JP3219467B2 - Audio decoding method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention is applied to, for example, a mobile radio system in which code errors frequently occur on a transmission line, and is provided with speech coded information subjected to a code error control technique (error correction / detection coding processing and the like). The present invention relates to a speech decoding method for reproducing speech with low quality degradation due to a code error from code error detection information.

[0002]

2. Description of the Related Art As a high-efficiency speech coding method of 8 kbit / s or less for digital mobile radio and voice storage services, Code Excited Linear Pr
edictionCoding: CELP), Vector Sum Excited Linear Prediction Coding
: VSELP). For each technology, see MRSchroeder and BSAtal: "Code
-Excited Linear Prediction (CELP): High-quality S
peech at Very Low Bit Rates ", Proc. ICASSP-85, 25.
1.1, pp. 937-940 (1985) and IAGerson an
d MA Jasiuk: "Vector Sum Excited Linear Predict
ion (VSELP) Speech Coding at 8kbps ", Proc.ICASSP-9
0, S9.3, pp. 461-464, (1990).

Here, CELP coding will be described. In CELP coding, a linear prediction parameter code, a power parameter code indicating the power of the frame, a pitch period code, a noise code, and a gain parameter code of a sound source are transmitted as speech coding information, with about 5 to 50 ms as one frame. I do. FIG. 1 shows a CELP decoder. The CELP decoder converts the decoded linear prediction parameter A into a linear prediction synthesis filter.
Set as a coefficient of 12. Using the excitation vector (residual waveform) of the immediately preceding (past) frame obtained up to that time and the decoded pitch cycle code L, the past excitation vector is cut out from the adaptive codebook 10 in that cycle, and this is extracted. A time-series vector repeated for the number of frames is output, and a code vector indicated by the input noise code C is read from the noise codebook 11 as a vector. These time-series vectors are weighted according to the gain codes G (= g1, g2) input by the multipliers 21 and 22, respectively, and then added by the adder 23, and the added output is used as an excitation vector as a synthesis filter. 12
And a power gain P is multiplied by a multiplier 24 to obtain a reproduced sound. Further, the post-filter 13 obtains a final sound output in which noise is reduced by formant emphasis and pitch emphasis.

[0004] In a voice transmission system applied to an application field in which a code error occurs, the quality of transmitted voice is degraded due to a code error by using an error correction coding technique. But,
In an application area such as mobile radio where burst errors frequently occur, even if a strong error correction code is used, a sufficient application effect cannot be obtained. This is because the error density in the burst is high, and it is difficult to remove all errors even if a strong error correction code is used.

[0005] When speech is decoded from coded information that leaves an error that cannot be corrected, the decoded speech is greatly distorted. In order to minimize distortion, error detection information is provided with an error detection function for checking whether or not the frame code has been correctly decoded. If an error is detected, processing different from a normal decoding method is performed. That is, a scheme is performed to perform a waveform restoration process to minimize the influence of code errors.

There are the following two types of restoration methods so far. One is, for example, MJMcLaughlin: "Channel c
oding for digital speech transmission in the Japan
esedigital cellular system ", IEICE Technical Committee on Radio Communication Systems, RCS90-27, pp.41-45, 1990, when the current frame is a frame with a code error, This is a method of repeating the parameter, or a method of repeating the parameter and further controlling the gain of the current frame according to the state of the past frame.

Another repair method is disclosed in Japanese Patent Laid-Open No. 4-30200.
As shown in the "speech decoding method", as shown, there is a method of interpolating a current erroneous frame from a past correct frame and a future correct frame when an error occurs. In this case, if the past correct frame and the future correct frame are close in time, interpolation with small distortion can be performed.

[0008]

However, the restoration from the past alone is to repeat the past speech and reduce the power of the erroneous frame to make the distortion difficult to hear. Since the current erroneous frame parameter is determined only from the past information, a delay longer than that of the current frame is not required for restoration, but the continuity with the next frame is deteriorated, and the discontinuity causes new quality deterioration. There is a disadvantage that.

Further, the interpolation from the past and future correct frames according to JP-A-4-30200 is applied to a transmission line in which burst errors frequently occur, and the frame following the current frame is not a correct frame. Not only that, past and future correct frames may be largely separated in time, and there is a drawback that a large distortion is generated by interpolation. In addition, if any number of errors continue after the current frame, a large delay time is required to use the correct frame information in the future, which is suitable for applications that emphasize real-time performance such as mobile radio. Absent.

Further, in CELP coding, since the memory of the adaptive codebook and the synthesis filter depends on the decoding result of the past frame, if there is no error in the past frame even if there is no code error in the current frame, the reproduced voice is reproduced. However, there is a problem that distortion occurs. SUMMARY OF THE INVENTION An object of the present invention is to solve such a problem and to provide a speech decoding method with less quality deterioration even on a transmission line having a burst error.

[0011]

In the present invention, there is provided a means for solving] receives the speech encoded information and error correction code for each frame the
Present, past, and future obtained by decoding audio encoded information
Frame based on at least one of the parameters
Subframe parameters using subframe parameters.
To the vectors from the adaptive and random codebooks
After weighting and adding, the added excitation vector is
Speech decoding that feeds to the Ruta and multiplies the power to reproduce the sound
The method includes decoding past, current,
Use the code error detection information for the current and future frames.
Flags for past, present, and future frames.
The flag S1 is set from 1 to M-1 (M is a verse) as S2, S1, S0, respectively.
Number of frames considered to have an error
And if no error is detected using the error correction code,
If the error is detected, the flag S0 is set to 0.
S1 + 1, and when the flag S1 is M, the error correction code is used.
And if no error is detected, the flag S0 is set to M + 1,
If an error is detected, the flag S0 is set to M, and the frame update
Update flag S2 with flag S1 and flag S1 with flag S0
And the flag S1 is 0 or the flag S1 is 1 to M
Either flag S0 is 0 or M + 1, or flag S1 is 1 to M
Whether flag S0 is neither 0 nor M + 1 or flag S1 is M + 1
Or the parameters of the subframe in the current frame
Sounds characterized by the process of playing or restoring
Decryption method. Here, if the flag S1 is 0,
If S2 is 0, the parameters used in the past for audio playback
Parameters of the current frame and parameters of the current frame.
Interpolate the meter and if flag S2 is any of 1 to M,
Use the parameters of the current frame, or set the flag S0 to 0
Then the parameters of the future frame and the parameters of the current frame
Interpolate the subframe parameters from the meter and set the flag
If S0 is any of 1 to M, the parameters of the subframe
Using the parameters of the current frame as the
Is either 1 to M and flag S0 is 0 or M + 1
The parameters of the past frame and the parameters of the future frame.
Interpolating the parameters of the sub-frame using data,
M + 1 even if flag S1 is 1 to M and flag S0 is 0
If not, the parameter
Using the parameters used for the playback of audio removed by the flag
If S1 is M + 1, the parameters used in the past for audio playback
The current frame with parameters that result in a data or flat spectrum
By interpolating the parameters of
Or if the flag S0 is 0,
From the parameters and the parameters of the current frame,
Frame parameters, and if flag S0 is M,
The parameters of the current frame as subframe parameters
And set it as a coefficient of the synthesis filter.
Alternatively, the audio coding information further includes a power parameter
Including a sign and a flag for power as said parameter
If S1 is 0, if the flag S2 is 0, it is used for audio playback in the past
From the parameters set and the parameters of the current frame.
Interpolate the subframe parameters and set the flag S2 from 1 to M
In either case, use the parameters of the current frame and
Parameters used to play audio in the past frame.
If the value is larger than the constant multiple of the meter,
Use a parameter multiplied by a smaller value, or use a flag
If S0 is 0, future frame parameters and current frame
Interpolate the sub-frame parameters from the parameters of
If the flag S0 is any of 1 to M,
Using the parameters of the current frame as parameters,
The lag S1 is 1 to M and the flag S0 is 0 or M + 1
In the case of, the parameters used for audio playback in the past frame
Using the parameters of the future frame
The parameters of the interpolation result are
Constant parameters used for audio playback in the last frame
If the interpolation result is larger than the threshold times,
Is multiplied by a value less than 1, and the flag S1 is any one of 1 to M
If the flag S0 is neither 0 nor M + 1,
Used in the past for audio playback
Multiply the parameter by a value less than 1 and set the flag S1 to M + 1
If the parameters used in the past to play the audio and the current
From the parameters of the frame,
Interpolated and multiplied by a value less than 1, or flag S0
If 0, the parameters of the future frame and the current frame are
Interpolating the parameters of the sub-frame from the parameters,
If the flag S0 is M, the
Multiply the parameter of the current frame by a value less than 1.
You. Alternatively, the audio coding information may further include a subframe.
If the flag S1 is 0, the pitch period code
Groups on the pitch delay indicated by the pitch period codes of standing frame
Flag S1 is between 1 and M and flag S0 is 0.
Or M + 1, used for audio playback in the past frame
With the pitch delay and the pitch period code of the future frame
The flag S1 is between 1 and M based on the indicated pitch delay.
If the flag S0 is neither 0 nor M + 1
Based on the pitch delay used to play the sound in the
When the lag S1 is M + 1, the pitch period code of the current frame
The past excitation vector is cut based on the indicated pitch delay.
From the adaptive codebook
Output. Alternatively, the audio coding information may further include a sub-file.
Including the gain code for each frame, if the flag S1 is 0,
Using a set of gains indicated by the gain sign of the current frame,
Flag S1 is 1 to M and flag S0 is 0 or M + 1
Is used for audio playback in the past frame.
Gain and gain indicated by the gain sign of the future frame
And the flag S1 is any of 1 to M and the flag S0 is 0
If it is not M + 1, it will be used for audio playback in the past frame
If the flag S1 is M + 1 using the set of gains
Using a set of gains indicated by the gain sign of the current frame,
Each vector from the adaptive codebook and the random codebook
Find out.

[0012]

[0013]

[Embodiment] In the following, CELP is used as a high-efficiency speech coding system.
Taking the scheme, an embodiment of the present invention for the CELP scheme will be described. Although the present embodiment describes a CELP decoder, the present invention can be applied to a similar scheme or a scheme including a similar configuration.

FIG. 2 is a block diagram showing a speech decoding process according to an embodiment of the present invention. In the present embodiment, decoding using information on past and present frames and up to one future frame will be described. First, the code sent from the transmission path is subjected to error correction processing in the error correction circuit 30 for each frame. The voice encoding information is sent from 30 to the buffer 31, and the error detection information is sent to the buffer. The buffers 31 and 34 take in past and present frames and at least information of at least one frame ahead. The error detection information from the buffer 34 enters the error state management unit 35, and the error states of past, present, and future frames are determined.

The speech coded information output from the buffer 31 and the error state from the error state management unit 35 are compared with the parameter decoding unit 32.
To decode and interpolate the parameters of the current frame.
Based on the output of the parameter decoding unit 32, the audio is reproduced by the audio decoder 33. The error state management unit 35, as shown in FIG. 3, uses a group of flags indicating the state of the past, present, and future frames based on the error detection information up to the next frame stored in the buffer 34 of FIG. 352 (S2, S1, S0) is updated.
S2 corresponds to the past frame, S1 corresponds to the current frame, and S0 corresponds to the future frame. The flags S2 and S1 are obtained by shifting the contents of the flags S1 and S0 in the previous frame when updating the frame for speech coding. The content of the S0 flag is determined in the status flag updating unit 351 according to the error detection information of the input frame from the corresponding buffer 34 and the content of the S1 flag (which corresponds to the S0 flag in the previous frame).

FIG. 4 shows an update table of S0. In FIG.
When the input of the buffer 34 = 0, the input frame indicates that the error correction has been completed completely, and the input of the buffer 34 =
When 1, it means that the frame could not be corrected, that is, a frame in which an error was detected. When S1 is 0, S0 = 0 if no error is detected in the input frame, and S0 = 1 if an error is detected. If S1 is from 1 to a predetermined number M-1, S0 = 0 if no error is detected in the input frame, and S0 = S1 + 1 if an error is detected, until the corresponding frame. The number of consecutive frames of error frames is held as the content of the flag. When S1 is M, if no error is detected in the input frame, S0 = M + 1, and if an error is detected, S0
= M. Here, M is a length that is considered to have a burst error added to consecutive frames, for example, 6,
And When S1 is M + 1, S0 = 0 if no error is detected in the input frame, and S0 = M if an error is detected in the input frame. Therefore, in the case of S0 = M + 1, there is no error detection of the corresponding frame, but since the burst error continues, it is not set to 0, and the error is detected in the next frame. It is set to 0 when it is no longer performed.

Regarding the transition of the state flag,
MJ McLaughlin's document described in the conventional technology (IEICE Technical Committee on Wireless Communication Systems, RCS90-27, pp.41-45,1
990), the current frame is used as an input, and S0 is used as a flag of the current frame together with the past flag. In the present invention, the current frame is S1, the future frame is S0, and the frame to be decoded is Is greatly different in that the flag of the frame that comes next to is used.

Next, the error management control unit 35 thus obtained
Using the output S (S2, S1, S0) of FIG.
At 32, parameter decoding of each piece of speech coded information is performed in consideration of error states of past, present and future frames. FIG. 5 shows an example of a flowchart of a processing procedure depending on the error state flag in the parameter decoding unit 32. First, in step 101 of FIG. 5, it is determined whether or not the error state flag S1 of the current frame is 0, and if it is 0, the process 1 of step 111 is executed. In the process 1, since S1 = 0, the parameters are basically decoded based on the speech coding information of the current frame. When S0 and S2 are checked, if S0 or S2 is not 0, the parameters are continuously changed. Modify to connect.

Next, in step 102, if S1 is not 0 and the flag is any of 1 to M, it indicates that an error has been detected in the current frame. Therefore, in this case, the audio coding information of the current frame includes:
Contains errors. Therefore, information of past frames and future frames is used. In step 103, if S0 = 0 or M + 1, the future frame is correct, so
As processing 2 in step 112, the parameters of the current frame are interpolated from both sides from the parameters used for decoding in the past frame and the decoding parameters of the future frame. At this time, the weight of interpolation can be changed depending on the degree of error continuation of the flag of S1. For example, if S1 is close to M,
The parameters of the past frame are the parameters obtained by the extrapolation method of the step 113 processing 3 described below in which the burst error continues, so that the weight of the past parameter is small and the weight of the future frame is increased. This makes it possible to obtain the parameters of the current frame with less distortion. In step 103, S0 ≠ 0 and S0
In the case of ≠ M + 1, an error has been detected in a future frame. In this case, in the process 3 of step 113, an interpolation process is performed by extrapolating the parameters used for decoding in the past frame to the current frame. At that time, the weight to the extrapolation result can be changed according to the degree of error continuation of the S1 flag.

In step 102, if NO, S1
= M + 1, which indicates that the current frame is correct after a burst error of M frames or more in the past frame. Therefore, in the processing 4 of step 114, the parameters decoded on the basis of the speech coding information of the current frame are used as the basis, and since S2 = M, there is a burst error in the past frame and S0 is 0. Or M
Therefore, the parameter is modified so as to be connected continuously according to the value.

Next, with respect to actual parameters, examples of processing 1 to processing 4 in FIG. 5 will be described. As shown in FIG. 1, the speech coding information of each frame in CELP coding includes a linear prediction parameter code A, a power parameter code P, a pitch period code L, a noise code vector code C of a noise codebook, and a gain of a sound source. It consists of a parameter code G. First, an example of the linear prediction parameter code A will be described. As the linear prediction parameter, at the time of interpolation, a line spectrum pair (LSP: Line Spectrum) having a good linear interpolation characteristic is used.
Pair) is used. Assuming that one frame is 20 ms, the linear prediction parameter actually requires a parameter for each subframe unit of about 5 ms to 10 ms. In that case, the parameters of the sub-frame are obtained by interpolation between the parameters of the previous frame (past frame) or the subsequent frame (future frame). Then, when there is a code error in the previous (or subsequent) frame, a large distortion occurs when directly interpolating with an incorrect parameter. Even if the parameters of the current frame are incorrect, if the immediately preceding and succeeding parameters have no error, the current frame can be interpolated from the preceding and succeeding frames.

The past (time n-1) and the present (time n
), LSP parameters that are not interpolated immediately after error correction in the future (time n + 1) frame are ω (n-1), ω (n),
ω (n + 1), and parameters corresponding to times n−1 and n after interpolation in consideration of the error state flag are Ω (n−1) and Ω (n).
Expressed by For t in the range of a ≦ t ≦ b, the time a
Define the interpolation parameter Ω (t) at time t when Xa is the parameter of Xb at time b and Ω (t) = f_A (t; (a, Xa), (b, Xb)) . The function f_A is a function for interpolating between two points (a, Xa) and (b, Xb), and outputs an interpolated value at time t.
For example, linear interpolation may be used.

In the processing 1 of step 111 in FIG. 5, since the error state flag S1 of the current frame is 0 and there is no error in ω (n), the following is performed according to the error state of S2 and S0. good. At t in the current frame, for n-1 ≦ t ≦ n, if S2 = 0, Ω (t) = f_A (t; (n-1, Ω (n-1)), (n, ω
(n))), S2 = 1, ..., M, then Ω (t) = f_A (t; (n-1, ω (n)), (n,
ω (n))). Further, when n ≦ t ≦ n + 1, if S0 = 0, Ω (t) = f_A (t; (n, ω (n)), (n + 1, ω (n +
1))), S0 = 1, ..., M, then Ω (t) = f_A (t; (n, ω (n)), (n + 1,
ω (n))).

In the process 2 of step 112 in FIG. 5, ω (n) is not used because the current frame parameter has an error.
Since S0 = 0 or M + 1, using Ω (n-1) and ω (n + 1), the parameter at time t in the current frame is expressed as Ω (t) = f_A (t; (n-1, Ω (n-1)), (n + 1, ω (n + 1))). In the process 3 of step 113, since S0 = 1,..., M, Ω (t) = f_A (t; (n-1, Ω (n−1)), (n + 1, Ω (n-1))) to interpolate the parameter at time t in the current frame. In this case, the output LSP is calculated using a function that extends the bandwidth of the Ω (t) spectrum according to the values of S2, S1, and S0.
Interpolation to prevent distortion may be performed by weakening the resonance characteristics of the parameters.

In process 4 of step 114 in FIG. 5, ω (n)
Can be used, but Ω (n-
Use either 1) or use Ω0 as a parameter that results in a flat spectrum. This is Ω ′. At t in the current frame, when n-1 ≤ t ≤ n, Ω (t) = f_A (t; (n-1, Ω ')), (n, ω (n))), n ≤ t ≤ For n + 1, if S0 = 0, Ω (t) = f_A (t; (n, ω (n)), (n + 1, ω (n +
1))), if S0 = M, Ω (t) = f_A (t; (n, ω (n)), (n + 1, ω (n))).

Next, consider each processing of the power parameter code shown in FIG. Basically, it is the same as the case of interpolation of the linear prediction parameter. In the case of power, the case of calculating power in subframe units will be described. Power parameters that are not interpolated immediately after error correction in the past (time n-1), current (time n), and future (time n + 1) frames are p (n-1), p (n), p (n + 1), and parameters corresponding to times n−1 and n after interpolation in consideration of the error state flag are P (n−1) and P (n).

Time when Xa at time a and Xb at time b
The power interpolation parameter P (t) at t is given by P (t) = f_P (t;
(a, Xa), (b, Xb)). The function f_P has two points (a, X
a), (b, Xb) linear interpolation between, or, Xa, may be a function such as linear interpolation of the logarithmic representation of Xb.

In the processing 1 of step 111 in FIG. 5, since S1 = 0, there is no error in p (n). Therefore, at t in the current frame, if n−1 ≦ t ≦ n, if S2 = 0, If P (t) = f_P (t;
(n-1, P (n-1)), (n, p (n))), S2 = 1, ..., M, then P (t) =
Let f_P (t; (n-1, p (n)), (n, p (n))). Then, the resulting P (t) is a constant multiple of P (n-1) (for example, α ≧ 1).
If it is greater than α′P (t), (0 ≦ α ′ <1) , the power is limited to a small value so that distortion due to a sudden change in power is not heard. Also, n ≦ t ≦ n + 1
Then, if S0 = 0, P (t) = f_P (t; (n, p (n)), (n + 1, p (n
+1))), S0 = 1, ..., M, then P (t) = f_P (t; (n, p (n)), (n
+ 1, p (n))).

In process 2 of step 112 in FIG. 5, since there is an error in the current frame parameter, p (n) is not used, and S0 = 0 or M + 1, so that P (n-1), p (n n + 1), the parameter at time t in the current frame is calculated as P (t) = f_
Interpolation is performed using P (t; (n-1, P (n-1)), (n + 1, p (n + 1))). Here, for a certain threshold value (for example, α (≧ 1)), P (t)>
αP (n-1), and if the power change is too large,
By substituting the power as α′P (t), (0 ≦ α ′ <1) and limiting the power to a small value, distortion is not heard. α ′ may be changed according to the number of continuous burst errors in S1.

In the process 3 in step 113, since S0 = 1,..., M, P (t) = f_P (t; (n-
1, P (n-1)), (n + 1, P (n-1))), and further, the volume of the error frame is weakened, and P (t) is set to α′P (t), (0 ≦ α '<1) . α ′ may be changed according to the number of continuous burst errors in S1. For example, if S1 = M, it means that a large burst error of M frames or more has occurred by the current frame. Therefore, α ′ = 0 is set as a silent section, and no distorted speech is output.

In process 4 of step 114 in FIG. 5, p (n) can be used, but the previous frame is a burst error frame.
Therefore, at t in the current frame, when n-1 ≦ t ≦ n, P (t) = f_P (t; (n-1, P (n-1)), (n, p (n))) And
Replace P (t) with α′P (t) (0 ≦ α ′ <1) . n ≤t
For ≦ n + 1, if S0 = 0, P (t) = f_P (t; (n, p (n)), (n
+ 1, p (n + 1))), if S0 = M, P (t) = f_P (t; (n, p (n)), (n
+ 1, p (n))). When S0 = M, P (t) is further set to α′P (t), (0 ≦ α ′ <1) .

Next, an example of parameter decoding of the pitch period code L will be described. The pitch period code L is a delay value indicating the pitch delay of the adaptive codebook 10 in FIG. This is 1
Divide the frame into k (for example, 2) subframes,
The value of L is transmitted as pitch information by the number. The pitch delay in CELP is not only the pitch period but also a double period, 3
It may take discrete values such as double periods. Therefore, in this embodiment, when k = 2, a method of replacing the value of the pitch period between the previous frame and the subsequent frame when the current frame is erroneous is shown. Of course, if the fluctuation of the pitch period is small and the pitch periods of the preceding and following frames can be interpolated, the pitch period may be obtained by interpolation.

The pitch delays immediately after error correction in the past (time n-1), present (time n), and future (time n + 1) frames are lk (n-1), lk (n), and lk (n, respectively. +1), (k = 1, 2), and parameters corresponding to times n−1, n after considering the error state flag are Lk (n−1), Lk (n), (k = 1, 2). ). FIG.
In the processing 1 of step 111, since S1 = 0, there is no error in lk (n), so that Lk (n) = lk (n), (k = 1,2).

In process 2 of step 112 in FIG. 5, since there is an error in the parameters of the current frame, lk (n) is not used, and L1 (n) = L2 (n-1), L2 (n) = l1 (n + 1). In process 3 of step 113, since S0 = 1,..., M, Lk (n) = L2 (n-1), (k = 1, 2) using the past parameters.

In the processing 4 of step 114, the current frame is correct, so that Lk (n) = lk (n), (k = 1, 2), as in the processing 1. Next, a decoding example of the sound source gain parameter G will be described. G is the adaptive codebook 10 and the noise codebook 11 in FIG.
Collectively refers to a set of gains g1 and g2 multiplied by each sound source signal selected in. For G, parameter decoding is performed in the same manner as for the pitch delay. Past (time n-1), present (time n
), The sound source gain immediately after error correction in the future (time n + 1) frame is σk (n-1), σk (n), σk (n + 1), (k = 1,
2) and time n-1 and n after considering the error status flag
Are defined as Gk (n-1), Gk (n), (k = 1,2).

In process 1 of step 111 in FIG. 5, since S1 = 0 and there is no error in σk (n), Gk (n) = σk (n), (k = 1, 2). In process 2 of step 112 in FIG. 5, since there is an error in the current frame parameter, σk (n) is not used, and G1 (n) = G2 (n-1) and G2 (n) = σ1 (n + 1)

In process 3 of step 113, since S0 = 1,..., M, Gk (n) = G2 (n-1), (k = 1,2) using the past parameters. In process 4 of step 114, the current frame is correct, so that Gk (n) = σk (n), (k = 1, 2) as in process 1.

Regarding the decoding of the noise vector code C in the noise codebook, since the correlation between frames is not large, in this embodiment, the code sent to the current frame is used as it is for reproduction regardless of the error state flag. . Thus, FIG.
The parameter decoding unit 32 decodes the parameters of the speech coded information based on the past, present, and future error state flags.
Will be repaired. The result is passed to the next audio decoder 33,
The voice is resynthesized.

In the speech decoder 33, the error state management unit
Past, present and future frame error status flag S from 35
By controlling the reproduced voice of the decoded voice using (= S2, S1, S0), distortion due to a transmission error can be reduced. That is, in the post filter 13 of FIG. 1, the coefficient of the post filter is controlled using S to suppress the auditory distortion. Usually, the post filter is constituted by H (z) as follows.

[0040]

(Equation 1)

Here, αi is a linear prediction coefficient of each frame, βi is a long-term prediction coefficient, and τ is a pitch delay. These are adaptively determined from transmission parameters or decoded speech. Formant enhancement is performed by A (z / γ1) / A (z / γ2), and pitch enhancement is performed by 1 / (1-δB (z)).

[0042]

(Equation 2)

Thus, the effect of spectral tilt correction is provided. In this embodiment, H (z) is adaptively changed for each frame by using η, δ, γ1, γ2 as a function of the error state flag S (= S2, S1, S0). When S2, S1, and S0 are all 0 or M + 1, the normal value is used. However, when any of S2, S1, and S0 is other than 0 and M + 1 and the frame is an error, Since there is a high possibility that the adaptive codebook and the noise code vector code contain errors, reduce η and δ, weaken the high-frequency spectral tilt and the pitch enhancement of the sound source, and adjust γ1 and γ2. By increasing the degree of formant emphasis, noise in the output sound can be suppressed. Here, by using the error status flags up to the future frame of S0, S2, S1
The switching of η, δ, γ1, and γ2 can be performed more continuously than in the case of only the above, and deterioration due to a change in sound quality due to switching between frames can be suppressed.

[0044]

As described above, according to the present invention,
When a code error of the transmission line remains in the audio coded bit, it is possible to suppress the annoying distortion superimposed on the decoded audio.

[Brief description of the drawings]

FIG. 1 is a block diagram of a CELP speech decoder.

FIG. 2 is a block diagram of a speech decoding process in the embodiment of the present invention.

FIG. 3 is an enlarged view of an error state management unit 35 of FIG. 2;

FIG. 4 is an update table of an S0 flag in a status flag updating unit 351 of FIG. 3;

FIG. 5 is a schematic diagram of a processing flow according to an error state flag.

[Explanation of symbols]

 10 is an adaptive codebook 11 is a noise codebook 12 is a linear prediction synthesis filter 13 is a post filter 21, 22, 24 is a multiplier 23 is an adder 30 is an error correction circuit 31, 34 is a buffer 32 is a parameter decoding unit 33 is a voice The decoder 35 is an error state management unit

────────────────────────────────────────────────── ─── Continuation of the front page (56) References JP-A-2-176800 (JP, A) JP-A-4-30200 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) G10L 13/00 G10L 19/00

Claims (5)

(57) [Claims] 1. An audio coding apparatus comprising:
Received by decoding the audio coded information
Current, past and / or future frames
Subframe parameters based on
Use adaptive codebook and random codebook for each subframe
After the vectors are weighted,
Supply the vibration vector to the synthesis filter and multiply the power
In speech decoding method for reproducing, in decoding the voice coded data, past, current use and the respective code error detection information for future frame <br/> over arm, past, present, correspond to future frames Each flag
The flag S1 is set from 1 to M-1 (M is bursty) as S2, S1, S0.
Number of frames considered to have an error)
If no error is detected using the error correction code,
The lag S0 is set to 0, and the flag S0 is set to S1 + 1 when an error is detected.
When the flag S1 is M, error detection is performed using the error correction code.
If not issued, the flag S0 is set to M + 1 and an error is detected.
Flag S0 is set to M , flag S2 is set to flag S1, and flag S1 is set to
Update with flag S0, flag S1 is 0, flag S1 is 1 to M, and flag S0 is 0 or M + 1
Or M + 1 even if flag S1 is 1 to M and flag S0 is 0
Or whether the flag S1 is M + 1 or
A speech decoding method, wherein a process of reproducing or restoring a parameter depends on the process. 2. The speech decoding method according to claim 1, wherein
hand, The audio encoding information includes a linear prediction parameter code,
For the line spectrum pair as the parameter, If the flag S1 is 0, and if the flag S2 is 0,
Raw parameters and current frame parameters
Interpolate the subframe parameters from If the flag S2 is any of 1 to M, the parameter of the current frame
Using a meter, Or if the flag S0 is 0, the parameters of the future frame
And the parameters of the current frame Data from the subframe
If the flag S0 is any of 1 to M,
The parameters of the current frame as subframe parameters
Data, Flag S1 is 1 to M and flag S0 is 0 or M + 1
 In the case of
The parameters of the subframe are supplemented using the parameters.
Wait M + 1 even if flag S1 is 1 to M and flag S0 is 0
If not, the parameter
Using the parameters that were previously used for audio playback, If the flag S1 is M + 1, it was used for audio playback in the past
Parameter or the parameter that results in a flat spectrum
By interpolating the frame parameters,
Interpolate the meter, Or if the flag S0 is 0, the parameters of the future frame
From the parameters of the current frame and the parameters of the current frame.
Parameters are interpolated, and if the flag S0 is M, the subframe
Using the parameters of the current frame as
And setting the coefficients as the coefficients of the synthesis filter.
Audio decoding method. 3. The speech decoding method according to claim 1, wherein
hand, The audio coding information includes a power parameter code,
About power as a parameter If the flag S1 is 0,
If the lag S2 is 0, the parameters used for audio playback in the past
Data and subframe parameters from the current frame parameters.
Interpolate the parameters If the flag S2 is any of 1 to M, the parameter of the current frame
Using a meter, The parameter was used to play audio in the past frame
If the parameter is larger than a constant multiple,
Using a parameter multiplied by a value less than 1, Or if the flag S0 is 0, the parameters of the future frame
And subframe parameters from the current frame parameters
If the flag S0 is any of 1 to M,
The parameters of the current frame as subframe parameters
Data, Flag S1 is 1 to M and flag S0 is 0 or M + 1
 In the case of, the parameters used for audio playback in the past frame
Data and the parameters of future frames. Buffle
The parameters of the interpolation
One of the parameters used for audio playback in the past frame
If the value is larger than a certain threshold value,
Multiplied by a value less than 1 M + 1 even if flag S1 is 1 to M and flag S0 is 0
If not, the parameter
The parameter used to play back the sound was less than 1
Multiply by If the flag S1 is M + 1, it was used for audio playback in the past
From the parameters and the parameters of the current frame,
Interpolate the frame parameters and multiply by a value less than 1. Or if the flag S0 is 0, the parameters of the future frame
From the parameters of the current frame and the parameters of the current frame.
Parameters are interpolated, and if the flag S0 is M, the subframe
1 for the parameter of the current frame
Sound reproduced by multiplying the sound by a smaller value
Voice decoding method. 4. The speech decoding method according to claim 1, wherein
hand, The speech coded information is a pitch period code for each subframe.
Including When the flag S1 is 0, the pitch period code of the current frame is used.
Based on the indicated pitch delay, Flag S1 is 1 to M and flag S0 is 0 or M + 1
 Is used for audio playback in the past frame.
Indicated by the pitch delay and the pitch period code of the future frame
Based on the pitch delay M + 1 even if flag S1 is 1 to M and flag S0 is 0
If not, the last frame used to play the audio
Based on pitch delay, If the flag S1 is M + 1, the pitch period code of the current frame
Based on the pitch delay shown by Time series vector obtained by extracting past excitation vectors and repeating
Output from the adaptive codebook.
Decryption method. 5. The speech decoding method according to claim 1, wherein
hand, The audio coding information includes a gain code for each subframe.
See When the flag S1 is 0, it is indicated by the gain code of the current frame.
The flag S1 is one of 1 to M
If the flag S0 is 0 or M + 1,
Gain used for audio playback and the gain of future frames
Using the gain indicated by the M + 1 even if flag S1 is 1 to M and flag S0 is 0
If not, the last frame used to play the audio
Using a set of gains, When the flag S1 is M + 1, it is indicated by the gain code of the current frame.
Using a set of gains Each vector from the adaptive codebook and the random codebook
A voice decoding method characterized by finding a voice.