Method and device for adaptive bandwidth pitch search in coding wideband signals
View Patent ↗A pitch search method and device for digitally encoding a wideband signal, in particular but not exclusively a speech signal, in view of transmitting, or storing, and synthesizing this wideband sound signal. The new method and device which achieve efficient modeling of the harmonic structure of the speech spectrum uses several forms of low pass filters applied to a pitch codevector, the one yielding higher prediction gain (i.e. the lowest pitch prediction error) is selected and the associated pitch codebook parameters are forwarded.
1. A method comprising:
receiving a pitch codevector and a target vector;
passing the pitch codevector through at least one frequency shaping filter to obtain at least one filtered version of the pitch codevector;
determining at least two pitch prediction errors, a prediction error being representative of a difference between the target vector and a product of the pitch codevector and a pitch gain value, at least one of the pitch prediction errors being determined from a filtered version of the pitch codevector;
comparing, via a processor, the at least two pitch prediction errors to identify the pitch prediction error having a lowest error energy value;
selecting the pitch gain associated with the lowest energy value; and
providing an indication of the selected pitch gain.
2. The method according to claim 1 , further comprising providing an indication of the at least one frequency shaping filter used to obtain the at least one filtered version of the pitch codevector.
3. The method according to claim 2 , wherein the indication of the at least one frequency shaping filter is an index representative of the at least one frequency shaping filter.
4. The method according to claim 1 , wherein passing the pitch codevector through the at least one frequency shaping filter comprises passing the pitch codevector through more than one frequency shaping filter to obtain more than one filtered version of the pitch codevector.
5. The method according to claim 1 , wherein passing the pitch code vector through the at least one frequency shaping filter comprises passing the pitch code vector through a low-pass filter.
6. The method according to claim 1 , wherein passing the pitch code vector through the at least one frequency shaping filter comprises passing the pitch code vector through a band-pass filter.
7. The method according to claim 1 , wherein determining the at least two pitch errors comprises:
convolving the pitch codevector with a weighted synthesis filter impulse response signal to obtain a convolved pitch codevector;
multiplying the convolved pitch codevector by the pitch gain value to produce an amplified convolved pitch codevector; and
subtracting the amplified convolved pitch codevector from the target vector.
8. The method according to claim 7 , further comprising calculating the pitch gain value b (j) from the convolved pitch codevector using the relation
b (j) =x t y (j) /∥y (j) ∥ 2
where j=0, 1, 2 . . . , K, and K corresponds to a number of pitch prediction errors to be determined, x is the target vector and y (j) is the convolved pitch codevector.
9. The method according to claim 1 , wherein the pitch codevector is an interpolated pitch codevector having sub-sample pitch resolution.
10. A device comprising:
at least one frequency shaping filter for obtaining at least one filtered version of a pitch codevector;
at least one subtractor for determining at least two pitch prediction errors, a prediction error being representative of a difference between a target vector and a product of the pitch codevector and a pitch gain value, at least one of the pitch prediction errors being determined from a filtered version of the pitch codevector;
a processor for comparing the at least two pitch prediction errors to identify the pitch prediction error having a lowest error energy value; and
a selector for selecting the pitch gain associated with the lowest energy value and for providing an indication of the selected pitch gain.
11. The device according to claim 10 , wherein the selector further provides an indication of the at least one frequency shaping filter used to obtain the at least one filtered version of the pitch codevector.
12. The device according to claim 11 , wherein the indication of the at least one frequency shaping filter is an index representative of the at least one frequency shaping filter.
13. The device according to claim 10 , comprising more than one frequency shaping filter to obtain more than one filtered version of the pitch codevector.
14. The device according to claim 10 , wherein the at least one frequency shaping filter comprises a low-pass filter for filtering the pitch code vector.
15. The device according to claim 10 , wherein the at least one frequency shaping filter comprises a band-pass filter for filtering the pitch code vector.
16. The device according to claim 10 , comprising:
a convolution module for convolving the pitch codevector with a weighted synthesis filter impulse response signal to obtain a convolved pitch codevector; and
an amplifier for multiplying the convolved pitch codevector by the pitch gain value to produce an amplified convolved pitch codevector;
wherein the at least one subtractor further subtracts the amplified convolved pitch codevector from the target vector.
17. The device according to claim 16 , further comprising a gain calculator for calculating the pitch gain value b (j) from the convolved pitch codevector using the relation
b (j) =x t y (j) /∥y (j) ∥ 2
where j=0, 1, 2 . . . , K, and K corresponds to a number of pitch prediction errors to be determined, x is the target vector and y (j) is the convolved pitch codevector.
18. The device according to claim 10 , wherein the pitch codevector is an interpolated pitch codevector having sub-sample pitch resolution.