JP2007310296A - Band spreading apparatus and method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To spread the band of input signal by making the spectral characteristics of the input signal reflected. <P>SOLUTION: The present invention is related to the band spreading apparatus for spreading the input signal whose frequency band is restricted, so so as to include the out-of-band signal components. The apparatus comprises an out-of-band signal creation means for creating the out-of-band signal component, based on the input signal; an addition ratio determination means for determining an addition ratio of the out-of-band signal component generated, according to the spectral characteristics of the input signal; and a compositing means in which the out-of-band signal component generated, is added to the input signal, after the determined addition ratio has been applied to it. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a band extending apparatus and method, and can be applied to, for example, a case where a wideband audio signal is obtained by extending a frequency band on a receiving side with respect to an audio signal having a narrow frequency band transmitted by communication, broadcasting, or the like. Is.

  For example, the frequency band of an audio signal transmitted by a telephone is 300 Hz to 3.4 kHz, which is inherently much narrower than the frequency band of audio emitted by humans. For this reason, the voice sent through the phone has a poor quality such as being muffled.

In order to solve such a problem, for example, as in the technique described in Patent Document 1, there is a technique of improving the quality by extending the frequency band of the audio signal to the high frequency side on the reception side. In the technology described in Patent Document 1, a higher frequency audio signal (high frequency signal) is generated from the received audio signal (low frequency signal), and the high frequency signal is added to the low frequency signal. Is getting a voice signal. Here, the technique described in Patent Document 1 is characterized in that the addition ratio of the high frequency signal to the low frequency signal can be varied by an external operation.
JP 2000-134162 A

  However, in the technique described in Patent Document 1, the addition ratio is for reflecting the user's preference and is irrelevant to the nature of the target audio signal. Originally, a voice uttered by a human has different spectral characteristics between, for example, voiced sound and unvoiced sound. In the case of voiced sound, there are more low-frequency components than in the high-frequency components. Conversely, in the case of unvoiced sounds, the high-frequency components are larger than in the low-frequency components. In the technique described in Patent Document 1, such characteristics are not reflected, and thus there is a problem in that sufficient quality cannot be improved in a signal after band expansion.

  The present invention has been made in view of the above-described problems, and provides a band expansion apparatus and method that can expand the band of an input signal by reflecting the spectral characteristics of the input signal and obtain a high-quality band expansion signal. It is what I tried to provide.

  According to a first aspect of the present invention, there is provided a band extending apparatus for extending an input signal whose frequency band is limited to include a signal component outside the band. (1) Based on the input signal, the signal component outside the band Out-of-band signal generating means for generating (2) addition ratio determining means for determining the addition ratio of the generated out-of-band signal component according to the spectral characteristics of the input signal, and (3) the generated band And combining means for adding to the input signal after applying the determined addition ratio to the external signal component.

  According to a second aspect of the present invention, an out-of-band signal generating unit, an addition ratio determining unit, and a synthesizing unit are provided in a band extending method for extending an input signal whose frequency band is limited to include a signal component outside the band. (1) The out-of-band signal generating means generates a signal component outside the band based on the input signal, and (2) the addition ratio determining means is generated according to the spectral characteristics of the input signal. And (3) the combining means applies the determined addition ratio to the generated out-of-band signal component and then adds the added ratio to the input signal. It is characterized by.

  ADVANTAGE OF THE INVENTION According to this invention, the spectrum characteristic of an input signal can be reflected, the band of an input signal can be expanded, and a high quality band expansion signal can be obtained.

(A) First Embodiment Hereinafter, a first embodiment in which a band extending apparatus and method according to the present invention are applied to band extension of an audio signal will be described in detail with reference to the drawings.

  FIG. 1 is a block diagram showing the overall configuration of the audio signal band extending apparatus according to the first embodiment.

  In FIG. 1, an audio signal band extending apparatus 100 according to the first embodiment includes an upsampling circuit 101, a spectrum characteristic analysis circuit 102, a high frequency component generation circuit 103, an addition circuit 104, and a multiplication circuit 105. For example, an audio signal s (n) arriving from a distance (not shown) is input to the audio signal band extending apparatus 100 of the first embodiment, and the audio signal s (n) is a narrow-band signal (digital signal). is there.

  The upsampling circuit 101 increases (for example, doubles) the sampling frequency of the input audio signal s (n) using a known method.

  The high frequency component generation circuit 103 generates a component (high frequency component) in a frequency band higher than the frequency band of the input audio signal s (n) from the input audio signal s (n). As a method for generating a high frequency component, for example, the method described in Patent Document 1 and the method described in Japanese Patent Laid-Open No. 9-258787 can be applied. According to the method described in Patent Document 1, the high frequency component generation circuit 103 increases the sampling frequency by inserting zero between each sample of the input audio signal s (n), and then increases the sampling frequency. A band component is generated, and the generated high band component is output as a high band signal su (n) after passing through a bandpass filter that passes the high band.

  The spectral characteristic analysis circuit 102 analyzes the spectral characteristic of the input voice signal s (n) to obtain an adaptive gain g.

  The multiplication circuit 105 multiplies the high frequency signal su (n) output from the high frequency component generation circuit 103 by the adaptive gain g output from the spectrum characteristic analysis circuit 102.

  The adder circuit 104 adds the up-sampled audio signal (low-frequency signal) sd (n) from the up-sampling circuit 101 and the gain-adjusted high-frequency signal su (n) × g from the multiplier circuit 105. The broadband audio signal so (n) is obtained and output.

  FIG. 2 is a block diagram illustrating a detailed configuration of the spectrum characteristic analysis circuit 102 according to the first embodiment.

  The spectral characteristic analysis circuit 102 according to the first embodiment includes a low-frequency bandpass filter circuit 201, a high-frequency bandpass filter circuit 202, power calculation circuits 203 and 204, and a first adaptive gain calculation circuit 205.

  The low-pass bandpass filter circuit 201 extracts, for example, components in the lower half frequency band (for example, 0 to 2 kHz) out of the frequency band (for example, 0 to 4 kHz) of the input audio signal s (n). On the other hand, the high-pass bandpass filter circuit 202 extracts the upper half frequency band (for example, 2 to 4 kHz). An LPF (low-pass filter) may be applied instead of the low-pass bandpass filter circuit 201, and an HPF (high-pass filter) may be applied instead of the high-pass bandpass filter circuit 202.

  The power calculation circuit 203 calculates the power Pl of the output signal from the low-frequency bandpass filter circuit 201, and the power calculation circuit 204 calculates the power Pu of the output signal from the high-frequency bandpass filter circuit 202. Is.

  The first adaptive gain calculation circuit 205 calculates an adaptive gain g from the powers Pl and Pu output from the power calculation circuits 203 and 204. The first adaptive gain calculation circuit 205 calculates the adaptive gain g according to, for example, the equation (1) and outputs it to the multiplication circuit 105 as described above.

g = √ (Pu / Pl) (1)
Here, the expression (1) indicates that the power relationship between the low frequency band and the high frequency band in the band of the input audio signal s (n) is the high frequency generated by the band of the input audio signal s (n) and the higher frequency band. The adaptive gain g is calculated on the assumption that the analog signal can be applied to the power relationship between the band signals.

  In the audio signal band extending apparatus 100 of the first embodiment, the input audio signal s (n) is given to the upsampling circuit 101, the spectrum characteristic analysis circuit 102, and the high frequency component generation circuit 103. The up-sampling circuit 101 up-samples the input audio signal s (n), gives the up-sampled audio signal sd (n) to the adder circuit 104, and the spectral characteristic analysis circuit 102 determines the input audio signal s (n). An adaptive gain g is calculated according to the spectral characteristics and applied to the multiplication circuit 105, and the high frequency component generation circuit 103 generates a high frequency signal su (n) from the input audio signal s (n) and supplies it to the multiplication circuit 105. It is done. As a result, the multiplication circuit 105 multiplies the generated high frequency signal su (n) by the adaptive gain g, and the addition circuit 104 multiplies the multiplication output su (n) × g and the up-sampled audio signal sd (n ) And the addition circuit 104 outputs a wideband audio signal so (n) obtained by extending the band of the input audio signal s (n).

  According to the first embodiment, according to the spectral characteristics of the input voice signal, the voice signal having the band expanded by adaptively changing the addition ratio of the generated high frequency signal is generated. It is possible to obtain a band-extended audio signal with higher quality than before.

(B) Second Embodiment Next, a second embodiment in which the band expansion apparatus and method according to the present invention are applied to the band expansion of an audio signal will be described in detail with reference to the drawings.

  The overall configuration of the audio signal band extending apparatus of the second embodiment can also be expressed in FIG. 1 described above. In the audio signal band extending apparatus of the second embodiment, the detailed configuration of the spectrum characteristic analysis circuit 102 is different from that of the first embodiment.

  FIG. 3 is a block diagram illustrating a detailed configuration of a spectrum characteristic analysis circuit (hereinafter, reference numeral 102A) according to the second embodiment. In FIG. 3, the spectral characteristic analysis circuit 102 </ b> A of the second embodiment includes an LPC analysis circuit 301, an LPC-LSP conversion circuit 302, and a second adaptive gain calculation circuit 303.

  The LPC analysis circuit 301 performs second-order LPC analysis (linear prediction analysis) using the input speech signal s (n), and outputs LPC coefficients a1 and a2.

  The LPC-LSP conversion circuit 302 converts the LPC coefficients a1 and a2 output from the LPC analysis circuit 301 into LSP (line spectrum pair) coefficients b1 and b2, and outputs them.

  The second adaptive gain calculation circuit 303 uses the LSP coefficients b1 and b2 output from the LPC-LSP conversion circuit 302 to calculate the adaptive gain g as described later. The second adaptive gain calculation circuit 303 calculates the adaptive gain g using the following characteristics of the LSP coefficients b1 and b2.

  The LSP coefficients b1 and b2 are parameters corresponding to the frequency and, as illustrated in FIG. 4, are values (frequency) of a portion where the power of the frequency included in the input audio signal s (n) is large. The difference between both LSP coefficients b1 and b2 corresponds to the degree of power concentration, as illustrated in FIG. 4, and a small difference indicates that the frequency component in the vicinity of the LSP coefficient b1 and b2 protrudes from other parts. (See FIG. 4 (C)), a large difference indicates a small difference from other parts (see FIG. 4 (D)). That is, the LSP coefficients b1 and b2 are parameters that can be used to estimate power in a higher frequency range than the band of the input audio signal s (n), and the second adaptive gain calculation circuit 303 calculates the LSP coefficients b1 and b2 The adaptive gain g is calculated using this.

  5 and 6 are explanatory diagrams of a method for calculating the adaptive gain g by the second adaptive gain calculation circuit 303. FIG.

  As shown in FIG. 5, the second adaptive gain calculation circuit 303 sets areas AR1 and AR2 in which the band of the input audio signal s (n) is divided into ½ (0.5) (note that 2 Instead of equally dividing, the two areas AR1 and AR2 may be set by equally dividing into two). Then, the second adaptive gain calculation circuit 303 determines how the two LSP coefficients b1 and b2 belong to the two areas AR1 and AR2.

(Case 1)
When the two LSP coefficients b1 and b2 both belong to the area AR1 (see, for example, FIGS. 4A and 4C), the concentration degree parameter t is calculated according to the equation (2) described later, It is determined whether or not the calculated concentration degree parameter t is smaller than a predetermined value (here, 3). When the calculated concentration degree parameter t is smaller than 3, 1 / t is output as the adaptive gain g, and when the concentration degree parameter t is 3 or more, 1/3 is output as the adaptive gain g.

t = 0.5 / (| b2-b1 |) (2)
The fact that the two LSP coefficients b1 and b2 both belong to the area AR1 is that the portion higher than the band of the input audio signal s (n) has a power higher than the average power of the band of the input audio signal s (n). It is considered to represent weakness. In the example of FIG. 4, FIGS. 4A and 4C correspond. A large concentration degree parameter t indicates that the concentration degree is high. In the example of FIG. 4, the degree of concentration is larger in FIG. 4C than in FIG. When the two LSP coefficients b1 and b2 both belong to the area AR1 and the degree of concentration is high, it is estimated that the power of the high frequency component is much smaller than when the degree of concentration is low. Therefore, the adaptive gain g having the larger concentration degree parameter t is reduced, and the lower limit value is set so that the generated high frequency signal su (t) does not become meaningless.

(Case 2)
When the LSP coefficient b1 and the LSP coefficient b2 belong to different areas, that is, when the LSP coefficient b1 belongs to the area AR1 and the LSP coefficient b2 belongs to the area AR2, the second adaptive gain calculation circuit 303 Outputs an adaptive gain g set to 1.

  When the LSP coefficient b1 and the LSP coefficient b2 belong to different areas, the input audio signal s (n) has a substantially flat spectral characteristic as illustrated in FIG. The band signal su (t) is regarded as having the same gain, and the adaptive gain g is set to 1 and output.

(Case 3)
When the two LSP coefficients b1 and b2 belong to the area AR2 (see, for example, FIG. 4B), the concentration degree parameter t is calculated according to the above-described equation (2), and the calculated concentration degree is calculated. It is determined whether or not the parameter t is smaller than a predetermined value (here, 3). When the calculated concentration degree parameter t is smaller than 3, t is output as the adaptive gain g, and when the concentration degree parameter t is 3 or more, 3 is output as the adaptive gain g.

  The fact that the two LSP coefficients b1 and b2 both belong to the area AR2 is that the portion higher than the band of the input audio signal s (n) has a power higher than the average power of the band of the input audio signal s (n). Consider it strong. In the example of FIG. 4, FIG. 4 (B) corresponds. A large concentration degree parameter t indicates that the concentration degree is high. When the two LSP coefficients b1 and b2 both belong to the area AR2 and the degree of concentration is high, it is estimated that the power of the high frequency component is much higher than when the degree of concentration is low. Therefore, the adaptive gain g having a larger concentration degree parameter t is increased, and an upper limit is set for the adaptive gain g so that the high-frequency signal su (t) × g after adaptation does not become excessively large.

  Also according to the second embodiment, the audio signal with the expanded band is generated by adaptively changing the addition ratio of the generated high frequency signal in accordance with the spectral characteristics of the input audio signal. It is possible to obtain a higher-bandwidth audio signal with higher quality.

  Here, according to the second embodiment, since the adaptive gain is controlled by capturing the spectral characteristics of the input audio signal including the power concentration level, it is generated more than the first embodiment. Higher quality of the band extension voice signal can be expected.

(C) Other Embodiments In the above embodiments, all the components of the spectrum characteristic analysis circuits 102 and 102A have been described as the dedicated components of the spectrum characteristic analysis circuits 102 and 102A. The components that realize the similar function in FIG. 6 may be used as the components of the spectrum characteristic analysis circuits 102 and 102A.

  In general, in an IP telephone using VoIP, an audio signal is encoded and transmitted, and decoded on the receiving side to obtain reproduced audio. Many of the speech coding methods used here use LPC coefficients and LSP parameters in their internal processing, and these LPC coefficients and LSP parameters may be used in the second embodiment. In this case, the LPC analysis circuit 301 and / or the LPC-LSP conversion circuit 302 can be omitted.

  In the first embodiment, the band of the input voice signal is divided into two to obtain the spectrum characteristic (the square root of the power ratio). In the second embodiment, the spectrum characteristic of the input voice signal is 2 Although one for obtaining one LSP parameter has been shown, the value used for calculating the adaptive gain is not limited to two, and may be three or more.

  For example, the power of each divided band is obtained by dividing the band of the input audio signal into three, and the adaptive gain is calculated by extrapolating the power in the band of the generated high frequency signal from the three power values. Also good. Further, for example, the adaptive gain may be calculated by extrapolating the component intensity in the band of the generated high frequency signal from the number of LSP parameters exceeding 2.

  In the second embodiment, the adaptive gain g is determined after converting the LPC coefficient to the LSP parameter. However, the adaptive gain g may be determined based on the relationship between a plurality of LPC coefficients. . Further, other analysis methods for obtaining the spectrum envelope may be used.

  In each of the above embodiments, the case where the signal to be expanded is an audio signal has been described, but other types of signals may be used. For example, an acoustic signal may be used. Further, the extension band is not limited to the high band, and the low band side may be extended.

  In addition, although the above embodiments have been described with an image in which each component is realized in hardware, it is needless to say that all or some of the components may execute processing in software. .

It is a block diagram which shows the whole structure of the audio | voice signal band expansion apparatus which concerns on 1st Embodiment. It is a block diagram which shows the detailed structure of the spectrum characteristic analysis circuit in 1st Embodiment. It is a block diagram which shows the detailed structure of the spectrum characteristic analysis circuit in 2nd Embodiment. It is explanatory drawing of the characteristic of the LSP coefficient in 2nd Embodiment. It is explanatory drawing (1) of the determination method of the adaptive gain in 2nd Embodiment. It is explanatory drawing (2) of the determination method of the adaptive gain in 2nd Embodiment.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 ... Audio signal band expansion apparatus, 101 ... Up-sampling circuit, 102, 102A ... Spectral characteristic analysis circuit, 103 ... High frequency component generation circuit, 104 ... Adder circuit, 105 ... Multiplication circuit, 201 ... Low frequency band pass filter circuit, DESCRIPTION OF SYMBOLS 202 ... High band pass filter circuit, 203, 204 ... Power calculation circuit, 205 ... 1st adaptive gain calculation circuit, 301 ... LPC analysis circuit, 302 ... LPC-LSP conversion circuit, 303 ... 2nd adaptive gain calculation circuit .

Claims (8)

In a band extension device that extends an input signal whose frequency band is limited to include a signal component outside the band,
Out-of-band signal generating means for generating a signal component out of the band based on the input signal;
An addition ratio determining means for determining an addition ratio of the generated out-of-band signal component according to a spectral characteristic of the input signal;
A band extending apparatus comprising: a combining unit that applies the determined addition ratio to the generated out-of-band signal component and then adds it to the input signal. The addition ratio determining means is
A band dividing circuit for dividing the frequency band of the input signal into a plurality of bands;
A signal power calculation circuit for obtaining the signal power of each of the divided bands;
The band extending apparatus according to claim 1, further comprising: a first addition ratio generation circuit that generates the addition ratio from the relationship between the signal powers of the bands. The addition ratio determining means is
An LPC calculation circuit for obtaining a linear prediction coefficient using the input signal;
The band extending apparatus according to claim 1, further comprising: a second addition ratio generation circuit that generates an addition ratio according to the characteristic of the obtained linear prediction coefficient.   4. The band extending apparatus according to claim 3, wherein the LPC calculation circuit is a circuit provided in a decoder that obtains the input signal by decoding processing. In a band extending method for extending an input signal having a limited frequency band to include a signal component outside the band,
An out-of-band signal generation means, an addition ratio determination means and a synthesis means;
The out-of-band signal generating means generates a signal component outside the band based on the input signal,
The addition ratio determining means determines the addition ratio of the generated out-of-band signal component according to the spectral characteristics of the input signal,
The synthesizing means applies the determined addition ratio to the generated out-of-band signal component and then adds it to the input signal.   The addition ratio determining means divides the frequency band of the input signal into a plurality of bands, obtains the signal power of each of the divided bands, and generates the addition ratio from the relationship of the signal power of each band. The band extending method according to claim 5, wherein:   6. The band extending method according to claim 5, wherein the addition ratio determining means obtains a linear prediction coefficient using the input signal, and generates an addition ratio according to characteristics of the obtained linear prediction coefficient. The said addition ratio determination means diverts what is provided in the decoder which obtains the said input signal by a decoding process as a structure which calculates | requires a linear prediction coefficient using the said input signal. The described bandwidth extension method.

Images