JP4783412B2 - Signal broadening device, signal broadening method, program thereof, and recording medium thereof - Google Patents

Description

  The present invention is used for voice communication in an IP network, for example, and relates to a signal broadening apparatus that obtains a pseudo wideband signal from a narrowband signal, a signal widening method, a program thereof, and a recording medium thereof.

  The frequency band of voice signals that can be transmitted by a conventional telephone system is about 300 Hz to 3.4 kHz. The purpose of the speech coding technology of the conventional telephone system is to minimize the amount of transmission parameters, and it is impossible to obtain speech exceeding the frequency band of the coded speech signal. By the way, with the recent development of acoustic technology and the development of digital signal processing technology, the voice quality of equipment used in daily life has improved. In such a situation, for example, there is a voice that demands high sound quality for telephone sound quality. For the purpose of responding to such demands, a speech pseudo-wideband apparatus and its method are used.

  Therefore, as described in Patent Document 1, a narrowband audio signal whose band is limited is replicated in a higher frequency band and is synthesized as a highband signal, thereby synthesizing the narrowband signal and the highband signal. A signal broadening method for creating a broadband signal has been proposed. FIG. 1 shows an example of the functional configuration of a signal broadbanding device used in the signal broadbanding method of Patent Document 1.

  As shown in FIG. 1, the signal broadening apparatus 100 of the prior art 1 includes an upsampling unit 101, a full wave rectification unit 102, STFT analysis units 103 and 105, a bandpass filter unit 104, a duplication unit 106, and a gain adjustment unit 107. And 108, an STFT synthesis unit 109, and an addition unit 110.

  The upsampling unit 101 upsamples, for example, a narrowband audio signal sampled at 8 kHz from an audio signal of 8 kHz sampling to 16 k sampling, for example. The full-wave rectification unit 102 performs full-wave rectification (nonlinear processing) on the upsampled audio signal, and obtains a wideband audio signal having a high correlation with the narrowband audio signal, for example, including a harmonic component of the basic frequency of the audio. The STFT analysis unit 103 performs a STFT (short-time Fourier) analysis of the full-wave rectified audio signal every time a fixed time (10 to 20 msec) elapses, and obtains a frequency domain audio signal S.

  The bandpass filter unit 104 performs bandpass filtering on the frequency domain audio signal S to extract a complex spectrum component of 300 Hz or less that does not exist in the narrowband audio signal, while removing a complex spectrum component of 60 Hz or less. That is, the band-pass filter unit 104 obtains a low-frequency signal (60 to 300 Hz) frequency-domain audio signal S ′ that is not included in the narrow-band audio signal. The gain adjustment unit 107 performs gain adjustment by multiplying the bandpass filtered low-frequency complex spectrum by a certain gain.

  On the other hand, the STFT analysis unit 105 performs STFT analysis on the up-sampled audio signal every predetermined time (10 to 20 msec) to obtain a frequency domain wideband signal T. The duplication unit 106 copies the low frequency band signal T as a complex spectrum of the high frequency band. For example, the 10th to 40th frequency domain wideband signals T are used as the 100th to 130th frequency domain wideband signals T. In addition, 0 is assigned to all other parts of the complex spectrum. As a result, a high frequency domain wideband signal T ′ not included in the narrowband audio signal is obtained. The gain adjusting unit 108 multiplies the high frequency domain wide band signal T ′ by a constant magnification, thereby performing gain adjustment.

  The STFT synthesis unit 109 performs STFT synthesis on the frequency domain audio signal S ′ from the gain adjustment unit 107 and the high frequency domain wideband signal T ′ from the gain adjustment unit 108. As a result, low-frequency and high-frequency audio signals that are not included in the narrowband audio signal can be obtained.

Adder 110 adds the audio signal from upsampling unit 101 and the audio signal synthesized by STFT from STFT synthesis unit 109. As a result, a pseudo-wideband audio signal having a low frequency and a high frequency not included in the narrowband audio signal can be obtained.
Japanese Patent No. 3301473

  A telephone used in a conventional telephone system encodes and transmits a narrowband audio signal collected by a handset having IRS characteristics standardized by ITU-T. Here, the IRS characteristic indicates a gentle high-pass filter type frequency characteristic as shown in FIG. That is, a signal having a large IRS characteristic indicates that the intensity on the low frequency side is small. A narrowband audio signal collected by a handset having IRS characteristics has a spectrum shape in which a signal on the low frequency side (hereinafter referred to as “low frequency side”) is attenuated. That is, a large IRS characteristic means that the low frequency side is attenuated. When a wideband sound is artificially created from the narrowband sound signal in which the low-frequency side signal is attenuated by the method shown in the related art 1, the created wideband sound is attenuated in the low frequency range and is unnatural for hearing. There was a problem of becoming.

The signal broadening apparatus of the present invention includes a frequency conversion unit, an analysis unit, a gain calculation unit, a multiplication unit, a duplication unit, a combination unit, and a frequency inverse conversion unit. The frequency conversion unit generates a low-frequency signal by converting the narrow-band signal into the frequency domain. The analysis unit performs first-order and second-order percoll coefficients by performing a percall analysis on the narrowband signal. When the first-order Percoll coefficient is less than or equal to a predetermined first threshold value or the second-order Percoll coefficient is less than or equal to a predetermined second threshold value , the gain calculation unit determines each frequency of the low-frequency signal. as the corresponding gain factor, than the value corresponding to the high frequency side of the low-frequency signal and outputs a low-frequency side higher gain factor value corresponding to the low frequency signal, in other cases, the low frequency signal high Gain output when the first-order Percoll coefficient is equal to or lower than a predetermined first threshold or the second-order Percoll coefficient is equal to or lower than a predetermined second threshold as a gain coefficient corresponding to each frequency on the band side A gain coefficient equal to the value of the coefficient is output, and as a gain coefficient corresponding to each frequency on the low frequency side of the low frequency signal, the first-order percoll coefficient is equal to or lower than a predetermined first threshold or the second-order percoll The coefficient is less than or equal to a predetermined second threshold A gain coefficient equal to the value of the gain coefficient output in a certain case is output, and the first-order percoll coefficient is equal to or lower than a predetermined first threshold as a gain coefficient corresponding to each frequency on the low frequency side of the low frequency signal. A gain coefficient smaller than the value of the gain coefficient that is output when the second-order Percoll coefficient is equal to or smaller than a predetermined second threshold is output . The multiplication unit multiplies the gain coefficient for each frequency of the low frequency signal to generate an enhanced low frequency signal. The replicating unit generates a high frequency signal by replicating part or all of the low frequency signal. The coupling unit arranges the emphasized low frequency signal on the low frequency side and arranges the high frequency signal on the high frequency side, and combines them to generate a pseudo wideband frequency signal. The frequency inverse conversion unit outputs the pseudo wideband signal by converting the pseudo wideband frequency signal into the time domain.

  According to the present invention, the low-frequency signal is multiplied by the low-frequency gain of the narrow-band signal in which the low-frequency signal is attenuated, and then a pseudo-wideband signal is created. As a result, the created wideband signal has a flat frequency characteristic and can be quasi-broadband naturally natural.

  The best mode for carrying out the invention will be described below. In addition, the same number is attached | subjected to the process which performs the same part and the process which has the same function, and duplication description is abbreviate | omitted.

  FIG. 3 shows a functional configuration example of the signal broadening apparatus 200 of the first embodiment, and FIG. 4 shows a processing flow. In this example, the signal broadening apparatus 200 includes a frequency conversion unit 210, a storage unit 290, a duplication unit 230, a multiplication unit 220, a combining unit 240, and a frequency inverse conversion unit 250. FIG. 5 shows a frequency spectrum of a signal generated by each component. 5A to 5D, the horizontal axis indicates the frequency, and the vertical axis indicates the amplitude power.

  First, the narrowband signal is a signal in the time domain, and is input to the input terminal 205 for each frame having D samples (D is an integer of 1 or more). Here, the narrow band signal is a signal having a narrow frequency band, and the signal is, for example, a voice signal of human voice. Subsequent processing is performed for each frame. The number of samples D may be a predetermined value or a value obtained as a result of analyzing the input narrowband signal. In the former case and the latter case, the number of samples D may be the same in all frames, or the number of samples D may differ from frame to frame. The input narrowband signal in the time domain is represented as In (t) (t = 0, 1,..., D-2, D-1). t is a time index.

  Next, the frequency conversion unit 210 converts the input narrowband signal In (t) from the time domain to the frequency domain for each frame, so that the lowband signal InF (k) (k = 0, 1,. , D-2, D-1) are generated and output (step S2). The low-frequency signal InF (k) is a frequency-domain signal, k represents a frequency index, and a smaller value of k represents a low-frequency signal, and a larger value represents a high-frequency signal. Further, as an example of the frequency conversion method, DCT (Discrete Cosine Transform) or MDCT (Modified Discrete Cosine Transform) may be used. However, when MDCT is used, D low-frequency signals are created by overlapping each frame from two frames (2D) of narrowband signals. Moreover, the frequency spectrum of the output signal of the frequency converter 210 is shown in FIG. 5A. As shown at a position α in FIG. 5A, the low frequency signal InF (k) is attenuated on the low frequency side.

The storage unit 290 stores a gain coefficient predetermined for each frequency of the low frequency signal InF (k) in advance, and the gain coefficient is output to the multiplication unit 220. Then, the multiplication unit 220 multiplies the low frequency signal InF (k) by the gain coefficient to generate the enhanced low frequency signal FLG (k) (step S4). Here, a description will be given of a method in which the multiplier 220 multiplies a gain coefficient for each frequency of all frequencies of the low-frequency signal InF (k). Here, assuming that the gain coefficient is g (k) (k = 0, 1,..., D-2, D-1), for example, the gain coefficient is set as follows.
g (0) = 1.0
g (1) = 3.0
g (2) = 2.5
g (3) = 2.2
g (4) = 1.8
g (5) = 1.6
g (6) = 1.2
g (7) = 1.0
g (8) = 1.0
g (9) = 1.0
・
・
・
g (D-2) = 1.0
g (D-1) = 1.0

That is, the multiplication unit 220 calculates FLG (k) = g (0) · InF (0) + g (1) · InF (1) +... + G (D−1) · InF (D−1). Do. However, the operator “·” represents multiplication. If the gain coefficient g (k) (k = 0, 1,..., D-2, D-1) is expressed as a vector, the gain coefficient vector G is expressed as follows. be able to.
G = (1.0, 3.0, 2.5, 2.2, 1.8, 1.6, 1.2, 1.0,
1.0, 1.0,. . . 1.0)
That is, the number of elements of the gain coefficient vector is D.

The low-frequency signal InF (k) (k = 0, 1,..., D-2, D-1) and the enhanced low-frequency signal FLG (k) (k = 0, 1,..., D-2, Assuming that the vector notation of D-1) is InF ^→ and FLG ^→ , the multiplication unit 220 calculates the following expression.

FLG ^→ = G * InF ^→
However, the operator “*” indicates an inner product. Thus, the gain coefficient g (k) or the gain coefficient vector G is set so that the value on the low frequency side (k value is small) is larger than that on the high frequency side (value of k is large). In the above example, if k = D−2 and D−1, the gain coefficients are all 1.0 (the gain coefficient is small), and if k = 1, 2, and 3, the gain coefficient is 3.0, 2.5 and 2.2 (the gain coefficient is large), that is, the gain coefficient has a larger value on the low frequency side than on the high frequency side.

  In the above example of g (k), g (0) = 1.0, and the reason will be described. Normally, the amplitude power of the lowest frequency part of the low frequency signal (hereinafter referred to as “lowest frequency part”, the part where k = 0 in the above example) may increase due to noise of the device. Therefore, it is preferable to set the gain coefficient (g (0) in the above example) to be multiplied by the lowest frequency portion of the low frequency signal to 1.0 or a value close to 1.0 (for example, 0.9). Multiplying the gain coefficient 1.0 is synonymous with not performing multiplication processing. That is, the gain coefficient to be multiplied by the lowest band part is not set, and the multiplication process of the multiplication unit 220 may not be performed for the lowest band part. That is, the multiplication unit 220 may perform multiplication processing so that the amplitude power of the lowest band portion is substantially equal to the amplitude power of the input low band signal. In the above example, only the part where k = 0 is described as the lowest part, but it is better to change the lowest part according to the situation such as the noise of the device. For example, it may be better to set k = 0, 1, and 2 as the lowest range portion.

  When there are many situations in which the amplitude power of the lowest frequency part of the low frequency signal increases due to noise of the above equipment (hereinafter referred to as “Case 1”), the gain coefficient of the lowest frequency part and the gain on the high frequency side are preliminarily set. The coefficient is substantially equal (in the above example, for example, g (0) = 1.0). In other words, in the case of “Case 1”, “low side” in the wording of “gain coefficient having a large value on the low side compared to the high side” means that the lowest part is excluded. The gain coefficient and the gain coefficient on the high frequency side are almost equal.

  In addition, when there is little situation where there is no noise or the like of the above apparatus and the amplitude power of the lowest frequency part of the low frequency signal is large (hereinafter referred to as “Case 2”), the gain coefficient of the lowest frequency part is maximized in advance. (For example, g (0) = 3.2). In other words, in case 2, “low frequency side” means that the lowest frequency part is included, and the gain coefficient in the lowest frequency part is the largest compared to other gain coefficients. In other words, “low side” for “gain factor with a large value on the low side compared to the high side” may or may not include the lowest part. It is predetermined according to the situation (in the above, case 1 or case 2), and can be said to be a change in design matters.

  FIG. 5B shows the frequency spectrum of the enhanced low-frequency signal FLG (k) that is the output signal of the multiplier 220. As shown in the position of β in FIG. 5B, the attenuated amplitude level is amplified by the operation of the multiplication unit 220.

In the above description, the process of multiplying the gain coefficient for every frequency of the low frequency signal InF (k) has been described. However, the gain coefficient need not be multiplied for the frequency at which the gain coefficient is 1.0. In that case, it stores the frequency corresponding to the gain factor is not 1.0 in the storage unit 290, information of the frequency F _a and the corresponding non-1.0 to the multiplier 220 a gain coefficient may be input. Then, the multiplication unit 220 may perform a multiplication process of _a gain coefficient other than 1.0 and a low frequency signal for the frequency Fa. By doing so, the amount of multiplication processing can be reduced.

On the other hand, the duplication unit 230 sets a part or all of the low-frequency signal InF (k) as an extension signal. Then, by duplicating the extended signal, a high frequency signal FH (k) (k = 0, 1,..., D-2, D-1) is generated in a pseudo manner (step S6). This will be specifically described below with reference to FIG. FIG. 5 shows an example in which the duplicating unit 230 duplicates “a part” of the low frequency signal InF (k). As shown in FIG. 5A, the head frequency index of the source extension signal (described in FIG. 5A) is D _L (where 0 ≦ D _L ≦ D−1), and the destination high frequency signal FH (k) and D _H a frequency index of the first (described in FIG. 5C), the number of coefficients of the extension signal to be copied to D _w. However, 1 ≦ D _w ≦ D (see FIG. 5A), D _L + D _w ≦ D (see FIG. 5A), and D _H + D _w ≦ D (see FIG. 5C). D _L , D _H , and D _w may be fixed values determined in advance or may be switched dynamically.

FIG. 6 shows a processing flow of the duplication unit 230. FIG. 6 shows a case where the low frequency range of the copy source is a fixed value. First, initial values D _L0 , D _W0 , and D _H0 of D _L , D _W , and _DH are set (step S131). Then, the frequency index k is set to k = 0, which is the lowest limit of the high frequency signal (step S132). MIN is written in the high frequency signal FH (k) from the frequency index k = D to k = D _H −1 (steps S133 to S135). Frequency index k is equal to or k = _{D H} A high-frequency signal FH (k), the amplitude of the copy source first frequency index k = _{D L} signal is duplicated (step S136). That _{is k-D H -D L = D} H -D H + D L = D L. Accordingly, the amplitude of the signal in the range of k = D _L to (D _L + D _W ) of the replication source is replicated in the range of k = D _H to (D _H + D _W ) of the replication destination (No in step S138). loop). MIN is written in the high frequency signal FH (k) in the range of the frequency index k = (D _H + D _W ) to (D−1) (steps S139 to S141). As a result, the high frequency signal FH (k) is as shown in the following equation.
FH (k) = MIN (0 ≦ k ≦ D _H −1)
= InF (k−D _H + D _L ) (D _H ≦ k ≦ D _H + D _w −1)
_{= MIN (D H + D w} ≦ k ≦ D-1)

  Here, the value of MIN may be 0 or a very small value. If a rounded shape is formed by applying a sine window or the like as shown by γ in FIG. 5C, the generated pseudo wideband signal (described below) is improved in audibility. Further, the portion ε in FIG. 5C may be rounded by the same method. In the example of FIG. 5, an example in which “a part” of the low-frequency signal InF (k) is duplicated has been described. However, the “all” of the low-frequency signal InF (k) is regarded as an extension signal, and the extension signal is duplicated. Thus, the high frequency signal FH (k) may be generated. Further, a plurality of portions of the extension signal may be divided and copied.

Next, as shown in FIG. 5D, the combining unit 240 arranges the emphasized low frequency signal FLG (k) on the low frequency side, arranges the high frequency signal FH (k) on the high frequency side, and combines them. A pseudo broadband frequency signal PsF (k) is generated (step S8). The pseudo wideband frequency signal PsF (k) is expressed by the following equation.
PsF (k) = FLG (k) (0 ≦ k ≦ D−1)
= FH (k−D) (D ≦ k ≦ 2D−1)
Here, although the maximum value of PsF (k) is 2D-1, the maximum value need not be limited to 2D-1 as long as the sound quality of the pseudo wideband signal is maintained.

  The frequency inverse transform unit 250 outputs the pseudo wideband signal out (u) (u = 0, 1,..., 2D-1) by transforming the pseudo wideband frequency signal PsF (k) into the time domain (Step 1). S10). Here, the frequency inverse transformation (transformation into the time domain) to be performed is an inverse transformation corresponding to the frequency transformation in the frequency transformation unit 210. However, the number of samples is different between the narrowband signal In (t) and the pseudo wideband signal out (u). Only the sampling frequency is different, and it corresponds completely in time. For example, in this embodiment, since the sampling frequency of the pseudo wideband signal out (u) is twice the sampling frequency of In (t), t = 0 and u = 0, t = D and t = 2D, respectively. It is the same time.

  As described above, the conventional signal broadening apparatus performs processing without amplifying the amplitude power on the low band side despite having the IRS characteristic (the low band side of the narrow band signal is attenuated). . That is, the multiplication process of the gain coefficient by the multiplication unit 220 shown in FIGS. 3 and 5 is not performed. Therefore, the pseudo wideband signal output from the conventional signal broadbanding device has an unnatural sound quality in terms of hearing. However, in the signal broadening apparatus 200 of this embodiment, the output pseudo-wideband signal has a flat frequency and a natural audible sound quality by multiplying the attenuated low frequency side by a gain coefficient. It came to have a good effect.

  For a narrowband signal whose low frequency is attenuated, such as an audio signal having IRS characteristics, a wideband signal with a flat frequency and natural audibility is artificially generated by the method described in the first embodiment. Can do. However, if a pseudo wideband signal is created by the method of the first embodiment for a narrowband signal collected by a handset having a flat frequency, such as a handset having no IRS characteristics, the amplitude power of the low-frequency signal is reduced. There is a problem that it becomes extremely large and a pseudo-generated broadband signal becomes a muffled sound. In order to solve this problem, in the second embodiment, the amplitude power of the low frequency signal becomes extremely large regardless of whether or not the low frequency of the input narrow band signal is attenuated. A method for artificially creating a high-quality wideband signal that does not become unnatural due to excessively small amplitude power will be described.

  FIG. 7 shows a functional configuration example of the signal broadening apparatus 300 of the second embodiment, and FIG. 8 shows a processing flow. The signal broadbanding device 300 is obtained by replacing the storage unit 290 of the signal broadbanding device 100 with a gain control unit 360. In this example, the gain control unit 360 includes an analysis unit 363 and a gain calculation unit 365.

The analysis unit 363 performs a percall (PARCOR coefficient: PARtial auto-CORrelation) analysis on the narrowband signal input to the input terminal 205 to obtain a percoll coefficient and outputs it as percoll coefficient information (step S12). The Percoll analysis is described in “Sadaoki Furui, Acoustic / Speech Engineering, Modern Science Co., Ltd., March 10, 1996, p131-p141” and the like. In the second embodiment, the Percoll coefficient information will be described as a Percoll coefficient. It is efficient to use a first-second order coefficient as the percoll coefficient, but a higher order percoll coefficient may be used. Also, i is the order of the Percoll coefficient, and the i-th order Percoll coefficient is p _i . Next, the Percoll coefficient is input to the gain calculation unit 365.

  The gain calculating unit 365 obtains a gain coefficient g (k) having a larger value on the low frequency side than on the high frequency side as the input coefficient information is small (step S14). Here, “input coefficient information” includes, in addition to the Percoll coefficient from the analysis unit 363, logarithmic cross-sectional area ratio described in Example 4 below, interpolation coefficient information described in Example 5 below, This is the difference coefficient information described in the sixth embodiment. As a preferable configuration example of the gain calculation unit 365, there is a configuration having a determination unit 3652 and a gain determination unit 3654. Hereinafter, the gain calculation unit 365 is described as being configured by the determination unit 3652 and the gain determination unit 3654. To do.

  The determination unit 3652 determines that the IRS characteristic is greater as the input coefficient information is smaller. In this embodiment, the coefficient information is Percoll coefficient information, but is also interpolation coefficient information and difference coefficient information. Moreover, the property that “the IRS characteristic is smaller as the Percoll coefficient as coefficient information is larger” is found as a result of repeated experiments by the inventors. Furthermore, the determination unit 3652 may determine that the IRS characteristic is smaller as the Percoll coefficient is larger. Then, the determination unit 3652 outputs information indicating the magnitude of the IRS characteristic, for example, an IRS characteristic determination flag F. Here, the gain calculation unit 365 may be configured to obtain a gain coefficient g (k) having a larger value on the low frequency side than on the high frequency side as the coefficient information input without using the IRS characteristic is small. For example, it is not necessary to generate the IRS characteristic determination flag F.

An example of a method for determining the IRS characteristic by the determination unit 3652 will be described. (I) As a first method, the determination unit 3652 determines that the IRS characteristic is large if the Percoll coefficient is smaller than a predetermined threshold value, and the IRS characteristic is small if the Percoll coefficient is larger than the predetermined threshold value. Is determined. For example, the first order PARCOR coefficient when the _{p 1,} and the threshold value Th1 for example 0.885,
When p ₁ > Th1: IRS characteristic is small p ₁ ≦ Th1: It is determined that the IRS characteristic is large, and as the IRS characteristic determination flag F, when the IRS characteristic is large, “1” is output and small In this case, “0” is output.

(Ii) As a second method, further a second-order PARCOR coefficient and _{p 2,} and the threshold Th2 for example 0.875,
In the case of p ₁ > Th1 and p ₂ > Th ₂ ... IRS characteristic is small or otherwise. In this way, more accurate determination of the IRS characteristic can be performed by determining the magnitude of the second or higher order Percoll coefficient.

(Iii) As a third method, the magnitude of the IRS characteristic can be divided into a plurality of stages. For example, the threshold Th3 is set to 0.825.
p _1> In the case of Th1 ··· IRS characteristic small Th1 ≧ _p 1> In the case of the case of Th3 ··· IRS characteristics in Th3 ≧ _{p 1} and ··· IRS characteristic large. As the IRS characteristic determination flag F, for example, “1” can be output when the IRS characteristic is large, “1” when it is “2”, and “0” when it is small. In this way, more detailed determination can be performed by determining the magnitude of the IRS characteristic using a plurality of threshold values.

  The gain determination unit 3654 determines from the IRS characteristic determination flag F, and obtains a gain coefficient having a larger value on the low frequency side than the high frequency side as the IRS characteristic is larger. Further, when the IRS characteristic is small, that is, when the low band side of the narrowband signal is not so attenuated, a gain coefficient having substantially the same values on the high band side and the low band side is obtained. An example of how to obtain the gain coefficient will be described. For example, a gain coefficient corresponding to the magnitude of the IRS characteristic is determined in advance and held in the gain determination means 3654. Then, the gain determination unit 3654 may determine the magnitude of the IRS characteristic from the input IRS characteristic determination flag F and select a gain coefficient.

For example, the gain determination unit 3654 retains the following gain coefficient vectors G1 and G2 and outputs G1 or G2 to the multiplication unit 220 in accordance with the IRS characteristic determination flag F. A case where the determination unit 3652 uses the first method (described in the part (i)) will be described.
When IRS characteristic is small (F = 0) G1 = (1.0, 1.2, 1.2, 1.0, 1.0, 1.0, 1.0, 1.0,
1.0, 1.0,. . . , 1.0) is output.
When IRS characteristic is large (F = 1) G2 = (1.0, 3.0, 2.5, 2.2, 1.8, 1.6, 1.2, 1.0,
1.0, 1.0,. . . , 1.0) is output.

  Further, the gain determination unit 3654 may not determine the gain coefficient vector in advance, but may dynamically determine the gain coefficient vector based on the IRS characteristic determination flag F and output the gain coefficient vector to the multiplication unit 220.

  In this manner, the Percoll analysis is performed on the input narrowband signal, and the gain coefficient is determined from the obtained Percoll coefficient. Thus, by dynamically determining and multiplying the gain coefficient (gain constant vector), the low frequency becomes too large regardless of whether the low frequency of the input narrowband signal is attenuated or not. It is possible to generate a pseudo-wideband signal that does not become unnatural due to the sound being too small.

In the second embodiment, the case where the Percoll coefficient information output from the analysis unit 363 is a Percoll coefficient has been described. However, when a plurality of threshold values are used as in the second method (explained at the location (ii)) and the third method (explained at the location (iii)), Th1 (= 0.885), It is experimentally known that the threshold values are close to each other (the interval between the threshold values is small) as in Th2 (= 0.875) and Th3 (= 0.825). Then, the analysis unit 363 has to calculate the decimal part of the Percoll coefficient in detail, and as a result, there is a problem that the calculation amount for obtaining the Percoll coefficient of the analysis unit 363 increases. Therefore, in the third embodiment, the analysis unit 363 does not output Percoll coefficient as PARCOR coefficient information, and outputs the logarithmic sectional area ratio n _i obtained from the Percoll coefficients. Then, in the determination process of the determination unit 3652, it has been experimentally found that the interval between the thresholds can be widened even when a plurality of threshold values are used, and the determination unit 3652 can easily perform the determination. The i-th logarithmic cross-sectional area ratio _ni is obtained, for example, by the following equation.

n _i = log [(1 + p _i ) / (1−p _i )]
Then, logarithmic cross-sectional area ratio _{n i} is input to the determining means 3652. When the determination unit 3652 uses the third method (using three threshold values Th1, Th2, and Th3), for example, Th1 = 2.9, Th2 = 2.2, and Th3 = 1.7 are used. In addition, the difference between the threshold values can be increased as compared with the case where the Percoll coefficient is used, and as a result, the determination unit 3652 can easily perform the determination process.

  Regardless of whether or not the low frequency band of the input narrowband signal is attenuated by the method described in the second or third embodiment, the low frequency band becomes a large sound or becomes too small and unnatural. It is possible to generate a pseudo broadband signal that does not become necessary. However, in the method described in the second or third embodiment, for example, when the narrowband signal is a voice signal, the change in the Percoll coefficient becomes large when the consonant part of the voice is changed to the vowel part. There is a problem that the gain coefficient to be multiplied with the area signal InF (k) is frequently switched, and noise that is harsh on hearing is generated. In order to solve this problem, the fourth embodiment will explain a technique for generating a high-quality pseudo-wideband signal without generating annoying noise by suppressing frequent switching of gain coefficient values.

  FIG. 9 shows a functional configuration example of the signal broadening device 400 of the fourth embodiment. The difference from the signal broadening apparatus 300 described in the second and third embodiments is that the gain control unit 360 is replaced with a gain control unit 460. The gain control unit 460 is provided with an interpolation coefficient calculation unit 464 before the gain calculation unit 365 (determination means 3652) of the gain control unit 360.

In order to suppress fluctuations in the input Percoll coefficient, the interpolation coefficient calculation unit 464 interpolates the currently input Percoll coefficient information using the previously input Percoll coefficient information, thereby obtaining the interpolation coefficient information p ^ _i . Generate. For example, the interpolation coefficient calculation unit 464 calculates a weighted sum p ^ _i as the interpolation coefficient information p ^ _i . The current input i following Percoll coefficient and p _i, when the i-th order PARCOR coefficient input immediately before the dp _i, the interpolation coefficient information interpolation coefficient calculation unit 464 is, for example, a weighted sum according to the following formula Calculate p ^ _i .
_{p ^ i = a 1i p i} + a 2i dp i

The fluctuation of the interpolation coefficient information p ^ _i can be made smaller than the fluctuation of the Percoll coefficient, and as a result, frequent switching of the gain coefficient value can be suppressed. Here, when i = 1, the weighting coefficient is preferably set to a _1i = a _2i = 0.5, for example. In addition, here, an example has been described in which the interpolation coefficient information p ^ _i is calculated for all inputted orders of percoll coefficients. However, only the percoll coefficients of a predetermined order (that is, orders with large fluctuations) are used for interpolation coefficient information. You may calculate. Further, as a PARCOR coefficient input immediately before the dp _i In the above example, performs the processing, Percoll coefficients also inputted in the past from the immediately preceding processing may be performed as dp _i. Further, the interpolation coefficient calculation unit 464 may perform processing on the logarithmic cross-sectional area ratio (described in the third embodiment) instead of the Percoll coefficient. Then, the gain calculation unit 365 (determination unit 3652) obtains a gain coefficient having a larger value on the low frequency side than on the high frequency side as the input interpolation coefficient information is larger.

  As in the fourth embodiment, the interpolation coefficient calculation unit 464 obtains the interpolation coefficient information (for example, the weighted sum of the input Percoll coefficient information and the previously input Percoll coefficient information), thereby changing the change in the Percoll coefficient. Since the fluctuation of the interpolation coefficient information can be reduced even when the signal is large, it is possible to suppress frequent switching of the gain coefficient value, and as a result, a high-quality pseudo-wideband signal without generating harsh noise Can be generated.

  Regardless of whether or not the low frequency band of the narrowband signal input by the method described in the second or third embodiment is attenuated, the low frequency becomes a loud sound or becomes too small to be unnatural. A pseudo broadband signal can be generated. However, for example, when the narrowband signal is a female voice signal, the value of the percall coefficient is smaller than that of other speakers on average, so the IRS characteristic is large even if the low frequency band is not attenuated. There is a problem that the low frequency is excessively emphasized by the gain coefficient multiplication processing of the multiplication unit 220. In order to solve this problem, the fifth embodiment will explain a method of generating a high-quality pseudo-wideband signal even for a narrowband signal having a small percoll coefficient value on average.

  FIG. 10 shows an example of a functional configuration of the signal broadening apparatus 500 of the fifth embodiment. A difference from the signal broadening apparatus 300 described in the second and third embodiments is that the gain control unit 360 is replaced with a gain control unit 560. The gain control unit 560 is provided with a difference coefficient calculation unit 565 in the preceding stage of the gain calculation unit 365 (determination means 3652) of the gain control unit 360.

Difference coefficient calculation unit 565, by calculating the difference between the coefficient information of the different orders of the same frame, the difference coefficient information p ^- generating a _i. In the fifth embodiment, the coefficient information input to the difference coefficient calculation unit 565 is a Percoll coefficient from the analysis unit 363 or a logarithmic cross-sectional area ratio. Assuming that the order of the input Percoll coefficient is PKR and the predetermined weighting coefficients are b _i and b _{i + 1} , the difference coefficient calculation unit 565 calculates the difference coefficient information p ⁻ _i by calculating the following equation, for example. Is generated.
p ⁻ _i = b _i p _i− b _{i + 1} p _{i + 1} (0 <i ≦ PKR−1)
p ⁻ _i = p ⁻ _i (i = PKR)

In this way, it is desirable to take the difference in the Percoll coefficient between adjacent orders (that is, i and i + 1). Then, the difference coefficient information p ^- _i gain calculation unit 365 (determination means 3652) is input, and the gain coefficient having a larger value on the low frequency side than the high frequency side is obtained as the difference coefficient information is large. Further, the difference coefficient calculation unit 565 may perform processing on the logarithmic cross-sectional area ratio.

As described above, the gain calculation unit 365 obtains the gain coefficient using the difference coefficient information p ^- _i (the determination unit 366 determines the magnitude of the IRS characteristic based on the difference coefficient information p ^- _i ), whereby the average value of the Percoll coefficient is obtained. even if small, the difference coefficient information p is the difference in Percoll coefficients between adjacent orders ^- _i by using, IRS characteristics even without the low-frequency is attenuated by the fact not determined to be greater, As a result, the low frequency is not excessively emphasized.

  FIG. 11 shows a functional configuration example of the signal broadening apparatus 600 of the sixth embodiment. A difference from the signal broadening apparatus 500 described in the fifth embodiment is that the gain control unit 560 is replaced with a gain control unit 660. The gain control unit 660 is different from the gain control unit 560 in that an interpolation coefficient calculation unit 464 is provided after the analysis unit 363 and before the gain calculation unit 365 (determination unit 3652) of the gain control unit 560. The coefficient information input to the difference coefficient calculation unit 565 of the sixth embodiment is interpolation coefficient information from the interpolation coefficient calculation unit 464. By adopting a configuration such as the signal broadening device 600, both of the effects described in the fourth and fifth embodiments can be obtained.

<Hardware configuration>
The present invention is not limited to the above-described embodiment. In addition, the various processes described above are not only executed in time series according to the description, but may be executed in parallel or individually according to the processing capability of the apparatus that executes the processes or as necessary. Needless to say, other modifications are possible without departing from the spirit of the present invention.

  Further, when the above-described configuration is realized by a computer, processing contents of functions that the signal broadband apparatus should have are described by a program. The processing function is realized on the computer by executing the program on the computer.

  The program describing the processing contents can be recorded on a computer-readable recording medium. The computer-readable recording medium may be any medium such as a magnetic recording device, an optical disk, a magneto-optical recording medium, or a semiconductor memory. Specifically, for example, the magnetic recording device may be a hard disk device or a flexible Discs, magnetic tapes, etc. as optical disks, DVD (Digital Versatile Disc), DVD-RAM (Random Access Memory), CD-ROM (Compact Disc Read Only Memory), CD-R (Recordable) / RW (ReWritable), etc. As the magneto-optical recording medium, MO (Magneto-Optical disc) or the like can be used, and as the semiconductor memory, EEP-ROM (Electronically Erasable and Programmable-Read Only Memory) or the like can be used.

  The program is distributed by selling, transferring, or lending a portable recording medium such as a DVD or CD-ROM in which the program is recorded. Furthermore, the program may be distributed by storing the program in a storage device of the server computer and transferring the program from the server computer to another computer via a network.

  A computer that executes such a program first stores, for example, a program recorded on a portable recording medium or a program transferred from a server computer in its own storage device. When executing the process, the computer reads a program stored in its own recording medium and executes a process according to the read program. As another execution form of the program, the computer may directly read the program from a portable recording medium and execute processing according to the program, and the program is transferred from the server computer to the computer. Each time, the processing according to the received program may be executed sequentially. Also, the program is not transferred from the server computer to the computer, and the above-described processing is executed by a so-called ASP (Application Service Provider) type service that realizes the processing function only by the execution instruction and result acquisition. It is good. Note that the program in this embodiment includes information that is used for processing by an electronic computer and that conforms to the program (data that is not a direct command to the computer but has a property that defines the processing of the computer).

  In this embodiment, the present apparatus is configured by executing a predetermined program on a computer. However, at least a part of these processing contents may be realized by hardware.

  The signal broadband device described in this embodiment includes a CPU (Central Processing Unit), an input unit, an output unit, an auxiliary storage device, a RAM (Random Access Memory), a ROM (Read Only Memory), and a bus. (Both not shown).

  The CPU executes various arithmetic processes according to the read various programs. The auxiliary storage device is, for example, a hard disk, an MO (Magneto-Optical disc), a semiconductor memory, or the like, and the RAM is an SRAM (Static Random Access Memory), a DRAM (Dynamic Random Access Memory), or the like. The bus connects the CPU, the input unit, the output unit, the auxiliary storage device, the RAM, and the ROM so that they can communicate with each other.

<Cooperation between hardware and software>
The signal broadening apparatus of this embodiment is constructed by reading a predetermined program into the hardware as described above and executing it by the CPU. The functional configuration of each device constructed in this way will be described below.

  An input unit and an output unit (both not shown) of the signal broadbanding device are communication devices such as a LAN card and a modem that are driven under the control of a CPU loaded with a predetermined program. The frequency conversion unit 210, the multiplication unit 220, the duplication unit 230, the combination unit 240, the frequency inverse conversion unit 250, and the gain control unit 360 to 660 are arithmetic units that are constructed by reading a predetermined program into the CPU and executing it. It is. The storage unit 290 functions as the auxiliary storage device.

The block diagram which showed the example of a function structure of the conventional signal wideband apparatus. The figure which showed the IRS characteristic. FIG. 3 is a block diagram illustrating an example of a functional configuration of the signal broadening device according to the first embodiment. The figure which showed the processing flow of the signal wideband apparatus of Example 1. FIG. The figure which showed the frequency spectrum of the signal output from each structure part. The figure which showed the processing flow of the duplication part. FIG. 6 is a block diagram illustrating an example of a functional configuration of a signal broadening device according to a second embodiment. The figure which showed the processing flow of the signal wideband apparatus of Example 1. FIG. FIG. 10 is a block diagram illustrating an example of a functional configuration of a signal broadening device according to a fourth embodiment. FIG. 10 is a block diagram illustrating an example of a functional configuration of a signal broadening device according to a fifth embodiment. FIG. 10 is a block diagram illustrating an example of a functional configuration of a signal broadening device according to a sixth embodiment.

Claims (12)

By converting the narrowband signal to the frequency domain, a frequency converter that generates a lowband signal,
An analysis unit for obtaining first-order and second-order Percoll coefficients by performing a Percall analysis on the narrowband signal;
When the first-order Percoll coefficient is equal to or lower than a predetermined first threshold value or when the second-order Percoll coefficient is equal to or lower than a predetermined second threshold value, a gain coefficient corresponding to each frequency of the low-frequency signal A gain coefficient having a value corresponding to the low frequency side of the low frequency signal larger than the value corresponding to the high frequency side of the low frequency signal ,
In other cases, as the gain coefficient corresponding to each frequency on the high frequency side of the low frequency signal, the first order Percoll coefficient is equal to or less than a predetermined first threshold, or the second order Percoll coefficient is A gain coefficient equal to the value of the gain coefficient output when it is equal to or less than a predetermined second threshold is output, and the first-order Percoll coefficient is used as a gain coefficient corresponding to each frequency on the low frequency side of the low frequency signal. A gain calculation unit that outputs a gain coefficient smaller than a value of a gain coefficient that is output when the second-order percoll coefficient is equal to or smaller than a predetermined second threshold ;
A multiplier for generating an enhanced low-frequency signal by multiplying the gain coefficient for each frequency of the low-frequency signal;
A duplicating unit that creates a high frequency signal by duplicating part or all of the low frequency signal;
A coupling unit that generates a pseudo wideband frequency signal by arranging and combining the enhanced low-frequency signal on the low-frequency side and the high-frequency signal on the high-frequency side, and
A frequency inverse transform unit that outputs a pseudo wideband signal by transforming the pseudo wideband frequency signal into a time domain; and
A signal broadening apparatus comprising: By converting the narrowband signal to the frequency domain, a frequency converter that generates a lowband signal,
An analysis unit for obtaining first-order and second-order Percoll coefficients by performing a Percall analysis on the narrowband signal;
When the first-order Percoll coefficient is less than or equal to a predetermined first threshold and when the second-order Percoll coefficient is less than or equal to a predetermined second threshold, the gain coefficient corresponding to each frequency of the low frequency signal The value corresponding to the low frequency side other than the lowest frequency part of the low frequency signal is larger than the corresponding value on the high frequency side of the low frequency signal, and corresponds to the lowest frequency part of the low frequency signal. value outputs a gain coefficient which is a value equal the value corresponding to the high frequency side of the low frequency signal,
In other cases, as the gain coefficient corresponding to each frequency on the high frequency side of the low frequency signal, the first order Percoll coefficient is equal to or less than a predetermined first threshold, or the second order Percoll coefficient is A gain coefficient equal to the value of the gain coefficient output when it is equal to or less than a predetermined second threshold is output, and the first-order Percoll coefficient is used as a gain coefficient corresponding to each frequency on the low frequency side of the low frequency signal. A gain calculation unit that outputs a gain coefficient that is equal to or less than a value of a gain coefficient that is output when the second-order Percoll coefficient is equal to or less than a predetermined second threshold ;
A multiplier for generating an enhanced low-frequency signal by multiplying the gain coefficient for each frequency of the low-frequency signal;
A duplicating unit that creates a high frequency signal by duplicating part or all of the low frequency signal;
A coupling unit that generates a pseudo wideband frequency signal by arranging and combining the enhanced low-frequency signal on the low-frequency side and the high-frequency signal on the high-frequency side, and
A frequency inverse transform unit that outputs a pseudo wideband signal by transforming the pseudo wideband frequency signal into a time domain; and
A signal broadening apparatus comprising: The signal broadening device according to claim 1 or 2,
The signal broadening apparatus characterized in that the analysis unit obtains a logarithmic cross-sectional area ratio from the obtained Percoll coefficient and outputs it as the Percoll coefficient information. The signal broadening device according to claim 1,
In addition, the interpolation coefficient calculation unit for generating the interpolation coefficient information by interpolating the currently input Percoll coefficient information using the previously input Percoll coefficient information,
The signal broadening apparatus according to claim 1, wherein the gain calculation unit obtains a gain coefficient having a larger value on the low frequency side than on the high frequency side as the interpolation coefficient information is small. The signal broadening device according to claim 1,
Furthermore, a difference coefficient calculation unit that generates difference coefficient information by calculating a difference between coefficient information of different orders in the same frame is provided,
The signal broadening apparatus according to claim 1, wherein the gain calculation unit obtains a gain coefficient having a larger value on the low frequency side than on the high frequency side as the difference coefficient information is larger. By converting a narrowband signal to the frequency domain, a frequency conversion process for generating a lowband signal,
An analysis process for obtaining first-order and second-order percoll coefficients by performing a percall analysis on the narrowband signal;
When the first-order Percoll coefficient is equal to or lower than a predetermined first threshold value or when the second-order Percoll coefficient is equal to or lower than a predetermined second threshold value, a gain coefficient corresponding to each frequency of the low-frequency signal A gain coefficient having a value corresponding to the low frequency side of the low frequency signal larger than the value corresponding to the high frequency side of the low frequency signal ,
In other cases, as the gain coefficient corresponding to each frequency on the high frequency side of the low frequency signal, the first order Percoll coefficient is equal to or less than a predetermined first threshold, or the second order Percoll coefficient is A gain coefficient equal to the value of the gain coefficient output when it is equal to or less than a predetermined second threshold is output, and the first-order Percoll coefficient is used as a gain coefficient corresponding to each frequency on the low frequency side of the low frequency signal. A gain calculation process of outputting a gain coefficient smaller than a gain coefficient value output when the second-order percoll coefficient is equal to or less than a predetermined second threshold ;
A multiplication process for generating an enhanced low-frequency signal by multiplying the gain coefficient for each frequency of the low-frequency signal;
A replication process for generating a high frequency signal by replicating one or all of the low frequency signals; and
A combination process of generating a pseudo wideband frequency signal by arranging and combining the enhanced low-frequency signal on the low-frequency side and the high-frequency signal on the high-frequency side;
A frequency inverse transform process of outputting the pseudo wideband signal by transforming the pseudo wideband frequency signal into the time domain;
A signal broadbanding method comprising: By converting a narrowband signal to the frequency domain, a frequency conversion process for generating a lowband signal,
An analysis process for obtaining first-order and second-order percoll coefficients by performing a percall analysis on the narrowband signal;
When the first-order Percoll coefficient is less than or equal to a predetermined first threshold and when the second-order Percoll coefficient is less than or equal to a predetermined second threshold, the gain coefficient corresponding to each frequency of the low frequency signal The value corresponding to the low frequency other than the lowest frequency part of the low frequency signal is larger than the corresponding value on the high frequency side of the low frequency signal, and corresponds to the lowest frequency part of the low frequency signal. value outputs a gain coefficient which is a value equal the value corresponding to the high frequency side of the low frequency signal,
In other cases, the first-order Percoll coefficient is equal to or lower than a predetermined first threshold or the second-order Percoll coefficient is a predetermined gain coefficient corresponding to each frequency on the high frequency side of the proposed signal. A gain coefficient that is equal to the value of the gain coefficient that is output when the frequency is equal to or lower than the second threshold value, and the first-order Percoll coefficient is used as a gain coefficient corresponding to each frequency on the low frequency side of the low frequency signal. A gain calculating step of outputting a gain coefficient that is equal to or smaller than a predetermined first threshold or a gain coefficient that is equal to or smaller than a value of a gain coefficient that is output when the second-order Percoll coefficient is equal to or smaller than a predetermined second threshold ;
A multiplication process for generating an enhanced low-frequency signal by multiplying the gain coefficient for each frequency of the low-frequency signal;
A replication process for generating a high frequency signal by replicating one or all of the low frequency signals; and
A combination process of generating a pseudo wideband frequency signal by arranging and combining the enhanced low-frequency signal on the low-frequency side and the high-frequency signal on the high-frequency side;
A frequency inverse transform process of outputting the pseudo wideband signal by transforming the pseudo wideband frequency signal into the time domain;
A signal broadbanding method comprising: The signal broadening method according to claim 6 or 7,
The signal broadening method according to claim 1, wherein the analysis step is to obtain a logarithmic cross-sectional area ratio from the obtained Percoll coefficient and output it as the Percoll coefficient information. A signal broadening method according to claim 6-8,
Furthermore, it also includes an interpolation coefficient calculation process for generating interpolation coefficient information by interpolating currently input Percoll coefficient information using previously input Percoll coefficient information,
The gain calculation process is characterized in that, as the interpolation coefficient information is smaller, a gain coefficient having a larger value on the low frequency side than on the high frequency side is obtained. The signal broadening method according to claim 6-9,
Furthermore, a difference coefficient calculation process for generating difference coefficient information by calculating a difference between coefficient information of different orders in the same frame is provided,
The signal broadening method characterized in that the gain calculation step is to obtain a gain coefficient having a larger value on the low frequency side than on the high frequency side as the difference coefficient information is larger.   A program for causing a computer to operate as the signal broadening device according to any one of claims 1 to 5.   The computer-readable recording medium which recorded the program of Claim 11.

Images