JP2002044793A - Method and apparatus for sound signal processing - Google Patents

Abstract

PROBLEM TO BE SOLVED: To achieve enhancement and suppression for only specified components of a positioned signal at a specified position without loss of stereo feelings for stereo sound signals. SOLUTION: In a method for sound signal processing, inputted sound signals in plural channels are frequency-analyzed for every channel and also the specified components to be enhanced and suppressed among the channels are designated. An operational processing among complex spectrum channels composing the components is processed only for the area of the components among sound signals in frequency areas of every channel, so that the components are enhanced and suppressed, finally the sound signal in the frequency area is respectively put back to time areas as to be outputted.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an audio signal processing method and apparatus for emphasizing or suppressing a specific component signal from a stereo audio signal.

[0002]

2. Description of the Related Art Conventionally, for example, for the purpose of canceling a center-localized vocal with respect to a stereo source, an arithmetic processing for subtracting a right channel audio signal from a left channel audio signal in a time domain (hereinafter referred to as "LR processing"). ) Is often performed. For example, a device such as a filter processing is used to limit the LR processing target to an audio band (for example, JP-A-5-35283).

[0003]

However, in such LR processing in the time domain, even if the input audio signal is a stereo signal, the output signal is monaural in principle at least in the audio band. For this, there is a technique (Japanese Patent Laid-Open No. 7-64577) in which a monaural signal is expanded by pseudo-stereo processing, for example, but there is a problem that the original stereo feeling of the source is impaired.

SUMMARY OF THE INVENTION The present invention has been made in view of such a problem, and does not greatly impair the sense of stereo for a stereo sound signal. It is an object of the present invention to provide an audio signal processing method and apparatus capable of emphasizing or suppressing (cancelling) only the component.

[0005]

A sound signal processing method according to the present invention is characterized in that a frequency analysis is performed for an input sound signal of a plurality of channels for each channel, and a sound signal in a frequency domain is output for each channel. A complex component that specifies a specific component to be emphasized or suppressed in the input audio signals of the plurality of channels, and configures the specific component only in a specified component region of the audio signal in the frequency domain of each channel. A specific component is emphasized or suppressed by performing the arithmetic processing between the channels, and the acoustic signal in the frequency domain of each channel in which the specific component is emphasized or suppressed is returned to the time domain by the inverse transform so as to be output. It is characterized by having done.

Further, the audio signal processing apparatus according to the present invention comprises a plurality of frequency analysis means for frequency-analyzing the input audio signals of a plurality of channels for each channel and outputting an audio signal in the frequency domain for each channel. A specific component specifying means for specifying a specific component to be emphasized or suppressed from the input audio signal, and a channel of a complex spectrum constituting the specific component only in a region of the specific component specified by the specific component specifying means. An inter-specific-component operation unit that emphasizes or suppresses a specific component by performing an operation process between
A plurality of inverse transforming means for returning the acoustic signal in the frequency domain of each channel in which the specific component is emphasized or suppressed by the computing means between the specific components to the time domain by inverse transform and outputting the signal. .

According to the present invention, only the audio signal of a specific component determined by, for example, the pitch or the like is emphasized or suppressed from the audio signal of each channel in the frequency domain, and the audio signals other than the component do not change significantly. So
The original stereo feeling is hardly lost.

In a typical embodiment of the present invention, when the input acoustic signal has a harmonic structure, a component determined by its pitch is designated as the specific component to be emphasized or suppressed. At this time, only when the pitch is a pitch in a preset range, by specifying a component determined by the pitch as a specific component to be emphasized or suppressed, an acoustic signal having a desired harmonic structure is obtained. Can be emphasized or suppressed, and the number of calculation targets can be reduced at the same time. Also, of the components determined by the pitch, the amount of calculation can also be reduced by designating only components included in a predetermined frequency range and / or components larger than a predetermined level as specific components.

When the localization position is known or can be estimated in advance, the operation between channels is performed by the localization position,
A coefficient to be assigned to each channel can be determined. In this case, the acoustic signals localized at other positions are hardly affected by the arithmetic processing. For this reason, there is an advantage that the influence when a signal localized at a different position happens to overlap with the specific component is small. When the localization positions of the audio signals of a plurality of channels are not known in advance, the localization position is estimated from the ratio of the audio signals between the channels of the specified specific component, and the coefficient in the calculation between the specific components is estimated from the estimation result. Should be adjusted.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a configuration of an audio signal processing device according to an embodiment of the present invention. The left and right channel stereo audio input signals S _IL and S _IR are input to frequency analysis units 10 and 20, respectively, where they are subjected to frequency analysis, and a specific component designation unit 30 is also provided.
The specific component to be input and designated is extracted. Analysis results of frequency analysis units 10 and 20 and specific component designation unit 30
Is input to the inter-specific-component calculation unit 40, and the emphasis / suppression processing by the calculation between the components in the frequency domain is executed only for the specified component specified here. The calculation results are input to the inverse converters 50 and 60, respectively, and
/ R is returned to a time domain waveform for each channel and output as stereo audio output signals S _OL and S _OR .

Here, the frequency analysis section 10 includes a time division section 11 for time-dividing the time waveform of the audio input signal _SIL into analysis units of a predetermined length, and a time-division time waveform such as a Hamming window. And a FFT operation unit 13 that performs frequency analysis on the windowed time waveform by FFT (fast Fourier transform) processing. Similarly, the frequency analysis unit 20 includes a time division unit 21, a windowing unit 22, and an FFT operation unit 23.

The specific component specifying section 30 includes an adder 31 for adding the audio input signals S _IL and S _IR ,
Pitch extraction unit 3 for extracting a pitch from the signal added in step 3
2 and a harmonic component specifying unit 33 for specifying a component based on the harmonic structure from the pitch extracted by the pitch extracting unit 32.

The inverse transform section 50 includes an inverse FFT section 51 for converting a signal in the frequency domain, which is a result of the operation in the inter-specific-component arithmetic section 40, into a signal in the time domain, and a combining section 52 for connecting analysis units. It is configured. Inversion unit 60
Also includes an inverse FFT unit 61 and a combining unit 62.

Next, the operation of the audio signal processing apparatus thus configured will be described. Here, an example will be described in which a sound having a harmonic structure localized at the center of the stereo audio input signals S _IL and S _IR is suppressed. The stereo audio input signals S _IL and S _IR are respectively _supplied to the frequency analysis unit 10 and
It is input to 20 and subjected to frequency analysis. FIG. 2 (a),
FIG. 2B is a diagram illustrating an example of a frequency analysis result by FFT, FIG. 2A is a diagram illustrating an L-channel frequency analysis result output from the frequency analysis unit 10, and FIG.
FIG. 9 is a diagram illustrating a frequency analysis result of an R channel output from the frequency analysis unit 20. The frequency analysis method is FF
Although not limited to T, in the case of FFT, it is actually a complex spectrum. In this example, for the sake of simplicity, it is shown that the stereo audio input signals S _IL and S _IR include only sounds having two harmonic structures, one of which is located at the center and the other is located at the center. Shows an example in which the components are localized from the left. However, of course, a component having no harmonic structure may be included.

On the other hand, the specific component specifying section 30 specifies a specific component based on the harmonic structure by the processing shown in FIG. First, the specification information of the previous specific component is cleared (S1), and the pitch frequency P is extracted from the newly input voice input signal of the analysis unit (S2). That is, L is used for pitch extraction.
The signals of the channel and the R channel are added in the time domain by the adder 31, and the frequency is analyzed by the pitch extraction unit 32. FIG. 2C is a diagram illustrating a result of frequency analysis of L + R. In this example, the result of the frequency analysis is such that the component of the center localization is emphasized. As a method of pitch extraction, for example, a method of extracting a pitch of a signal that is most dominant to a monaural signal (for example, Japanese Patent Laid-Open No. 5-19793) can be used. In this method, a prediction residual signal is obtained by an inverse filter, pitch frequency candidates based on spectral products are extracted by a set number, and a pitch frequency is determined based on a frequency having the maximum frequency.
Further, for example, a method of detecting a plurality of pitches in which all frequency components exceeding a predetermined fundamental tone level and / or harmonic level are set as pitch frequencies (for example, Japanese Patent Laid-Open No. 6-20262).
No. 7), the same processing can be performed for a plurality of pitches.

When the pitch is extracted, each harmonic position can be estimated as a position of an integral multiple of the fundamental frequency from its harmonic structure. Therefore, as shown in FIG. 2D, the region of the fundamental frequency and each harmonic position ( Specific component) can be set. In the case where the pitch is not extracted because the target sound (for example, vocal) does not actually exist (S3), since the area is not set, it is not necessary to perform the subsequent arithmetic processing between the specific components ( S12). Also, by performing processing only when the pitch is within a certain range, it is possible to prevent components other than a specific component (for example, audio) from being removed, and at the same time, reduce the amount of calculation by limiting the processing target. (S4). If multiple pitches can be detected,
The processing target may be limited in consideration of the level of each overtone and its time change.

The width of each component in the frequency direction is as follows:
The determination can be made in consideration of the main lobe width determined by the window function used in the windowing units 12 and 22 of the frequency analysis units 10 and 20. The number of overtones to be processed may be determined based on prior information about the target and in consideration of actually extracted pitch information, the frequency of each overtone, the actual level, and the like. For example, FIG.
In the processing of (1), the upper limit value of the number of harmonics and the upper limit value and the lower limit value of the frequency are determined in advance, i is used as a parameter of the number of harmonics (S5, S6), and f = i × p is calculated (S8). Until the frequency exceeds the upper limit (S7), or after the frequency f becomes equal to or higher than the predetermined lower limit (S9), and exceeds the predetermined upper limit (S10). Around the position specified as a specific component (S
11). If i exceeds a preset upper limit, or f
When the value exceeds the preset upper limit, the designation of the specific component ends (S13).

The inter-specific-component operation unit 40 performs the processing only on the area set in this manner (the area shown in FIG. 2D) as conceptually shown in FIGS. 2A and 2B. Performs arithmetic processing between spectra. Performing the operation between specific components in the frequency domain can be described as follows when FFT is used. That is, the complex spectrum of the i-th component of the L channel is XL [i], the complex spectrum of the i-th component of the R channel is XL [i], and the complex spectra of the respective channels of the operation result are YL [i] and YL, respectively.
If [i], the components contained in the specific components are as follows:

[0019]

## EQU1 ## YL [i] = αLL × XL [i] + αRL × XR
[I] YR [i] = αLR × XL [i] + αRR × XR [i]

In addition, for components not included in the specific components,

[0021]

## EQU2 ## YL [i] = β × XL [i] YL [i] = β × XR [i]

It is assumed that However, αLL, αRL, αLR, αRR,
β is a coefficient determined by the type of suppression / emphasis and localization information. β can be used to make overall level adjustments to avoid clipping in enhancement. In the case of suppression, β may be 1.0. A set of coefficients (αLL,
αRL, αLR, αRR, β) can be determined as follows.

[0023]

## EQU3 ## When a specific component of the central location is emphasized (0.5, 0.5, 0.5, 0.5, 0.5) A specific component of the central location is suppressed by LR (1.0, -1.0, 1.0, -1.0, 1.0) When the specific component of the central location is suppressed by RL (-1.0, 1.0, -1.0, 1.0, 1.0) When mixed (1.0, -1.0, -1.0, 1.0, 1.0) or (-1.0, 1.0, 1.0) 1.0, −1.0, 1.0)

In the suppression, when the localization is not exactly at the center, and when the localization is known in advance, the value of each coefficient may be adjusted according to the information. For example, L: R
= 7: 8, it is known that the sound is localized from R.

[0025]

ΑLL / αRL = αLR / αRR = −8 / 7 β = 1.0

It is sufficient to set a value such that Furthermore,
When the localization is not notified in advance, when the localization information is calculated based on the ratio included in the L / R of the specific target component included in the actual data, and the set of α values is known in advance accordingly. It may be set in the same manner as described above. actually,
For each region designated as a specific component, the amplitude of the peak of the spectrum is determined for each of the L / R, and in the case of the harmonic component, for example, the amplitudes of, for example, the main 1 to n harmonics are determined and averaged. Localization estimation is possible. n
May be determined based on information on the target sound source.

Now, in the example of FIG. 2, when the coefficient of or 3 is used, the target component localized at the center is canceled, and the component localized leftward is not canceled. In the case of this example, among the sounds having the harmonic structure from the left, the tenth harmonic is accidentally included in the region of the seventh harmonic of the set specific component. However, since this component is not centralized, the result of LR is not zero. Therefore, the components are affected by the calculation but are not offset. In such a point, it is different from the fact that the offset result becomes 0 and that the set area is made 0. In addition, since the components other than the component designated as the specific component have basically the same value in each of L / R, the result of performing the calculation between the specific components in this manner is as follows.
As shown in FIG. 8E for L, and FIG. 10F for R, only the dominant component localized in the center is suppressed, and the component located from the left is almost It remains as it is.

The audio signals in the frequency domain of each channel in which the audio signal of the specific component is emphasized or suppressed are inversely converted into audio signals in the time domain by the inverse converters 50 and 60, and only the specific component is emphasized or suppressed. Suppressed audio output signal S
_OL and S _OR can be obtained.

[0029]

As described above, according to the present invention,
From the audio signal of each channel in the frequency domain, only the audio signal of a specific component determined by, for example, the pitch or the like is emphasized or suppressed, and the audio signal other than the component does not change at all, so that the original stereo feeling is impaired. The effect that there is little is produced.

[Brief description of the drawings]

FIG. 1 is a block diagram of an audio signal processing device according to an embodiment of the present invention.

FIG. 2 is a diagram showing a processing result of each unit of the apparatus.

FIG. 3 is a flowchart showing a method for specifying a specific component in the apparatus.

[Explanation of symbols]

10, 20 ... frequency analysis unit, 11, 21 ... time division unit, 1
2, 22: windowing unit, 13, 23: FFT operation unit, 30
... Specific component designator, 31 ... Adder, 32 ... Pitch extractor, 33 ... Harmonic component designator, 40 ... Calculator between specific components,
50, 60: inverse transform section, 51, 61: inverse FFT section, 5
2,62 ... Synthesis unit.

Claims (7)

[Claims] 1. An input audio signal of a plurality of channels is subjected to frequency analysis for each channel to output an audio signal in a frequency domain for each channel, and emphasis or suppression of the input audio signal of the plurality of channels is performed. Specifying a specific component to be performed, and for only the specified specific component region of the audio signal in the frequency domain of each channel, the specific component is obtained by performing an arithmetic process between the channels of the complex spectrum constituting the specific component. A sound signal processing method characterized by emphasizing or suppressing the sound signal in the frequency domain of each channel in which the specific component has been emphasized or suppressed and returning the sound signal to the time domain by inverse transform. 2. The audio signal according to claim 1, wherein when the input audio signal has a harmonic structure, a component determined by the pitch is designated as the specific component to be emphasized or suppressed. Processing method. 3. The component according to claim 2, wherein a component determined by the pitch is designated as the specific component to be emphasized or suppressed only when the pitch is within a predetermined range. Sound signal processing device. 4. A component determined by the pitch,
4. The acoustic signal processing method according to claim 2, wherein only components included in a predetermined frequency range and / or components larger than a predetermined level are designated as specific components. 5. When the localization positions of the audio signals of the plurality of channels are not known in advance, the localization positions are estimated from the ratio of the audio signals between the channels of the specified specific component, and the specific components are determined based on the estimation result. 5. A coefficient according to claim 1, wherein the coefficient is adjusted.
The sound signal processing method according to any one of the preceding claims. 6. A plurality of frequency analysis means for frequency-analyzing an input audio signal of a plurality of channels for each channel and outputting an audio signal in a frequency domain for each channel; A specific component specifying means for specifying a specific component to be suppressed; and specifying only a region of the specific component specified by the specific component specifying means by performing an arithmetic processing between channels of a complex spectrum constituting the specific component. Calculating means for emphasizing or suppressing the component; and a plurality of means for returning the sound signal in the frequency domain of each channel in which the specific component has been emphasized or suppressed by the calculating means between the specific components to the time domain by inverse transform, and outputting the signal. An acoustic signal processing device comprising: an inverse conversion means. 7. An adder for adding the input audio signals of the plurality of channels, and a pitch extracting means for extracting a pitch based on a harmonic structure from the audio signals added by the adder. 7. A harmonic component specifying means for obtaining a harmonic component based on the pitch extracted by the pitch extracting means and designating the harmonic component as the specific component. Sound signal processing device.