JPH08292799A - Adpcm coded voice decoding method - Google Patents

Abstract

PURPOSE: To make the auditive distortion just after a chipped frame small. CONSTITUTION: When a frame chipping is generated, a code string for chipped frame (codes are all '1's) is supplied to a decoding part 14 and when the frame chipping is restored, the decoding part 1 4 is initialized at the time of starting an ADPCM decoding. Then, a decoded accoustic signal output is made by applying a gain mulitiplication to a decoded waveform in a level adjusting part 35 so that power of the frame just after the chipped frame becomes the order of two third with respect to power of the frame just before the chipped frame and thereafter after 19 frames, the gain becomes '1'.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention decodes an acoustic signal such as voice which is received from a transmission line in which frame loss may occur, such as mobile radio, and which is ADPCM (adaptive differential PCM) encoded. The present invention relates to an ADPCM coded speech decoding method.

[0002]

2. Description of the Related Art In speech coding and decoding, which is applied to an application field in which significant phase fluctuations and reception level fluctuations occur on a transmission line such as mobile radio and code errors frequently occur, error compensation technology is generally used. In many cases, measures are taken to reduce the deterioration of the quality of transmitted speech due to code errors.

However, when ADPCM coding / decoding is used as in a digital cordless telephone and voice coding / decoding is performed without error correction, the quality of transmitted voice is often deteriorated due to a code error. Therefore, in a digital cordless telephone using ADPCM voice coding / decoding,
Utilizing the error detection function of the transmission frame unit provided in the transmission channel, a technique called muting that replaces a part of the received bit string of the frame in which an error is detected,
An attempt has been made to suppress auditory quality deterioration of a decoded speech signal by using a technique called attenuation, which suppresses the power of a decoded waveform of a frame in which an error is detected.

However, the above method cannot be applied when a frame loss occurs due to an extreme decrease in received input. As a processing method in this case (when a frame is lost), several methods have been proposed in the past, and as a method widely used because of its simplicity and effectiveness, a code sequence of a lost frame is processed by an ADPCM decoding unit. There is a method of replacing with a separately prepared code sequence before inputting into. This method will be described with reference to FIG. 15A. The received ADPCM code sequence from the input terminal 11 is input to the frame loss detection unit 12, and when no frame loss occurs, passes through the frame loss detection unit 12 and is further input to the ADPCM decoding unit 14 through the changeover switch 13. Normal decoding is performed to obtain a decoded acoustic signal. When a frame loss occurs, this is the frame loss detection unit 12
The detection output is supplied to the switching control unit 15, and the switching control unit 15 switches the changeover switch 13 to the side of the lost frame code sequence generation unit 16 so that the predetermined lost frame code sequence is ADPCM.
The decoded acoustic signal is supplied to the decoding unit 14 to perform decoding and obtain a decoded acoustic signal. For example, the ITU-T standard G.264. 726 32 kbit
In the case of / s ADPCM, the code sequence for the lost frame uses a code "1111" that converges the decoded signal at the time of loss to 0 so that click noise does not occur, or a received code before the lost frame. A method of generating a decoded signal by repeatedly using a part of the sequence is used.

An example of the configuration of the ADPCM decoding unit 14 is shown in FIG. 15B.
Shown in The input code sequence is inversely quantized by the adaptive dequantization unit 18 to decode the difference signal, and the difference signal is added with the prediction signal from the adaptive prediction unit 21 by the addition circuit 19 to obtain a decoded acoustic signal. To be The decoded acoustic signal and the dequantized differential signal are input to the adaptive prediction unit 21 and temporarily stored in the internal state storage unit thereof, and the same method as that on the encoding side is used by the previous plural samples and the current input. The next predicted signal is predicted and output, and a preferable quantization width is estimated, and the estimated quantization width is set in the adaptive dequantization unit 18 by the adaptive control unit 22.

[0006]

The prior art is intended to compensate for the perceptual deterioration of quality when a frame is lost, and no attempt has been made to suppress the perceptual deterioration of quality after the loss of a frame. That is, in the prior art, the loss frame is replaced with a separately prepared lossy frame code sequence to reduce quality deterioration at the time of frame loss.
Since the internal state of the decoding unit is different from the internal state of the encoding side, no consideration is given to the waveform distortion that occurs after the lost frame, and it is possible to prevent the auditory quality from deteriorating due to noise generation. There is a problem that there is no.

An object of the present invention is to provide an ADPCM coded speech decoding method for suppressing auditory quality deterioration after frame loss, which cannot be realized by the conventional technique.

[0008]

According to the invention of claim 1, a part of the code sequence of the frame next to the defective frame is replaced with a code sequence having a large difference value by a code replacement table to correct the code sequence. This correction code string is ADPCM decoded. According to the invention of claim 2, a part of the difference value generated at the time of decoding in the frame next to the lost frame is multiplied by a gain to perform correction to increase the difference value, and the corrected difference value Is ADPCM decoded.

According to the third aspect of the present invention, the ratio between the decoded waveform power immediately before the lost frame and the decoded waveform power immediately after the lost frame is obtained, and a gain corresponding to this power ratio is obtained and the decoded signal immediately after the lost frame is obtained. Is multiplied by and corrected to be almost equal to the decoded waveform power immediately before the lost frame. Claim 4
According to the invention, the internal state of ADPCM decoding is initialized and corrected at the start of decoding of the frame following the lost frame.

According to the invention of claim 5, claims 1 to 3
In any one of the above aspects, the internal state of ADPCM decoding is initialized and corrected at the time of starting the decoding of the frame following the lost frame. According to the invention of claim 6, claims 1 to 5
In any one of the above aspects, the correction is performed only when the power of the decoded waveform immediately before the lost frame is larger than the threshold value.

[0011]

Embodiment 1 FIG. 1A shows an embodiment of the invention according to claim 1, and the portions corresponding to those in FIG. 15A are designated by the same reference numerals. Code replacement unit 25
And the code sequence from the switch 13 to the ADPCM decoding unit 1
4 and the code replacement unit 25, a replacement control unit 27 that controls the switch 26, and a replacement table 28 are added. The code sequence switching control when a frame loss occurs is the same as the conventional method. The frame loss information is input to the replacement control unit 27 from the frame loss detection unit 12. When the frame loss is recovered, the replacement control unit 27 controls the switch 26 to input the received code sequence to the code replacement unit 25. The received code sequence is replaced by the replacement table 28 in the code replacement unit 25.
Is replaced with a code having a large difference value, and the replaced code sequence is supplied to the ADPCM decoding unit 14. This replacement is performed from one to several samples, for example, about two samples from the beginning of the frame following the defective frame. That is, the switch 26 is connected to the code replacement unit 25 side only for one to several samples (codes) immediately after the lost frame.
In FIG. 1B, ITU-T standard G.264. 72632 kbit / s
An example of the substitution table 28 in the case of ADPCM (sampling frequency 8 kHz) decoding and using the code “1111” that converges the decoded signal to 0 in the defective frame is shown. The table 28 is converted into an output code according to the input code and output. With respect to the code 1111 (0000), a value obtained by adding 1 to the least significant bit so that the difference value is large for each of the positive side code and the negative side code is set as the output code, and the code 0001 indicating the maximum positive difference value is given. , The reference numeral 1000 indicating the maximum negative difference value cannot increase the difference value indicated by each, and is therefore used as the output code as it is.

When the code "1111" that converges the decoded signal to 0 in the defective frame is used, the difference value indicated by the code is large with respect to the input code in the code replacement unit. Therefore, the decoded signal rapidly approaches the true decoded signal. Since the output code has a large difference value with respect to the input code, a code sequence with a difference value of zero is input in the missing frame, and the difference value from the decoded signal to a true decoded signal is large. However, it will be approaching rapidly. Embodiment 2 FIG. 2A shows an embodiment of the invention of claim 2 and the same reference numerals are given to the portions corresponding to FIG. 15B. In this embodiment, a differential gain generator 31 is provided and the differential gain generator 31 is provided.
Is multiplied by the multiplier 32 with respect to the dequantized difference value from the adaptive dequantization unit 18, and the difference value is increased to be supplied to the addition circuit 19 and the adaptive prediction unit 21.

The code sequence switching control when a frame loss occurs is the same as the conventional method. When the loss of the frame is recovered, the received code sequence is input to the ADPCM decoding unit 14, and the gain is generated in the differential gain generation unit 31 immediately after the loss frame and is multiplied by the output of the adaptive dequantization unit 18. ADPCM decoding is performed.
The gain multiplication for the dequantized difference value immediately after the defective frame is initially large, the gain is small with the passage of time, and the multiplication gain is gradually set to 1 over several frames. For example, ITU-T standard 32 kbit / s G.I. 72
When the frame length is set to 40 samples (5 ms) at 6 ADPCM (sampling frequency 8 kHz) and all “1111” is inserted as the code sequence for the lost frame, for example, as shown in FIG. 2B, the third frame is lost and the lost frame # 3. Immediately after frame # 4, the multiplication gain with respect to the difference value is set to about 5, and the gain becomes 1 at the end of frame # 6, which is three frames after the lost frame # 3. Also in this case, the difference value is made large immediately after the lost frame, and this is brought into the normal state in a few frames, and the normal decoded signal is rapidly obtained. Embodiment 3 FIG. 3 shows an embodiment of the invention of claim 3. The decoded signal from the ADPCM decoding unit 14 is temporarily stored in the storage unit 33 and is also supplied to the gain control unit 34. The output from the storage unit 33 is a level adjusting unit 35 including an amplifier or an attenuator.
The level is adjusted with and output. The code sequence switching control when a frame loss occurs is the same as the conventional method.

When the loss of the frame is recovered, the gain control unit 34 causes the level adjusting unit 35 to correspond to the ratio of the decoded waveform power of the frame immediately before the lost frame to the decoded waveform power of the frame immediately after the lost frame. The level of the decoded signal from the storage unit 33 is controlled. Below, ITU-T standard 32
kbit / s G.I. 726 ADPCM (sampling frequency 8
An example of a concrete implementation of the present invention and a decoded waveform in ADPCM decoding when the frame length is 40 samples (5 ms) at (kHz) is shown. FIG. 3B is a waveform output from the ADPCM decoding unit 14 when no frame loss has occurred. In the waveform of the figure, at some point ADPC
When the input of the M decoding unit 14 loses one frame (5 ms) and the code sequence “1111” is input to the ADPCM decoding unit 14 at the time of the loss, the decoded waveform at that time is as shown in FIG. 4A.
It becomes as shown in. Immediately after the frame loss is recovered in this way, the AD of the corresponding portion is compared with that in FIG. 3B in the conventional method.
The level of the PCM voice waveform is quite low.

In the present invention, as shown in FIG. 4B, the decoded signal immediately after the defective frame is multiplied by the gain to increase the decoded waveform level. That is, the gain control unit 34 compares the decoded waveform power immediately before the frame loss with the decoded waveform power immediately after the frame loss recovery, and controls the gain of the level adjustment unit 35 so that the levels are about the same. Level adjustment for the decoded signal after the lost frame takes time to zero the adjustment, an example of which is shown in FIG. 4C. This is an example of the gain value to be multiplied when frame # 3 is lost. With respect to the total power P _O of the frame # 2 immediately before the lost frame, the gain is set so that the total power of the frame # 4 immediately after the lost frame is ⅔ of P _O , and the gain is set in the subsequent 19 frames. 19 is determined for each sample by triangle interpolation.
The gain is reduced to 1 over the frame. As a result of this, FIG.
Compared with the decoded waveform when the conventional method shown in A is used,
It can be seen that when the third embodiment of the present invention is used, the time until the level of the decoded waveform becomes equivalent to that in FIG. 3B is significantly shortened. Embodiment 4 FIG. 5A shows an embodiment of the invention of claim 4. The initialization control unit 37 is different from the conventional technique shown in FIG. 15A. The code sequence switching control when a frame loss occurs is the same as the conventional method. When the loss of the frame is recovered, the initialization control unit 37 sets A
The DPCM decoding unit 14 is instructed to perform initialization and is initialized to initialize the reception code sequence.
To ADPCM decoding. Embodiment 5 FIG. 5B shows an embodiment of the invention of claim 5. This is Figure 1A
5 is a combination of the first embodiment shown in FIG. 5 and the fourth embodiment shown in FIG. 5A. When the loss of the frame is recovered, the initialization control section 37 initializes the ADPCM decoding section 14 and receives it. The code sequence is input to the code replacement unit 25, one to several samples are referred to the replacement table 28, converted so that the difference value is large, and input to the ADPCM decoding unit 14. Embodiment 6 FIG. 6 shows another embodiment of the invention of claim 5. This is Figure 2
It is a combination of the second embodiment shown in A and the fourth embodiment shown in FIG. 5A, and when the loss of the frame is recovered, the ADPCM decoding unit 14 is initialized by the instruction of the initialization control unit 37, The differential gain generator 31 generates a gain, multiplies the output difference value of the adaptive inverse quantizer 18 and supplies it to the adder circuit 19 to perform ADPCM decoding. Embodiment 7 FIG. 7 shows still another embodiment of the invention of claim 5. This is a combination of Example 3 of FIG. 3A and Example 4 of FIG. 5A.

When the loss of the frame is recovered, the ADPCM decoding unit 14 is initialized by the instruction of the initialization control unit 37, and the received code sequence is input to the initialized ADPCM decoding unit 14 to perform the ADPCM decoding. Doing and AD
The output of the PCM decoding unit 14 is level-adjusted by the level adjusting unit 35 controlled by the gain control unit 34. Embodiment 8 FIG. 8 shows an embodiment of the invention of claim 6. A frame power calculation unit 39 is added to the first embodiment of FIG. 1A,
The calculated frame power information is also the replacement control unit 27.
Is input to Only when the power of the decoded waveform of the immediately preceding frame at the time of frame loss calculated by the frame power calculation unit 39 is larger than the threshold value, the code replacement is performed for several samples immediately after the lost frame, but the decoding of the immediately preceding frame at the time of frame loss is performed. When the power of the waveform is less than or equal to the threshold value, code replacement immediately after the lost frame is not performed. In other words, if the decoded waveform power is originally low, the level rises from a small level even if frame loss does not occur, and if the frame loss occurs, a sudden rise causes unnaturalness. As a threshold value for whether or not to perform sign replacement,
The value is set to a value slightly lower than the normal voice power during low voice. Embodiment 9 FIG. 9 shows another embodiment of the invention of claim 6. This is Figure 2
The frame power calculation unit 39 in FIG. 8 is added to the second embodiment shown in A, and the calculated frame power information is also input to the differential gain generation unit 31. In this case as well, when the power of the decoded waveform of the immediately preceding frame when the frame is lost is larger than the threshold value, gain multiplication is performed on the frame immediately after the lost frame by the difference value, and the power of the decoded waveform of the immediately preceding frame when the frame is lost. If is less than or equal to the threshold value, gain multiplication is not performed on the difference value. Embodiment 10 FIG. 10 shows still another embodiment of the invention of claim 6. The frame power calculation unit 39 is added to the third embodiment shown in FIG. 3A, and this calculation information is also input to the gain control unit 34. When the power of the decoded waveform of the immediately preceding frame at the time of frame loss is greater than the threshold, the gain multiplication is performed on the frame immediately after the lost frame, but the power of the decoded waveform of the immediately preceding frame at the time of frame loss is less than or equal to the threshold. In the case of, the gain multiplication is not performed. Embodiment 11 FIG. 11 shows still another embodiment of the invention of claim 6. A frame power calculation unit 39 is added to the fourth embodiment shown in FIG. 5A, and this calculation information is also input to the initialization control unit 37. When the power of the decoded waveform of the immediately preceding frame when the frame is lost is larger than the threshold value, the ADPCM decoding unit 14 is initialized at the start of decoding immediately after the lost frame. If it is less than the threshold value, the initialization is not performed. Embodiment 12 FIG. 12 shows still another embodiment of the invention of claim 6. A frame power calculation unit 39 is added to the fifth embodiment shown in FIG. 5B, and this calculation information is input to the replacement control unit 27 and the initialization control unit 37. When the power of the decoded waveform of the immediately preceding frame when the frame is lost is larger than the threshold, code replacement and initialization immediately after the lost frame are performed, but the power of the decoded waveform of the immediately preceding frame when the frame is lost is less than or equal to the threshold. Does not perform the code replacement and initialization. Embodiment 13 FIG. 13 shows still another embodiment of the invention of claim 6. A frame power calculation unit 39 is added to the sixth embodiment shown in FIG. 6, and this calculation information is input to the initialization control unit 37 and the differential gain generation unit 31.

When the power of the decoded waveform of the immediately preceding frame when the frame is lost is larger than the threshold value, initialization after the lost frame and gain multiplication of the difference value are performed. If below the threshold,
The initialization and the gain multiplication of the difference value are not performed. Embodiment 14 FIG. 14 shows still another embodiment of the invention of claim 6. The frame power calculation unit 39 is added to the seventh embodiment of FIG.
This calculation information is input to the initialization control unit 37 and the gain control unit 34.

When the power of the decoded waveform of the immediately preceding frame when the frame is lost is larger than the threshold value, initialization and gain multiplication are performed after the lost frame, but the power of the decoded waveform of the immediately preceding frame when the frame is lost is reduced. If the threshold value or less, the initialization and the gain multiplication are not performed. Each drawing of each of the above-described embodiments is shown functionally, and in practice, a microprocessor or the like is used, and there are some portions that are not configured as hardware.

[0019]

As described above, according to the present invention, even when a frame loss occurs, the frame after the frame loss rises faster than before and the waveform distortion is reduced.

[Brief description of drawings]

1A is a block diagram showing a configuration of a first embodiment of an ADPCM voice decoding method to which the invention of claim 1 is applied, and B is a diagram showing an example of a replacement table 28 in FIG. 1A.

FIG. 2A is a block diagram showing a configuration of a second embodiment of an ADPCM voice decoding method to which the invention of claim 2 is applied, and B is a diagram showing an example of a gain for multiplying a difference value immediately after a lost frame. is there.

FIG. 3A is a block diagram showing a configuration of a third embodiment of an ADPCM voice decoding method to which the invention of claim 3 is applied, and B is a diagram showing a waveform of an output of the ADPCM decoder in a state without frame loss. is there.

FIG. 4A is a diagram showing a waveform of an output of an ADPCM decoder according to a conventional method when a frame loss occurs;
FIG. 3 is a diagram showing a waveform of an output of an ADPCM decoder to which the invention of FIG. 2 is applied, and C is a diagram showing an example of a gain for multiplying a decoded waveform immediately after a lost frame.

5A is a block diagram showing a configuration of an embodiment 4 of an ADPCM speech decoding method to which the invention of claim 4 is applied, and FIG. 5B is an embodiment of an ADPCM speech decoding method to which the invention of claim 5 is applied. 5 is a block diagram showing the configuration of FIG.

FIG. 6 is a block diagram showing a configuration of a sixth embodiment of an ADPCM voice decoding method to which the invention of claim 5 is applied.

FIG. 7 is a block diagram showing a configuration of a seventh embodiment of an ADPCM voice decoding method to which the invention of claim 5 is applied.

FIG. 8 is a block diagram showing a configuration of an eighth embodiment of an ADPCM voice decoding method to which the invention of claim 6 is applied.

FIG. 9 is a block diagram showing the configuration of an embodiment 9 of the ADPCM voice decoding method to which the invention of claim 6 is applied.

FIG. 10 is a block diagram showing the configuration of embodiment 10 of the ADPCM voice decoding method to which the invention of claim 6 is applied.

FIG. 11 is a block diagram showing the configuration of an eleventh embodiment of an ADPCM voice decoding method to which the invention of claim 6 is applied.

FIG. 12 is a block diagram showing the configuration of a twelfth embodiment of an ADPCM voice decoding method to which the invention of claim 6 is applied.

FIG. 13 is a block diagram showing a configuration of an thirteenth embodiment of an ADPCM voice decoding method to which the invention of claim 6 is applied.

FIG. 14 is a block diagram showing the configuration of embodiment 14 of the ADPCM voice decoding method to which the invention of claim 6 is applied.

FIG. 15 is a block diagram showing a conventional ADPCM speech decoding method, and B is a block diagram showing a specific example of the ADPCM decoding unit.

Claims (6)

[Claims] 1. An ADPCM coded voice for decoding a frame-structured ADPCM code sequence to obtain a decoded acoustic signal and inserting a code sequence for a defective frame into the defective frame to perform decoding when a frame loss occurs. In the decoding method, a part of the code sequence of the frame next to the lost frame is
The code replacement table is used to replace the code series with a large difference value for correction, and the corrected code series is used for ADPC.
ADPCM coded speech decoding method characterized by performing M decoding. 2. An ADPCM coded speech which decodes a frame-structured ADPCM code sequence to obtain a decoded acoustic signal, and when a frame loss occurs, inserts a code sequence for a lost frame into the lost frame for decoding. In the decoding method, a part of the difference value generated at the time of decoding in the frame next to the above-mentioned lost frame is multiplied by a gain to make a correction to make the difference value large, and this corrected difference value is used. ADPCM coded speech decoding method, characterized in that the decoding is performed. 3. An ADPCM-coded speech for decoding a frame-structured ADPCM code sequence to obtain a decoded acoustic signal, and inserting a code sequence for a defective frame into the missing frame to perform decoding when a frame loss occurs. In the decoding method, the ratio of the power of the decoded waveform immediately before the lost frame to the power of the decoded waveform of the frame immediately after the lost frame is obtained, and the gain corresponding to this power ratio is set to the decoded signal immediately after the lost frame. The multiplication is performed to correct the power so that the power is almost equal to the power of the decoded waveform immediately before the lost frame.
DPCM coded speech decoding method. 4. An ADPCM-coded voice for decoding a frame-structured ADPCM code sequence to obtain a decoded acoustic signal, and inserting a code sequence for a defective frame into the lost frame to perform decoding when a frame loss occurs. In the decoding method, when the decoding of the frame next to the above-mentioned lost frame is started, the ADP
An ADPCM coded speech decoding method characterized by initializing and correcting an internal state of CM decoding. 5. The ADPC according to claim 1, wherein an internal state of ADPCM decoding is initialized when decoding of a frame next to the lost frame is started.
M-coded speech decoding method. 6. The ADPCM coded speech according to claim 1, wherein the correction is performed when the decoded waveform power immediately before the lost frame is larger than a preset threshold value. Decryption method.