JPH10149187A - Audio information extracting device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To exactly extract the audio signal to be extracted from an input signal even when the target audio signal is changed. SOLUTION: The input signal is supplied through a filter 1 with an audio band as a passing band to a frequency analytic part 2 and a volume analytic part 3. At the frequency analytic part 2, a frequency fc of the audio signal to be extracted is extracted and musical scale data are generated from the frequency fc and outputted. At the volume analytic part 3, the volume of audio signal to be extracted is analyzed and volume A of audio signal to be extracted is extracted. Based on the frequency fc and volume A fed back from these parts, the filter 1 changes its filter coefficient so that the passing band can include the frequency of audio signal to be extracted. Even when the frequency of audio signal extracted by the frequency analytic part 2 is changed, the filter coefficient is changed so as to narrow the passing band while following that change so that the audio signal can be exactly extracted from the input signal.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an audio information extracting apparatus for extracting a target audio signal from an input signal.
In particular, the present invention relates to an audio information extracting apparatus that is optimal for efficiently extracting a target sound when several sounds (pitches) are mixed and the changes in the pitches are almost continuous.

[0002]

2. Description of the Related Art Hitherto, for example, in a karaoke apparatus, the pitch, volume, etc. of a specific vocal have been extracted from a song. An audio information extraction device used in such an application usually extracts a frequency band around a target audio signal from a digital signal having a voice band as a pass band from an input signal, and extracts a signal in the extracted band from a signal in the extracted band. The pitch data and the volume data of the target audio signal are extracted using a technique such as pitch extraction.
The pass band of the digital filter used here is determined in consideration of the frequency of the audio signal to be extracted and the amount of change thereof, and the filter coefficient is fixed so as to obtain such a pass band.

[0003]

However, in the above-mentioned conventional audio information extracting apparatus, if the pass band width is set narrow in order to prevent erroneous detection of signals other than the target audio signal included in the pass band of the filter. When the frequency of the audio signal to be extracted changes, there is a problem that the level of the audio signal is reduced and a signal other than the audio signal to be originally extracted is erroneously detected. Also, if the passband is set wide, the frequency of the audio signal to be extracted cannot be specified,
There is a problem that the audio signal to be extracted cannot be accurately extracted from the input signal.

[0004] The present invention has been made in view of such a problem, and even when a target audio signal changes,
It is an object of the present invention to provide an audio information extracting device capable of accurately extracting an audio signal to be extracted from an input signal.

[0005]

A voice information extracting apparatus according to the present invention sets a predetermined frequency band including a frequency of a target voice signal from which voice information is to be extracted as a pass band, and determines from an input signal a frequency band other than the pass band. Filter means for reducing the signal amplitude, and frequency analysis means for analyzing the frequency of the signal in the pass band supplied from the filter means and extracting the frequency of the target audio signal, the filter means and the frequency analysis In the audio information extracting apparatus for extracting the audio information from at least one output of the means, the filter means sets the predetermined frequency band fixed in an initial state as a pass band, and the frequency of the audio signal is extracted. After that, the frequency that follows the change in the frequency of the extracted audio signal and changes the filter coefficient so as to narrow the pass band Characterized in that it is a tracking type filter.

In addition, the apparatus further comprises a sound volume analyzing means for analyzing the sound volume of the audio signal supplied from the filtering means, wherein the filtering means, when the sound volume of the audio signal analyzed by the sound volume analyzing means is lowered. The filter coefficient may be changed so that a passband narrowed following the frequency of the audio signal is initialized to the predetermined frequency band.

Further, an upper limit and a lower limit of the pass band are provided, and the filter means, when exceeding the range of the upper limit and the lower limit, sets the pass band narrowed to follow the frequency of the audio signal to the predetermined value. The filter coefficient may be changed so as to initialize to a frequency band.

According to the audio information extracting apparatus according to the present invention, even if the frequency of the audio signal extracted by the frequency analyzing means changes, the filter means can follow the change and narrow the pass band. Since the filter coefficient changes, the probability of erroneously extracting an audio signal other than the audio signal to be extracted is reduced, and the audio signal to be extracted can be accurately extracted from the input signal.

The change in the audio signal is generally considered to be continuous. However, if the change occurs discontinuously, there is a possibility that the target audio signal may not be extracted by the filter means whose pass band is narrowed. . Therefore, a sound volume analyzing means is further provided, and when the sound volume analyzing means detects that the sound volume of the audio signal has dropped, the sound to be extracted is changed by changing the filter coefficient so as to return the pass band of the filter means to the initial state. Even if the change in the frequency of the signal is not followed, the pass band of the filter means is expanded to include the frequency of the audio signal to be extracted, and the audio signal to be extracted from the input signal is accurately extracted again. Can be.

Further, an upper limit and a lower limit are set in the pass band of the filter means, the filter coefficient is changed so that the pass band falls within this range, and the filter coefficient is initialized when the range exceeds this range. . For example, when the upper limit and the lower limit are made to coincide with the upper limit and the lower limit of the audio band, the filter means operates only in the audio band, thereby preventing erroneously following a signal outside the audio band. be able to.

[0011]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a configuration of a voice information extracting device according to one embodiment of the present invention. An input signal including an audio signal to be subjected to pitch and volume extraction is input to a filter 1 having an audio band as a pass band, where amplitude components other than the audio band are removed, and a frequency analysis unit 2 and a volume analysis unit 3 are used. Supplied to

[0012] frequency analysis unit 2 analyzes the frequency of the output signal of the filter 1 extracts the frequency f _c of the audio signal to be extracted based on the analysis result, the pitch data from the extracted frequency f _c Generate and output. The frequency analysis unit 2, the frequency f _c of the extracted audio signal is fed back to the filter 1, changes the filter coefficient so narrow down the peripheral portion of the frequency f _c of the audio signal to be extracted passband of the filter 1 Let it.

On the other hand, the sound volume analyzing unit 3 analyzes the sound volume of the audio signal to be extracted from the output signal of the filter 1 and feeds back the sound volume A, which is the result of the analysis, to the filter 1.
The filter coefficient is changed so that the pass band of the filter 1 includes the frequency of the audio signal to be extracted.

FIG. 2 is a block diagram showing a configuration of the filter 1. The filter 1 includes, for example, a second-order IIR (InR) in which a delay circuit 11, a multiplier 12, and an adder 13 are combined.
finite Inpulse Response) filter. Each filter coefficient a ₀ , a ₁ ,
a _2, b _1, b ₂ is controlled by the filter coefficient setting unit 14, so that the filter 1 functions as follows shaped filter to follow the frequency f _c. That is, these filter coefficients a ₀ , a ₁ , a ₂ , b ₁ , b ₂ are represented by the following equations, respectively.

[0015]

(Equation 1)

[0016] In this case, T is the sampling period, ω ₀ is the cut-off angular _{frequency, K = ω 0 / Q,} T = 1 / f s,
Q = f _c / B, to satisfy the ω ₀ = 2πf _c. Since the sampling frequency f _s is a fixed value (e.g., 8 kHz),
It is given a center frequency f _c and the width B of the pass band of the filter 1, the filter coefficients a ₀ in the filter 1 having such pass _{band, a 1, a 2, b} 1, b 2 are determined. Therefore,
In the initial state, it is set at a center frequency f _c = 400 Hz, the filter coefficient setting unit 14 so that the width B = 600 Hz pass band of the filter 1 filters 1 as shown in FIG. 3 (a).

The filter coefficient a in the initial state₀, A₁, A_Two,
b₁, B_TwoIs set in the initialization coefficient register 21.
You. The extracted frequency f is stored in the filter coefficient memory 22.
_cIs the center frequency, and the bandwidth is set to B = 200 Hz, for example.
Filter coefficient a to be set₀, A₁ , A_Two, B₁, B_TwoIs each f_c
It is tabulated corresponding to. This filter coefficient method
Filter coefficient a read from memory 22₀, A₁,
a_Two, B₁, B_TwoAs a result, the filter 1 becomes as shown in FIG.
Sea urn f_cAnd the passband has a narrow width that follows. Week
The wave number comparator 23 calculates the extracted frequency f_cIs f_cL<F_c<
f_cHOutputs “1” when the condition is satisfied, and “0” otherwise.
Is output. The volume comparator 24 outputs the extracted volume data.
A is A_LIf <A is satisfied, “1” is output.
Outputs “0”. Frequency comparator 23 and volume comparator 2
4 is "1", the AND gate 25 is activated.
Set switch 26 to a side to set filter 1 to follow the frequency
In other cases, set the switch 26 to the b side.
To set the filter 1 to the initial state. In addition, such
Filter 1 is based on DSP (Digital Signal Processor)
Can also be configured.

FIG. 4 shows the frequency analysis unit 2 and the volume analysis unit 3
FIG. 3 is a block diagram showing the configuration of FIG. The pitch of the output signal of the filter 1 supplied to the frequency analysis unit 2 is extracted by the pitch extraction unit 31. As a method of pitch extraction,
A known zero-crossing interval detection method, an autocorrelation method, a combination method thereof, or the like can be used. After the pitch-extracted signal is appropriately removed of the vibrato component by the LPF 32, the signal is converted into pitch data by the frequency pitch converter 33, and is fed back to the filter 1 as described above. Frequency f _c is the reciprocal of the extracted pitch is expressed by the following equation using the octave data P and chromatic data Q as a deviation from the fundamental frequency 32.7032Hz.

[0019]

F _c = 32.7032 × 2 ^{(P + Q / 12)} [Hz]

Therefore, the frequency pitch conversion unit 33 calculates pitch data based on the following equation.

[0021]

## EQU3 ## P + (Q / 12) = log ₂ (f / 32.77032)

The octave data P is an integer and is determined so that the absolute value of Q is minimized. The frequency fc and pitch data of the audio signal thus obtained are stored in registers (REG) 34 and 35, respectively.

The output signal of the filter 1 supplied to the volume analysis unit 3 is converted to an absolute value by an absolute value circuit (ABS) 41 and smoothed by a smoothing unit 42. The volume A of the audio signal obtained in this manner is stored in the register (RE
G) 43 is stored.

Next, the operation of the audio information extracting apparatus thus configured will be described. First, in the initial state, the filter coefficient setting unit 14 converts the filter coefficients a ₀ , a ₁ , a ₂ , b ₁ , and b ₂ in the initialization coefficient register 21 into the respective multipliers 12.
To give, as shown in FIG. 3 (a), the center frequency f _c = 400 Hz pass band of the filter 1, the bandwidth B = 600H
Set to z. As a result, the frequency component having the largest amplitude in this pass band is extracted by the frequency analysis unit 2 as the target audio signal. If the frequency f _c of the audio signal is extracted, the condition of the frequency comparator 23 and volume comparator 24 is satisfied, the filter coefficients from the filter coefficient memory 22 switch 26 I switched _{a 0, a 1, a 2} , b ₁ , b ₂
Is supplied to each multiplier 12. As a result, FIG.
As shown in, the pass band to follow the frequency f _c of the band width B is narrowed and extraction is changed.

As a result of the following, when the extracted frequency f _c exceeds a preset upper limit f _cH , or when the frequency f _{c
When c} becomes smaller than the preset lower limit value _fcL , the output of the frequency comparator 23 becomes "0", the switch 26 is switched, and the filter 1 is initialized. As a result, it is possible to prevent the audio signal to be extracted again from being extracted in the initial state and to keep following the signal other than the audio signal by mistake.

Further, there is no problem when the frequency f _c is continuously changed, for example, as shown in FIG. 5 (a), as the frequency is suddenly changed to f _c 'from f _c In this case, the volume A falls below the predetermined level A _L. In such a case, the output of the volume comparator 24 may be “0” because there is a possibility that the change in the pitch of the audio signal to be extracted may not be followed.
And the switches are switched so that the filter coefficients a ₀ , a ₁ , a ₂ ,
b ₁ and b ₂ are initialized. As a result, FIG.
As shown in (b), the frequency f _c ′ after the change is extracted as the frequency of the audio signal to be extracted again.
As shown in FIG. 5 (c), the filter 1 again may be accurately extracted audio signal to follow the frequency f _c '.

The audio information to be extracted is a filter 1
The audio signal passing through a specific band and its volume data may be output from the frequency analysis unit 2 and may be pitch data or audio frequency, and may be appropriately selected according to the purpose of use.

[0028]

As described above, according to the present invention,
Even if the frequency of the audio signal extracted by the frequency analysis means changes, the filter coefficient of the filter means changes so as to follow the change and narrow the pass band. The probability of erroneously extracting a signal is reduced, and an audio signal to be extracted can be accurately extracted from the input signal.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of a voice information extraction device according to an embodiment of the present invention.

FIG. 2 is a block diagram showing a configuration of a filter in the same device.

FIG. 3 is a diagram for explaining filter coefficients in the same filter.

FIG. 4 is a block diagram illustrating a configuration of a frequency analysis unit and a volume analysis unit in the embodiment.

FIG. 5 is a diagram showing a pass band following when the frequency of an audio signal to be extracted suddenly changes.

[Explanation of symbols]

 1 ... Filter, 2 ... Frequency analysis unit, 3 ... Volume analysis unit.

 ──────────────────────────────────────────────────続 き Continued on the front page (72) Inventor Ken Daishoji 10-1 Nakazawacho, Hamamatsu City, Shizuoka Prefecture Yamaha Corporation

Claims (3)

[Claims] 1. A filter means for setting a predetermined frequency band including the frequency of a target audio signal from which audio information is to be extracted as a pass band, and reducing signal amplitudes other than the pass band from an input signal from the input signal, and supplying from the filter means. Frequency analysis means for analyzing the frequency of the signal of the passband obtained and extracting the frequency of the target audio signal, wherein the audio for extracting the audio information from the output of at least one of the filter means and the frequency analysis means In the information extraction device, the filter unit may be configured such that, in an initial state, the predetermined fixed frequency band is a pass band, and after the frequency of the audio signal is extracted, a change in the frequency of the extracted audio signal is performed. Speech information characterized by being a frequency-following filter for following and changing a filter coefficient so as to narrow the pass band. Extraction device. 2. The sound processing apparatus according to claim 1, further comprising: a sound volume analyzing unit configured to analyze a sound volume of the audio signal supplied from the filtering unit, wherein the filtering unit includes: 2. The audio information extracting apparatus according to claim 1, wherein a filter coefficient is changed so that a passband narrowed following the frequency of the audio signal is initialized to the predetermined frequency band. 3. An upper limit and a lower limit of the pass band are set, and the filter means, when exceeding the range of the upper limit and the lower limit, sets the pass band narrowed in accordance with the frequency of the audio signal to the predetermined value. 3. The audio information extracting apparatus according to claim 1, wherein the filter coefficient is changed so as to initialize to a frequency band.