of the dither generation means whose addition ratio is controlled is added to the input signal. Accordingly, the present invention can generate a pseudo signal having a spectrum distribution similar to the spectrum distribution of the input signal, and add a high-frequency component of the pseudo signal to the input signal to generate a more natural sound. This has the effect that an extension device can be realized.

[0033]

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram showing a preferred embodiment of the present invention. << Embodiment 1 >> FIG. 1 is a block diagram of a voice band extending apparatus according to Embodiment 1 of the present invention. In FIG. 1, 101 is an input terminal, 102 is an oversampling LPF (oversampling low-pass filter), 103 is a spectrum analysis circuit, 104 is a non-linear circuit, 105 is a dither generation circuit, 106 is a filter, 107 is a filter, 108 is a level control circuit, 109 is a level control circuit, 110 is an addition circuit, 111 is an addition circuit, 112 is an output terminal, 113
Is a delay circuit.

The voice band extending apparatus of the first embodiment is
By adding a high-frequency component of 20 kHz to 40 kHz (corresponding to band B) to an input signal in a band of 20 kHz (corresponding to band A), an audio signal having a more natural sound is output. The input signal is a digital audio signal having a sampling frequency fs, and in the processing after the oversampling circuit in the oversampling type low-pass filter 102, all sampling clocks for the signal processing are p · fs (the dither generation circuit 105 and the like). Also included). In this embodiment, fs = 44.1 kHz.
z, p = 2, and p · fs = 88.2 kHz.

FIG. 2 schematically shows the frequency characteristics of the spectrum intensity of the input signal and FIG. 3 shows the frequency characteristics of the spectrum intensity of the output signal of the voice band extending apparatus of the first embodiment. The audio band extending apparatus adds an extended signal (high-frequency component of band B) to the input audio digital signal (FIG. 2) so as to be smoothly connected on a spectrum characteristic (FIG. 3). That is, the spectrum of the digital signal output from the audio signal band extending apparatus of the present embodiment has a spectrum 20 at the highest frequency within the band A of the input digital signal and a spectrum 20 at the highest frequency within the band B of the added extended signal. Spectrum 21 at frequency
After that, the slope 22 of the spectrum in the band B is made continuous like the slope of the spectrum in the band A.

A digital audio signal is input through the input terminal 101. This signal is a compact disc (CD)
From the sampling frequency (f
s) This is a signal of 44.1 kHz and a word length of 16 bits. The oversampling low-pass filter 102 receives a signal input to the input terminal 101, multiplies the sampling frequency of the input signal by p (p is a positive integer), and attenuates unnecessary bands. p is 2 or more (usually an even number), and attenuates a band of 1/2 or more of the sampling frequency fs of the input signal by 60 dB or more.

In this embodiment, the oversampling type low-pass filter 102 has an oversampling rate p = 2 (fs = 44.1 kHz). That is, the input signal is oversampled at a frequency of 2fs = 88.2 kHz. For example, in a case where a digital-to-analog converter is connected after the audio band extender of the embodiment and a digital-to-analog converted audio signal is output, an oversampling rate p that is a larger value may be adopted (for example, p = 64). In such a case, the band B of the extended audio signal is narrower than fs / 2 to pfs / 2, for example, fs / 2 to fs. The oversampling signal whose high frequency has been cut is transmitted to the delay circuit 113, the spectrum analysis circuit 103, and the nonlinear circuit 104.

The spectrum analysis circuit 103 receives the oversampling signal from which the high frequency band has been cut, analyzes the spectrum thereof, and outputs an analysis result signal. The analysis result signal is supplied to the filters 106 and 107, the level control circuit 108
And 109. The non-linear circuit 104 receives the oversampling signal from which the high frequency has been cut, performs non-linear processing on the input signal, and outputs a signal containing a high-frequency component.
The output signal of the nonlinear circuit 104 is input to the filter 106.

The filter 106 receives the output signal of the nonlinear circuit 104 and the analysis result signal and passes the output signal of the nonlinear circuit 104 to a filter selected based on the analysis result signal. Specifically, the filter 106 includes the nonlinear circuit 1
04 cuts the frequency component of fs / 2 or less of the output signal (high-pass filter), and the spectrum analysis circuit 103
The output signal of the nonlinear circuit 104 is passed through a filter (low-pass filter) having a 1 / f characteristic or a 1 / f ² characteristic in a band of fs / 2 or more based on the output signal (analysis result signal). The level control circuit 108
6 and the analysis result signal, and amplifies the output signal of the filter 106 at an amplification factor selected based on the analysis result signal.

The dither generation circuit 105 generates a noise signal (bell-shaped dither signal) having a bell-distribution (bell-shaped) probability density. The filter 107 receives the noise signal and the analysis result signal, and passes the noise signal to a filter selected based on the analysis result signal. Specifically, the filter 107 cuts a frequency component of fs / 2 or less from the output signal of the dither generation circuit 105 (high-pass filter), and outputs the signal (analysis result signal) of the spectrum analysis circuit 103.
And outputs the output signal of the dither generation circuit 105 through a 1 / f characteristic or 1 / f ² characteristic filter (low-pass filter) in a band equal to or higher than fs / 2. The level control circuit 109 receives the output signal of the filter 107 and the analysis result signal, and amplifies the output signal of the filter 107 at an amplification factor selected based on the analysis result signal.

The adder circuit 110 includes a level control circuit 108
And 109 are input, added, and the result of addition is output. The delay circuit 113 receives the oversampling signal from which the high frequency has been cut, delays the signal, and outputs the delayed signal. The path where the oversampling signal whose high frequency has been cut is input to the adder 111 through the spectrum analysis circuit 103, the nonlinear circuit 104, the adder 110 and the like, and the oversampling signal whose high frequency has been cut out passes through the delay circuit 113. The delay time in the delay circuit 113 is set so that the delay time with the path input to the adder 111 is equal.

The adder circuit 111 adds the output signal of the delay circuit 113 and the output signal of the adder 110, and outputs an addition result. The output terminal 112 outputs the addition result of the adder 111. As described above, the voice band extending apparatus according to the present embodiment generates a high frequency having a spectrum equal to or higher than the band of the input signal, and converts the generated high frequency component to the input signal in accordance with the high frequency spectrum intensity of the input signal. The voice band is extended by adding.

Hereinafter, each block will be described in detail.
The oversampling low-pass filter 102 (p = 2 in the embodiment) includes an oversampling circuit and a digital low-pass filter. First, the oversampling circuit performs the sampling cycle TS / 2 (=
An audio digital signal having a sampling frequency of 2 fs is generated by alternately outputting an input signal and zero data at 1 / (2 fs)).

Next, the audio digital signal having a sampling frequency of 2 fs is passed through a digital low-pass filter.
The digital low-pass filter has (a) frequency 0 to 0.4
5 fs pass band and (b) frequency 0.54 fs to fs
And (c) an attenuation of 60 dB or more at a frequency of fs or more. Digital low-pass filters
The input audio digital signal is low-pass filtered to remove aliasing noise generated by the oversampling process. With the above configuration, the oversampling type low-pass filter 102 oversamples the input signal at twice the frequency, and substantially converts only the effective band (frequency 0 to 0.45 fs) of the input audio digital signal. Let it pass.

FIG. 4 shows a configuration of the spectrum analysis circuit 103 in the first embodiment. The spectrum analysis circuit 103 includes an FFT circuit (fast Fourier Transf).
orm circuit) 41, a data selection circuit 42, and a weighting addition circuit 43. The FFT circuit 41
The input audio digital signal is subjected to fast Fourier transform processing using an arithmetic method. For example, if the frequency resolution number is 1024
Then, based on the data of every 2048 pieces, that is, the unit time TB of the spectrum analysis = 2048 / (p · fs) = 2
048 / (2 × 44.1 kHz) = Calculate 1024 spectral intensities every 23.2 ms and output to the data selection circuit 42 (sampling frequency fs = 44.1 k)
Hz. Oversampling rate p = 2).

Next, the data selection circuit 42, based on the input spectrum intensity for each frequency fs / 1024, for example, frequency fs / 4 to 1.26fs / 4 (11k
Hz to 13.9 kHz), data of the spectrum intensity corresponding to the band of frequency 1.26 fs / 4 to 1.59 fs /
4 (13.9 kHz to 17.5 kHz), data of the spectrum intensity corresponding to the band, and a frequency of 1.59 fs /
The spectrum intensity data corresponding to the band of 4 to 2 fs / 4 (17.5 kHz to 22 kHz) is selectively extracted and output to the weighting addition circuit 43. 1.26 = 3rd root of 2; 1.59 = (3rd root of 2).

The weighting and adding circuit 43 multiplies the extracted data of each spectrum intensity by a predetermined weighting coefficient and adds each data resulting from the multiplication to obtain a frequency fs / 4 to 1.26f
s / 4 (11 kHz to 13.9 kHz) entire spectrum intensity (energy level) L1, frequency 1.2
6 fs / 4 to 1.59 fs / 4 (13.9 kHz to 1
7.5 kHz) entire spectrum intensity (energy level) L2 and frequency 1.59 fs / 4 to 2 fs
Calculate and output the spectrum intensity (energy level) L3 of the entire band of / 4 (17.5 kHz to 22 kHz). The spectral intensities L1, L2, and L of these three bands
Based on the size and ratio of 3, filters 106 and 1
07 and the level control circuits 108 and 109 are controlled.

Calculate L2 / L1 and L3 / L2. This calculation result is expressed by four attenuation curves (a frequency distribution of the spectrum intensity. The spectrum intensity is attenuated as the frequency becomes higher), that is, −6 dB in Expression 1 (or Expression 2) which is the output signal of the nonlinear circuit 104. / Oct (the first frequency characteristic of the filter 106 described later), and the attenuation curve obtained by multiplying Equation 1 (or Equation 2) by -12 dB / oct (the second curve of the filter 106 described later).
Frequency characteristic), an attenuation curve obtained by multiplying the output signal of the dither generation circuit 105 (FIG. 10) by −6 dB / oct (a first frequency characteristic of the filter 107 described later), and an output signal of the dither generation circuit 105. 12dB / oct
In the same band (fs / 4 to 1.26) of the attenuation curve (second frequency characteristic of the filter 107 described later)
fs / 4 band, 1.26fs / 4 to 1.59fs / 4
, And the ratios of the respective spectral intensities (L2 / L1 and L2 / L1 / 4).
3 / L2).

The spectrum analysis circuit 103 calculates the ratio (L2 / L1 and L2 / L1) of the spectrum intensities of the four attenuation curves.
3 / L2). The spectrum analysis circuit 103 selects an attenuation curve closest to L2 / L1 and L3 / L2 among the above four attenuation curves,
The output signal (the nonlinear circuit 104 or the dither generation circuit 105) having the attenuation curve is controlled to be output.
By comparing the spectral intensities of the three bands, 4
From these attenuation curves, it is possible to select an optimal attenuation curve that approximates the first and second derivative characteristics of the attenuation curve of the spectrum intensity of the input signal. Band A
The spectrum intensity of the input signal at the contact point (for example, 20 kHz) between the (input signal band) and the band B (band of the added high-frequency component), and the selected output signal (non-linear circuit 104 or dither generation circuit 105) at the contact point Level control circuit 108 so that the spectrum intensity of the
Alternatively, the amplification factor of 109 is controlled.

An example in which the spectrum analysis circuit 103 decides to use an attenuation curve obtained by multiplying the output signal of the nonlinear circuit 104 by -6 dB / oct will be described. The spectrum analysis circuit 103 controls the output signal of the non-linear circuit 104 to pass through a filter of −6 dB / oct of the filter 106, and the adder 111 controls the spectrum intensity of the input signal at 20 kHz and the output signal of the filter 106. Level control circuit 10 so that the spectrum intensity matches
8 is controlled. The amplification factor of the level control circuit 109 is controlled to zero.

In the above embodiment, the optimum attenuation curve is selected from the four attenuation curves. However, in the voice band extending apparatus of another embodiment, the nonlinear circuit 1 is selected.
04 and the output signal of the dither generation circuit 105 are passed through filters having four attenuation curves. The output signals of the four filters are each multiplied by an appropriate coefficient and then added (generating an added signal). The above four coefficients are determined so that the added signal has a frequency characteristic approximating the attenuation curve of the input signal. By adding the added signal (having a high-frequency component) to the input signal, the voice band extending apparatus outputs an output signal giving a natural impression.

For example, in FIG. 1, the spectrum analysis circuit 103 outputs
06 so as to pass through a -6 dB / oct filter, and the output signal of the dither generation circuit 105 is controlled by the filter 10.
7 so that the attenuation curve of the synthesized high-frequency component approximates the attenuation curve of the input signal.
The amplification factors of the level control circuits 108 and 109 are controlled so that the spectrum intensity of the input signal at 0 kHz matches the spectrum intensity of the added high frequency component.

FIG. 5 is a block diagram showing a configuration of a spectrum analysis circuit according to another embodiment. 5 includes a high-pass filter 51 and an absolute value calculation circuit 52.
, A low-pass filter 53, a subtractor 54, a low-pass filter 55, an absolute value calculation circuit 56, a low-pass filter 57, and a determination circuit 58. In FIG. 5, the oversampling type low-pass filter 1 of FIG.
The audio digital signal that has been low-pass filtered at 02 is input to a high-pass filter 51 and a subtractor 54. The high-pass filter 51 receives the output signal of the oversampling-type low-pass filter 102 and outputs the frequency fs / 4 to fs.
After high-pass filtering to pass only the band component of s / 2 (11 kHz to 22 kHz), the signal after high-pass filtering is input to the absolute value calculation circuit 52.

The absolute value calculation circuit 52 receives the output signal of the high-pass filter 51, performs full-wave rectification, and outputs a rectified signal. The low-pass filter 53 receives the rectified signal, performs time integration, and outputs a smoothed signal yah. The smoothed signal yah has a frequency fs / 4 to fs / 2 (11 kHz to 2 kHz) of the input audio digital signal.
2 shows a spectrum intensity in a band of 2 kHz). The signal yah indicating the spectrum intensity is input to the determination circuit 58. On the other hand, the subtracter 54 receives the output signal of the oversampling type low-pass filter 102, subtracts the output signal of the high-pass filter 51, and outputs a signal of the subtraction result.

The low-pass filter 55 receives the result of the subtraction, passes a component in a frequency band of 0 to fs / 4 (0 to 11 kHz), and outputs the signal. The absolute value calculation circuit 56 inputs the extracted components in the frequency band from 0 to fs / 4, rectifies and outputs. The low-pass filter 57 receives the rectified signal, performs time integration, and outputs a smoothed signal yal. The smoothed signal yal indicates the spectrum intensity of the input audio digital signal in the frequency band of 0 to fs / 4 (0 to 11 kHz). The signal yal indicating the spectrum intensity is input to the determination circuit 58.

Then, the judgment circuit 1408 compares the spectrum intensity yal of the input audio digital signal at frequencies 0 to fs / 4 with the spectrum intensity yah at frequencies fs / 4 to fs / 2, as follows. Filters 106 and 107 and level control circuits 108 and 1
09 is controlled. (A) when the spectrum intensity yal is equal to or higher than a predetermined threshold level and the spectrum intensity yah is lower than the threshold level, or (b) when the spectrum intensity yal is lower than the predetermined threshold level and the spectrum intensity yah is equal to the predetermined threshold level. When the above is true, the level control circuits 108 and 109
To 0. That is, the output signals of the nonlinear circuit 104 and the dither generation circuit 105 are not added to the input signal.

On the other hand, in cases other than the above (a) and (b),
The amplification factors of the level control circuits 108 and 109 are set to values other than 0, and the output signals of the non-linear circuit 104 and the dither generation circuit 105 are oversampled by a low-pass filter 10.
2 is added to the output signal. That is, the input audio digital signal has a frequency band of 0 to fs / 4 and a frequency f
When each of the two bands of s / 4 to fs / 2 has a spectrum intensity equal to or greater than a predetermined threshold, the output signals of the nonlinear circuit 104 and the dither generation circuit 105 are added to the input signal, and the input audio digital signal is output. The bandwidth of the signal is extended.

On the other hand, if the spectrum intensity yal is equal to or higher than the predetermined threshold level and the spectrum intensity yah is lower than the predetermined threshold level, the band component of the frequency fs / 4 to fs / 2 does not substantially exist. If not expanded, the frequency spectrum distribution is more natural. Also, the frequency fs / 4
The fact that the band component of fs / 2 substantially does not exist can be determined that the input audio digital signal is a signal of a single spectrum in a band of fs / 4 or less or a non-sound signal generated intentionally. Therefore, in such a case, the amplification factors of the level control circuits 108 and 109 are set to 0 as described above.

When the spectrum intensity yal is less than the predetermined threshold level and the spectrum intensity yah is equal to or higher than the predetermined threshold level, it is considered that there is no fundamental component and only harmonic components. That is, it is determined that the input audio digital signal is not a musical sound but a high-frequency single-spectrum sound or an intentionally generated non-musical sound. Therefore, in such a case, as described above, the level control circuits 108 and 109
To 0. Thus, when a single spectrum signal or a non-tone signal is detected, the voice band extending apparatus outputs the input signal of FIG. 2 as it is.

In cases other than the above (a) and (b), yah
An optimum attenuation curve is selected from four attenuation curves (frequency distribution of spectrum intensity) based on the ratio of / yal. That is, Equation 1 which is an output signal of the nonlinear circuit 104
Attenuation curve (filter 106) obtained by multiplying (or Expression 2) by -6 dB / oct, and Expression 1 (or Expression 2) by-
Attenuation curve (filter 106) multiplied by 12 dB / oct, −6 dB to the output signal of dither generation circuit 105
/ Oct multiplied by an attenuation curve (filter 10
7) and -12d is applied to the output signal of the dither generation circuit 105.
B / oct multiplied by an attenuation curve (filter 10
Among 7), an attenuation curve having a frequency characteristic closest to the ratio of yah / yal is selected. Judgment circuit 58
Is the ratio of the spectral intensities of the four attenuation curves (y
ah / yal).

The decision circuit 58 selects an attenuation curve closest to yah / yal among the above-mentioned four attenuation curves, and outputs an output signal having the attenuation curve (non-linear circuit 10).
4 or the dither generation circuit 105). That is, the judgment circuit 58 is, for example, the dither generation circuit 10
When the output signal of No. 5 is selected, the adder 111 controls the amplification factor of the level control circuit 109 so that the spectrum intensity of the input signal at 20 kHz and the spectrum intensity of the added high-frequency component coincide with each other. The amplification factor of the level control circuit 108 for amplifying the output signal of the circuit 104 is 0
To control.

In the above embodiment, the spectrum analysis circuit 6 calculates the spectrum intensities of the two bands to determine whether the input audio digital signal is a single spectrum or a non-tone signal. . Similarly, the spectrum analysis circuit 103 of the first embodiment calculates each spectrum intensity of the three bands to determine whether the input audio digital signal is a single spectrum or a non-tone signal. I can do it. In the above embodiment, the audio signal band extending device is configured by a digital signal processing circuit of hardware. However, the present invention is not limited to this. For example, the configuration of FIG. 1, FIG. 4 or FIG. It can also be realized by software using a processing program. Further, the signal processing program may be executed by hardware using a DSP (Digital Signal Processor).

The non-linear circuit 104 has non-linear input / output characteristics. The non-linear circuit 104 performs non-linear processing on the input output signal of the oversampling type low-pass filter 102 to distort the audio digital signal to generate harmonics. And outputs an audio digital signal having a harmonic component. FIG. 6 illustrates the configuration of the nonlinear processing circuit 104. 6, the non-linear circuit 104 includes an absolute value calculation circuit 61.
And a DC offset removal circuit 62. The DC offset removal circuit 62 includes a subtractor 63 and an averaging circuit 6
4, a 1/2 multiplier 65, and a delay circuit 66.

The absolute value calculation circuit 61 performs a nonlinear process such as a full-wave rectification process on the input audio digital signal, and then applies the nonlinear digital audio signal to the delay circuit 66 and the averaging circuit 64. Output to The absolute value calculation circuit 61 outputs a signal having a positive amplitude as it is, and converts a signal having a negative amplitude into a positive amplitude having the same absolute value as the negative amplitude and outputs the same. For this reason, a harmonic component is generated when a signal having a negative amplitude is turned to the positive side at the boundary of the zero level. For example, the output signal of the absolute value calculation circuit 61, which is a full-wave rectifier circuit to which the input signal sinx is input, is represented by Expression 1 above. Similarly, the output signal of the absolute value calculation circuit 61, which is a half-wave rectification circuit to which the input signal sinx is input, is represented by the above-described equation (2).

The averaging circuit 64 receives and smoothes the output signal of the absolute value calculation circuit 61 and outputs it. Averaging circuit 64
Is a very low cut-off frequency compared to the sampling frequency fs, for example, about 0.001 fs (fs = 44.1 k
Hz, a low-pass filter having a cutoff frequency of about 2.3 ms). Specifically, the averaging circuit 6
4 is a sampling period (Ts = 1 / fs = 1 / 44.1).
kHz) for a sufficiently long period (eg, 0.001 kHz).
An average value of the amplitude of the output signal of the absolute value arithmetic circuit 61 during a period of about fs) is calculated, and the average value is output.

The 乗 算 multiplier 65 inputs the average value,
The value of the multiplication result is output by multiplying by 値 of the average value. The output signal of the 1/2 multiplier 65 is supplied to the absolute value arithmetic circuit 6
1 corresponds to the DC offset amount of the output signal. The output signal of the 乗 算 multiplier 65 is input to the subtractor 63. The delay circuit 66 receives, delays, and outputs the output signal of the absolute value calculation circuit 61. The processing time of the path in which the output signal of the absolute value operation circuit 61 is input to the subtractor 63 via the averaging circuit 64 and the 乗 算 multiplier 65 and the output signal of the absolute value operation circuit 61 The delay time of the delay circuit 66 is set so that the delay time of the path input to the subtractor 63 via the relay circuit coincides. The subtractor 63 removes the DC offset of the output signal of the absolute value calculation circuit 61 by subtracting the output signal of the 乗 算 multiplier 65 from the output signal of the absolute value calculation circuit 61, and outputs a subtraction result.
In the embodiment, the absolute value calculation circuit 61 is used for the configuration of the nonlinear circuit 104. However, the same effect can be obtained by a circuit that passes only a signal having a positive amplitude and makes a signal having a negative amplitude zero. Can be

Since the digital audio signal input to the input terminal 101 is a signal based on a zero level (does not include a DC component), the signal output from the output terminal 112 is also configured so as not to include a DC component. I have. Nonlinear circuit 10
4 (the output signal of the oversampling type low-pass filter 102) is a signal based on the zero level, but the absolute value calculation circuit 61 performs nonlinear processing on the input signal to the nonlinear circuit 104 (absolute value). Is calculated and output.), A DC offset occurs. Therefore, the DC offset included in the output signal of the absolute value calculation circuit 61 is removed by the subtractor 63.

A digital signal containing a harmonic component generated by the input of the output signal of the oversampling type low-pass filter 102 by the nonlinear circuit 104 is input to the filter 106 as shown in FIG. Filter 106
As will be described later, only high-frequency components having a frequency of at least about fs / 2 in the input digital signal are high-pass filtered and output.

The dither generation circuit 105 shown in FIG. 1 generates an audio digital signal having a random amplitude level with respect to the time axis. That is, the dither generation circuit 105 outputs a dither signal generated without correlation with the audio digital signal input to the input terminal 101. Dither signal is filter 1
07. The filter 107, as described later,
Only high-frequency components having a frequency of at least fs / 2 in the input dither signal are high-pass filtered and output.

The dither signal generation circuit 105 has the structure shown in FIG. 7 in the first embodiment. In FIG. 7, a dither signal generation circuit 105 includes a plurality of N pseudo noise sequence noise signal generation circuits (hereinafter referred to as “PN sequence noise signal generation circuits”) 70-n (n = 1, 2,...). N (N
Is an arbitrary integer of 2 or more)), an adder 71, a delay circuit 72, a DC offset removal constant signal generator 73,
And a subtractor 74.

Each PN sequence noise signal generation circuit 70-n
Generates a pseudo noise signal having uniformly random amplitude levels, for example, an M-sequence noise signal, having initial values independent of each other. The output signal of each PN sequence noise signal generation circuit 70-n is input to the adder 71. The adder 71 includes a plurality of PN sequence noise signal generation circuits 70-n
(N = 1, 2,..., N), the N pseudo noise signals that are output signals are added, and a sum (a dither signal) of the pseudo noise signals that are the addition results is output. The output signal of the adder 71 (sum of pseudo noise signals; dither signal) is supplied to a delay circuit 72.
And a DC offset removal constant signal generator 73.

DC offset removing constant signal generator 73
Inputs the output signal (sum of the pseudo noise signal) of the adder 71, and is preferably a certain period (the certain period is a period sufficiently longer than the output clock of the adder 71. For example, 1000 / fs). = 22.7 ms.) Is calculated as の of the average value of the output signal of the adder 71. 1/2 of the average value corresponds to the DC offset value of the output signal of the adder 71 (the sum of the pseudo noise signals). The DC offset removal constant signal generator 73 outputs 1/1 of the average value.
2 (DC offset value; DC offset removal constant signal) is output. The half of the average value is input to the subtractor 74.

The delay circuit 72 receives the output signal of the adder 71 (sum of pseudo noise signals; dither signal), delays the signal, and outputs the delayed signal. The delayed output signal is input to the subtractor 74. The delay time of the delay circuit 72 is equal to the delay time of the signal in the DC offset removal constant signal generator 73 (most of the delay time is generated by the average value calculation circuit). This eliminates the time difference between the output signal of the adder 71 and the output signal of the DC offset removal constant signal generator 73. The subtractor 74 subtracts the DC offset removal constant signal from the sum of the pseudo noise signals output from the delay circuit 74 to generate and output a dither signal having no DC offset.

FIG. 8 shows each PN sequence noise signal generating circuit 7.
The configuration of 0-n (n = 1, 2,..., N) is illustrated.
In FIG. 8, a PN sequence noise signal generation circuit 70-n
Comprises a 32-bit shift register 81, an exclusive OR gate 82, a clock signal generator 83, and an initial value data generator 84. 32-bit shift register 8
After the initial value data generator 84 sets different initial values for each of the PN sequence noise signal generating circuits 70-n by the initial value data generator 84, the initial value is shifted by one step by the clock signal generated by the clock signal generator 83. .

The 32-bit shift register 81 (0 to 31)
Bit) (MSB, 31st bit)
Is input to the input terminal of the exclusive OR gate 82.
Based on the clock signal from the clock signal generator 83, the exclusive OR gate 82 outputs 1-bit data that is the result of the exclusive OR operation, and outputs the least significant bit (LSB) of the 32-bit shift register 81. 0th bit)
To enter. The lower 8 bits of data (0th to 7th bits) of the 32-bit shift register 81 are output as a PN sequence noise signal. By configuring the PN sequence noise signal generation circuit 70-n in this manner, each PN
The PN series noise signal output from the series noise signal generation circuit 70-n is an independent 8-bit PN series noise signal.

In the example of FIG. 8, each PN-sequence noise signal generating circuit 70-n generates an 8-bit PN-sequence noise signal independent of each other by the above-described configuration. However, the present invention is not limited to this. It may be configured as (1), (2) or (3). (1) For each PN sequence noise signal generation circuit 70-n,
From the output signals at different bit positions of the 32-bit shift register 81 (output signals of different D flip-flops)
The bit data is extracted to generate a PN sequence noise signal. That is, the PN sequence noise signal generation circuit 70-1 extracts the PN sequence noise signal of 8 bits from the least significant 8 bits, and the PN sequence noise signal generation circuit 70-2 extracts the PN sequence noise from the 8 bits immediately above the least significant 8 bits. The PN sequence noise signal is extracted in the same manner as described above. (2) The bit positions of the 32-bit shift register 81 for extracting the input signal to the exclusive OR gate 82 are made different from each other in each PN sequence noise signal generation circuit 70-n. (3) The above (1) and (2) are combined.

The adder 71 (FIG. 7) adds a plurality of PN sequence noises independent of each other, and
Alternatively, a PN sequence noise signal having a probability density with respect to the amplitude level as shown in FIG. 11 can be generated. For example, when N = 1 (only one PN sequence noise signal generation circuit), a white noise signal having a probability density of a uniform distribution with respect to the amplitude level is generally generated as shown in FIG. . When N = 12 (12 PN sequence noise signal generation circuits), the respective PN sequence noise signals from the PN sequence noise signal generation circuit 70-n that generates 12 uniform random numbers are added. As a result, as shown in FIG. 11, a Gaussian noise signal having a probability density of Gaussian distribution with respect to the amplitude level is generated. This is because the variance of the Gaussian distribution becomes 1 / (√12) by the central limit theorem.

Further, when N = 3, as shown in FIG. 10, the variance is close to the Gaussian distribution and slightly larger than the Gaussian distribution. A bell distribution type (bell-shaped) noise signal is generated. As described above, FIGS.
With the circuit configuration described above, for example, the noise signal of FIG. 10 or FIG. 11 can be generated. In this way, a small-scale circuit can generate a dither signal close to a natural sound or a musical sound signal.

FIG. 12 shows a detailed block diagram of the filters 106 and 107 (FIG. 1). The filters 106 and 107 have the same configuration. In FIG. 12, 12
01 is a high-pass filter, 1202 is a 1 / f characteristic filter (first frequency characteristic filter), 1203 is 1 / f
^2- characteristic filter (filter of second frequency characteristic), 12
04 is a switch.

FIG. 13A shows a 1 / f characteristic filter 120.
2 illustrates a 1 / f frequency characteristic (−6 dB / oct). The 1 / f characteristic filter 1202 has a band A of the input signal.
(The cutoff frequency is fs / 2. In the embodiment, fs = 4
(4.1 kHz), and has a 1 / f attenuation characteristic in a frequency higher than a cutoff frequency. Assuming that the oversampling frequency of the oversampling type low-pass filter 102 is pfs (p = 2 in the embodiment), the band through which the signal passes with an attenuation characteristic of 1 / f is a band B (fs / 2 to pfs / 2). is there.
At frequencies higher than pfs / 2, it has an attenuation of 60 dB or more.

FIG. 13B shows 1 / f²Characteristic filter 12
03 1 / f²The characteristics (-12 dB / oct) are shown.
You. 1 / f²The characteristic filter 1203 has a band A of the input signal.
A low-pass filter having the same pass band as
1 / f above the cutoff frequency ²It has the following damping characteristics.
1 / f²The band through which the signal passes with the attenuation characteristic of band B (fs /
2 to pfs / 2). at frequencies higher than pfs / 2
Has an attenuation of 60 dB or more.

As described above, the spectrum analysis circuit 103 outputs the spectrum intensity (for example, fs / 4 to fs /
2), and whether the high-frequency spectrum structure is closer to the 1 / f characteristic or the 1 / f ² characteristic. Details of the operation of the filters 106 and 107 will be described with reference to FIGS. High pass filter 1201
Removes the components of the output signal of the nonlinear circuit 104 and the dither generation circuit 105 in the band of fs / 2 or less. The high-pass filter 1201 corresponds to the first filter described in the claims.

The output signal of the high-pass filter 1201 is
The signals are input to the 1 / f characteristic filter 1202 and the 1 / f ² characteristic filter 1203. The switch 1204 switches between the output of the 1 / f characteristic filter 1202 and the output signal of the 1 / f ² characteristic filter 1203 in accordance with the output signal (analysis result signal) of the spectrum analysis circuit 103, and outputs one selected output signal. I do. 1 / f characteristic filter 12
02, 1 / f ² characteristic filter 1203 and switch 1
204 corresponds to the second filter in the claims.

As described above, in the voice band extending apparatus according to the first embodiment of the present invention, a non-linear circuit, a dither generating circuit and a filter generate a high frequency higher than the band of the input signal, and the spectrum analyzing circuit and the level control circuit input the high frequency. It has a configuration in which the level and the spectrum structure are controlled according to the high-band spectrum intensity of the signal, added to the input signal and output. This allows
As shown in FIG. 3, the input signal of the band A of fs / 2 or less
The band is extended by adding a band B signal of s / 2 to pfs / 2.

The spectrum structure of most music signals is 1 /
It has f- ¹ / f2 characteristics. Spectrum analysis circuit 1
03 indicates whether the input signal of band A is close to the 1 / f characteristic or 1 / f
^The voice band is extended in a form close to the spectrum structure of the input signal by analyzing whether the characteristic is close to the ^two characteristics and determining the spectrum structure of the signal to be band-extended based on the output. Therefore, since the band is extended at a high frequency close to the spectrum structure of the input signal, the specific band is not emphasized, and the band can be extended with a natural feeling. In addition, since a dither signal close to a Gaussian distribution is used, the band is extended with a sound close to a natural sound.

Since the signal processing is digital processing, there is no variation in performance due to variations in components constituting the circuit or temperature characteristics. In addition, the sound quality does not deteriorate every time the audio signal passes through the circuit. Furthermore, even if the accuracy of the filter configured is pursued, it is possible to realize a voice band expansion method and apparatus which does not increase the circuit scale as compared with the analog circuit configuration and does not lead to an increase in cost.

In the first embodiment of the present invention, the nonlinear circuit 10
4 and the output signal of the dither generation circuit 105 are independently processed by filters 106 and 107 and level control circuits 108 and 109, respectively, and these signals are added by addition circuits 110 and 111. 10
It is needless to say that the same effect can be obtained by performing the filtering process and the level control process after adding the output of the dither generating circuit 105 and the output of the dither generating circuit 105. Further, when the spectrum analysis circuit determines that the input signal is a signal of a single spectrum, the amplification factor of the level control circuit is set to zero. This solves the problem that the signal degradation is caused by the voice band extending device because the high band signal is not added even if the sine wave is input.

<< Embodiment 2 >> The overall configuration of a voice band extending apparatus according to Embodiment 2 of the present invention is basically the same as that of the voice band extending apparatus according to Embodiment 1 of the present invention (FIG. 1). The input signal is a digital audio signal having a sampling frequency fs, and in the processing after the oversampling circuit in the oversampling type low-pass filter 102, all sampling clocks for the signal processing are p · fs (the dither generation circuit 105 and the like). Also included). In this embodiment, fs = 4
4.1 kHz, p = 2, p · fs = 88.2 kHz. The audio band extending apparatus according to the second embodiment differs from the audio band extending apparatus according to the first embodiment in the configuration of the filters 106 and 107 (FIG. 12). Second Embodiment Filters 106 and 107
Is different from the filters 106 and 107 of the first embodiment in the configuration of the high-pass filter 1201 (FIG. 12).

FIG. 14 is a detailed block diagram of the high-pass filter 1201 of the filters 106 and 107 of the voice band extending apparatus according to the second embodiment of the present invention. In FIG. 14, reference numeral 1401 denotes high-pass filters 1 to m (m is an arbitrary integer of 2 or more), and 1402 denotes a switch. FIG. 15 shows frequency characteristics of a plurality of high-pass filters 1 to m (1401) included in the high-pass filters 1201 of the filters 106 and 107 of the voice band extending apparatus according to the second embodiment of the present invention. The operation of the thus configured voice band extending apparatus according to the second embodiment of the present invention will be described below. Since the basic operation is the same as that of the voice band extending apparatus according to the first embodiment of the present invention, the differences will be described in detail.

The high-pass filter 1201 of the first embodiment is
Although a single high-pass filter that cuts frequency components equal to or lower than fs / 2 is used, the high-pass filter 1201 according to the second embodiment has a plurality of high-pass filters 1 and 2 as described above.
To m (1401) and a switch 1402.
The frequency characteristic of each high-pass filter is the characteristic shown in FIG. 15, and each cut-off frequency or cut-off characteristic is different. The switch 1402 selects and outputs one output signal from the output signals of the plurality of high-pass filters 1 to m (1401) according to the output signal of the spectrum analysis circuit 103.
The output signal of switch 1402 is input to 1 / f characteristic filter 1202 and 1 / f ² characteristic filter 1203.

The low-pass side of the high-pass filter 1201 (fs
/ 2) cut-off frequency or cut-off characteristic in accordance with the output of the spectrum analysis circuit 103. The low-frequency characteristic (characteristics near fs / 2) of the high-frequency signal (band B) added.
To change. This allows fine adjustment of the level of the output signal of the adder 111 near fs / 2, which is the point where the input signal (band A) is in contact with the added high-frequency signal.
It is possible to suppress undulations such as a swelling of the spectrum near s / 2, and to expand the band with a more natural spectrum distribution.

A digital signal is input through the input terminal 101. If this signal is reproduced from a compact disc (CD), the sampling frequency (fs)
It is a signal of 44.1 kHz and a word length of 16 bits. The oversampling low-pass filter 102 multiplies the sampling frequency of the signal input via the input terminal 101 by p (p is a positive integer) and attenuates unnecessary bands. p is 2 or more (usually an even number), and the band of 1/2 (fs / 2) or more of the sampling frequency of the input signal is 60d.
Decrease by B or more. The spectrum analysis circuit 103 analyzes the spectrum distribution of the output signal of the oversampling type low-pass filter 102 and determines the spectrum intensity of the high-frequency component (for example, the spectrum intensity at fs / 4 to fs / 2).
It detects which of the 1 / f characteristic filter 1202 and the 1 / f ² characteristic filter 1203 is suitable based on the spectrum distribution of the high frequency band.

The non-linear circuit 104 performs non-linear processing on the output signal of the oversampling type low-pass filter 102 to generate harmonics. The filter 106 optimally cuts a band of fs / 2 or less of the output signal of the nonlinear circuit 104 by the high-pass filter 1201, and fs / 2
For the above bands, the 1 / f characteristic filter 1202 or 1 / f ² based on the output signal of the spectrum analysis circuit 103
By passing through the characteristic filter 1203, a signal of the band B having a spectrum distribution suitable for the input signal is output.

On the other hand, the dither generation circuit 105 generates a bell-shaped dither whose probability density has a bell-shaped distribution close to a substantially Gaussian distribution within a predetermined amplitude. The filter 107 optimally cuts the band of fs / 2 or less of the output signal of the dither generation circuit by the high-pass filter 1201. by passing the filter 1202 or 1 / ^{f 2} characteristic filter 1203 outputs a signal of band B having a spectral distribution suitable for the input signal.

Each of level control circuits 108 and 109 amplifies the output signals of filters 106 and 107 at an amplification factor according to the output signal of spectrum analysis circuit 103. For example, when the spectrum intensity of the input signal at fs / 4 to fs / 2 is large, filters 106 and 107 are used.
Is operated to increase the output level, and to reduce the output level when the spectrum intensity is low. Adder circuit 11
0 adds the outputs of the two level control circuits 108 and 109.

The delay circuit 113 receives the output signal of the oversampling type low-pass filter 102, delays the signal, and outputs the delayed signal. The output signal of the oversampling type low-pass filter 102 is passed through a delay circuit 113 to the adder 1.
And processing until the output signal of the oversampling type low-pass filter 102 is input to the adder 111 via the spectrum analysis circuit 103, the nonlinear circuit 104, the adder 110, etc. The delay time of the delay circuit 113 is determined so that the time matches. The addition circuit 111 receives the output signal of the delay circuit 113 and the output signal of the addition circuit 110, adds the two, and outputs the addition result. The addition result is output from the output terminal 112.

As described above, the voice band extending apparatus according to the second embodiment generates a high frequency having a spectrum of the band B higher than the band A of the input signal, and generates the high frequency in accordance with the high band spectrum intensity of the input signal. The voice band is extended by adding a high frequency component to the input signal. By controlling the low-frequency characteristics of the high-frequency signal to be added according to the input signal, the spectrum distribution of the input signal and the added signal can be made continuous, so that natural band expansion is possible.

Further, since the signal processing is digital processing, there is no performance variation due to the variation of components constituting the circuit and the temperature characteristics. In addition, the sound quality does not deteriorate every time the audio signal passes through the circuit. Furthermore, even if the accuracy of the filter configured is pursued, it is possible to realize a voice band expansion method and apparatus which does not increase the circuit scale as compared with the analog circuit configuration and does not lead to an increase in cost.

<< Embodiment 3 >> The overall configuration of a voice band extending apparatus according to a third embodiment of the present invention is basically the same as that of the voice band extending apparatus (FIG. 1) according to the first embodiment of the present invention. The input signal is a digital audio signal having a sampling frequency fs, and in the processing after the oversampling circuit in the oversampling type low-pass filter 102, all sampling clocks for the signal processing are p · fs (the dither generation circuit 105 and the like). Also included). In this embodiment, fs = 4
4.1 kHz, p = 2, p · fs = 88.2 kHz. The voice band extending apparatus according to the third embodiment differs from the voice band extending apparatus according to the first embodiment in the configuration of the filters 106 and 107 (FIG. 12). Since the basic operation is the same as that of the voice band extending apparatus according to the first embodiment of the present invention, the differences will be described in detail.

FIG. 16 is a block diagram of the filters 106 and 107 of the voice band extending apparatus according to the third embodiment. The configurations of the filter 106 and the filter 107 are the same. FIG.
, 1601 is a low-pass filter, 1602 is a high-pass filter, 1603 is an addition circuit, and 1604 is 1 /
f characteristic filter, 1605 is 1 / f ² characteristic filter, 1
606 is a switch. FIG. 17 shows Embodiment 3 of the present invention.
6 shows the overall frequency characteristics of the filters 106 and 107 of FIG. The filters 106 and 107 of the first embodiment have the same frequency characteristics in band B as in FIG.
Are all removed. On the other hand, the filters 106 and 107 according to the third embodiment have the same characteristics as the filter according to the first embodiment in the band B, and pass the signal component in the band A though the attenuation factor is large.

The high-pass filter 1602 receives an output signal of the non-linear circuit 104 and outputs a signal component of a band B which is a band higher than the band A of the input signal. Does not decay.) The low-pass filter 1601 receives an output signal of the nonlinear circuit 104 and the like, and outputs a signal component of a band A of the input signal.
The signal components higher than the band A are cut. The low-pass filter 1601, the high-pass filter 1602, and the adder 1603 correspond to the first filter in the claims. Since the low-pass filter 1601 has a large attenuation rate even in the pass band (band A), the signal level of the output signal of the low-pass filter 1601 is very small and smaller than the LSB level of the input signal. Therefore,
In the third embodiment, the word length of each block of the voice band extending apparatus is made larger than the word length of the input signal, and the output signal of the low-pass filter 1601 is added to the added lower bits.

The adder 1603 includes the low-pass filter 16
01 is multiplied by a predetermined value d, and the resulting signal is added to the output signal of the high-pass filter 1602.
The predetermined value means that the energy level of the output signal (signal component of band A) of the low-pass filter 1601 is L4, and the output signal of the high-pass filter 1602 (signal component of band B).
Is the value d such that the ratio of d · L4 / L5 becomes a constant value c when the energy level of L is L5. Therefore, d.
From L4 / L5 = c, d is calculated by the formula of d = c · L5 / L4. Also, a signal (signal component of band A) obtained by multiplying the output signal of low-pass filter 1601 by a predetermined value d.
Is not more than the expanded word length (the difference between the word length of the output signal of the output terminal 112 and the word length of the input signal).

The output signal of adder 1603 is input to 1 / f characteristic filter 1604 and 1 / f ² characteristic filter 1605. The switch 1606 controls the 1 / f characteristic filter 16 according to the output signal of the spectrum analysis circuit 103.
04 of the output signal or 1 / ^{f 2} characteristic by selecting one of the output signal of the filter 1605 and outputs. The 1 / f characteristic filter 1604, the 1 / f ² characteristic filter 1605, and the switch 1606 correspond to the second filter in the claims.

In the third embodiment, the signal components of the bands A and B are added in accordance with the high-band spectrum intensity of the input signal, so that the data word length and the band are extended. I have. By setting the energy ratio between the band A and the band B to a predetermined value, the spectrum structure of the signal to be added does not change, so that a specific band is not emphasized, and the data word length can be expanded and the band can be expanded. . The operation of the thus-configured voice band extending apparatus according to the third embodiment of the present invention will be described below.

A digital signal is input through the input terminal 101. If this signal is reproduced from a compact disc (CD), the sampling frequency (fs)
It is a signal of 44.1 kHz and a word length of 16 bits. The oversampling low-pass filter 102 multiplies the sampling frequency of the signal input via the input terminal 101 by p (p is a positive integer) and attenuates unnecessary bands. p is 2 or more (usually an even number), and the band of 1/2 (fs / 2) or more of the sampling frequency of the input signal is 60d.
Decrease by B or more.

The spectrum analysis circuit 103 analyzes the spectrum distribution of the output signal of the oversampling type low-pass filter 102, and determines the spectrum intensity of the high frequency component (for example, the spectrum intensity at fs / 4 to fs / 2) and the high frequency component. 1 / f characteristic filter 160 based on spectral distribution
Any of the filters 4 and 1 / ^{f 2} characteristic filter 1605 detects whether suitable. The nonlinear circuit 104 performs nonlinear processing on an output signal of the oversampling type low-pass filter 102 to generate a harmonic.

The filter 106 outputs fs / fs of the output signal of the nonlinear circuit 104 by the low-pass filter 1601.
The signal component of band A of 2 or less is attenuated and extracted, and the high-pass filter 1602 outputs a signal component of band B that is equal to or more than fs / 2 of the output signal of the nonlinear circuit 104 (band B has no attenuation). ), Adder 1603 adds both. Then, based on the output signal of the spectrum analysis circuit 103, the 1 / f characteristic filter 1604 or 1 / f ²
By passing through the characteristic filter 1605, the signal components of the band A and the band B having the spectrum distribution suitable for the input signal are output.

On the other hand, the dither generating circuit 105 generates a bell-shaped dither whose probability density has a bell-shaped distribution close to a substantially Gaussian distribution within a predetermined amplitude. The filter 107 attenuates and extracts the signal component of the band A equal to or less than fs / 2 of the output signal of the dither generation circuit 105 by the low-pass filter 1601, and outputs the fs of the output signal of the dither generation circuit 105 by the high-pass filter 1602. / 2 or more signal components in band B are output (band B is not attenuated),
An adder 1603 adds the two. Thereafter, a switch 1606 is performed based on the output signal of the spectrum analysis circuit 103.
To 1 / f characteristic filter 1604 or 1 / f ²
By passing through the characteristic filter 1605, the signal components of the band A and the band B having the spectrum distribution suitable for the input signal are output.

Each of the level control circuits 108 and 109 amplifies the output signals of the filters 106 and 107 at an amplification factor corresponding to the output signal of the spectrum analysis circuit 103. For example, when the spectrum intensity of the input signal at fs / 4 to fs / 2 is large, filters 106 and 107 are used.
Is operated to increase the output level, and to reduce the output level when the spectrum intensity is low. Adder circuit 11
0 adds the outputs of the two level control circuits 108 and 109.

The delay circuit 113 receives the output signal of the oversampling type low-pass filter 102, delays the signal, and outputs the delayed signal. The output signal of the oversampling type low-pass filter 102 is passed through a delay circuit 113 to the adder 1.
And processing until the output signal of the oversampling type low-pass filter 102 is input to the adder 111 via the spectrum analysis circuit 103, the nonlinear circuit 104, the adder 110, etc. The delay time of the delay circuit 113 is determined so that the time matches. The addition circuit 111 receives the output signal of the delay circuit 113 and the output signal of the addition circuit 110, adds the two, and outputs the addition result. The addition result is output from the output terminal 112.

As described above, the voice band extending apparatus according to the third embodiment generates a noise signal having the spectrum of the band A and the band B of the input signal, and generates the noise signal in accordance with the high-band spectrum intensity of the input signal. By adding a noise signal to the input signal, the voice band is extended and the data word length is extended. Since the band is extended with a high-band signal close to the spectrum structure of the input signal, a specific band is not emphasized, and the band can be extended with a natural feeling.

Further, since the signal processing is digital processing, there is no variation in performance due to variations in components constituting the circuit and temperature characteristics. In addition, the sound quality does not deteriorate every time the audio signal passes through the circuit. Furthermore, even if the accuracy of the filter configured is pursued, it is possible to realize a voice band expansion method and apparatus which does not increase the circuit scale as compared with the analog circuit configuration and does not lead to an increase in cost.

<< Embodiment 4 >> The overall configuration of a voice band extending apparatus according to Embodiment 4 of the present invention is basically the same as that of the voice band extending apparatus according to Embodiment 3 of the present invention (FIG. 1). The input signal is a digital audio signal having a sampling frequency fs, and in the processing after the oversampling circuit in the oversampling type low-pass filter 102, all sampling clocks for the signal processing are p · fs (the dither generation circuit 105 and the like). Also included). In this embodiment, fs = 4
4.1 kHz, p = 2, p · fs = 88.2 kHz. In the voice band extending apparatus according to the fourth embodiment, the filters 106 and 107 (FIG. 18) include the level control circuits 1803 and 181.
It differs from the filters 106 and 107 (FIG. 16) of the voice band extending apparatus of the third embodiment in having 0.

FIG. 18 is a block diagram of the filters 106 and 107 of the voice band extending apparatus according to the fourth embodiment. The configurations of the filter 106 and the filter 107 are the same. FIG.
, 1801 and 1808 are low-pass filters,
1802 and 1809 are high-pass filters, 1803 and 1810 are level control circuits, 1804 and 1811 are addition circuits, 1805 and 1812 are 1 / f characteristic filters, 1806 and 1813 are 1 / f ² characteristic filters, 1
807 and 1814 are switches.

The filters 106 and 1 according to the fourth embodiment of the present invention
The overall frequency characteristics of the filter 07 are basically the same as those of the filters 106 and 107 of the third embodiment, and are shown in FIG. The frequency characteristics of the low-pass filters 1801 and 1808 of the fourth embodiment are the same as the frequency characteristics of the low-pass filter 1601 of the third embodiment (band A in FIG. 17), and the frequency characteristics of the high-pass filters 1802 and 1809 of the fourth embodiment are the same. 3
Is the same as the frequency characteristic of the high-pass filter 1602 of FIG. The operation of the thus configured voice band extending apparatus according to the fourth embodiment of the present invention will be described below. Since the basic operation is the same as that of the voice band extending apparatus according to the third embodiment of the present invention, the differences will be described in detail.

The high-pass filters 1802 and 1809 are
An output signal or the like of the nonlinear circuit 104 is input, and a signal component of a band B which is a band higher than the band A of the input signal is output (the signal is not attenuated in a band B which is a pass band). The low-pass filters 1801 and 1808 receive the output signal of the non-linear circuit 104 and the like and input signal band A
Is output. The signal components higher than the band A are cut. Since the low-pass filters 1801 and 1808 have a large attenuation rate even in the pass band (band A), the signal levels of the output signals of the low-pass filters 1801 and 1808 are very small, and are smaller than the level of the LSB of the input signal.

Therefore, in the fourth embodiment, the word length of each block of the voice band extending apparatus is made larger than the word length of the input signal, and the output signals of the low-pass filters 1801 and 1808 are added to the added lower bits. . Low-pass filters 1801, 1808, high-pass filter 1802,
1809, level control circuits 1803 and 1810, and adders 1804 and 1811 correspond to the first filter in the claims.

The level control circuits 1803 and 1810 receive the output signals of the high-pass filters 1802 and 1809, amplify the output signals of the high-pass filters 1802 and 1809 at an amplification factor based on the output signal of the spectrum analysis circuit 103, and output the amplified signals. Gain α of level control circuit 1803
1 and the amplification factor α2 of the level control circuit 1810,
The values of α1 and α2 are set such that, under the condition of α1 + α2 = constant, the spectrum analysis circuit 1 determines that the spectrum distribution of the noise signal to be added most closely approximates the spectrum distribution of the input signal.
03 is determined by the output signal.

The adders 1804 and 1811 multiply the output signals of the low-pass filters 1801 and 1808 by a predetermined value d, and apply the multiplied signals to the level control circuits 1803 and 1811.
And 10 output signals. The predetermined value means that the energy level of the output signal (signal component of band A) of the low-pass filters 1801 and 1808 is L4 and the level control circuit 18
03, 1810, when the energy level of the output signal (signal component of the band B) is L5, the value d is such that the ratio of d · L4 / L5 becomes a constant value c. Therefore, d · L4 /
From L5 = c, d is calculated by the formula of d = c · L5 / L4. The value of d of the adder 1804 and the value of d of 1811
Are independent and independent. Also, a low-pass filter 180
1, the word length of the signal (the signal component of band A) resulting from multiplying the output signal of 1808 by a predetermined value d (the word length of the output signal of the output terminal 112 and the word length of the input signal) Difference).

The output signals of the adders 1804 and 1811 are
The 1 / f characteristic filters 1805 and 1812 and the 1 / f ² characteristic filters 1806 and 1813 are input. The switches 1807 and 1814 provide 1 / f characteristic filters 1805 and 18 according to the output signal of the spectrum analysis circuit 103.
12 output signals or 1 / f ² characteristic filters 1806, 1
813 is selected and output. 1 /
f characteristic filters 1805, 1812, 1 / f ² characteristic filters 1806, 1813, and switches 1807, 18
14 corresponds to the second filter in the claims.

In the fourth embodiment, the signal components of each of the bands A and B are added in accordance with the high-band spectrum intensity of the input signal, so that the data word length and the band are extended. I have. By setting the energy ratio between the band A and the band B to a predetermined value, the spectrum structure of the signal to be added does not change, so that a specific band is not emphasized, and the data word length can be expanded and the band can be expanded. . The operation of the thus configured voice band extending apparatus according to the fourth embodiment of the present invention will be described below.

A digital signal is input through the input terminal 101. If this signal is reproduced from a compact disc (CD), the sampling frequency (fs)
It is a signal of 44.1 kHz and a word length of 16 bits. The oversampling low-pass filter 102 multiplies the sampling frequency of the signal input via the input terminal 101 by p (p is a positive integer) and attenuates unnecessary bands. p is 2 or more (usually an even number), and the band of 1/2 (fs / 2) or more of the sampling frequency of the input signal is 60d.
Decrease by B or more.

The spectrum analysis circuit 103 analyzes the spectrum distribution of the output signal of the oversampling type low-pass filter 102, and determines the spectrum intensity of the high-frequency component (for example, the spectrum intensity at fs / 4 to fs / 2) and the high-frequency component. 1 / f characteristic filter 180 based on spectral distribution
5, 1812 and 1 / f ² characteristic filters 1806, 18
It detects which of the filters 13 is suitable. The nonlinear circuit 104 performs nonlinear processing on an output signal of the oversampling type low-pass filter 102 to generate a harmonic.

The filter 106 attenuates and extracts the signal component in the band A of fs / 2 or less of the output signal of the nonlinear circuit 104 by the low-pass filter 1801 described above. Further, a signal component in band B of fs / 2 or more of the output signal of the nonlinear circuit 104 is extracted by the high-pass filter 1802 (the band B is not attenuated), and is amplified by the level control circuit 1803. The two are added by an adder 1804. Thereafter, the switch 1807 is switched based on the output signal of the spectrum analysis circuit 103 to switch 1 / f
Characteristic filter 1805 or 1 / f ² characteristic filter 180
6 to output signal components in bands A and B having a spectrum distribution suitable for the input signal.

On the other hand, the dither generation circuit 105 generates a bell-shaped dither having a bell-shaped distribution with a probability density close to a substantially Gaussian distribution within a predetermined amplitude. The filter 107 attenuates and extracts the signal component in the band A of fs / 2 or less of the output signal of the dither generation circuit 105 by the low-pass filter 1808. Further, the above-mentioned high-pass filter 1809 extracts a signal component in band B of fs / 2 or more of the output signal of the dither generation circuit 105 (band B is not attenuated), and amplifies it in a level control circuit 1810. Adder 1
According to 811, both are added. After that, based on the output signal of the spectrum analysis circuit 103, the switch 1814 is switched and passed through the 1 / f characteristic filter 1812 or 1 / f ² characteristic filter 1813, so that the band A and the band B having the spectrum distribution suitable for the input signal. Is output.

Each of the level control circuits 108 and 109 amplifies the output signals of the filters 106 and 107 at an amplification factor corresponding to the output signal of the spectrum analysis circuit 103. For example, when the spectrum intensity of the input signal at fs / 4 to fs / 2 is large, filters 106 and 107 are used.
Is operated to increase the output level, and to reduce the output level when the spectrum intensity is low. Adder circuit 11
0 adds the outputs of the two level control circuits 108 and 109.

The delay circuit 113 receives an output signal of the oversampling type low-pass filter 102, delays the signal, and outputs the delayed signal. The output signal of the oversampling type low-pass filter 102 is passed through a delay circuit 113 to the adder 1.
And processing until the output signal of the oversampling type low-pass filter 102 is input to the adder 111 via the spectrum analysis circuit 103, the nonlinear circuit 104, the adder 110, etc. The delay time of the delay circuit 113 is determined so that the time matches. The addition circuit 111 receives the output signal of the delay circuit 113 and the output signal of the addition circuit 110, adds the two, and outputs the addition result. The addition result is output from the output terminal 112.

As described above, the voice band extending apparatus according to the fourth embodiment generates the noise signal having the spectrum of the band A and the band B of the input signal, and generates the noise signal according to the high-band spectrum intensity of the input signal. By adding a noise signal to the input signal, the voice band is extended and the data word length is extended. Since the band is extended with a high-band signal close to the spectrum structure of the input signal, a specific band is not emphasized, and the band can be extended with a natural feeling. In particular, the total energy level of the high-frequency signal generated by the non-linear circuit and the high-frequency signal generated by the dither generation circuit can be controlled according to the high-frequency spectrum structure of the input signal while the total energy level remains constant, enabling natural band expansion. It is.

Further, since the signal processing is digital processing, there is no variation in performance due to variations in components constituting the circuit and temperature characteristics. In addition, the sound quality does not deteriorate every time the audio signal passes through the circuit. Furthermore, even if the accuracy of the filter configured is pursued, it is possible to realize a voice band expansion method and apparatus which does not increase the circuit scale as compared with the analog circuit configuration and does not lead to an increase in cost.

<< Embodiment 5 >> The overall configuration of a voice band extending apparatus according to Embodiment 5 of the present invention is basically the same as that of the voice band extending apparatus (FIG. 1) according to Embodiment 3 of the present invention. The input signal is a digital audio signal having a sampling frequency fs, and in the processing after the oversampling circuit in the oversampling type low-pass filter 102, all sampling clocks for the signal processing are p · fs (the dither generation circuit 105 and the like). Also included). In this embodiment, fs = 4
4.1 kHz, p = 2, p · fs = 88.2 kHz. In the voice band extending apparatus according to the fifth embodiment, the filters 106 and 107 (FIG. 19) include the level control circuits 1903 and 191.
It differs from the filters 106 and 107 (FIG. 16) of the voice band extending apparatus of the third embodiment in having 0.

FIG. 19 is a block diagram of the filters 106 and 107 of the voice band extending apparatus according to the fifth embodiment. The configurations of the filter 106 and the filter 107 are the same. FIG.
, 1901 and 1908 are low-pass filters,
1902 and 1909 are high-pass filters, 1903 and 1910 are level control circuits, 1904 and 1911 are addition circuits, 1905 and 1912 are 1 / f characteristic filters, 1906 and 1913 are 1 / f ² characteristic filters, 1
Reference numerals 907 and 1914 are switches.

The filters 106 and 1 according to the fifth embodiment of the present invention
The overall frequency characteristics of the filter 07 are basically the same as those of the filters 106 and 107 of the third embodiment, and are shown in FIG. The frequency characteristics of the low-pass filters 1901 and 1908 of the fifth embodiment are the same as the frequency characteristics of the low-pass filter 1601 of the third embodiment (band A in FIG. 17), and the frequency characteristics of the high-pass filters 1902 and 1909 of the fifth embodiment are the same. 3
Is the same as the frequency characteristic of the high-pass filter 1602 of FIG. The operation of the thus configured voice band extending apparatus according to the fifth embodiment of the present invention will be described below. Since the basic operation is the same as that of the voice band extending apparatus according to the third embodiment of the present invention, the differences will be described in detail.

The high-pass filters 1902 and 1909 are
An output signal or the like of the nonlinear circuit 104 is input, and a signal component of a band B which is a band higher than the band A of the input signal is output (the signal is not attenuated in a band B which is a pass band). The low-pass filters 1901 and 1908 receive the output signal of the non-linear circuit 104 and the like, and input signal band A
Is output. The signal components higher than the band A are cut. Since the low-pass filters 1901 and 1908 also have a large attenuation rate in the pass band (band A), the signal levels of the output signals of the low-pass filters 1901 and 1908 are very small, and are smaller than the level of the LSB of the input signal.

Therefore, in the fifth embodiment, the word length of each block of the voice band extending apparatus is made larger than the word length of the input signal, and the output signals of the low-pass filters 1901 and 1908 are added to the added lower bits. . Low-pass filters 1901, 1908, high-pass filter 1902,
1909, level control circuits 1903 and 1910, and addition circuits 1904 and 1911 constitute a first filter according to the invention.

The level control circuits 1903 and 1910 receive the output signals of the low-pass filters 1901 and 1908, amplify the output signals of the low-pass filters 1901 and 1908 at an amplification factor based on the output signal of the spectrum analysis circuit 103, and output the amplified signals. Amplification rate α of level control circuit 1903
3 and the amplification factor α4 of the level control circuit 1910,
The values of α3 and α4 are set so that the spectrum distribution of the noise signal to be added is most similar to the spectrum distribution of the input signal under the condition of α3 + α4 = constant.
03 is determined by the output signal.

The adders 1904 and 1911 multiply the output signals of the level control circuits 1903 and 1910 by a predetermined value d, and multiply the multiplied signals by the high-pass filters 1902 and 1902.
909 are added. The predetermined value means that the energy level of the output signal (signal component of band A) of the level control circuits 1903 and 1910 is L4, the high-pass filter 1
When the energy level of the output signals of 902 and 1909 (the signal component of the band B) is L5, the value d is such that the ratio d · L4 / L5 becomes a constant value c. Therefore, d · L4
From / L5 = c, d is calculated by the formula of d = c.L5 / L4. The value of d of the adder 1904 and the value of d of 1911 are independent and independent. Also, the level control circuit 190
3, the word length of the signal (signal component of band A) resulting from multiplying the output signal of 1910 by a predetermined value d (the word length of the output signal of the output terminal 112 and the word length of the input signal) Difference).

The output signals of the adders 1904 and 1911 are
The 1 / f characteristic filters 1905 and 1912 and the 1 / f ² characteristic filters 1906 and 1913 are input. The switches 1907 and 1914 switch the 1 / f characteristic filters 1905 and 19 according to the output signal of the spectrum analysis circuit 103.
12 output signals or 1 / f ² characteristic filters 1906, 1
913 is selected and output. 1 /
f characteristic filters 1905, 1912, 1 / f ² characteristic filters 1906, 1913, and switches 1907, 19
14 corresponds to the second filter in the claims.

In the fifth embodiment, the signal components of the bands A and B are added in accordance with the high-band spectrum intensity of the input signal, thereby expanding the data word length and the band. I have. By setting the energy ratio between the band A and the band B to a predetermined value, the spectrum structure of the signal to be added does not change, so that a specific band is not emphasized, and the data word length can be expanded and the band can be expanded. . The operation of the thus configured voice band extending apparatus according to the fifth embodiment of the present invention will be described below.

A digital signal is input through the input terminal 101. If this signal is reproduced from a compact disc (CD), the sampling frequency (fs)
It is a signal of 44.1 kHz and a word length of 16 bits. The oversampling low-pass filter 102 multiplies the sampling frequency of the signal input via the input terminal 101 by p (p is a positive integer) and attenuates unnecessary bands. p is 2 or more (usually an even number), and the band of 1/2 (fs / 2) or more of the sampling frequency of the input signal is 60d.
Decrease by B or more.

The spectrum analysis circuit 103 analyzes the spectrum distribution of the output signal of the oversampling type low-pass filter 102, and obtains the spectrum intensity of the high frequency component (for example, the spectrum intensity at fs / 4 to fs / 2) and the high frequency component. 1 / f characteristic filter 190 based on spectral distribution
5,1912 and 1 / ^{f 2} characteristic filter 1906,19
It detects which of the filters 13 is suitable. The nonlinear circuit 104 performs nonlinear processing on an output signal of the oversampling type low-pass filter 102 to generate a harmonic.

The filter 106 attenuates and extracts the signal component in the band A of fs / 2 or less of the output signal of the nonlinear circuit 104 by the low-pass filter 1901. Further, the above-mentioned high-pass filter 1902 extracts a signal component in band B of fs / 2 or more of the output signal of the nonlinear circuit 104 (band B is not attenuated), and amplifies it in the level control circuit 1903. The two are added by an adder 1904. After that, the switch 1907 is switched based on the output signal of the spectrum analysis circuit 103 and 1 / f
Characteristic filter 1905 or 1 / f ² characteristic filter 190
6 to output signal components in bands A and B having a spectrum distribution suitable for the input signal.

On the other hand, the dither generation circuit 105 generates a bell-shaped dither whose probability density has a bell-shaped distribution close to a substantially Gaussian distribution within a predetermined amplitude. The filter 107 attenuates and extracts the signal component in the band A of fs / 2 or less of the output signal of the dither generation circuit 105 by the low-pass filter 1908. Further, the signal component of band B of fs / 2 or more of the output signal of the dither generation circuit 105 is extracted by the high-pass filter 1909 (the band B is not attenuated), and is amplified by the level control circuit 1910. Adder 1
According to 911, both are added. After that, based on the output signal of the spectrum analysis circuit 103, the switch 1914 is switched to pass through the 1 / f characteristic filter 1912 or 1 / f ² characteristic filter 1913, so that the band A and the band B having the spectrum distribution suitable for the input signal. Is output.

Each of level control circuits 108 and 109 amplifies the output signals of filters 106 and 107 at an amplification factor corresponding to the output signal of spectrum analysis circuit 103. For example, when the spectrum intensity of the input signal at fs / 4 to fs / 2 is large, filters 106 and 107 are used.
Is operated to increase the output level, and to reduce the output level when the spectrum intensity is low. Adder circuit 11
0 adds the outputs of the two level control circuits 108 and 109.

The delay circuit 113 receives an output signal of the oversampling type low-pass filter 102, delays the signal, and outputs the delayed signal. The output signal of the oversampling type low-pass filter 102 is passed through a delay circuit 113 to the adder 1.
And processing until the output signal of the oversampling type low-pass filter 102 is input to the adder 111 via the spectrum analysis circuit 103, the nonlinear circuit 104, the adder 110, etc. The delay time of the delay circuit 113 is determined so that the time matches. The addition circuit 111 receives the output signal of the delay circuit 113 and the output signal of the addition circuit 110, adds the two, and outputs the addition result. The addition result is output from the output terminal 112.

As described above, the voice band extending apparatus of the fifth embodiment generates a noise signal having the spectrum of the band A and the band B of the input signal, and generates the noise signal in accordance with the high band spectrum intensity of the input signal. By adding a noise signal to the input signal, the voice band is extended and the data word length is extended. Since the band is extended with a high-band signal close to the spectrum structure of the input signal, a specific band is not emphasized, and the band can be extended with a natural feeling. In particular, the total energy level of the high-frequency signal generated by the non-linear circuit and the high-frequency signal generated by the dither generation circuit can be controlled according to the high-frequency spectrum structure of the input signal while the total energy level remains constant, enabling natural band expansion. It is.

Further, since the signal processing is digital processing, there is no performance variation due to the variation of the components constituting the circuit and the temperature characteristics. In addition, the sound quality does not deteriorate every time the audio signal passes through the circuit. Furthermore, even if the accuracy of the filter configured is pursued, it is possible to realize a voice band expansion method and apparatus which does not increase the circuit scale as compared with the analog circuit configuration and does not lead to an increase in cost.

Embodiment 6 The overall configuration of a voice band extending apparatus according to Embodiment 6 of the present invention is basically the same as that of the voice band extending apparatus according to Embodiment 3 of the present invention (FIG. 1). The input signal is a digital audio signal having a sampling frequency fs, and in the processing after the oversampling circuit in the oversampling type low-pass filter 102, all sampling clocks for the signal processing are p · fs (the dither generation circuit 105 and the like). Also included). In this embodiment, fs = 4
4.1 kHz, p = 2, p · fs = 88.2 kHz. In the voice band extending apparatus according to the sixth embodiment, the filters 106 and 107 (FIG. 20) include the level control circuits 2003 and 200.
4, 2011 and 2012 are different from the filters 106 and 107 (FIG. 16) of the voice band extending apparatus of the third embodiment.

FIG. 20 is a block diagram of filters 106 and 107 of the voice band extending apparatus according to the sixth embodiment. The configurations of the filter 106 and the filter 107 are the same. FIG.
, 2001 and 2009 are low-pass filters,
2002 and 2010 are high-pass filters, 2003,
2004, 2011 and 2012 are level control circuits,
005 and 2013 are addition circuits, and 2006 and 2014
Is a 1 / f characteristic filter, and 2007 and 2015 are 1 / f characteristic filters.
^{The two} characteristic filters, 2008 and 2016, are switches.

The filters 106 and 1 according to the sixth embodiment of the present invention
The overall frequency characteristics of the filter 07 are basically the same as those of the filters 106 and 107 of the third embodiment, and are shown in FIG. The frequency characteristics of the low-pass filters 2001 and 2009 of the sixth embodiment are the same as the frequency characteristics of the low-pass filter 1601 of the third embodiment (band A in FIG. 17), and the frequency characteristics of the high-pass filters 2002 and 2010 of the sixth embodiment are the same. 3
Is the same as the frequency characteristic of the high-pass filter 1602 of FIG. The operation of the voice band extending apparatus according to the sixth embodiment of the present invention thus configured will be described below. Since the basic operation is the same as that of the voice band extending apparatus according to the third embodiment of the present invention, the differences will be described in detail.

The high-pass filters 2002 and 2010 are:
An output signal or the like of the nonlinear circuit 104 is input, and a signal component of a band B which is a band higher than the band A of the input signal is output (the signal is not attenuated in a band B which is a pass band). The low-pass filters 2001 and 2009 receive the output signal of the non-linear circuit 104 and the like, and input signal band A
Is output. The signal components higher than the band A are cut. Since the low-pass filters 2001 and 2009 have a large attenuation rate even in the pass band (band A), the signal levels of the output signals of the low-pass filters 2001 and 2009 are very small and smaller than the LSB level of the input signal.

Therefore, in the sixth embodiment, the word length of each block of the voice band extending apparatus is made larger than the word length of the input signal, and the output signals of the low-pass filters 2001 and 2009 are added to the added lower bits. . Low-pass filter 2001, 2009, high-pass filter 2002,
2010, level control circuits 2003, 2004, 201
1, 2012 and the adders 2005 and 2013 correspond to the first filter in the claims.

The level control circuits 2003 and 2011 receive the output signals of the low-pass filters 2001 and 2009, amplify the output signals of the low-pass filters 2001 and 2009 at an amplification factor based on the output signal of the spectrum analysis circuit 103, and output the amplified signals. Amplification rate α of level control circuit 2003
5 and the amplification factor α6 of the level control circuit 2011,
The values of α5 and α6 are set so that the spectrum distribution of the noise signal to be added is most similar to the spectrum distribution of the input signal under the condition of α5 + α6 =.
Is determined by the output signal.

The level control circuits 2004 and 2012
Inputs the output signals of the high-pass filters 2002 and 2010, amplifies the output signals of the high-pass filters 2002 and 2010 with an amplification factor based on the output signal of the spectrum analysis circuit 103, and outputs the amplified signals. Assuming that the amplification factor α7 of the level control circuit 2004 and the amplification factor α8 of the level control circuit 2012 are, the values of α7 and α8 are as follows:
It is determined by the output signal of the spectrum analysis circuit 103 so that the spectrum distribution of the noise signal to be added is most similar to the spectrum distribution of the input signal.

The adders 2005 and 2013 multiply output signals of the level control circuits 2003 and 2011 by a predetermined value d, and multiply the multiplied signals by the level control circuits 2004 and 20.
And twelve output signals. The predetermined value means that the energy level of the output signal (signal component of band A) of the level control circuits 2003 and 2011 is L4, and the level control circuit 200
4, when the energy level of the output signal of 2012 (signal component of band B) is L5, the value d is such that the ratio of d · L4 / L5 becomes a constant value c. Therefore, d · L4 / L
From 5 = c, d is calculated by the formula of d = c · L5 / L4. Note that the value of d of the adder 2005 and the value of d of 2013 are independent and independent. Also, the level control circuits 2003, 2
011 is multiplied by a predetermined value d, the word length of the signal (the signal component of band A) is extended to the word length (output terminal 11
2) (the difference between the word length of the output signal and the word length of the input signal).

The output signals of the adders 2005 and 2013 are:
The 1 / f characteristic filters 2006 and 2014 and the 1 / f ² characteristic filters 2007 and 2015 are input. The switches 2008 and 2016 switch the 1 / f characteristic filters 2006 and 20 according to the output signal of the spectrum analysis circuit 103.
14 output signal or 1 / f ² characteristic filter 2007, ²
015 is selected and output. 1 /
f characteristic filters 2006 and 2014, 1 / f ² characteristic filters 2007 and 2015, and switches 2008 and 20
16 corresponds to the second filter in the claims.

In the sixth embodiment, the signal components of the bands A and B are added in accordance with the high-band spectrum intensity of the input signal, so that the data word length and the band are extended. I have. By setting the energy ratio between the band A and the band B to a predetermined value, the spectrum structure of the signal to be added does not change, so that a specific band is not emphasized, and the data word length can be expanded and the band can be expanded. . The operation of the voice band extending apparatus according to the sixth embodiment of the present invention thus configured will be described below.

A digital signal is input through the input terminal 101. If this signal is reproduced from a compact disc (CD), the sampling frequency (fs)
It is a signal of 44.1 kHz and a word length of 16 bits. The oversampling low-pass filter 102 multiplies the sampling frequency of the signal input via the input terminal 101 by p (p is a positive integer) and attenuates unnecessary bands. p is 2 or more (usually an even number), and the band of 1/2 (fs / 2) or more of the sampling frequency of the input signal is 60d.
Decrease by B or more.

The spectrum analysis circuit 103 analyzes the spectrum distribution of the output signal of the oversampling type low-pass filter 102, and determines the spectrum intensity of the high frequency component (for example, the spectrum intensity at fs / 4 to fs / 2) and the high frequency component. 1 / f characteristic filter 200 based on spectral distribution
6, 2014 and 1 / f ² characteristic filters 2007, 20
It detects which of the 15 filters is suitable. The nonlinear circuit 104 performs nonlinear processing on an output signal of the oversampling type low-pass filter 102 to generate a harmonic.

In the filter 106, the signal component of the band A of fs / 2 or less of the output signal of the nonlinear circuit 104 is attenuated and extracted by the low-pass filter 2001, and is amplified by the level control circuit 2003. Further, the above-described high-pass filter 2002 extracts a signal component of band B of fs / 2 or more of the output signal of the nonlinear circuit 104 (the band B has no attenuation), and the level control circuit 2004.
Amplify with The two are added by the adder 2005. After that, the switch 2008 is switched based on the output signal of the spectrum analysis circuit 103 to switch the 1 / f characteristic filter 20.
By passing 06 or 1 / f ² characteristic filter 2007, and outputs the signal component of the band A and band B having a spectral distribution suitable for the input signal.

On the other hand, the dither generating circuit 105 generates a bell-shaped dither whose probability density has a bell-shaped distribution close to a substantially Gaussian distribution within a predetermined amplitude. The filter 107 attenuates and extracts the signal component in the band A of fs / 2 or less of the output signal of the dither generation circuit 105 by the low-pass filter 2009, and amplifies the signal by the level control circuit 2011. In addition, the dither generation circuit 1
A signal component of the band B of fs / 2 or more of the output signal of No. 05 is extracted (the band B is not attenuated) and amplified by the level control circuit 2012. The two are added by the adder 2013. Thereafter, based on the output signal of the spectrum analysis circuit 103, the switch 2016 is switched to pass through the 1 / f characteristic filter 2014 or 1 / f ² characteristic filter 2015, so that the band A and the band B having the spectrum distribution suitable for the input signal. Is output.

Each of the level control circuits 108 and 109 amplifies the output signals of the filters 106 and 107 at an amplification factor corresponding to the output signal of the spectrum analysis circuit 103. For example, when the spectrum intensity of the input signal at fs / 4 to fs / 2 is large, filters 106 and 107 are used.
Is operated to increase the output level, and to reduce the output level when the spectrum intensity is low. Adder circuit 11
0 adds the outputs of the two level control circuits 108 and 109.

The delay circuit 113 receives the output signal of the oversampling type low-pass filter 102, delays the signal, and outputs the delayed signal. The output signal of the oversampling type low-pass filter 102 is passed through a delay circuit 113 to the adder 1.
And processing until the output signal of the oversampling type low-pass filter 102 is input to the adder 111 via the spectrum analysis circuit 103, the nonlinear circuit 104, the adder 110, etc. The delay time of the delay circuit 113 is determined so that the time matches. The addition circuit 111 receives the output signal of the delay circuit 113 and the output signal of the addition circuit 110, adds the two, and outputs the addition result. The addition result is output from the output terminal 112.

As described above, the voice band extending apparatus according to the sixth embodiment generates a noise signal having the spectrum of the band A and the band B of the input signal, and generates the noise signal in accordance with the high-band spectrum intensity of the input signal. By adding a noise signal to the input signal, the voice band is extended and the data word length is extended. Since the band is extended with a high-band signal close to the spectrum structure of the input signal, a specific band is not emphasized, and the band can be extended with a natural feeling. In particular, the total energy level of the high-frequency signal generated by the non-linear circuit and the high-frequency signal generated by the dither generation circuit can be controlled according to the high-frequency spectrum structure of the input signal while the total energy level remains constant, enabling natural band expansion. It is.

Further, since the signal processing is digital processing, performance variations do not occur due to variations in components constituting the circuit and temperature characteristics. In addition, the sound quality does not deteriorate every time the audio signal passes through the circuit. Furthermore, even if the accuracy of the filter configured is pursued, it is possible to realize a voice band expansion method and apparatus which does not increase the circuit scale as compared with the analog circuit configuration and does not lead to an increase in cost.

[0165]

As described above, a high-frequency signal having a spectrum similar to the spectrum of the input signal (a signal having a frequency higher than the audible band) is generated by using the non-linear circuit, the dither generation circuit and the filter, and the spectrum is analyzed. A circuit and a level control circuit control the level of the high frequency signal according to the high frequency spectrum intensity of the input signal and the like, add the high frequency signal to the input signal, and output the added signal. Since a high-frequency signal having a high-frequency spectrum of 1 / f characteristic or 1 / f ² characteristic close to a natural sound is added to expand the band of the audio signal, natural sound quality without emphasizing a specific band. Is obtained. According to the present invention, since the high-frequency component of the spectrum distribution according to the spectrum distribution of the input signal is added at an appropriate level, an advantageous effect that a voice band extending apparatus that generates a natural sound can be realized is obtained. Can be

According to the present invention, the cutoff frequency and cutoff frequency characteristics near the upper limit of the audible band of the high-frequency signal to be added are switched according to the analysis result signal, thereby suppressing the disturbance of the frequency characteristics near the upper limit of the audible band. Therefore, an advantageous effect of being able to realize a voice band extending device that generates a more natural sound can be obtained.

Further, since the signal processing is digital processing, there is obtained an effect that performance variations do not occur due to variations in components constituting the circuit and temperature characteristics. Further, there is obtained an effect that the sound quality does not deteriorate every time the audio signal passes through the circuit. Further, even if the accuracy of the filter is pursued, an effect that does not increase the cost without increasing the circuit scale as compared with the analog circuit configuration can be obtained. In addition, since a Gaussian distribution type dither having a 1 / f characteristic close to a natural sound is used, an effect of obtaining a natural sound quality without emphasizing a specific band even when a voice band is extended can be obtained.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating an overall configuration of a voice band extending apparatus according to a first embodiment of the present invention.

FIG. 2 is a schematic frequency characteristic of an input signal of the voice band extending apparatus according to the first embodiment of the present invention.

FIG. 3 is a schematic frequency characteristic of an output signal of the voice band extending apparatus according to the first embodiment of the present invention.

FIG. 4 is a block diagram of a spectrum analysis circuit of the voice band extending apparatus according to the first embodiment of the present invention.

FIG. 5 is a block diagram of a spectrum analysis circuit of a voice band extending apparatus according to another embodiment of the present invention.

FIG. 6 is a block diagram of a nonlinear circuit of the voice band extending apparatus according to the first embodiment of the present invention.

FIG. 7 is a block diagram of a dither generation circuit of the voice band extending apparatus according to the first embodiment of the present invention.

FIG. 8 is a block diagram of a PN sequence noise signal generation circuit of the voice band extending apparatus according to the first embodiment of the present invention.

FIG. 9 is a probability density of a white noise signal output from a PN sequence noise signal generation circuit.

FIG. 10 is a probability density of a bell distribution type noise signal output from a PN sequence noise signal generation circuit.

FIG. 11 is a probability density of a Gaussian distribution type noise signal output from a PN sequence noise signal generation circuit.

FIG. 12 is a block diagram of a filter of the voice band extending apparatus according to the first embodiment of the present invention.

FIG. 13 is a characteristic diagram showing characteristics of a filter of the voice band extending apparatus according to the first embodiment of the present invention.

FIG. 14 is a block diagram of a high-pass filter of a filter constituting the voice band extending apparatus according to the second embodiment of the present invention.

FIG. 15 is a characteristic diagram illustrating characteristics of a high-pass filter of a filter included in the voice band extending apparatus according to the second embodiment of the present invention.

FIG. 16 is a block diagram of a filter constituting a voice band extending apparatus according to a third embodiment of the present invention.

FIG. 17 is a characteristic diagram showing characteristics of a filter included in the voice band extending apparatus according to the third embodiment of the present invention.

FIG. 18 is a block diagram of a filter constituting a voice band extending apparatus according to a fourth embodiment of the present invention.

FIG. 19 is a block diagram of a filter constituting a voice band extending apparatus according to a fifth embodiment of the present invention.

FIG. 20 is a block diagram of a filter constituting a voice band extending apparatus according to a sixth embodiment of the present invention;

FIG. 21 is a block diagram showing a configuration of a conventional audio signal reproducing device.

[Explanation of symbols]

Reference Signs List 101 input terminal 102 oversampling low-pass filter 103 spectrum analysis circuit 104 nonlinear circuit 105 dither generation circuit 106, 107 filter 108, 109 level control circuit 110, 111 addition circuit 112 output terminal 41 FFT circuit 42 data selection circuit 43 weighted addition Circuits 51, 1201, 1401, 1602, 1802, 18
09, 1902, 1909 High-pass filters 52, 56, 61 Absolute value calculation circuits 53, 55, 57,
1601, 1801, 1808, 1901, 1908, 2001, 2009 Low pass filter 58 Judgment circuit 62 DC offset removal circuit 63, 74 Subtractor 64 Averaging circuit 65 1/2 multiplier 66, 72 Delay circuit 70-n (n = 1 , 2,..., N) PN sequence noise signal generator 71, 1603, 1804, 1811, 1904, 19
11, 2005, 2013 Adder 73 DC offset removal constant signal generator 81 32-bit shift register 82 Exclusive OR circuit 83 Clock signal generator 84 Initial data generator 1202, 1604, 1805, 1812, 1905,
1912, 2006, 2014 1 / f characteristic filter 1203, 1605, 1806, 1813, 1906,
1913, 2007, 2015 1 / f ² characteristic filter 1204, 1402, 1606, 1807, 1814,
1907, 1914, 2008, 2016 Switchers 1803, 1810, 1903, 1910, 2003,
2004, 2011, 2012 Level control circuit

Claims (10)

[Claims] 1. A low-pass filter for oversampling an input audio signal of a band A (A is an arbitrary band) at a sampling frequency of twice or more and removing aliasing noise; The output signal of the pass filter is separated, the spectrum of the output signal of the low-pass filter in the band A is analyzed at each time TB (TB is an arbitrary value), and an analysis result signal indicating the analysis result is output. A spectrum analysis means, a non-linear means for receiving an output signal of the low-pass filter and outputting an output signal subjected to non-linear processing, and an output having a probability density distribution of an arbitrary distribution between a Gaussian distribution and a bell-shaped distribution A dither generating means for generating a signal; and a first filter for passing a component of a band B (B is an arbitrary band higher than the band A) which is a frequency band higher than the band A. Data and, 1 / f characteristic according to the result of said analysis signal (f
Has a ^second filter that changes a frequency characteristic in a range between a frequency) and a 1 / f2 characteristic, and receives an output signal of the non-linear means and an output signal of the dither generation means, and Filter means for passing through a filter and the second filter; level control means for controlling an output level of an output signal of the filter means in accordance with the analysis result signal; and output of the filter means controlled by the level control means An adder for adding a signal to an output signal of the low-pass filter and outputting an added signal. 2. The level control means according to claim 1, wherein a component based on an output signal of said non-linear means in an output signal of said filter means and a component based on an output signal of said dither generation means in an output signal of said filter means. 2. The voice band extending apparatus according to claim 1, wherein level control is performed separately for each of the following. 3. The filter according to claim 1, wherein the first filter is a band-pass filter or a high-pass filter including the band B in a pass band, and at least one of a lower cut-off frequency and a cut-off frequency characteristic is analyzed. 3. The voice band extending device according to claim 1, wherein the switching is performed according to a result signal. 4. The first filter includes the band A and the band B in a pass band, and includes an energy level of a component of the band A included in an output signal of the first filter and an energy level of the first filter. As the ratio with the energy level of the component of the band B included in the output signal approaches a constant value,
4. The voice band extending apparatus according to claim 1, wherein a characteristic of the first filter changes. 5. The non-linear means, the dither generation means,
The filter means, the level control means, or the addition means, wherein the band A or the band of a component of the addition signal based on an output signal of the non-linear means and a component of the addition signal based on an output signal of the dither generation means The voice band extending apparatus according to any one of claims 1 to 4, wherein a ratio of an energy level in at least one of B is changed according to the analysis result signal. 6. A low-pass filter step for oversampling an input audio signal in a band A (A is an arbitrary band) at a sampling frequency of twice or more and removing aliasing noise; The output signal of the band-pass filter is separated, and the spectrum of the output signal of the low-pass filter in the band A at each time TB (TB is an arbitrary value) is analyzed to generate an analysis result signal indicating a result of the analysis. A nonlinear analysis step of nonlinearly processing an output signal of the low-pass filter step; and a dither generation step of generating an output signal having a probability density distribution of an arbitrary distribution between a Gaussian distribution and a bell-shaped distribution. And a first filter that passes a component of a band B (B is an arbitrary band higher than the band A), which is a frequency band higher than the band A. And motor steps, 1 / f in response to the analysis result signals
A second filter step of processing a signal with a filter having a frequency characteristic that changes in a range from a characteristic (f is frequency) to a 1 / f ² characteristic, wherein the signal generated in the non-linear step and the dither generation are processed. A filter step of processing the signal generated in the step by the first filter step and the second filter step; and a level control step of controlling an output level of the signal generated in the filter step in accordance with the analysis result signal And adding an output signal of the filter step controlled by the level control step to an output signal of the low-pass filter step to generate an addition signal. Method. 7. A component based on an output signal of the non-linear step in an output signal of the filter step, and a component based on an output signal of the dither generation step in an output signal of the filter step, wherein the level control step includes: 7. The voice band extending method according to claim 6, wherein level control is separately performed for each of the following. 8. The filter of the first filter step is a band-pass filter or a high-pass filter including the band B in a pass band, and at least one of a cut-off frequency and a cut-off frequency characteristic on a low-pass side thereof. 8. The method according to claim 6, wherein the method is switched according to the analysis result signal. 9. The filter of the first filter step includes the band A and the band B in a pass band, and
So that the ratio of the energy level of the band A component included in the output signal of the filter step to the energy level of the band B component included in the output signal of the first filter step approaches a constant value. 9. The voice band extending method according to claim 6, wherein a frequency characteristic of the filter in the first filter step changes. 10. The non-linear step, the dither generation step, the filter step, the level control step or the addition step, wherein the component of the addition signal based on the output signal of the non-linear step and the output signal of the dither generation step 10. A ratio of an energy level in at least one of the band A and the band B to a component of the addition signal based on the sum is changed according to the analysis result signal. A voice band extension method according to claim 1.