Speech decoding and encoding apparatus for lost frame concealment using predetermined number of waveform samples peripheral to the lost frame

• US 8,812,306 B2
• Filed: 07/11/2007
• Issued: 08/19/2014
• Est. Priority Date: 07/12/2006
• Status: Active Grant

• Current Assignee: III Holdings 12, LLC (Intellectual Ventures LLC)
• Original Assignee: Panasonic Intellectual Property Corporation of America (Panasonic Holdings Corporation)
• Inventors: Kawashima, Takuya, Ehara, Hiroyuki, Yoshida, Koji
• Primary Examiner(s): Kazeminezhad, Farzad
• Application Number:
  US12/307,974
• Publication Number:
  US 20090326930A1
• Time in Patent Office: 2,596 Days
• Field of Search: 704/228, 704/201, 370/201
• US Class Current: 704/207
• CPC Class Codes: G10L 19/005 Correction of errors induced by the transmission channel, if related to the coding algorithm
• CPC Class Codes: G10L 2019/0011 Long term prediction filters, i.e. pitch estimation
• CPC Class Codes: G10L 25/90 Pitch determination of speech signals
• CPC Class Codes: H04L 47/10 Flow control; Congestion control
• CPC Class Codes: H04L 47/43 Assembling or disassembling of packets, e.g. segmentation and reassembly [SAR]

Abstract

An audio decoding device capable of suppressing an information amount for a lost flame compensation process and encoding efficiency is provided. A decoded sound source generator generates a lost frame'"'"'s CELP decoded sound source signal. A pitch pulse information decoder CELP decodes a pitch pulse position information and a pitch pulse amplitude information. A pitch pulse waveform learner learns a pitch pulse learning waveform in a past frame in advance from the lost frame. A convolution adjuster amplitude-adjusts the pitch pulse learning waveform according to the pitch pulse amplitude information by considering a predetermined number of waveforms peripheral to a peak position of the lost frame'"'"'s CELP decoded excitation signal, and convolutes a pitch pulse waveform into a time axis which has been amplitude-adjusted according to the pitch pulse position information. A sound source signal corrector adds or replaces the pitch pulse waveform convoluted into the time axis to the lost flame decoded sound source signal.

Claims

1. A speech decoding apparatus, comprising:

a receiver that receives frame loss information identifying a first frame that is a lost frame and speech encoded data of a second frame subsequent to the first frame wherein the speech encoded data includes an encoded excitation signal parameter of the second frame and, if encoded pitch pulse information of the first frame exists, the encoded pitch pulse information of the first frame;

a first decoder that decodes the encoded pitch pulse information, if the encoded pitch pulse information exists, to acquire a pitch pulse position and pitch pulse amplitude information of the first frame;

a second decoder that decodes the encoded excitation signal parameter to acquire excitation signal parameters;

a decoded excitation generation section that performs CELP decoding and concealment processing of a lost frame, and generates a decoded excitation signal using the excitation signal parameters;

a pitch pulse waveform learning section that analyzes whether a steady-state is present in the decoded excitation signal, and if the steady-state is present, detects peak position in the decoded excitation signal and extracts a predetermined number of samples peripheral to the peak position and performs learning of a pitch pulse waveform by smoothing the extracted samples with a past pitch pulse learning waveform and generates a learned pitch pulse waveform;

an excitation signal compensation section that adjusts the amplitude of the learned pitch pulse waveform according to the pitch pulse amplitude information of the first frame and adds the learned pitch pulse waveform to the decoded excitation signal according to the pitch pulse position of the first frame to generate a compensated excitation signal; and

an excitation signal selection section that selects the decoded excitation signal if the frame loss information does not indicates a frame loss and selects the compensated excitation signal if the frame loss information does indicate a frame loss.

2. The speech decoding apparatus according to claim 1, wherein:

the second decoder further decodes an encoded parameter transmitted in the second frame subsequent to the first frame to acquire an excitation power of the first frame as first excitation power; and

a first adjuster that adjusts power of the compensated excitation signal to the first excitation power.

3. The speech decoding apparatus according to claim 2, wherein the first adjuster adjusts power of the compensated excitation signal to the first excitation power while maintaining an amplitude of the first pitch pulse waveform.

4. The speech decoding apparatus according to claim 3, further comprising:

a first calculator that calculates excitation power of the first pitch pulse waveform as a second excitation power;

a second calculator that calculates a difference between the first excitation power and the second excitation power as a third excitation power; and

a difference generator that makes an amplitude of a section corresponding to the first pitch pulse waveform zero in an excitation signal of the first frame, to generate a differential excitation signal,

wherein the first adjuster adjusts amplitude of the differential excitation signal and adjusts excitation power of the differential excitation signal to the third excitation power.

5. The speech decoding apparatus according to claim 1, further comprising:

a detector that eliminates the first pitch pulse waveform in a search section of a predetermined length backward in time from a rear-end of the first frame and detects a pitch pulse waveform whose amplitude is at least equal to a threshold value as a pseudo-pitch-pulse-waveform;

a noise generator that generates noise of a length of the pseudo-pitch-pulse-waveform; and

a replacer that replaces the pseudo-pitch-pulse-waveform with the noise in an excitation signal of the first frame.

6. The speech decoding apparatus according to claim 5, wherein the predetermined length of the search section is equal to a pitch lag length of one frame ahead in time from the first frame.

7. The speech decoding apparatus according to claim 5, wherein the noise is generated using one of random noise, Gaussian noise, and a past steady-state excitation.

8. The speech decoding apparatus according to claim 1, wherein:

the pitch pulse information further includes the pitch lag; and

the excitation signal compensation section further generates a second pitch pulse waveform by multiplying the pitch pulse learned waveform by an attenuation coefficient, calculates an addition position of the second pitch pulse waveform using the pitch pulse position information and the pitch lag, and further adds the second pitch pulse waveform to the excitation signal at the calculated addition position to generate a compensated excitation signal.

9. The speech decoding apparatus according to claim 8, wherein the excitation signal compensation section calculates a position separated by the pitch lag in a past direction of a time domain from a position indicated by the pitch pulse position information as the addition position.

10. The speech decoding apparatus according to claim 8, wherein the excitation signal compensation section detects a last pitch pulse position of a frame immediately preceding the first frame that is the lost frame, and calculates a position, which is located at a distance equal to the pitch lag of the lost frame from the last pitch pulse position, as the addition position of the second pitch pulse.

11. The speech decoding apparatus according to claim 1, wherein the excitation signal compensation section detects a last pitch pulse position of a frame immediately preceding the first frame that is the lost frame, calculates a position, which is located at a distance equal to the pitch lag of the lost frame from the last pitch pulse position, as an addition position of the second pitch pulse, extracts, from the excitation signal, a predetermined section centered at the addition position of the second pitch pulse and performs windowing for the extracted section from the excitation signal so that an amplification coefficient of a window at an endpoint of a pitch pulse waveform of the extracted section becomes a value close to 1.0, to obtain a third pitch pulse waveform, and adds the third pitch pulse waveform to the addition position of the second pitch pulse of the excitation signal to generate a compensated excitation signal.