APPARATUS AND METHOD FOR AUDIO FRAME LOSS RECOVERY

• US 20140088974A1
• Filed: 09/26/2012
• Published: 03/27/2014
• Est. Priority Date: 09/26/2012
• Status: Active Grant

• Current Assignee: Google Technology Holdings LLC (Alphabet Inc.)
• Original Assignee: Motorola Mobility LLC (Lenovo Group Ltd.)
• Inventors: Mittal, Udar, Ashley, James P.
• Granted Patent: US 9,123,328 B2
• Time in Patent Office: Days
• US Class Current: 704/500
• CPC Class Codes: G10L 19/005 Correction of errors induced by the transmission channel, if related to the coding algorithm
• CPC Class Codes: G10L 19/18 Vocoders using multiple modes

Abstract

A method and apparatus provides for frame loss recovery following a loss of a frame in an audio codec. The lost frame is identified. Estimated linear predictive coefficients of a previous transform frame are generated based on a decoded audio of the previous transform frame. An estimated residual of the previous transform frame is generated based on the estimated linear predicative coefficients and the decoded audio. A pitch delay is determined from frame error recovery parameters received with the previous transform frame. An extended residual is generated based on the pitch delay and the estimated residual. A first synthesized signal is generated based on the extended residual and the linear predicative coefficients. A decoded audio output of at least the lost frame is generated based on the first synthesized signal. The frame error recovery parameters are generated by an encoder.

Claims

1. A method for generating a decoded frame in response to a loss of a frame in an audio codec, comprising:

identifying that a frame is lost;

generating a set of estimated linear predictive coefficients corresponding to a previous transform frame based on a decoded set of audio samples from the previous transform frame;

generating an estimated residual of the previous transform frame based on the set of estimated linear predicative coefficients and the decoded set of audio samples;

determining a pitch delay from a set of frame error recovery parameters received with the previous transform frame;

generating an extended residual based on the pitch delay and the estimated residual;

generating a first synthesized signal based on the extended residual and the set of linear predicative coefficients; and

generating a decoded audio output corresponding to at least the lost frame based on the first synthesized signal.

2. The method according to claim 1, wherein the previous transform frame is a most recent previous transform frame that is able to be successfully decoded.

3. The method according to claim 1, wherein the pitch delay is a best estimate of a pitch delay of a next frame.

4. The method according to claim 1, further comprising:

estimating a plurality of transform domain coefficients for the lost frame based on the transform domain coefficients of the previous transform frame;

generating a second synthesized signal based on the plurality of transform domain coefficients; and

generating the decoded audio output of the at least one lost frame based on a first weighted sum of the first synthesized signal and the second synthesized signal.

5. The method of claim 1, further comprising:

receiving a plurality of coded parameters for a next good frame following the lost frame, wherein the good frame is a successfully decoded frame;

generating a third synthesized signal for the next good frame further based on the plurality of coded parameters; and

generating the decoded audio output of the next good frame based on a second weighted sum of the first synthesized signal and the third synthesized signal.

6. The method of claim 5, wherein the plurality of coded parameters are transform coefficients.

7. The method of claim 5, wherein the plurality of coded parameters are LPC and excitation parameters.

8. The method according to claim 5, wherein the second weighted sum comprises at least two weight boundaries, and wherein the at least two weight boundaries are determined based on a pitch epoch of the first synthesized signal, a pitch epoch of decoded audio of the next good frame, and the pitch delay.

9. The method of claim 4, further comprising:

determining that the next good frame is a time domain frame;

generating an extended decoded audio output by extending the length of the decoded audio output of the at least one lost frame by a quantity of samples that is predetermined based on a bandwidth of an audio signal being conveyed by the frame;

coupling the extended decoded audio output to a CELP state generator; and

generating the decoded audio for the next good frame based at least upon the output of the CELP state generator.

10. The method of claim 1, further comprising:

receiving the frame and other audio information associated with the frame from a frame encoder;

determining frame error recovery parameters that include at least a pitch delay from the other audio information, wherein the pitch delay is a best estimate of a pitch delay of a next frame; and

transmitting the frame and the frame error recovery parameters to the codec.

11. A method of encoding a frame, comprising:

receiving the frame and other audio information associated with the frame from a frame encoder;

determining from the other audio information at least one of a classification as to whether the frame is a speech or music frame, a high frequency energy, and a low frequency energy;

setting a frame error recovery (FER) indicator to one of a pitch FER state and a MDCT FER state based on the at least one of a classification as to whether the frame is a speech or music frame, a high frequency energy, and a low frequency energy;

determining at least a pitch delay from the other audio information, wherein the pitch delay is a best estimate of a pitch delay of a next frame; and

transmitting the frame, the FER indicator, and the pitch delay.

12. The method according to claim 11, wherein setting the frame error recovery indicator comprises:

setting an open loop pitch gain threshold to values that are set to initialized values, wherein the initialized values depend on whether the frame is classified as a speech or music frame; and

modifying the initialized values using a ratio of the high frequency energy to the low frequency energy.

13. An apparatus for generating a decoded frame following a loss of a frame, wherein the frames are a sequence of encoded audio frames, comprising:

a receiver that receives the sequence of audio frames; and

at least one processor that executes program instructions stored in memory, wherein the executed program instructions

identify that a frame is lost,

generate a set of estimated linear predictive coefficients corresponding to a previous transform frame based on a decoded set of audio samples from the previous transform frame,

generate an estimated residual of the previous transform frame based on the set of estimated linear predicative coefficients and the decoded set of audio samples,

determine a pitch delay from frame error recovery parameters received with the previous transform frame,

generate an extended residual based on the pitch delay and estimated residual,

generate a first synthesized signal based on the extended residual and the linear predicative coefficients, and

generate a decoded audio output of at least the lost frame based on the first synthesized signal.

14. The apparatus according to claim 13, wherein the previous transform frame is a most recent previous transform frame that is able to be successfully decoded.

15. The apparatus according to claim 13, wherein the pitch delay is a best estimate of a pitch delay of a next frame.

16. The apparatus according to claim 13, wherein the executed program instructions further:

estimate a plurality of transform domain coefficients for the lost frame based on the transform domain coefficients of the previous transform frame;

generate a second synthesized signal based on the plurality of transform domain coefficients; and

generate the decoded audio output of the at least one lost frame based on a first weighted sum of the first synthesized signal and the second synthesized signal.

17. The apparatus according to claim 16, wherein the executed program instructions further:

receive a plurality of coded parameters for a next good frame following the lost frame, wherein the good frame is a successfully decoded frame;

generate a third synthesized signal for the next good frame further based on the plurality of coded parameters; and

generate the decoded audio output of the next good frame based on a second weighted sum of the first synthesized signal and the third synthesized signal.

18. The apparatus according to claim 17, wherein the second weighted sum comprises at least two weight boundaries, and wherein the at least two weight boundaries are determined based on a pitch epoch of the first synthesized signal, a pitch epoch of decoded audio of the next good frame, and the pitch delay.

19. The apparatus according to claim 16, wherein the executed program instructions further:

determine that the next good frame is a time domain frame;

generate an extended decoded audio output by extending the length of the decoded audio output of the at least one lost frame by a quantity of samples that is predetermined based on a bandwidth of an audio signal being conveyed by the frame;

couple the extended decoded audio output to a CELP state generator; and

generate the decoded audio for the next good frame based at least upon the output of the CELP state generator.