JP3147208B2 - Quantization code 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 a code error occurs, PCM coding, D
The present invention relates to a quantization code decoding method for decoding quantization codes such as PCM coding and ADPCM coding.

[0002]

2. Description of the Related Art A quantization code decoding method according to the present invention quantizes a sample value of a signal using a discretization table and decodes the information indicating the quantization, that is, quantization code transmission for transmitting a quantization code. Although it can be used generally,
In the following description, a case where the present invention is applied to wireless transmission of an audio signal waveform will be described.

[0003] In speech coding applied to an application field in which remarkable phase fluctuations and reception level fluctuations occur on a transmission path such as mobile radio and code errors frequently occur, error compensation techniques such as error correction / detection coding techniques are used. Is often devised to reduce the quality degradation of transmission voice due to code errors. However, when performing coded transmission of voice without using error correction coding as in a digital cordless telephone, the quality of transmitted voice is greatly degraded due to a code error. For this reason,
In a digital cordless telephone using ADPCM speech coding, a technique called muting that replaces a part of a received bit string of a frame in which an error is detected by using an error detection function of a transmission frame unit provided in a transmission channel, and an error-correcting technique. Attempts have been made to suppress the auditory quality degradation of the decoded speech signal by using a technique called attenuation, which suppresses the power of the decoded waveform of the frame in which is detected.

The technique called muting is described in
In the decoding of DPCM speech coding, when an error is detected in a quantized bit string representing a large difference value, the bit string is replaced with a small difference value. By using this technique, it is possible to suppress impulsive abnormal noise peculiar to ADPCM speech coding. The replacement table used in the actual system is experimentally determined so as to reduce the auditory quality deterioration.

[0005] In addition, the above-mentioned technique called "attenuation" is a technique of decoding an ADPCM speech codec in which, for a frame in which an error is detected, a frame having a large fluctuation in a decoded signal waveform is obtained by suppressing the power of the decoded signal waveform. This is a technique to make a signal waveform. By using this technique in combination with the muting technique, the perceived quality of abnormal sounds that could not be completely eliminated by the muting technique or, conversely, abnormal sounds generated by inadvertent replacement of bit strings by the muting technique are reduced. It is possible to suppress the influence of. The power suppression ratio used in an actual system has been experimentally determined so that auditory quality degradation is reduced.

[0006] In general, a decoder for speech coding in wireless transmission includes decision information (0 or 1) for each transmission bit,
Information (reliability information) indicating the magnitude of the reliability can be input. However, in speech coding that does not use error correction coding, decoding is performed using only the determination information of each transmission bit, and reliability information is not used. FIG. 6A
1 shows a configuration of a conventional quantization signal decoding method for waveform encoding. A signal received from the input terminal 11 is used as determination information of each transmission bit by the wireless demodulation unit 12. That is, the decoded value of each bit is “0” or “1”. This determination information sequence, that is, the decoded bit sequence is supplied to the inverse quantization unit 13. The inverse quantization unit 13 extracts a quantization value corresponding to the supplied decoded bit string from the quantization table 14, performs inverse quantization, decodes the output using the inverse quantization output, and outputs the decoded sample value.

FIG. 6B shows a configuration of a conventional quantized signal decoding method for waveform coding to which a muting technique and an attenuation technique are applied. The determination information sequence (decoded bit sequence) from the wireless demodulation unit 12 is supplied to the muting unit 16 and is also supplied to the frame error detection unit 17. The frame error detection unit 17 performs error detection decoding for each transmission frame, generates error detection information according to whether or not there is a bit error in the frame, and the error detection information is supplied to the muting unit 16. You. The muting unit 16 receives the error detection information and the bit string determination information, performs the above-described muting process, and supplies the muted decoded bit sequence to the inverse quantization unit 13. In the inverse quantization unit 13, a quantization value corresponding to the supplied decoded bit string is extracted from the quantization table 14 and supplied to the attenuation unit 18. The attenuation unit 18 performs the attenuation process according to the error detection information, and outputs the result as a decoded sample value.

[0008]

In the conventional method, a decoding method using only the decision information (decoded bit sequence) is used for decoding the waveform coding transmission of the voice. There is a disadvantage that waveform distortion occurs. Also, even when a transmission channel error detection function is used as in a digital cordless telephone, even if abnormal noise suppression techniques such as muting and attenuation are used, the waveform distortion of a decoded voice caused by a transmission error may occur. No attempt has been made to reduce this.

Further, when the above-mentioned allophone suppression technology is used,
Code replacement and waveform compression are also performed on the waveform portion where no code error occurs within the range (frame) where the error is detected, and the actual waveform distortion will increase for this portion, Deterioration of hearing quality may occur. When the number of transmission errors is small, a certain amount of hearing quality can be obtained by the conventional method, but when the number of transmission errors is large, the perceived quality of the decoded voice signal of the conventional method deteriorates greatly and the waveform distortion of the decoded signal is large.

[0010]

According to the first aspect of the present invention, the reliability information of the bit is obtained for the decoded bit sequence, and the dequantized output of the corresponding bit is multiplied by the gain corresponding to the reliability information. , The multiplication result is used as an inverse quantization output to obtain a decoded output. According to the second aspect of the present invention, as the quantization decoding table (inverse quantization table), for each code, one storing the weight information of the configuration name bit is used, and the bit reliability of each bit of the decoded bit string is used. Gain information corresponding to each bit reliability information, a gain corresponding to the bit and the corresponding gain are multiplied by each weight information read from the quantization decoding table by a code, and the multiplication result is added. , And a decoded output is obtained by using this addition output as a conventional inverse quantization output.

[0011]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The quantization code decoding method of the present invention is applicable to all waveform coding such as PCM, DPCM, ADPCM, which quantizes a sample value and transmits it, and generates a decoded sample value from the received signal. However, the following embodiment describes a case where a PCM coded signal is decoded.

FIG. 1A shows an embodiment of the present invention, and FIG.
The same reference numerals are given to portions corresponding to. In the present invention, the determination information (decoded bit) output from the wireless demodulation unit 12
, Reliability information indicating the likelihood of the determination is also detected and output. The reliability information is, for example, the reception level of the bit. This reliability information is supplied to the gain generation unit 21, and a gain corresponding to the reliability information is obtained. This gain is generated, for example, as follows.

That is, the threshold value A is set in advance, and the gain a is set to 1 when the inputted reliability information x is larger than A, and is set to 0 when the inputted reliability information x is smaller than A. Alternatively, a threshold value A and a smaller value B are set in advance, and a = 1 when the reliability information x is larger than A, a = 0 when smaller than B, and a = (x −A) / (AB).

The thresholds A and B are determined, for example, as follows. That is, when a characteristic curve of bit reliability information and an error rate in a system to be actually applied is actually measured and obtained, and this characteristic curve becomes, for example, as shown in FIG. Is used, the error rate is 20
% Reliability information value L ₂ become a threshold A, 2 pieces of threshold A, the case of using B, and reliability information value L ₃ the error rate of 2% is A, the error rate reliability information value L ₁ comprising 48% to B.

The gain generated in this way is obtained by multiplying the dequantized output read from the quantization decoding table 14 by the multiplier 2
The product is multiplied by 2, and in this case, since the signal is PCM, the multiplied output is output as a decoded (sample value) output. In this case, the same quantization decoding table 14 as the conventional one is used. When the determination information is "1", E is read from the quantization table 14, and when the determination information is "0", the quantization is performed. -E is read from the decoding table 14. In the present invention, the read E or -E is multiplied by the bit reliability information and the corresponding gain a, and this is output as a decoded output. Note that 1-bit PCM can be applied to a scalar quantization unit in quantization combining, for example, vector quantization and scalar quantization.

FIG. 2A shows an embodiment of the second aspect of the present invention.
Parts corresponding to those in FIG. 1A are denoted by the same reference numerals. In this embodiment, each sample is quantized by 2 bits. The decision information (bit string) from the radio demodulation unit 12 is output to a code extracting unit 23, where a plurality of bits corresponding to each sample, in this example, 2 bits are sequentially transmitted. The quantization decoding table 24 is read out with these two bits.

The quantization decoding table 24 is different from the conventional quantization decoding table, as shown in FIG. 2B, in which weight information represented by each bit constituting a code is stored for each code. In the example, each code is composed of 2 bits, the sample value is quantized to any of 3E, E, -E, and -3E, and the quantized value is represented by a 2-bit code. For (0 1), weight information 2E is stored for the upper bit “0”, and for lower bit “1”, weight information E is stored. For code (0 0), 2E is stored for the upper bit “0”. But,
-E is stored for the lower bit "0", and the code (1
For (0), -2E is stored for the upper bit "1" and E is stored for the lower bit "0". For code (11), -2E is stored for the upper bit "1", and the lower bit "1" is stored. -E is stored.

For reference, the decoded quantization value of each code is also shown. For the input code, higher order weight information corresponding to the higher order bit and lower order weight information corresponding to the lower order bit are read out from the quantization decoding table 24 and supplied to the multipliers 22 ₁ and 22 ₂ . On the other hand, the bit reliability information corresponding to each bit string is supplied to the gain generation unit 21, and the corresponding gain is generated as in the case of FIG. 1A. This gain sequence is divided by the distribution unit 25 into two gains a corresponding to the upper and lower bits of the code of the code extracting unit 23, respectively.
₁ and a ₂ are taken out and supplied to the multipliers 22 ₁ and 22 ₂ to be multiplied by the read weight information. That is, a is set to the read output 2E or -2E of the upper bit.
_1, a ₂ are respectively multiplied to the read output E or -E the lower bits. The multiplication results of these multipliers 22 ₁ and 22 ₂ are added by the adder 26, and the addition result is output as a decoded sample value because it is PCM in this case.

In the case of 3-bit PCM, the configuration is the same.
FIG. 3A shows an example of the quantization decoding table 24 in this case. That is, each sample is represented by 7E, 5E, 3E, E, -E,-
When quantized to any of the eight values of 3E, -5E, and -7E, each quantized value is represented by a 3-bit code, and weight information is assigned to the most significant bits "0" and "1". Is 4
E, -4E, the weight information is 2E or -2E for the next upper bits "0" and "1", and the weight information is E or 2E for the lower bits "0" and "1". -E.
Gain a ₁ , a for 3 bits constituting each code
₂ and a ₃ are read by the three bits, respectively.
The corresponding one of the two pieces of weight information is multiplied, and the results of these multiplications are added to obtain a decoded sample output.

FIG. 4A shows a configuration in which the present invention is applied to a quantization signal decoding method of DPCM coding. As in the case of FIG. 1A, the decoded output of the determination information is multiplied by the gain from the gain generation unit 21 by the multiplier 22 to decode the difference value, and the decoded difference value and the immediately preceding decoded sample from the delay unit 31 are decoded. The values are added by an adder 32, output as a decoded sample value, and supplied to a delay unit 31.

FIG. 5B shows an example in which the present invention is applied to a quantization signal decoding method of ADPCM encoding. An inverse quantization value is obtained from the multiplier 22 in the same manner as in FIG. 4A. On the other hand, the applied gain calculator 33 accumulates a plurality of decoded sample values decoded before that, and adaptively calculates a gain value from the decoded sample values. Is multiplied by the inverse quantized value of .times. A result obtained by adding the decoded difference value and the immediately preceding decoded sample value from the delay unit 31 becomes a decoded sample value.

SN between the decoding method according to the present invention and the conventional method
3B, 5A and 5B show R-bit error rate characteristics comparisons. From the reliability information-error rate characteristics in the mobile radio transmission path, threshold values A and B are set so that the decoded waveform distortion is minimized.
FIG. 3B is a result of computer simulation in the case where is selected by trial and error, FIG. 3B shows 1-bit PCM encoding, and FIG.
Fig. 5B shows the case of 2-bit PCM coding, and Fig. 5B shows the case of 3-bit PCM coding. In any case, the method of the present invention improves the SN ratio with respect to the same error rate under a poor code error condition in the mobile radio transmission line, and can reduce the decoded waveform distortion accordingly.

[0023]

As described above, according to the present invention, the inverse quantization value of each bit is corrected in accordance with the bit reliability information of each decoded bit, so that the decoded signal is less susceptible to bit errors. Waveform distortion can be reduced.

[Brief description of the drawings]

FIG. 1A is a block diagram showing an example of a decoder to which the embodiment of the first aspect of the present invention is applied, and FIG. 1B is a diagram showing an example of a relationship between a bit error rate and bit reliability information of a transmission path.

FIG. 2A is a block diagram showing an example of a decoder to which the embodiment of claim 2 is applied, and FIG. 2B is a diagram showing an example of storage contents of a quantization decoding table.

FIG. 3A is a diagram showing an example of stored contents of a quantization decoding table used in the present invention method for a 3-bit quantized PCM code signal, and FIG. 3B is a comparison of characteristics between the present invention method and a conventional method in 1-bit quantized PCM. FIG.

FIG. 4A is a block diagram showing an example of a decoder in which the present invention is applied to a DPCM decoding method, and FIG.
It is a block diagram which shows the example of the decoder applied to the decoding method.

5A and 5B are 2-bit and 3-bit quantized P, respectively.
It is a figure which shows the characteristic comparison of this invention method and conventional method in CM.

FIG. 6A is a block diagram illustrating a conventional quantization code decoder, and FIG. 6B is a block diagram illustrating a conventional quantization code decoder to which an abnormal noise suppression process using a transmission channel error detection function is applied.

Claims (2)

(57) [Claims] An input signal is bit-judged to form a bit sequence, each bit of the bit sequence is dequantized with reference to a quantization decoding table, and a decoded output is obtained using the dequantized output. In the method, for each bit of the bit string, reliability information of the bit is obtained, a gain corresponding to the reliability information is multiplied by the dequantized output of the corresponding bit, and a result of the multiplication is output to the dequantized output. A decoding output method, wherein a decoded output is obtained as 2. A bit string is formed by bit-determining an input signal, a dequantized output is obtained by referring to a quantization decoding table using a code composed of a plurality of bits in the bit string, and the dequantized output is used by using the dequantized output. In the quantization code decoding method for obtaining a decoded output, the quantization decoding table stores, for each code, weight information of each bit constituting the code, and uses reliability information of the bit for each bit of the bit sequence. The gain corresponding to each bit reliability information is obtained, each weight information of the reference result of the quantization decoding table by the code is multiplied by the gain corresponding to the bit, and the multiplication result is added. And a decoding output obtained by using the added output as the inverse quantization output.