JPH06318876A - Method and device for signal conversion and recording medium - Google Patents

Abstract

PURPOSE:To attain the conversion of a voice and an acoustic signal significant with respect to sound quality based on a human audible sense by controlling a harmonious relation between frequency components within a critical band backed up in an audible sense. CONSTITUTION:Data processed by respective MDCT circuits 5a-5d are fed to a frequency moving peak detection circuit 12, a dissonance frequency detection circuit 11 and a masking curve detection circuit 16, from which a moving peak value curve, a dissonance frequency region being frequencies exceeding a threshold level having a difference of the moving peak value, and a masking threshold curve are obtained. A mask circuit 10 uses an audible effect to revise no frequency component in an unnecessary frequency band. The circuit 10 provides an output of component information from which an effective operation to improve the audible sound quality is obtained among frequency components having a dissonance relation with the local peak component. Then a frequency component revision circuit 6 revises the magnitude of a frequency component being an object based on the information.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention is applied to, for example, digital audio equipment, and changes the sound quality of an input audio signal, which is a time signal, by using the auditory property (that is, the characteristic of time signal information is converted). The present invention relates to a signal conversion method and device, and a recording medium on which information in which the characteristics of time signal information are converted by the method or device is recorded.

[0002]

2. Description of the Related Art Conventionally, as a method of changing the sound quality of acoustic signal information, for example, a method of changing frequency characteristics by filtering, a method of generating higher harmonics, or a so-called compressor is used to change the dynamic range. The method of doing is used.

[0003]

However, in the case of the method using the above-mentioned filter, the sound quality is changed by changing the way of using the filter such as increasing the presence by increasing the midrange. In the case of the method of generating high-order harmonics, it is used for effective sound rather than obtaining a sound that is easy to hear. In addition, the method of changing the dynamic range by the above compressor is
A loud sound will hurt your ears and a quiet sound will not be masked by ambient noise. With these methods, it is difficult to optimally control the sound so that the sound can be heard audibly comfortably in response to a change in the acoustic signal information that changes instantaneously.

Therefore, the present invention has been proposed in view of the above situation, and a signal conversion method and apparatus capable of converting a voice signal and an acoustic signal that are significant in terms of sound quality in the light of human hearing. Another object is to provide a recording medium.

That is, the problem to be solved by the present invention is to provide a method for instantaneously and instantaneously producing a sound which is pleasant to the human being in a high quality sound quality by using an acoustic principle. . Another object of the present invention is to improve the quality by reducing the auditory influence of the quantization noise from the acoustic signal information that has already been digitized and added with the quantization noise. . Another object of the present invention is to reduce the auditory influence of this quantization noise from the audio signal information that has already been digitized and added with the quantization noise. A technique for reducing the perceptual noise level by changing the spectrum of the quantization noise so as to match the so-called equal loudness characteristics and masking characteristics, which is proposed as a technology for improving the sound quality of audio equipment such as so-called compact discs ( Hereinafter, this technique will be referred to as, for example, Super Bit Mapping technique), that is, for example, Japanese Patent Laid-Open No. 20812/1990.
Japanese Patent Laid-Open No. 2-185552, Japanese Patent Laid-Open No. 2-1
When the technique disclosed in Japanese Patent No. 85556 is used to record data on a compact disc having a word length of 16 bits, it is intended to produce data with improved sound quality by aural processing. The super bit mapping technique can improve the sound quality when a digital signal having a word length exceeding 16 bits is requantized for a compact disc having a 16 bit length. Further, one of the problems of the present invention is that the audio quality of audio signal information to which quantization noise has already been added is equivalently 16
By improving once more than a bit and then requantizing it to 16 bits again, by improving the sound quality by keeping the S / N of the frequency band that is acoustically important to 16 bits or more, it is possible to improve the sound quality. is there.

[0006]

DISCLOSURE OF THE INVENTION The signal conversion method and apparatus of the present invention have been proposed in order to achieve the above-mentioned object, and a frequency component obtained from acoustic time signal information is within a substantially critical band. Transforming the characteristic of the acoustic temporal signal information by changing the difference in the size of the attribute between the frequency component exceeding the minimum audible level or the masking threshold level among the other frequency components of Among the other frequency components in the substantially critical band, the above-mentioned acoustic time signal information is obtained by changing the difference in attribute size between the minimum audible limit level and the frequency component exceeding the larger masking threshold level. Transforming the characteristics of the above, and changing the size of the attribute between other frequency components within a substantially critical band and frequency components within a limited level range. It is obtained so as to convert the characteristics of the acoustic time signal information.

Here, in the characteristic conversion, the difference in attribute size between the frequency component within the level range limited by the quantization noise level is changed, and a plurality of values obtained from the acoustic time signal information are changed. The difference in attribute size between at least one local peak of the frequency components of the above and other frequency components within the substantially critical band, and a frequency having a frequency difference of 10% to 50% of the substantially critical bandwidth. Increase the difference in the size of the attribute from other frequency components in the region, and determine the frequency region in which the difference in the size of the attribute of the frequency component is changed by the difference between the two moving peak values obtained from the frequency component. That is, the value obtained by subtracting the moving peak value of 10% width of the substantially critical bandwidth from the moving peak value of 50% width of the substantially critical bandwidth reduces or deletes the frequency component in the negative frequency region. Adjusting the magnitude of the frequency component so as to save the short time energy of the time signal information, and adjusting the magnitude of the frequency component of at least one local peak so as to save the short time energy of the time signal information , Re-quantizing the re-synthesized time signal information on the time axis with a noise shape characteristic. At this time, the noise shape characteristic may depend on at least one of the minimum audibility limit, equal loudness and masking characteristic, and the attribute may be the magnitude of the frequency component.

That is, the signal conversion method and apparatus of the present invention obtain frequency components by filtering the input acoustic time signal or using orthogonal transformation. Next, the moving peak value for each adjacent component of these frequency components is obtained with two different frequency widths related to the critical band, and the magnitude of the frequency component of the frequency band in which the difference between these two types of moving peak values occurs. By reducing, the degree of dissonance between the local peak frequency component and other frequency components is reduced. In developing the input acoustic time signal on the frequency axis,
After obtaining the components on the time axis of multiple frequency bands with a filter, etc., use a blocked frequency analysis method such as orthogonal transformation, or use so-called QMF (Quadrature Mirror Filter).
By connecting band division filters such as CQF (Conjugate Quadrature Filter) in a tree structure,
The frequency resolution gradually decreases from the low range to the high range,
On the contrary, band division is performed to improve time resolution.

At this time, the low frequency band may be divided into blocks having a longer time than the high frequency band, and orthogonal transformation may be performed or peak values of a plurality of samples on the time axis may be taken. The frequency bandwidth and time width of the block are designed to give a frequency resolution sufficiently satisfying the critical bandwidth so as to be acoustically optimal. In each block, it is judged whether the spectrum obtained by the analysis is above the masking threshold (masking threshold) or not based on its magnitude and frequency, and if it is below the masking threshold, the strength is determined. , Make sure that attributes such as phase are not changed. This also applies to the minimum audible limit, and frequency components below the minimum audible limit are not changed even if the difference between the moving peak values is not zero.

Furthermore, when the level of quantization noise that has already been added can be identified or predicted, it is not possible to perform processing different from other components on the frequency component of this level. It is effective in effectively removing the digitization noise. For example, it is effective to give a larger attenuation rate than other components or completely remove it.
Further, reducing the bit length by performing the super-bit mapping process on the acoustic signal information processed as described above means that when recording / reproducing transmission is performed with a limited word length,
This is effective in preventing auditory deterioration of sound quality as much as possible.
As described above, the present invention solves the above-mentioned problems by controlling the frequency component of acoustic signal information by an auditory method.

Further, the recording medium of the present invention is one in which compressed data obtained by being processed by the above-mentioned signal conversion method or apparatus is recorded.

[0012]

According to the present invention, by controlling the harmonic relationship between frequency components within the critical band that is acoustically supported, it is possible to adjust the sound quality of voice and sound signals to be beneficial to humans. In addition, masking thresholds and frequency components below the minimum audible limit should not be changed in order to prevent unnecessary side effects such as connection distortion by not performing unnecessary processing that is unrelated to sound quality. It is valid. Further, even though the digital sample data recorded on the compact disc has only a 16-bit word length resolution, 16-bit acoustic signal information is created by combining auditory frequency component change and super bit mapping processing. It is effective to record the acoustic signal information to which the quantization noise has been added and the acoustic signal information including aurally undesired frequency components on a compact disc, a digital audio tape or the like.

[0013]

Embodiments of the present invention will be described below with reference to the drawings.

The signal conversion apparatus of this embodiment to which the signal conversion method of the present invention is applied is, as shown in FIG. 1, a band division filter which will be described later as conversion means for converting acoustic time signal information into a plurality of frequency components. 2, 3, 4 and MDCT circuit 5
a, 5b, 5c, 5d and a plurality of frequency components obtained from the converting means, among other frequency components within the substantially critical band, a frequency component exceeding the minimum audible limit level or the masking threshold level. A frequency component changing circuit 6 and a mask circuit 10, which will be described later, as attribute changing means for changing the difference in the size of the attribute between them, a frequency shift peak detecting circuit 12, a dissonance frequency detecting circuit 11, a masking threshold curve detecting circuit 16, And a minimum audible curve generating circuit 17.

First, FIG. 1 is a block circuit diagram showing a schematic configuration of an embodiment of a signal conversion apparatus of the present embodiment which realizes a signal conversion method according to the present invention. Hereinafter, the specific configuration of FIG. 1 will be described in detail.

That is, the signal conversion apparatus of this embodiment divides an input digital signal such as voice or acoustic signal information (acoustic time signal information) into a plurality of frequency bands, and also divides the lowest two adjacent bands. The width is the same, but in the higher frequency band, the bandwidth is selected wider for the higher frequency band, orthogonal transformation is performed for each frequency band, and the frequency domain shifts from the obtained spectrum data on the frequency axis. Obtain information about the peak curve and the masking curve in the frequency domain.

From the information of the moving peak curve in the frequency domain, a frequency band in which a preferable change in sound quality can be expected can be obtained by changing the frequency component from the harmonic relationship between the frequency components. Also, from the information of the masking curve in the frequency domain, it is possible to expect a change in the desired sound quality by changing the frequency component obtained from the information of the moving peak curve in the frequency domain. The impossible frequency region is obtained and excluded from the frequency band in which the frequency component is changed. Frequency components below the minimum audible limit are also excluded from changes. The magnitude of the frequency component within the frequency band that changes the frequency component thus obtained is reduced or eliminated.

Next, the frequency components are inversely orthogonally transformed to obtain time signal information, and all bands are combined by a synthesis filter to obtain all band time signal information. Further, in performing the quantization, a super bit mapping process for optimizing the quantized noise spectrum in the band of 20 kHz or less perceptually is performed.

Referring to FIG. 1 in more detail, the input terminal 1 has, for example, a sampling frequency of 44.1 kHz.
When z, an audio PCM signal of 0 to 22 kHz is supplied. This input signal is, for example, the so-called CQ mentioned above.
It is divided into a band of 0 to 11 kHz and a band of 11 to 22 kHz by a band dividing filter 2 such as F, and 0 to 11 kHz.
Similarly, the band signal is 0 to 5.5 kHz band and 5.5 k to 11 kH by the band division filter 3 such as CQF filter.
and z-band. Further, signals in the 0 to 5.5 kHz band are also set to 0 by the band division filter 4 such as CQF.
It is divided into a 2.75 kHz band and a 2.75 to 5.5 kHz band.

11k to 22k from the band division filter 2
A signal in the Hz band is an MDCT which is an example of an orthogonal transform circuit.
(Modified Discrete Cosine Transform) circuit 5a sends a signal of band 5.5k to 11kHz from band division filter 3 to MDCT circuit 5b and sends signal of band division filter 4 from band 2.75kHz to 5.5kHz. Is sent to the MDCT circuit 5c, and the signal in the band of 0 kHz to 2.75 kHz from the band division filter 4 is sent to the MDCT circuit 5c.
Then, the MDCT processing is performed on each of them. Of course, these orthogonal transform circuits are
Besides CT, orthogonal transform such as fast Fourier transform (FFT) and discrete cosine transform (DCT) can be used.

Here, as a method of dividing the input digital signal by the band division filter as described above into a plurality of frequency bands, for example, there is a method of using a filter such as the above CQF, which is, for example, Mark JT Smit.
h and Thomas P. Barnwell, "Exact Reconstruction Tec
hniques for Tree-Structured Subband Coders, "IEEE T
rans. ASSP, Vol ASSP-34 No 3, June 1986, pp. 434-4
41. Also, 1976 REC Crochiere Dig
ital coding of speech in subbands BellSyst.Tech.
J. Vol.55, No.8 1976 describes a method using a filter such as QMF. Furthermore, ICASSP 83, BOSTON Pol
yphase Quadrature filters-A newsubband coding tech
Nique Joseph H. Rothweiler describes a filter partitioning method with equal bandwidth.

As the above-mentioned orthogonal transform, for example, the input audio signal is divided into blocks in a predetermined unit time (frame), and for example, fast Fourier transform (FFT), cosine transform (DCT), modified DCT transform ( There is an orthogonal transformation in which the time axis is transformed into the frequency axis by performing MDCT) or the like. MDCT above
About ICASSP 1987 Subband / Transform Coding Usin
g Filter Bank DesignsBased on Time Domain Aliasing
Cancellation JPPrincen ABBradley Univ.of Surr
ey Royal Melbourne Inst. of Tech.

Here, each of the MDCT circuits 5a, 5b,
FIG. 2 shows a specific example of a standard input signal for blocks for each band supplied to 5c and 5d.

In the specific example of FIG. 2, the four filter output signals described above have different orthogonal transform block sizes for each band, and frequency analysis is performed so as to sufficiently satisfy the critical bandwidth at each frequency. To do. As a result, the higher the frequency, the lower the frequency resolution, but instead the time resolution improves. In this embodiment, the frequency decomposition is selected so that the critical band is divided into about 10 parts. This makes it possible to control the magnitude of the frequency component within the critical band by limiting the frequency within the critical band quite freely.

That is, in this embodiment, from 0 Hz to 2.
In the band up to 75 kHz, the time block size of orthogonal transformation is set to 46.4 msec so that a frequency resolution of approximately 10 Hz, which is one tenth of the narrowest critical bandwidth of 100 Hz of this band, can be obtained. Similarly, 2.7
The 5 kHz to 5.5 kHz band uses a time block size of orthogonal transformation of 11.6 msec to obtain a frequency resolution of 40 Hz and the 5.5 kHz to 11 kHz band has 5.8 m.
80H using the time block size of the orthogonal transform of sec
As for the frequency resolution of z, the frequency resolution of 160 Hz is obtained by using the time block size of the orthogonal transformation of 2.9 msec in the band from 11 kHz to 22 kHz. In addition, 11k
Since the critical bandwidth in Hz is approximately 3 kHz, it is effective to further halve the orthogonal transform block size to obtain a frequency resolution of 320 Hz in order to further increase the time resolution. Table 1 shows the center frequency and bandwidth of the critical band.

[0026]

[Table 1]

Returning to FIG. 1 again, each MDCT circuit 5a,
The frequency component or MDCT coefficient data obtained by MDCT processing at 5b, 5c, and 5d determines the frequency shift peak detection circuit 12 that determines the frequency region where the frequency component having a dissonance relation with the local peak frequency component exists. It is supplied to the dissonance frequency detection circuit 11 and the masking threshold curve detection circuit 16 for obtaining the masking threshold curve.

Here, the frequency shift peak detection circuit 1 described above is used.
The operation of No. 2 will be described below. In FIG. 3, how to obtain the moving peak value regarding three adjacent frequency components is described for easy understanding.

First, the moving peak value centered on the component s1 is the component s including the component s1 and the components on both sides thereof.
The moving peak value is defined by the size of the component having the maximum size among 0, s1, and s2. Next, the moving peak value centering on the component s2 is the size of the component having the largest size among the components s1, s2, and s3 including the component s2 and the components on both sides thereof, and the moving peak value is Is defined. In this way, a moving peak curve is obtained by successively obtaining peak values.

For easy understanding in FIG. 3, all the frequency components have the same bandwidth, and the frequency widths for obtaining the moving peaks are also shown to be equal, but in this embodiment, as shown in FIG. In addition, as the frequency becomes higher, the bandwidth of frequency components expands and 10% of the critical bandwidth at that frequency
Alternatively, the moving peak value in the frequency width of 50% width is obtained. In addition, in FIG. 4, BE1 to BE4 in the figure each represent a band, a curve P _{10 in} the figure represents a moving peak curve having a width of 10% of the critical bandwidth, and a curve P ₅₀ represents 50% of the critical bandwidth.
The moving peak curve of the width is shown, the curve SD shows the frequency component distribution, and CB shows the critical bandwidth at each frequency. Here, in the frequency band in which the peak values are defined overlappingly, the one having the larger peak value is selected.

The critical bandwidth is a physical quantity that allows human auditory characteristics such as consonance, noise level sensation, and masking characteristics to be well understood. With respect to the present invention, explanation of consonance will be given with reference to FIG. Explain. FIG. 5 shows that when the frequency difference between the two frequency components is only the frequency shown on the horizontal axis (the axis showing the frequency normalized by the critical bandwidth), the degree of reciprocity or incongruity these two frequency components have. The vertical axis indicates whether or not it shows consonance. According to this result, 2
When the frequency difference between the two frequency components is between 10% and 50% of the critical bandwidth (dissonance band NHB), a sense of dissonance occurs (dissonance level NHL), 0% to 10% and 50% to 100%. A sense of consonance occurs at the frequency difference (consonant band HB) (consonant level HL). As shown in Table 1 above, the critical bandwidth has a wider bandwidth as it goes higher.

Next, a concrete means for detecting the dissonance band will be described with reference to FIG. Each MDCT circuit 5 in FIG.
The frequency component or MDCT coefficient data obtained by the MDCT processing at a, 5b, 5c, and 5d is taken as an absolute value, and then the dissonance frequency detection circuit 11 shown in FIG. It is given to the input terminal 41. Here, the low-frequency side characteristic having a longer time width is commonly used for each high-frequency time.

From the frequency components given to the input terminal 41, moving peak characteristics having two different frequency widths can be obtained. That is, a moving peak detection circuit 42 having a moving bandwidth of 10% of the critical bandwidth and a moving peak detecting circuit 42 having a moving bandwidth of 10% of the critical bandwidth, and a moving peak value of 50% of the moving bandwidth having a moving peak value of 50% of the critical bandwidth. The moving peak detection circuit 43 can obtain moving peak characteristics having two different frequency widths.

A moving peak detection circuit 42 having a moving peak value of 10% of the critical bandwidth and a moving peak detecting circuit 42 having a moving peak value of 10% of the critical bandwidth and 50% of the critical bandwidth giving a moving peak value of 50% of the critical bandwidth. The difference of the moving peak curve obtained by the width moving peak detecting circuit 43 is obtained by the difference detecting circuit 44 and taken out from the output terminal 45.

A frequency region in which the difference between the moving peak values thus obtained exceeds a certain threshold is defined as a dissonance frequency region.

However, when considering other auditory effects, that is, masking effect, equal loudness, and minimum audible limit, it is not necessary to target all the frequency components included in the dissonance frequency region obtained as described above. Absent. That is, considering the masking effect, the equal loudness, and the minimum audible limit, the frequency component that is determined not to be heard auditorily has almost no effect even if it is excluded from the operation target, and the equal loudness is considered. Sometimes, targeting the operation only to the effective band helps reduce the amount of calculation.

The mask circuit 10 having the mask function, the masking threshold curve detecting circuit 16 having the masking curve calculating function, and the minimum audible curve generating circuit 17 for storing the minimum audible limit information in FIG. 1 are as described above. Considering the masking effect and minimum audibility,
It is used to remove the frequency component that is judged not to be audibly heard from the operation target.

The masking function in the masking circuit 10, the masking curve calculating function in the masking threshold curve detecting circuit 16 and the minimum audible limit storing function in the minimum audible curve generating circuit 17 will be described in more detail below. .

FIG. 7 is a block circuit diagram showing a schematic configuration of a specific example of the masking curve calculation function in the masking threshold curve detection circuit 16. 7, the input terminal 71 is supplied with frequency component data from the MDCT circuits 5a, 5b, 5c and 5d in FIG.

The input data on the frequency axis is sent to the energy calculating circuit 72 for each critical band, where the energy of each critical band calculates the sum of the amplitude values of the frequency components within each critical band. Required by Instead of the energy for each critical region, the peak value, the average value, etc. of the amplitude value may be used. This energy calculation circuit 72
As an output from, for example, the spectrum of the sum value of each band is shown as SB in the figure. However, in FIG. 8, the number of divided bands is expressed by 12 bands (B1 to B12) to simplify the illustration.

Here, in order to take into consideration the influence of the spectrum SB on so-called masking, a convolution process is performed such that the spectrum SB is multiplied by a predetermined weighting function and added. Therefore, the output of the energy calculation circuit 72 for each band, that is, each value of the spectrum SB is sent to the convolution filter circuit 73. The convolution filter circuit 73 includes, for example, a plurality of delay elements that sequentially delay input data, and a plurality of multipliers that multiply outputs from these delay elements by a filter coefficient (weighting function) (for example, 25 corresponding to each band). Number of multipliers), and a sum adder that sums the outputs of the multipliers. By this convolution processing, for the spectrum SB of the band shown by B6 in FIG. 8, for example, the sum of the parts shown by the dotted line in the drawing is taken. Note that the masking is a phenomenon in which one signal is masked by another signal and becomes inaudible due to human auditory characteristics, and this masking effect includes continuous masking by an audio signal on the time axis. There are an effect and a simultaneous masking effect by a signal on the frequency axis. Due to these masking effects, even if there is signal information or noise in the masked portion, these will not be heard. Therefore, in the actual audio signal, the signal information and noise within the masked range do not need to be operated.

A specific example of the multiplication coefficient (filter coefficient) of each multiplier of the convolution filter circuit 73 will be described below.
When the coefficient of the multiplier M corresponding to an arbitrary band is 1, the coefficient M of the multiplier M-1 is 0.15, the coefficient of the multiplier M-2 is 0.0019, and the coefficient of the multiplier M-3 is 0. 00000
By multiplying the output of each delay element by 86, the multiplier M + 1 by a coefficient 0.4, the multiplier M + 2 by a coefficient 0.06, and the multiplier M + 3 by a coefficient 0.007.
B convolution processing is performed. However, M is an arbitrary integer of 1 to 25.

Next, the output of the convolution filter circuit 73 is sent to the subtractor 74. The subtractor 74 obtains a level α corresponding to signal information or noise level that can be removed from an operation target to be described later in the convoluted area. The level α corresponding to the signal information or noise level that can be removed from the operation target is
As will be described later, the level is such that the signal information or noise level can be removed from the operation target for each band of the critical band (critical bandwidth) by performing the inverse convolution process. Here, the subtractor 74 is supplied with an allowance function (function expressing a masking level) for obtaining the level α. The level α is controlled by increasing or decreasing this allowance function. The permissible function is (n-
ai) It is supplied from the function generating circuit 75.

That is, the level α corresponding to the signal information or noise level that can be removed from the operation target is
When the number given in order from the lower band of the critical band is i, it can be calculated by the following equation (1). α = S- (n-ai) (1) In this equation (1), n and a are constants, a> 0, S is the intensity of the convolution-processed Bark spectrum, and (1)
In the formula, (n-ai) is the allowable function. In this embodiment, n = 38,
It is assumed that a = 1.

In this way, the level α is obtained, and this data is transmitted to the divider 76. The divider 76 is for inverse convolution of the level α in the convolved area. Therefore, by performing this inverse convolution process,
The masking spectrum can be obtained from the level α. That is, this masking spectrum becomes signal information or noise spectrum that can be removed from the operation target.

The inverse convolution process is
Although a complicated operation is required, in this embodiment, the inverse convolution is performed using the simplified divider 76.

Next, the masking spectrum is transmitted to the subtractor 78 via the synthesizing circuit 77. Here, the output from the energy detection circuit 72 for each critical band, that is, the spectrum SB described above, is input to the subtracter 78.
It is supplied via the delay circuit 79. Therefore, the subtractor 78 subtracts the masking spectrum from the spectrum SB, so that the spectrum SB becomes the masking spectrum M as shown in FIG.
The level below the level indicated by S will be masked.

The output from the subtractor 78 is taken out through an output terminal 81 via signal information or a noise level correction circuit (not shown) which can be removed from the operation target, and the mask circuit. The frequency domain that can be excluded from the operation target in the dissonance frequency domain is excluded.

The delay circuit 79 takes the amount of delay in each circuit before the synthesis circuit 77 into consideration, and the energy detection circuit 72.
Is provided to delay the spectrum SB from

By the way, at the time of synthesizing by the synthesizing circuit 77 described above, data supplied from the minimum audible curve generating circuit 17 and showing the so-called minimum audible limit curve RC which is the human auditory characteristic as shown in FIG. , And the masking spectrum MS can be combined. In this minimum audible curve, if the signal or noise absolute level is below this minimum audible limit curve, the signal and noise will not be heard. The minimum audible limit curve varies depending on, for example, the reproduction volume at the time of reproduction. However, in a realistic digital system, there is not much difference in how to enter music into a 16-bit dynamic range.
For example, if the quantization noise in the most audible frequency band around 4 kHz is not heard, it is considered that the quantization noise below the level of the minimum audible limit curve is not heard in other frequency bands. Therefore, assuming that the system is used in such a manner that noise near the word length of the system of 4 kHz cannot be heard, and the minimum audible limit curve RC and the masking spectrum MS are combined together, it can be removed from the operation target. By obtaining the appropriate signal information or noise level, the signal information or noise level that can be removed from the operation target in this case can be up to the shaded portion in FIG. 10.

In the present embodiment, the level of 4 kHz of the minimum audible limit curve is set to the minimum level equivalent to 20 bits, for example. Further, FIG. 10 also shows the signal spectrum SS at the same time.

As another method of limiting the frequency components to be operated, there is a case where the frequency components are limited according to the level of the quantization noise included in the input digital signal information. When the spectrum is almost white, the quantization noise level is almost determined by the word length. Therefore, by limiting the frequency components in this level range to limited operations, the quantization noise component that effectively causes dissonance. Can be reduced or eliminated. In FIG. 1, by storing the quantization noise level in the quantization noise level storage function,
The frequency components to be operated are limited to this level range. Of course, this quantization level may be adjusted to be optimum.

In the signal information or noise level correction circuit that can be removed from the operation target, the subtractor 78 is based on, for example, equal loudness curve information sent from a correction information output circuit (not shown). The signal information or noise level that can be removed from the operation target in the output from is corrected. Here, the equal loudness curve is a characteristic curve relating to human auditory characteristics, and is obtained by, for example, obtaining the sound pressure of sound at each frequency heard at the same loudness as a pure tone of 1 kHz and connecting them with a curve. Also called sensitivity curve. Further, this equal loudness curve draws a curve substantially the same as the minimum audible curve RC shown in FIG. In this equal loudness curve, for example, in the vicinity of 4 kHz, even if the sound pressure drops by 8 to 10 dB from 1 kHz, it sounds as loud as 1 kHz. It doesn't sound the same. Therefore, it is understood that the magnitude of the signal or noise exceeding the level of the minimum audible curve should be evaluated by the frequency characteristic given by the curve corresponding to the equal loudness curve. Therefore, in consideration of the equal loudness curve, it is necessary to select signal information or noise that can be removed from the operation target in order to reduce the calculation amount.
It can be seen that it matches human hearing characteristics.

Returning to FIG. 1, the mask circuit 10 uses the auditory effect described above to prevent the frequency component from being changed in an unnecessary frequency band. The mask circuit 10 outputs, as an output, component information of frequency components having a dissonant relationship with the local peak component, which enables an operation effective for auditory sound quality improvement. The frequency component changing circuit 6 of FIG. 1 changes the magnitude of the frequency component of interest based on this information.

FIG. 11 shows how the frequency component changing circuit 6 changes the magnitude of the frequency component. In FIG. 11, Band1 to Band4 in the figure are frequency regions in which the magnitude of the frequency component designated by the mask circuit 10 is changed, and the degree of change becomes larger toward the center of each band. This utilizes the fact that the degree of dissonance shown in FIG. 5 differs depending on the frequency difference. In addition, Sp1 to Sp4 in the figure represent the gain at the position of each local peak spectrum, and in order to compensate for the decrease in the overall energy due to the decrease in the frequency component in the dissonance frequency band, this frequency is compensated. This shows that the magnitude of the position spectrum is increased.

The output of the frequency component changing circuit 6 in which the magnitude of the frequency component is changed in this way is moved from the frequency axis to the time axis by the IMDCT circuits 9a, 9b, 9c and 9d for performing the inverse transform of the MDCT. Is converted to. IMDC from these IMDCT circuits 9a, 9b, 9c, 9d
The T output signal has a frequency synthesis (ICQ) opposite to that of the CQF.
F) Frequency synthesis is performed by the band synthesis filters 13, 14 and 15 having the function to become the full-band time signal.

These band synthesizing filters 13, 14, 15
The full-band signal due to may have a larger dynamic range compared to the original input signal information due to the change of the frequency component. Therefore, when recording on a compact disc, requantization to 16 bits is required. May be needed. The applicant of the present invention previously performed re-quantization on the input digital audio signal by noise shaping in the audio band to give a noise frequency characteristic close to an equal loudness characteristic, and then re-quantized it on a compact disc.
A technique for recording a bit requantized signal is disclosed in, for example, the above-mentioned Japanese Patent Laid-Open Nos. 20812/1990 and 1855/1990.
No. 52 and Japanese Patent Application Laid-Open No. 2-185556.

In the present invention, in such a case, the signal processed by the present invention is further subjected to the above noise shaping to obtain a compact disc recording signal having a characteristic exceeding 16 bits.

The operation of the noise shaper for performing the above noise shaping will be described below with reference to FIG. The signal supplied from the band synthesis filter 15 to the adding circuit 18 is
The difference from the output signal of the feedback filter 21 can be taken. The output of the adder circuit 18 is the requantizer 19 and the second adder circuit 20.
Is supplied to. The re-quantizer 19 outputs a signal with a word length shorter than the input signal word length, and thus tries to transmit and record the signal with a small amount of information. This requantizer 1
The output of 9 is the output terminal 22 and the second terminal of the noise shaper.
Is supplied to the adder circuit 20. The second addition circuit 20 obtains the difference between the input and output signals of the requantizer 19, and the quantization error is extracted as the output. The output of the second adder circuit 20 is supplied to the feedback filter 21.

Here, the feedback filter 21 will be described in detail with reference to FIG. In FIG. 12, the terminal 5
The signal supplied to the feedback filter 21 via 0 is sequentially shifted to the series circuit of the delay elements 52, 53, 54 and 55. The output of each delay element 52, 53, 54, 55 is
The multipliers 56, 57, 58 and 59 are connected to each other, and the product of the multipliers 56, 57, 58 and 59 and the filter coefficients supplied from the corresponding coefficient input terminals 62, 63, 64 and 65 is obtained. To be These multiplication elements 5
The outputs of 6, 57, 58, and 59 are added by the adder element 60 and guided to the output terminal 61 of the feedback filter.

The digital audio signal to which the noise frequency characteristic close to the equal loudness characteristic is given by the noise shaper including the adder circuit 18, the requantizer 19, the second adder circuit 20, and the feedback filter 21 is output. It is output from the terminal 22. This output signal is subjected to predetermined error correction processing and the like, and is recorded on a recording medium (magneto-optical disk, semiconductor memory, IC memory card, optical disk).
Recorded in.

As described above, according to the first signal conversion method of the embodiment of the present invention, with respect to the frequency component obtained from the acoustic time signal information, the minimum audible level or the masking among the other frequency components within the substantially critical band. It is a time signal information characteristic conversion method characterized by changing the difference in attribute size between frequency components exceeding the threshold level.

In the second signal conversion method of the embodiment of the present invention, with respect to the frequency component obtained from the acoustic time signal information, the minimum audible limit level and the masking threshold among the other frequency components within the substantially critical band. It is a time signal information characteristic conversion method characterized in that the difference in the size of the attribute is changed between the frequency component exceeding the higher level of the Shored level and the frequency component.

Further, according to the third signal conversion method of the embodiment of the present invention, the frequency component obtained from the acoustic time signal information is within a limited level range among other frequency components within the substantially critical band. It is a time signal information characteristic conversion method characterized by changing a difference in attribute size between frequency components.

Further, the first to third signal conversion methods of the embodiment of the present invention are time signal information characteristic conversion methods characterized in that the attribute is the magnitude of the frequency component.

The converted data formed by the embodiment of the present invention can be recorded on a recording medium or can be transmitted through a transmission path. For example, as the recording medium, a magneto-optical disk, an optical disk, a semiconductor recording medium, especially an IC memory card can be used.

Furthermore, the present invention is not limited to the above embodiments, but can be applied to image signal information and the like.

[0068]

As described above, according to the present invention, the sound time signal information can be instantly and instantly produced to a human being in a high quality and pleasant sound by using the auditory principle. it can. In addition, quality can be improved by reducing the auditory influence of the quantization noise from the acoustic time signal information that has already been digitized and added with the quantization noise. As a technique to improve the sound quality of audio equipment such as a compact disc after subtracting the auditory influence of the quantization noise from the audio signal information obtained, for example, the spectrum of the quantization noise is adjusted to match the equal loudness characteristics and masking characteristics. By using a technique of reducing the noise level on the sense of hearing by changing the data, it is possible to create data with improved sound quality by aural processing when recording on a compact disc having a word length of 16 bits. As a result, when requantizing a digital signal having a word length exceeding 16 bits for a compact disc having a 16-bit length, it is possible to improve the sound quality. Further, according to the present invention, for audio signal information to which quantization noise has already been added, the sound quality is auditorily improved once equivalently to 16 bits or more, and then requantized to 16 bits again. It is possible to improve the sound quality by setting the S / N of the important frequency band to 16 bits while keeping the S / N to 16 bits or more.

[Brief description of drawings]

FIG. 1 is a block circuit diagram showing a schematic configuration example of a signal conversion apparatus of the present embodiment that realizes a method of converting characteristics of time signal information (signal conversion method) of the present invention.

FIG. 2 is a diagram showing a time block for each band according to the present invention.

FIG. 3 is a diagram showing frequency shift peaks according to the present invention.

FIG. 4 is a diagram showing an example of frequency shift peak frequency characteristics according to the present invention.

FIG. 5 is a diagram showing a relationship between a degree of concordance and a critical band.

FIG. 6 is a block circuit diagram showing a configuration example of a dissonance band detection circuit of the device according to the embodiment of the present invention.

FIG. 7 is a block circuit diagram showing a configuration example of a masking threshold curve detection circuit of the device of this embodiment.

FIG. 8 is a diagram showing a total value of signal components in each critical band.

FIG. 9 is a diagram showing a sum total value of signal components in each critical band and a masking threshold.

FIG. 10 is a diagram showing a sum total value of signal components in each critical band, a masking threshold, and a minimum audible limit.

FIG. 11 is a diagram showing an example of changing the magnitude of frequency components.

FIG. 12 is a diagram showing a configuration example of a feedback filter for noise shaping.

[Explanation of symbols]

1, 41, 71 ... Input terminal 2, 3, 4 ... Band division filter (CQF) 5a, 5b, 5c, 5d ... MDCT circuit 6 ... Component change circuit 10 ... Mask circuit 11 ... Dissonance frequency detection circuit 12 ... Frequency shift peak detection circuit 13, 14, 15 ... Band Synthesis filter 16 ... Masking threshold curve detection circuit 16 17 ... Minimum audible curve generation circuit 18, 20 ... Addition circuit 19 ... ..Requantizer 21 ... Feedback filter 22, 45, 61, 81 ... Output terminal 42 ... Moving peak detection circuit 43 with 10% width of critical band 43・ ・ ・ ・ ・ ・ ・ ・ Moving peak detection circuit with 50% width of critical band 44・ ・ ・ ・ ・ ・ ・ ・ Difference detection circuit 52, 53, 54, 55 ・ ・ ・ Delay element 56, 57, 58, 59 ・ ・ ・ Multiplication element 60 ・ ・ ・ ・ ・ ・ Addition element 72 ・ ・ ・・ ・ ・ Critical band energy calculation circuit 73 ・ ・ ・ ・ ・ ・ Convolution filter circuit 75 ・ ・ ・ ・ ・ ・ Function generation circuit 74 ・ ・ ・ ・ ・ ・ ・ ・ Subtractor 76 ・ ・・ ・ ・ ・ ・ ・ Divider 77 ・ ・ ・ ・ ・ ・ Synthesis circuit 78 ・ ・ ・ ・ ・ ・ Subtraction circuit

Claims (27)

[Claims] 1. The frequency component obtained from the acoustic time signal information has a large attribute between the frequency component exceeding the minimum audible level or the masking threshold level among other frequency components within the substantially critical band. A signal conversion method, characterized in that the characteristics of the acoustic time signal information are converted by changing the difference in height. 2. A frequency component obtained from the acoustic time signal information is a frequency component that exceeds a minimum audible limit level and a masking threshold level, which are larger than other frequency components within a substantially critical band. A signal conversion method, characterized in that the characteristics of the acoustic time signal information are converted by changing the difference in attribute size between them. 3. Regarding the frequency component obtained from the acoustic time signal information, among the other frequency components within the substantially critical band, the difference in attribute size between the frequency component within a limited level range is changed. By so doing, the characteristic of the acoustic time signal information is converted. 4. The signal conversion method according to claim 3, wherein a difference in attribute size between the frequency component and a frequency component within a level range limited by the quantization noise level is changed. 5. A difference in attribute size between at least one local peak of a plurality of frequency components obtained from the acoustic time signal information and another frequency component within a substantially critical band is changed. Claims 1, 2, 3,
4. The signal conversion method described in 4. 6. The difference in the size of the attribute between the frequency component and another frequency component having a frequency difference of 10% to 50% of the substantially critical bandwidth is increased.
The signal conversion method described in 2, 3, 4, and 5. 7. The frequency region in which the difference in the magnitude of the attribute of the frequency component is changed is determined by the difference between the two moving peak values obtained from the frequency component.
The signal conversion method described in 2, 3, 4, 5, and 6. 8. A value obtained by subtracting a moving peak value of 10% width of the substantially critical bandwidth from a moving peak value of 50% width of the substantially critical bandwidth reduces or eliminates a frequency component in a negative frequency region. 8. The signal conversion method according to claim 7, wherein: 9. The signal according to claim 1, wherein the magnitude of the frequency component is adjusted so as to preserve the short-term energy of the time signal information. How to convert. 10. The signal conversion method according to claim 9, wherein the magnitude of the frequency component of at least one local peak is adjusted so as to preserve the short-term energy of the time signal information. 11. A requantization process for re-synthesizing time signal information on a time axis, which has a noise shape characteristic, is performed. 9
The signal conversion method described. 12. The signal conversion method according to claim 11, wherein the noise shape characteristic depends on at least one of the minimum audibility limit, the equal loudness and the masking characteristic. 13. The attribute according to claim 1, 2, 3, 4, 5, 6, 7,
8. The signal conversion method according to 8, 9, 10, 11, and 12. 14. A conversion means for converting acoustic time signal information into frequency components, and a minimum audible level or a masking threshold among other frequency components within a substantially critical band for the frequency components obtained from the conversion means. A signal conversion device for converting the characteristic of the acoustic time signal information, which has attribute changing means for changing the difference in the size of the attribute with the frequency component exceeding the drive level. 15. A converting means for converting acoustic time signal information into frequency components, and a minimum audible limit level and a masking threshold among other frequency components within a substantially critical band for the frequency components obtained from the converting means. A signal conversion device for converting the characteristics of the above-mentioned acoustic time signal information, comprising attribute changing means for changing the difference in the size of the attribute between the frequency component having a larger level and the frequency component having a larger level. 16. A conversion means for converting acoustic time signal information into frequency components, and a frequency component obtained from said conversion means, within other frequency components within a substantially critical band, within a limited level range. A signal conversion apparatus, comprising: an attribute changing unit that changes a difference in attribute size between a component and the characteristic of the acoustic time signal information. 17. The signal conversion apparatus according to claim 16, wherein the attribute changing means changes a difference in attribute size between a frequency component within a level range limited by a quantization noise level. . 18. The attribute changing means has a difference in attribute size between at least one local peak of a plurality of frequency components obtained from the acoustic time signal information and other frequency components within a substantially critical band. 18. The signal conversion device according to claim 14, 15, 16, or 17, wherein: 19. The attribute changing means increases the difference in attribute size between other frequency components having a frequency difference of 10% to 50% of the substantially critical bandwidth. The signal conversion device according to claim 14, 15, 16, 17, or 18. 20. The attribute changing means determines a frequency region in which a difference in attribute size of a frequency component is changed, based on a difference between two moving peak values obtained from the frequency component. , 15, 16, 17, 18,
19. The signal conversion device according to item 19. 21. In the attribute changing means, from a moving peak value of 50% width of the substantially critical bandwidth, 10% of the substantially critical bandwidth is obtained.
21. The signal conversion apparatus according to claim 20, wherein a value obtained by subtracting the moving peak value of the width reduces or eliminates a frequency component in a negative frequency region. 22. The attribute changing means adjusts the magnitude of the frequency component so as to save the short time energy of the time signal information.
The signal conversion device described in 17, 18, 19, 20, and 21. 23. The signal conversion apparatus according to claim 22, wherein the attribute changing unit adjusts the magnitude of the frequency component of at least one local peak so as to preserve the short-term energy of the time signal information. . 24. The requantization processing means having a noise shape characteristic for re-synthesized time signal information on a time axis is provided.
The signal conversion device according to 8, 19, 20, 21, 22, 23. 25. The signal conversion device according to claim 24, wherein the noise shape characteristic in said requantization processing means depends on at least one of minimum audibility limit, equal loudness and masking characteristic. 26. The attribute of the attribute changing means is the magnitude of a frequency component.
5, 16, 17, 18, 19, 20, 21, 22, 2
The signal conversion device according to 3, 24, 25. 27. A recording medium on which conversion data converted according to the signal conversion method according to claim 13 is recorded.