JP2974059B2 - Pitch post filter device - 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 post-filter used in a speech coding / decoding device.

[0002]

2. Description of the Related Art Conventionally, a pitch post-filter device of this kind has been disclosed in, for example, "ITU-T Recommendation."
on G. 723-Dual Rate Speech
Coder for Multimedia Com
communications Transmission
at 5.3 and 6.3 kbit / s "(3.
6 pp. 23-25, July 1995), a decoder of a speech encoding / decoding apparatus is used for improving the sound quality of reproduced speech by improving the pitch expression of a decoded excitation signal.

FIG. 6 is a block diagram showing an example of a conventional pitch post filter device.

The forward shift signal generating circuit 121 inputs the decoding excitation signal e (n) through the input terminal 10, inputs the optimum forward pitch lag M _fopt through the input terminal 31, and outputs the decoding excitation signal e (n). ) _Is shifted to a time length M _{fopt in the} future (optimal forward shift signal e (n +
M _fopt )) is generated and output.

The backward shift signal generating circuit 221 inputs the decoding excitation signal e (n) through the input terminal 10, inputs the optimum backward pitch lag M _bopt through the input terminal 32, and outputs the decoding excitation signal e (n). ) _Is shifted in the past by the time length M _bopt (the optimal backward shift signal e (n−
M _bopt )) is generated and output.

The switch 20 receives a control signal through an input terminal 23 and switches the input according to the control signal. Thereby, the optimal forward shift signal e (n + M
_fopt ) or the optimal backward shift signal e (n−
M _bopt ) is _converted to a pitch post-filter signal generation circuit 180,
The signals are input to the cross-correlation circuit 150 and the auto-correlation circuit 160.

There are the following two cases depending on the switching of the input by the switch 20.

[0008] 1. The optimal forward shift signal e (n +
When inputting M _fopt ) When connected to the output terminal 21, the switch 20 _{converts the} optimal forward shift signal e (n + M _fopt ) into a pitch post-filter signal generation circuit 180, a cross-correlation circuit 150,
Input to autocorrelation circuit 160.

[0009] The cross-correlation circuit 150 includes the decoding excitation signal e (n) and the optimal forward shift signal e (n + _Mfopt ).
, _And calculates and outputs the cross-correlation C _fopt shown in equation (1). N is the analysis section length (for example, subframe length).

[0010]

(Equation 1) The autocorrelation circuit 160 calculates the optimum forward shift signal e
(N + M _fopt ) and the autocorrelation D shown in equation (2)
Calculate and output _fopt .

[0011]

(Equation 2) Optimum gain calculation circuit 170, the said cross correlation C _fopt enter the autocorrelation D _fopt, calculate the forward optimum gain g _fopt by the formula (3), and outputs.

[0012]

(Equation 3) The pitch post-filter signal generation circuit 180 generates the decoded excitation signal e (n) and the optimal forward shift signal e (n +
M _fopt ) and the forward optimal gain g _fopt are input, and the pitch post-filter signal e _ppff (n) is calculated by equation (4) and output. The gain g _pf is calculated by Expression (6).

[0013]

(Equation 4) 2. When the optimal backward shift signal e (n-M _bopt ) is input When connected to the output terminal 22, the switch 20 _{converts the} optimal backward shift signal e (n-M _bopt ) into a pitch post-filter signal generation circuit 180, Cross-correlation circuit 150,
Input to autocorrelation circuit 160.

The cross-correlation circuit 150 includes the decoding excitation signal e (n) and the optimal backward shift signal e (n-M _bopt ).
, _And calculates and outputs the cross-correlation C _bopt shown in equation (7).

[0015]

(Equation 5) The autocorrelation circuit 160 calculates the optimum backward shift signal e.
(N−M _bopt ) and the autocorrelation D shown in equation (8)
Calculate and output _bopt .

[0016]

(Equation 6) Optimum gain calculating circuit 170, the inputted correlation C _Bopt and the autocorrelation D _Bopt, calculate the backward optimum gain g _Bopt by Equation (9), and outputs.

[0017]

(Equation 7) The pitch post-filter signal generation circuit 180 generates the decoded excitation signal e (n) and the optimal backward shift signal e (n−
M _bopt ) and the backward optimum gain g _bopt are input, and the pitch post-filter signal e _ppfb (n) is _obtained by Expression (10).
Is calculated and output. The gain g _pb is calculated by Expression (12).

[0018]

(Equation 8)

[0019]

(Equation 9)

[0020]

However, in the conventional device, the performance of pitch emphasis in a transient portion where the pitch period or phase changes is insufficient. The reason is that in the transient part where the pitch period and phase change, the pitch period and phase included in both the past and future signals coexist with the current analysis section. This is because only one of the past excitation signal and the future excitation signal is considered.

In such a case, it is necessary to perform pitch emphasis in consideration of the pitch cycle and phase of both the past signal and the future signal.

It is therefore an object of the present invention to provide a pitch post-filter device capable of obtaining a signal having a good pitch structure.

[0023]

According to a first aspect of the present invention, a pitch post-filter device receives an input signal, a first delay time and a second delay time, and converts the input signal to the first delay time. Means for calculating a first cross-correlation from the first signal obtained by shifting the time in the future and the input signal (370 in FIG. 1), and shifting the input signal to the second delay time in the past Means (38 in FIG. 1) for calculating a second cross-correlation from the second signal obtained in
0), means for calculating a third cross-correlation from the first signal and the second signal (390 in FIG. 1), and means for calculating a first auto-correlation from the first signal (FIG. 1 of 40
0), means for calculating a second autocorrelation from the second signal (410 in FIG. 1), the first cross-correlation, the second cross-correlation, the third cross-correlation and the second Means for calculating a first gain and a second gain from the autocorrelation of the first signal and the second autocorrelation (420 in FIG. 1), the first signal, the second signal, and the first gain. Means (430 in FIG. 1) for performing pitch enhancement on the input signal using the second gain.

A pitch post-filter device according to a second aspect of the present invention is characterized in that a first signal obtained by shifting an input signal and the input signal to a first delay time having a predetermined first range of values is obtained. And the second signal obtained by shifting the input signal in the past by a second delay time having a value in a predetermined second range, and the delay time in the first range and the second signal Means (350 in FIG. 2) for calculating an evaluation value for all or some of the combinations with the delay time in the range of 2;
Means for selecting the delay time and the second delay time (FIG. 2)
360), the input signal and the selected first signal
Means (370 in FIG. 2) for calculating a first cross-correlation with the first signal corresponding to the delay time of the input signal and the second signal corresponding to the selected second delay time.
Means (38 in FIG. 2) for calculating a second cross-correlation with the signal
0) and calculating a third cross-correlation between the first signal corresponding to the selected first delay time and the second signal corresponding to the selected second delay time. Means (390 in FIG. 2); means (400 in FIG. 2) for calculating a first autocorrelation of the first signal corresponding to the selected first delay time; Means for calculating a second autocorrelation of the second signal corresponding to the second delay time (410 in FIG. 2), the first cross-correlation, the second cross-correlation, and the third Means for calculating a first gain and a second gain from the cross-correlation, the first auto-correlation, and the second auto-correlation (420 in FIG. 2), corresponding to the selected first delay time The second signal corresponding to the selected second delay time of the first signal And means (430 in FIG. 2) that performs pitch enhancement on the input signal using the second gain and the first gain.

A pitch post-filter device according to a third aspect of the present invention provides a pitch post-filter device which is obtained by shifting an input signal and the input signal to a first delay time future having a predetermined first range of values. And a second signal obtained by shifting the previously obtained pitch post-filter output signal held in the storage means to a second delay time past having a value in a predetermined second range. Means for calculating an evaluation value for all or a part of the combination of the delay time in the first range and the delay time in the second range (350 in FIG. 3); Means for selecting the first delay time and the second delay time (360 in FIG. 3), and the input signal and the first signal corresponding to the selected first delay time Compute the first cross-correlation of Means for calculating a means (370 in FIG. 3), the second cross-correlation between the second signal corresponding to said input signal and said selected second delay time (Fig. 3
380) and a third cross-correlation between the first signal corresponding to the selected first delay time and the second signal corresponding to the selected second delay time. Means for calculating (390 in FIG. 3); means for calculating a first autocorrelation of the first signal corresponding to the selected first delay time (400 in FIG. 3); Means for calculating a second autocorrelation of the second signal corresponding to the second delay time (410 in FIG. 3);
Means for calculating a first gain and a second gain from the cross-correlation, the second cross-correlation, the third cross-correlation, the first auto-correlation, and the second auto-correlation (420 in FIG. 3). )
The first signal corresponding to the selected first delay time, the second signal corresponding to the selected second delay time, the first gain, and the second Means (430 in FIG. 3) for performing pitch emphasis on the input signal using a gain.

A pitch post-filter device according to a fourth aspect of the present invention is characterized in that the input signal and the input signal are shifted to a first delay time future having a predetermined first range of values. And the second signal obtained by shifting the input signal in the past by a second delay time having a value in a predetermined second range, and the delay time in the first range and the second signal Means (350 in FIG. 4) for calculating the error energy for all or a part of the combination of the delay time in the range 2 and selecting a plurality of the error energies from the smallest value to the smallest value. Means for selecting the first delay time and the second delay time for each of the plurality of error energies (360 in FIG. 4); Corresponding multiple Means for calculating a plurality of first cross-correlation between each said input signal of the serial first signal (37 in FIG. 4
0) and means for calculating a plurality of second cross-correlations between each of the plurality of second signals corresponding to the selected plurality of second delay times and the input signal (380 in FIG. 4). )
And a plurality of the first signals corresponding to the selected plurality of first delay times and a plurality of the second signals corresponding to the selected plurality of second delay times. Means (390 in FIG. 4) for calculating a third cross-correlation of
Means (400 in FIG. 4) for calculating a plurality of first autocorrelations of the plurality of first signals corresponding to the selected plurality of first delay times, and Means for calculating a plurality of second autocorrelations of the plurality of second signals corresponding to two delay times (410 in FIG. 4); and a plurality of first cross-correlations and a plurality of second cross-correlations. Means for calculating a plurality of first gains and a plurality of second gains from the correlation, the plurality of third cross-correlations, the plurality of first autocorrelations, and the plurality of second autocorrelations (FIG. 4) 420)
A plurality of the first signals corresponding to the selected plurality of the first delay times, a plurality of the second signals corresponding to the selected plurality of the second delay times, and the plurality of Means for emphasizing the pitch of the input signal using the first gain and the plurality of second gains (43 in FIG. 4).
0).

[0027] A pitch post-filter device according to a fifth aspect of the present invention is configured such that a first signal obtained by shifting an input signal and the input signal to a first delay time having a predetermined first range of values is obtained. And a second signal obtained by shifting the previously obtained pitch post-filter output signal held in the storage means to a second delay time past having a value in a predetermined second range. Means for calculating the error energy for all or some combination of the delay time in the first range and the delay time in the second range (350 in FIG. 5); Means (360 in FIG. 5) for selecting a plurality of the first and second delay times, each of which selects a plurality of the error energies in ascending order of the value and gives each of the selected plurality of the error energies; Means for calculating a plurality of first cross-correlations between each of the plurality of first signals corresponding to the selected plurality of first delay times and the input signal (370 in FIG. 5); Means for calculating a plurality of second cross-correlations between each of the plurality of second signals corresponding to the selected plurality of second delay times and the input signal (380 in FIG. 5); A plurality of third signals of a plurality of the first signals corresponding to the plurality of the first delay times and a plurality of the second signals corresponding to the selected plurality of the second delay times Means for calculating the cross-correlation (39 in FIG. 5)
0); means for calculating a plurality of first autocorrelations of the plurality of first signals corresponding to the selected plurality of first delay times (400 in FIG. 5); Means for calculating a plurality of second autocorrelations of the plurality of second signals corresponding to the plurality of second delay times (41 in FIG. 5)
0), the plurality of first cross-correlations, the plurality of second cross-correlations, the plurality of third cross-correlations, and the plurality of first cross-correlations.
From the plurality of second autocorrelations and the plurality of first autocorrelations.
Means (420 in FIG. 5) for calculating the gains and the plurality of second gains, the plurality of first signals corresponding to the selected plurality of first delay times, and the plurality of selected Means for performing pitch emphasis on the input signal using a plurality of the second signals corresponding to the second delay time, the plurality of first gains, and the plurality of second gains (FIG. 5) 430).

[0028]

By expressing the pitch post-filter signal using both the signal shifted in the future by the forward pitch lag and the signal shifted in the past by the backward pitch lag,
The pitch period and phase of both past and future signals can be reflected in the pitch post-filter signal. Therefore, both the pitch period and the phase included in the past signal and the pitch period and the phase included in the future signal with respect to the current analysis section, such as the transient portion of the audio signal, are included in the signal coexisting in the analysis section. On the other hand, good pitch emphasis becomes possible.

[0029]

FIG. 1 is a block diagram showing a pitch post-filter device according to a first embodiment of the present invention.

The forward shift signal generating circuit 121 inputs the decoding excitation signal e (n) through the input terminal 10, inputs the optimum forward pitch lag M _fopt through the input terminal 31, and outputs the decoding excitation signal e (n). ) _Is shifted to a time length M _{fopt in the} future (optimal forward shift signal e (n +
M _fopt )) is generated and output.

The backward shift signal generating circuit 221 inputs the decoding excitation signal e (n) through the input terminal 10, inputs the optimum backward pitch lag M _bopt through the input terminal 32, and outputs the decoding excitation signal e (n). ) _Is shifted in the past by the time length M _bopt (the optimal backward shift signal e (n−
M _bopt )) is generated and output.

The first cross-correlation calculation circuit 370 calculates the decoded excitation signal e (n) and the optimal forward shift signal e (n + M
_fopt ), and calculates and outputs the cross-correlation C _fopt shown in equation (1). N is the analysis section length (for example, subframe length).

The second cross-correlation calculation circuit 380 calculates the decoded excitation signal e (n) and the optimal backward shift signal e (n−M
_bopt ), and calculates and outputs the cross-correlation C _bopt shown in equation (7).

The third cross-correlation calculating circuit 390 receives the optimum forward shift signal e (n + M _fopt ) and the optimum backward shift signal e (n−M _bopt ), and _obtains a cross-correlation C _fopt shown in equation (13). Is calculated and output.

[0035]

(Equation 10) The first autocorrelation calculation circuit 400 receives the optimum forward shift signal e (n + _Mfopt ), calculates the autocorrelation _Dfopt shown in Expression (2), and outputs the result.

The second autocorrelation calculation circuit 410 receives the optimum backward shift signal e (n- _Mbopt ), calculates the autocorrelation _Dbopt shown in equation (8), and outputs the result.

The optimum gain calculation circuit 420 calculates the cross correlations C _fopt , C _bopt , C _fbopt and the auto correlations D _fopt , D _fopt
bopt is input, and the expression (1) is obtained from the expressions (14) and (15).
The gain (optimal gain g _fopt , g _bopt ) when the error energy E _fb shown in 6) is minimum is calculated and output.

[0038]

[Equation 11] Here, g _f and g _{b are} gains and N is an analysis section length.

Pitch post filter signal generation circuit 430
_{Are the} decoding excitation signal e (n), the optimal forward shift signal e (n + M _fopt ), and the optimal backward shift signal e
(N−M _bopt ) and the optimum gains g _fopt and g _bopt are input, and the pitch post-filter signal e is _calculated based on Expression (17).
_ppffb (n) is calculated and output from the output terminal 11. Here, g _pfb is a gain, which is calculated by Expression (19).

[0040]

(Equation 12) This concludes the description of the pitch post-filter device according to the first embodiment of the present invention.

FIG. 2 shows a pitch post filter device according to a second embodiment of the present invention.

Here, error energy is used as an evaluation value for calculating pitch lag.

The forward shift circuit 110 is connected to the input terminal 10
, The decoding excitation signal e (n) is input through the lower limit M ₁ of the pitch lag search range and the upper limit M _{2 of the} pitch lag search range
A signal (forward shift signal e (n + _Mf )) shifted in the future by the forward pitch lag M _f (M ₁ < M _f < M ₂ ) within the pitch lag search range determined by is generated and output.

The backward shift circuit 210 is connected to the input terminal 10
, The decoding excitation signal e (n) is input through the lower limit M ₁ of the pitch lag search range and the upper limit M _{2 of the} pitch lag search range
Pitch lag and generates a backward pitch lag _{M b (M 1 <M b} <M 2) only shifts in the past signal in the search range (backward shift signal e (n-M _b)) determined by outputs.

The error energy calculation circuit 350 receives the decoded excitation signal e (n) and receives the forward shift signal e (n).
(N + M _f ) and the backward shift signal e (n−
Enter the M _b), the lower limit value of the pitch lag search range M ₁
For forward pitch lag M _f and backward pitch lag M _b is in the range of the upper limit value M ₂ from the error energy E _fb for the current analysis intervals was calculated by the equation (16), and outputs. Here, it is assumed that M _f > 0 and M _b > 0.

The pitch lag selection circuit 360 minimizes the error energy E _fb which exists in the range from the lower limit value M ₁ to the upper limit value M ₂ of the pitch lag search range and exists by the number of combinations of the forward pitch lag M _f and the backward pitch lag M _b. forward the optimum value of pitch lag M _f select (best forward pitch lag M _fopt) the optimum value of the backward pitch lag M _b (optimum backward pitch lag M _Bopt), and outputs to.

The forward shift signal selection circuit 220 receives the forward shift signal e (n + M _f ) and the optimal forward pitch lag M _fopt, and receives the optimal forward shift signal e (n + M
_fopt ) Select and output.

The backward shift signal selection circuit 220 determines the backward forward direction e (n−M _b ) and the optimal backward pitch lag M
_bopt is input and the optimal backward shift signal e (n-M _bopt )
Select and output.

First cross-correlation circuit 370, second cross-correlation circuit 380, third cross-correlation circuit 390, first auto-correlation circuit 400, second auto-correlation circuit 410, optimum gain calculation circuit 420, pitch The configuration of the post-filter signal generation circuit 430 is the same as that of the pitch post-filter device according to the first embodiment, and a description thereof will be omitted.

The description of the pitch post-filter device according to the second embodiment of the present invention has been completed.

FIG. 3 shows a pitch post-filter device according to a third embodiment of the present invention.

Here, error energy is used as an evaluation value for calculating the pitch lag.

The difference from FIG. 2 is that the storage means 440 is added, and the input of the backward shift circuit 210 is the output of the storage means 440.

The storage means 440 holds the pitch post-filter signal e _ppffb2 (n).

The backward shift circuit 210 reads out the past pitch post-filter signal e _ppffb2 (n) for the current analysis section from the storage means 440 and inputs it.
A signal shifted backward in the past by a backward pitch lag M _b (M ₁ < M _b < M ₂ ) within the pitch lag search range determined by the lower limit value M ₁ of the pitch lag search range and the upper limit value M ₂ of the pitch lag search range (backward shift signal e _ppffb2) (N- _Mb ))
Is generated and output. Here, when M _b <N (N is the analysis section length), the section N−M _b <n <N of the backward shift signal e _ppffb2 (n) held in the storage unit 440 is extracted, and By repeating, a sequence of length N can be generated.

In the forward shift circuit 110, the configuration of the forward shift signal selection circuit 120 is the same as that of the pitch post filter device according to the second embodiment.

The error energy calculation circuit 350 receives the decoded excitation signal e (n) and inputs a forward shift signal e (n).
+ M _f ) and the backward shift signal e _ppffb2 (n
−M _b ) and the lower limit M of the pitch lag search range.
For forward pitch lag M _f and backward pitch lag M _b in the range ₁ to the upper limit value M _2, calculates and outputs an error energy E _fb2 for the current analysis intervals by the formula (20). Here, it is assumed that M _f > 0 and M _b > 0.

[0058]

(Equation 13) The backward shift signal selection circuit 220 _{calculates the} backward shift signal e _ppffb2 (n−M _b ) and the optimal backward pitch lag M
_bopt, and input the optimal backward shift signal e _ppffb2 (n−M
_bopt ) and output.

Hereinafter, in the pitch post-filter device according to the second embodiment of the present invention, e (n-M _bopt ) is _replaced by e _ppffb2 (n-M _bopt ), _whereby the second embodiment of the present invention is realized. The pitch post-filter signal e _ppffb2 (n) shown in Expression (21) is obtained by the same procedure as that of the pitch post-filter device according to the embodiment.

[0060]

[Equation 14]

[0061]

(Equation 15) This concludes the description of the pitch post-filter device according to the third embodiment of the present invention.

FIG. 4 shows a pitch post-filter device according to a fourth embodiment of the present invention. Forward shift circuit 11
The configurations of the 0, backward shift circuit 210 and error energy calculation circuit 350 are the same as those of the pitch post filter device according to the second embodiment.

The pitch lag selection circuit 360 calculates the error energy in the range of the lower limit value M ₁ to the upper limit value M ₂ of the pitch lag search range by the number of combinations of the forward pitch lag M _f and the backward pitch lag M _b represented by the equation (16). For the upper L values in ascending order of the values from the smallest E _fb , the forward pitch lag M _{f, m} (m = 0,..., L−1) and the backward pitch lag M _{b, m} (m = 0,. ) And output.

The forward shift signal selection circuit 120 outputs the forward shift signal e (n + M _f ) and the forward pitch lag M
_{f, m} (m = 0,..., L−1), and selects a forward shift signal e (n + M _{f, m} ) (m = 0,.
Output.

The backward shift signal selection circuit 120 provides the backward shift signal e (n−M _b ) and the backward pitch lag M
_{b, m} (m = 0,..., L−1), and selects a backward shift signal e (n−M _{b, m} ) (m = 0,.
Output.

[0066]

(Equation 16) The first cross-correlation calculation circuit 370 calculates the decoded excitation signal e
(N) and the forward shift signal e (n + M _{f, m} ) (m =
0,..., L−1) _, and calculates and outputs the cross-correlation C _{f, m} shown in equation (24).

The second cross-correlation calculating circuit 380 calculates the decoded excitation signal e (n) and the backward shift signal e (n−
M _{b, m} ) (m = 0,..., L−1), and the equation (25)
Cross correlation C _b shown in _{computes m,} and outputs.

[0068]

[Equation 17] The third cross-correlation calculating circuit 390 calculates the forward shift signal e (n + M _{f, l} ) (m = 0,..., L−1) and the backward shift signal e (n−M _{b, m} ) (m = 0 , ..., L-1)
And calculates and outputs the cross-correlation C _{fb, l, m} shown in equation (26).

[0069]

(Equation 18) The first autocorrelation calculating circuit 400 receives the forward shift signal e (n + M _{f, m} ) (m = 0,..., L−1),
The autocorrelation D _{f, m} shown in equation (27) is calculated and output.

[0070]

[Equation 19] The second autocorrelation calculating circuit 410 receives the backward shift signal e (n + _{Mb, m} ) (m = 0,..., L-1),
The autocorrelation D _{b, m} shown in equation (28) is calculated and output.

[0071]

(Equation 20) The optimum gain calculation circuit 420 calculates the first cross-correlation C
_{f, m} , the second cross-function C _{b, m} , the third cross-correlation C _{fb, l, m} (l = 0,..., L-1, m = 0,..., L-1)
And the first auto-correlation D _{f, m} and the second auto-correlation function D _{b, m} (m = 0,..., L−1), and the error energy E _fb3 shown in equation (29) is minimized. (Optimum gains g _{f, m} , g _{b, m} (m = 0,..., L−1)) are calculated and output. It is obtained from equations (30) and (31). Equations (32) and (33) are replaced by g _{f, m} , g _{b, m} (m = 0,..., L
-1) is obtained by solving.

[0072]

(Equation 21)

[0073]

(Equation 22) The pitch post-filter signal generation circuit 430 calculates the decoded excitation signal e (n) and the forward shift signal e (n + M
_{f, m} ) (m = 0,..., L−1), the backward shift signal e (n−M _{b, m} ) (m = 0,..., L−1), and the optimal gain g _{f, m} , g _{b, m} (m = 0, ..., L-1)
Based on the equation (34), the pitch post-filter signal e
_ppffb3 is calculated and output from the output terminal 11. Where g
_pfb3 is a gain, which is calculated by equation (36).

[0074]

(Equation 23)

[0075]

(Equation 24) This concludes the description of the pitch post-filter device according to the fourth embodiment of the present invention.

FIG. 5 shows a pitch post-filter device according to a fifth embodiment of the present invention. 4 is different from FIG. 4 in that a storage means 440 is added.
Is the output of the storage means 440.

The storage means 440 holds the pitch post-filter signal e _ppffb4 (n).

The backward shift circuit 210 reads out the past pitch post-filter signal e _ppffb4 (n) for the current analysis section from the storage means 440, and inputs it.
Pitch lag search range backward pitch lag _{M b (M 1 <M b} <M 2) only shifts in the past signal in the pitch lag search range determined by the upper limit value M ₂ of the lower limit value M ₁ and the pitch lag search range (backward shift signal e _Ppffb4 (N- _Mb ))
Is generated and output. Here, when M _b <N (N is the analysis section length), the section N−M _b <n <N of the backward shift signal e _ppffb4 (n) held in the storage unit 440 is extracted, and By repeating, a sequence of length N can be generated.

The error energy calculation circuit 350 receives the decoded excitation signal e (n) and inputs the forward shift signal e (n).
(N + M _f ) and input the backward shift signal e _ppffb4
(N−M _b ) and forward pitch lag M in the range from the lower limit M ₁ to the upper limit M ₂ of the pitch lag search range.
_{For f} and the backward pitch lag M _b , the error energy E _fb4 for the current analysis section is calculated and output _according to equation (37). Here, it is assumed that M _f > 0 and M _b > 0.

[0080]

(Equation 25) Forward shift circuit 110, forward shift signal selection circuit 1
The configuration of 20 is the same as that of the pitch post-filter device according to the fourth embodiment.

The pitch lag selection circuit 360 is in the range from the lower limit M ₁ to the upper limit M ₂ of the pitch lag search range. The error energy E _fb4 shown in the equation (37) having the number of combinations of the forward pitch lag M _f and the backward pitch lag M _b is obtained from the smallest L in the order from the smallest error energy E _fb4 .
Select forward pitch lag M _{f, m} (m = 0,..., L-1) and backward pitch lag M _{b, m} (m = 0,..., L-1),
Output.

The backward shift signal selection circuit 220 receives the backward shift signal e _ppffb4 (n−M _b ) and the backward pitch lag M _{b, m} (m = 0,..., L−1), and outputs the backward shift signal e _ppffb4. (N−M _{b, m} ) is selected and output.

Hereinafter, in the pitch post-filter device according to the fourth embodiment of the present invention, e (n−M _{b, m} ) is _replaced by e _ppffb4 (n−M _{b, m} ), whereby The pitch post-filter signal e _ppffb4 (n) shown in Expression (38) is obtained by the same procedure as that of the pitch post-filter device according to the fourth embodiment.

[0084]

(Equation 26) This concludes the description of the pitch post-filter device according to the fifth embodiment of the present invention.

[0085]

As described above, according to the present invention,
In the transient part of the audio signal in which the pitch period or phase changes, better pitch emphasis can be achieved as compared with the conventional method. The reason is that in the transient part of the audio signal, both the pitch period and the phase included in the past signal and the pitch period and the phase included in the future signal coexist in the analysis section with respect to the current analysis section. By expressing the pitch post-filter signal using both the signal shifted in the future by the pitch lag and the signal shifted in the past by the backward pitch lag, the pitch period and phase of both the past signal and the future signal are converted to the pitch post-filter signal. This is because it is possible to reflect the

[Brief description of the drawings]

FIG. 1 is a block diagram showing a pitch post-filter device according to a first embodiment of the present invention.

FIG. 2 is a block diagram showing a pitch post-filter device according to a second embodiment of the present invention.

FIG. 3 is a block diagram showing a pitch post-filter device according to a third embodiment of the present invention.

FIG. 4 is a block diagram showing a pitch post filter device according to a fourth embodiment of the present invention.

FIG. 5 is a block diagram showing a pitch post filter device according to a fifth embodiment of the present invention.

FIG. 6 is a block diagram showing a conventional pitch post filter device.

[Explanation of symbols]

 10, 23, 31, 32 input terminal 11, 21, 22, output terminal 110 forward shift circuit 210 backward shift circuit 120 forward shift signal selection circuit 220 backward shift signal selection circuit 121 forward shift signal generation circuit 221 backward shift signal generation circuit 350 error Energy calculation circuit 360 Pitch lag selection circuit 20 Switch 170, 420 Optimal gain calculation circuit 180, 430 Pitch post filter signal generation circuit 150, 370, 380, 390 Cross-correlation calculation circuit 160, 400, 410 Autocorrelation calculation circuit 440 Storage means

Continued on the front page (58) Fields surveyed (Int.Cl. ⁶ , DB name) G10L 3/00-9/20 H03H 17/00-17/02 H03M 7/30 H04B 14/04 JICST file (JOIS)

Claims (5)

(57) [Claims] 1. An input signal, a first delay time, and a second delay time are input, and a first signal obtained by shifting the input signal to the first delay time in the future and the input signal are used. Calculating a first cross-correlation and converting the input signal to the second
Calculating a second cross-correlation from a second signal obtained by shifting the delay time in the past and the input signal,
Calculating a third cross-correlation from the second signal and the second signal,
Calculating a first autocorrelation from the first signal;
From the first cross-correlation, the second cross-correlation, the third cross-correlation, the first auto-correlation, and the second auto-correlation. Calculating a gain and a second gain, and performing pitch emphasis on the input signal using the first signal, the second signal, the first gain, and the second gain. Pitch post filter device. 2. A first signal obtained by shifting an input signal and the input signal to a first delay time future having a value of a predetermined first range, and a second signal obtained by shifting the input signal to a predetermined second time. All or part of the delay time in the first range and the delay time in the second range using a second signal obtained by shifting in the past a second delay time having a value in the range An evaluation value is calculated for the combination of the above, the first delay time and the second delay time at which the evaluation value is the best are selected, and the input signal and the selected first delay time are selected. Calculating a first cross-correlation with the corresponding first signal, and calculating the input signal and the selected second signal;
Calculating a second cross-correlation with the second signal corresponding to the selected delay time, and calculating the first signal corresponding to the selected first delay time and the selected second delay Calculating a third cross-correlation with the second signal corresponding to time; calculating a first autocorrelation of the first signal corresponding to the selected first delay time; Calculating a second autocorrelation of the second signal corresponding to the calculated second delay time, and calculating the first crosscorrelation, the second crosscorrelation, the third crosscorrelation, and the first autocorrelation. Calculating a first gain and a second gain from the autocorrelation and the second autocorrelation, and calculating the first signal and the selected second signal corresponding to the selected first delay time. Using the second signal, the first gain, and the second gain corresponding to the delay time of Pitch post filter device and performs pitch enhancement on the input signal. 3. An input signal and a first signal obtained by shifting said input signal to a first delay time future having a value in a predetermined first range and a past signal held in a storage means. Using the second signal obtained by shifting the obtained pitch post-filter output signal in the past by a second delay time having a value in a predetermined second range, the delay time in the first range and Calculating an evaluation value for all or a part of the combination of the delay time in the second range, selecting the first delay time and the second delay time in which the evaluation value is the best, Calculating a first cross-correlation between an input signal and the first signal corresponding to the selected first delay time, and calculating a first cross-correlation corresponding to the input signal and the selected second delay time; The second with the second signal
And calculates the first signal corresponding to the selected first delay time and the selected second signal.
Calculating a third cross-correlation with the second signal corresponding to the delay time of, and calculating a first autocorrelation of the first signal corresponding to the selected first delay time; Calculating a second autocorrelation of the second signal corresponding to the selected second delay time, and calculating the first cross-correlation, the second cross-correlation, the third cross-correlation, A first gain and a second gain are calculated from a first autocorrelation and the second autocorrelation, and the first signal corresponding to the selected first delay time and the selected signal are calculated. A pitch post-filter device, wherein pitch enhancement is performed on the input signal using the second signal corresponding to a second delay time, the first gain, and the second gain. 4. A first signal obtained by shifting an input signal and the input signal to a first delay time future having a value in a predetermined first range and a second signal obtained by shifting the input signal to a predetermined second time All or part of the delay time in the first range and the delay time in the second range using a second signal obtained by shifting in the past a second delay time having a value in the range The error energy is calculated for the combination of the above, a plurality of the error energies are selected in ascending order of the value from the minimum value, and the first delay time and the Selecting a second delay time and calculating a plurality of first cross-correlations between each of the plurality of first signals corresponding to the selected plurality of first delay times and the input signal; The selected plurality of second delays Calculating a plurality of second cross-correlations between each of the plurality of second signals corresponding to the delay time and the input signal; and calculating a plurality of second cross-correlations corresponding to the selected plurality of first delay times; Calculating a plurality of third cross-correlations between one signal and a plurality of the second signals corresponding to the selected plurality of the second delay times, and calculating the selected plurality of the first delays Calculating a plurality of first autocorrelations of the plurality of first signals corresponding to time, and calculating a plurality of second autocorrelations of the plurality of second signals corresponding to the selected plurality of second delay times; Of the plurality of first cross-correlations, the plurality of second cross-correlations, the plurality of third cross-correlations, the plurality of first autocorrelations, and the plurality of second autocorrelations. Calculating a plurality of first gains and a plurality of second gains from the correlation; A plurality of the first signals corresponding to one delay time, a plurality of the second signals corresponding to the selected plurality of the second delay times, the plurality of first gains, and the plurality of first signals. A pitch post-filter device, wherein pitch enhancement is performed on the input signal using a gain of 2. 5. A first signal obtained by shifting the input signal and the input signal to a first delay time having a predetermined first range value in the future, and a first signal obtained by shifting the input signal in the past stored in the storage means. Using the second signal obtained by shifting the obtained pitch post-filter output signal in the past by a second delay time having a value in a predetermined second range, the delay time in the first range and Calculating the error energy for all or a part of the combination with the delay time in the second range, selecting a plurality of the error energies in ascending order from a minimum value, The first delay time providing each of the energy and the second
And calculating a plurality of first cross-correlations between each of the plurality of first signals corresponding to the selected plurality of first delay times and the input signal; Calculating a plurality of second cross-correlations between each of the plurality of second signals corresponding to the plurality of determined second delay times and the input signal, and calculating the selected plurality of first delays Calculating a plurality of third cross-correlations between the plurality of first signals corresponding to time and the plurality of second signals corresponding to the selected plurality of second delay times; Calculating a plurality of first autocorrelations of a plurality of the first signals corresponding to the plurality of the first delay times, and calculating a plurality of the first autocorrelations corresponding to the selected plurality of the second delay times. Calculating a plurality of second autocorrelations of the two signals;
A plurality of first gains and a plurality of second gains from the plurality of second cross correlations, the plurality of third cross correlations, the plurality of first autocorrelations, and the plurality of second autocorrelations. 2 and calculates a plurality of first signals corresponding to the selected plurality of first delay times and a plurality of second signals corresponding to the selected plurality of second delay times. 2
Pitch emphasizing is performed on the input signal by using the signal (i), the plurality of first gains, and the plurality of second gains.