JP3483998B2 - Pitch enhancement method and apparatus - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a pitch enhancement method and apparatus for improving the subjective quality of coded speech and speech including background noise.

[0002]

2. Description of the Related Art A technique for highly efficiently encoding a voice signal is
It is now an indispensable technology in mobile communications, where the available radio band is limited, and in storage media such as voice mail that requires effective use of memory. In the encoding of this audio signal, the bit rate is getting lower.

As the encoding bit rate decreases, the encoding distortion generated in the encoding / decoding process increases,
The quality of the reproduced sound will deteriorate. Therefore, a post filter is generally used for the purpose of improving the subjective quality of reproduced voice. This post filter is used at the final stage of the decoding device, and emphasizes the peaks and attenuates the valleys according to the shape of the spectrum of the speech signal to reduce the noise associated with encoding in the speech spectrum. Has the effect of masking and improving the sound quality.

At present, a post filter which is most widely used in various coding systems and has excellent characteristics is disclosed in Japanese Patent Laid-Open No. 64-132.
00 is disclosed. FIG. 6 is a block diagram of this conventional post filter, which includes a pitch emphasis filter 1, a formant emphasis filter 2, a high-frequency emphasis filter 3, a gain adjusting unit 4, and a multiplier 5.

The pitch emphasizing filter 1 is used for attenuating frequency components between pitch harmonic frequencies, and has a transfer function of the following equation (1) set based on the decoded pitch parameter.

[0006]

[Equation 1]

Here, p is the pitch period, and γ and λ are calculated by the following equations (2) and (3) based on the pitch gain. γ = C _z f (x) (2) λ = C _p f (x) (3) C _z and C _p are constants for controlling the degree of pitch enhancement and are empirically determined. Further, f (x) is a control factor used to avoid unnecessary pitch enhancement when the signal to be processed is unvoiced sound that does not include periodicity. x corresponds to the pitch gain, and when x is less than a certain threshold value (typically 0.6), it is determined as unvoiced sound and f (x)
= 0. When x is greater than or equal to the threshold value, f (x) = x. If x exceeds 1, f (x) will be used to maintain stability.
= 1. Cg is for absorbing the fluctuation of the gain of the pitch enhancement filter 1 between unvoiced sound and voiced sound,
It is calculated by the following equation (4).

Cg = (1−λ / x) / (1 + γ / x) (4) The formant enhancement filter 2 is used to enhance the formant peaks and attenuate the valleys, and the transfer function of the following equation (5) is used. have.

[0009]

[Equation 2]

The denominator of this equation corresponds to the transfer function 1 / (1-P (z)) of the LPC synthesis filter, and the spectral inclination of the all-pole type post filter is corrected to improve the feeling of sound muffled. Used for. α and β are constants that are empirically determined from the balance between noise suppression and muffled feeling,
α = 0.8 and β = 0.5 are typical values. The high-frequency emphasis filter 3 is used to improve the feeling of muffled that cannot be corrected only by the numerator of Expression (5). The gain adjusting unit 4 is for correcting the gain of the entire post filter via the multiplier 5 so that the power of the signal does not change between the input and output of the post filter.

This post filter can suppress the noise sensation generated by the encoding of the audio signal and is effective in improving the sound quality of the reproduced audio. However, if the pitch period changes greatly in the voiced sound section, or if the pitch period is not detected correctly, such as double pitch or half pitch, the harmonics of the pitch of the decoded speech are shown in FIG. As shown in FIG. 8, there is a problem in that the sound is greatly fluctuated and disturbed, and the disturbance is increased by the pitch enhancement filter 1 and the sound quality of the reproduced voice is greatly deteriorated.

In particular, when the pitch information is analytically and synthetically obtained by using the adaptive code length as in the CELP coding system, the correct pitch period is not necessarily obtained, and the pitch period is originally set. Even when the value is changing slowly, the obtained pitch period changes greatly,
Since the double pitch or the half pitch is rare, the deterioration of the sound quality due to the pitch enhancement filter 1 becomes a problem.

[0013]

As described above, in the conventional post filter, the pitch period is not correctly detected when the pitch period is greatly changed in the voice sound section or when the pitch period is double pitch or half pitch. When the pitch harmonics of the decoded voice are disturbed as in the case, there is a problem that the pitch enhancement filter increases this disturbance and the sound quality is greatly deteriorated.

It is an object of the present invention to provide a pitch emphasizing method and apparatus capable of suppressing the disturbance of harmonics and improving the subjective quality even when the pitch period is largely changed or is not correctly detected.

[0015]

In order to solve the above-mentioned problems, the present invention detects a temporal change in at least one of a pitch period and a pitch gain of an input voice signal, and based on this change, detects a change in the voice signal. It is characterized in that it comprises changing the degree of pitch enhancement.

More specifically, when the temporal change of at least one of the pitch period and the pitch gain of the input voice signal exceeds a predetermined threshold value, the degree of pitch enhancement for the voice signal is lowered. It is a thing.

When the pitch signal of the audio signal decoded and reproduced from the coded data obtained by the CELP coding method for obtaining the pitch information analytically and synthetically using the adaptive code length changes greatly on the coding side. Or, if the pitch period is not correctly detected, the pitch harmonics are disturbed. When such a disturbance occurs, the pitch enhancement filter has the property of further increasing it. Since human hearing is sensitive to the disturbance of harmonics, the sound quality is greatly deteriorated. According to the invention,
By monitoring the change in the pitch period, and when it exceeds a certain level, the degree of emphasis is reduced by changing the coefficient of the pitch emphasis filter, or by stopping the emphasis itself, such a disturbance of harmonics is caused. It is suppressed and the sound quality is improved.

In other words, the conventional post filter utilizes the static property of the masking effect of human hearing. That is, noise is made difficult to be perceived by masking the noise spectrum with the signal spectrum. For this reason, the conventional post filter focuses on how to shape the spectrum of the frame in which the sound is to be reproduced, and has not been impressed with the temporal variation of the spectrum. In contrast, the present invention focuses on the fact that human hearing is sensitive to changes in the spectrum, and in particular, by reducing the degree of emphasis of the post filter when the pitch period changes with time, that is, when the harmonics fluctuate significantly. , Which improves the subjective quality of speech.

[0019]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing an embodiment in which the present invention is applied to a speech decoding device. In the figure, a portion surrounded by a broken line is a post filter 130, which constitutes a voice decoding device together with the parameter coding unit 110 and the voice reproducing unit 120.

Coded data transmitted from a speech coder (not shown) is input to the input terminal 100 and sent to the parameter decoding unit 110. This parameter decoding unit 1
In 10, the parameters used in the audio reproducing unit 120 are decoded. The audio reproduction unit 120 reproduces an audio signal using the input parameters. The parameter decoding unit 110 and the audio reproduction unit 120 can have various configurations depending on the configuration of the encoding system device. The post filter according to the present invention is not restricted by the configurations of the parameter decoding unit 110 and the audio reproduction unit 120, and can be widely applied to various audio decoding devices. Therefore, detailed description of the parameter decoding unit 110 and the audio reproducing unit 120 is omitted here.

The post filter 130 includes a pitch emphasis filter 131, a pitch control section 132, a formant emphasis filter 133, a high frequency emphasis filter 134, and a gain adjusting section 13.
5 and a multiplier 136.

FIG. 2 is a flow chart showing a schematic procedure of main processing in this embodiment. First, when encoded data is input (step S1), parameters such as an LPC coefficient, a frame gain, a pitch period, a pitch gain, a noise vector and a driving gain are decoded by the parameter decoding unit 110 (step S2). The original audio signal is reproduced by the audio reproducing unit 120 based on the parameter (step S3).

Next, based on the pitch period and pitch gain which are pitch parameters among the parameters decoded by these parameter decoding units 110, the pitch control unit 1
The transfer function of the pitch enhancement filter 131 is set by 32 (step S4), and then the pitch enhancement filter 1
Pitch emphasis processing is performed by 31 (step S
5). Here, in the pitch control unit 132, as will be described later, the degree of pitch enhancement changes based on the temporal change of the pitch cycle. Specifically, the greater the temporal change of the pitch cycle, the greater the degree of pitch emphasis. The characteristic is that the transfer function of the pitch enhancement filter 131 is controlled so as to decrease.

The formant emphasis filter 133, the high-frequency emphasis filter 134, and the gain adjusting section 135 are used for the voice signal subjected to the pitch emphasis processing by the pitch emphasis filter 131.
And further processed by multiplier 136. The formant emphasis filter 133 is a filter used to emphasize the peaks of the formants and attenuate the valleys, and various known structures can be used. Further, the high-frequency emphasis filter 134 is provided in order to improve the muffled feeling of voice accompanying the introduction of the post filter. further,
The gain adjusting unit 135 is for correcting the gain of the entire post filter via the multiplier 135 so that the signal power does not change between the input and output of the post filter.
The high-frequency emphasis filter 134 and the gain adjusting unit 13
Any of the 5 can be configured by using various known techniques like the formant enhancement filter 133.

Various configurations can be considered as the configuration of the pitch enhancement filter 131. In the case of the all-pole type, the pitch enhancement filter 131 can be defined by the transfer function H (z) shown in Expression (6).

H (z) = 1 / (1-εαz ^−T ) (6) Here, T is a pitch period, and ε and α are filter coefficients determined by the pitch control unit 132.

In this case, the pitch control section 132 sets the transfer function of the pitch enhancement filter 131 according to the procedure shown in FIG. That is, the pitch control unit 135 first determines the pitch gain b as in Expression (7), calculates the filter coefficient α based on that, and further determines the temporal change of the pitch period T, and based on that, calculates Expression ( Calculate the filter coefficient ε as in 8).

[0028]

[Equation 3]

Where b is the decoded pitch gain, b
_th is a threshold value for voiced / unvoiced determination, ε ₁ and ε ₂ are parameters for controlling the degree of pitch enhancement, T _p is a pitch period of the previous frame, and T _th is a temporal change of the pitch period T | T. It is a threshold value for determining −T _p |. As a typical value, a threshold value b _th of 0.6, a parameter ε ₁ of 0.8, a parameter ε ₂ of 0.4 or 0.0, and a threshold value T _th of 10 can be used. In this way, the filter coefficients ε and α are set, and the transfer function H (z) shown in Expression (6) is set. on the other hand,
When the pitch enhancement filter 131 is defined by a pole-zero type transfer function, its transfer function H (z) is expressed by equation (9).

[0030]

[Equation 4]

In this case, the pitch control unit 132 sets the transfer function of the pitch enhancement filter 131 according to the procedure shown in FIG. That is, in the pitch control unit 135, first, the equation (10)
Then, the pitch gain b is determined, the parameter α is calculated based on the pitch gain b, the time change of the pitch period T is determined, and the parameters C ₁ and C are calculated based on the determination as shown in equations (11) and (12). Calculate ₂ .

[0032]

[Equation 5]

Then, these parameters α, C ₁ , C
Based on ₂ , the filter coefficients γ and λ of the pitch enhancement filter 131 are calculated by the equations (13) and (14). γ = C ₁ α (13) λ = C ₂ α (14) where c11, c12, c21, c22 are 0 ≦ c ₁₁ , c ₁₂ , c ₂₁ , c ₂₂ <1 (15) c ₁₁ > c ₁₂ (16) c ₂₁ > c ₂₂ Can be determined empirically under the constraint of (17). Typical values are c11 = 0.4, c12 = 0.0, c21 = 0.8,
c22 = 0.0 can be used.

Cg is for absorbing the voiced sound and the variation of the gain of the pitch enhancement filter 131 between the voiced sound and is calculated by the equation (18). Cg = (1−λ / b) / (1 + γ / b) (18) As is clear from the above, the pitch enhancement filter 131 is a temporal change of the pitch period T in any configuration | T. When −T _p | becomes equal to or greater than the threshold value T _th , the pitch control unit 132 controls the filter coefficient so that the degree of pitch enhancement with respect to the input voice signal is reduced.

In the above description, when | T−T _p | becomes _{equal to} or greater than the threshold value T _th , the pitch emphasizing process itself is performed and the degree of emphasizing is reduced, but the pitch emphasizing process itself is not performed. It is also possible to

In the above description, the degree of pitch enhancement is reduced when the temporal change of the pitch cycle becomes equal to or more than the threshold value. However, the temporal change of the pitch gain cycle is set to the threshold value. Even if the degree of pitch enhancement is reduced when the above is reached, the same effect can be obtained.

Furthermore, in the above embodiment, an example in which the present invention is applied to a speech decoding apparatus has been described, but it is applied to a technique called enhanced processing applied to speech signals containing various noises for the purpose of improving subjective quality. You can also do it. Such an embodiment is shown in FIG.

In FIG. 5, the same parts as those in FIG. 1 are designated by the same reference numerals, and only the differences from the previous embodiment will be described. In this embodiment, an audio signal is input to the input terminal 200.
This input audio signal is, for example, an audio signal reproduced by the audio reproducing unit 120 in FIG. 1 or an audio signal synthesized by a speech synthesizer. The pitch emphasizing filter 131 and the formant emphasizing having the same configurations as those in the previous embodiment. Filter 1
33, the high-frequency emphasis filter 134, the gain adjusting unit 135, and the multiplier 136 perform enhanced processing.

However, in this embodiment, the input signal is a voice signal, and the input signal does not include a parameter such as pitch gain as in the embodiment of FIG. By passing through the pitch analysis unit 220, the pitch control unit 132 generates information on the pitch period and pitch gain necessary for setting the transfer function of the pitch enhancement filter 131. The other points are similar to those of the previous embodiment, and thus the description thereof will be omitted.

The present invention is not limited to voice signals generated by humans, but can be applied to various audio signals such as tone signals. The audio signals referred to in the present invention collectively refer to these signals.

[0041]

As described above, according to the present invention, the change in the pitch period is monitored, and when it exceeds a certain level, it is emphasized by changing the degree of pitch emphasis, that is, the coefficient of the pitch emphasis filter. It is possible to suppress the disturbance of the harmonics by lowering the degree of, and in some cases, stopping the emphasis itself, and effectively improve the sound quality of the reproduced voice signal and the synthesized voice signal.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a speech decoding apparatus according to a first embodiment of the present invention.

FIG. 2 is a flowchart showing an outline of main processing in the same embodiment.

FIG. 3 is a flowchart showing an example of a transfer function setting procedure of the pitch enhancement filter in the embodiment.

FIG. 4 is a flowchart showing another example of a transfer function setting procedure of the pitch enhancement filter in the embodiment.

FIG. 5 is a block diagram showing a configuration of an enhancement processing device according to a second embodiment of the present invention.

FIG. 6 is a block diagram showing a configuration of a conventional post filter.

FIG. 7 is a diagram showing a spectrum of voiced sound.

FIG. 8 is a diagram showing a spectrum in which harmonics are changed.

[Explanation of symbols]

100 ... Encoded data input terminal 110 ... Parameter decoding unit 120 ... Voice playback unit 130 ... Post filter 131 ... Pitch enhancement filter 132 ... Pitch control unit 133 ... Formant enhancement filter 134 ... High-frequency emphasis filter 135 ... Gain adjusting section 136 ... Multiplier 210 ... LPC analysis unit 220 ... Pitch analysis section

─────────────────────────────────────────────────── ─── Continued Front Page (58) Fields surveyed (Int.Cl. ⁷ , DB name) G10L 11/04 G10L 19/06 G10L 21/04 JISST file (JOIS)

Claims (4)

(57) [Claims] 1. A time change of at least one of a pitch period and a pitch gain of an input audio signal is detected,
A pitch emphasizing method characterized in that when the change exceeds a predetermined threshold value, the degree of pitch emphasizing for the voice signal is reduced. 2. A pitch emphasizing means for emphasizing a pitch of an input voice signal, and a time change of at least one of a pitch period and a pitch gain of the voice signal is detected, and this change is predetermined.
A pitch emphasizing device , comprising: a control means for reducing the degree of pitch emphasizing in the pitch emphasizing means when the threshold value is exceeded . 3. A parameter decoding means for decoding a parameter including at least one of a pitch period and a pitch gain of the speech signal from encoded data of the speech signal, and a speech signal reproduced by using the parameter decoded by the parameter decoding means. Audio reproducing means, a pitch emphasizing means for emphasizing the pitch of the audio signal reproduced by the audio reproducing means, and detecting a temporal change in at least one of the pitch period and the pitch gain decoded by the parameter decoding means. , When this change exceeds a predetermined threshold value , the degree of pitch enhancement in the pitch enhancing means is lowered.
And a control means for controlling the pitch enhancement device. 4. A parameter extracting means for extracting a parameter including at least one of a pitch period and a pitch gain of an input voice signal, a pitch enhancing means for enhancing the pitch of the voice signal, and a parameter extracting means for extracting the parameter. When a change in at least one of the pitch period and the pitch gain with time is detected, and the change exceeds a predetermined threshold value , the degree of pitch enhancement in the pitch enhancing means is reduced.
And a control means for controlling the pitch enhancement device.