JP3183306B2 - Error protection processing method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an error protection processing method for protecting a code error occurring on a transmission line, and more particularly to a low-speed voice transmission method in a mobile radio system in which a code error frequently occurs on a transmission line. A mapping process for reducing the bit error sensitivity of a quantized bit that is applied and not protected by the error correction / detection code is applied to the quantized bit, and even if a bit error of the transmission path is superimposed on the bit, the quality of the decoded speech is reduced. The present invention relates to an error protection processing method capable of suppressing deterioration.

[0002]

2. Description of the Related Art The error protection processing method of the present invention is generally applicable to a signal to which vector quantization is applied, but the following description will be made on a case where the error protection processing method is applied to a speech signal.

[0003] In a voice transmission system applied to an application field in which code errors frequently occur, such as mobile radio, the quality of transmitted voice is degraded due to code errors by using an error correction coding technique. In a mobile radio transmission line, a burst error frequently occurs, so that a transmission line code error may remain in bits protected by an error correction code, and an error detection code such as a CRC is used to detect an overlooked code error. Used with a sign. When the residual error is detected, the location of the residual error bit cannot be specified. Therefore, a method of discarding the entire residual error possible range bit and decoding the speech from other information is adopted. Therefore, if an error remains even for one bit, the information of the entire error detection target bit is rejected, and speech is decoded from other information. Therefore, it is necessary to use a strong error correction code in order to minimize the probability of occurrence of residual errors and to improve quality.

On the other hand, in the application to a mobile radio system in which the effective use of frequency is the most important issue, it is required to reduce the number of redundant bits used for the error correction code as much as possible in order to suppress the required band expansion. Therefore, conventionally, when trying to protect all the encoded information bits by error,
Since the required redundant bits increase and the required bandwidth increases,
A device for intensively protecting only some important information bits is used. This scheme is a method in which the quantized bits of the audio signal are ordered according to their importance, and about 1/3 to 1/2 of the important bits among all the bits are intensively protected. Furthermore, attention is paid to the fact that the importance of the information bits to be error-corrected is also different, and methods of protecting the more important bits more strongly have been studied. These include BS-FEC and RC.
There are methods such as PC.

FIG. 4 is a block diagram showing the configuration of such a conventional error protection processing device. In the figure, the bits representing the quantization information quantized by the quantization unit 1 are combined with a first bit group to be subjected to the error correction / detection encoding process by the grouping unit 5 and the error correction / detection encoding process. It is divided into a second bit group that is not performed. And
The bits of the first bit group are subjected to error correction / detection encoding processing by an error correction / detection encoding processing section 7 and supplied to a multiplexing section 9 where the second bit group not subjected to error correction / detection encoding processing is processed. Is multiplexed with the bits of
1 has been transmitted.

[0006]

In the conventional method, an error protection method for a bit having a higher importance in quantization bits of speech is taken up, but an error protection method for a bit having a lower importance is not much considered. Absent. However, since the importance is low, the ratio of bits for which code errors are not protected is about 1/2, and
With bits, the influence on the quality is small but the effect increases as the number increases.

[0007] The present invention has been made in view of the above,
An object of the present invention is to provide an error protection processing method that can further reduce the error sensitivity of bits that are not error-protected by an error correction / detection code and improve the quality of decoded speech.

[0008]

In order to achieve the above object, an error protection processing method according to the present invention divides a signal into frames, and converts a signal waveform of the frame or analysis information of the signal waveform into a predetermined number of bits. In a signal transmission method for quantizing, error correcting / detecting and encoding and transmitting, a first bit group consisting of bits for protecting the quantized bits by an error correcting / detecting code and a second bit consisting of unprotected bits The quantized bits of the predetermined number of bits are converted by a mapping process for making the error sensitivity of the bits of the second bit group smaller, and the converted bits are combined with the bits of the first bit group and the second bit group. , The bits of the second bit group are transmitted as they are, and the converted first bits are separated.
The gist of the present invention is that the bits of the bit group are transmitted with error correction / detection encoding.

[0009]

According to the error protection processing method of the present invention, after a signal divided into frames is quantized into a predetermined number of bits, and before being multiplexed, each of the quantized quantized bits is subjected to an error correction / detection code. And a second bit group consisting of unprotected bits and a second bit group consisting of unprotected bits. Of these, only the bits of the second bit group are converted by a mapping process that reduces error sensitivity, Thereafter, the bits of the first bit group are separated into the bits of the first bit group and the bits of the second bit group, the bits of the second bit group are transmitted as they are, and the converted bits of the first bit group are subjected to error correction / detection. Encode and transmit.

[0010]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of an error protection processing device for implementing an error protection processing method according to one embodiment of the present invention. The error protection processing device shown in FIG.
The mapping processing unit 3 is provided between the quantization unit 1 and the grouping unit 5 in the conventional error protection processing device shown in FIG.
Thus, the mapping process is performed before the quantization bits are grouped. Then, the mapped bits are divided by the grouping unit 5 into a first bit group for which error correction / detection encoding processing is performed and a second bit group for which error correction / detection encoding processing is not performed, as described above. And divided into The bits of the first bit group are subjected to error correction / detection encoding processing by an error correction / detection encoding processing unit 7 and supplied to a multiplexing unit 9. On the other hand, the bits of the second bit group are directly supplied to the multiplexing unit 9 without being subjected to the error correction / detection coding processing, and are subjected to the error correction / detection coding processing here. The data is multiplexed with bits and transmitted to the transmission path 11.

The mapping processing unit 3 is provided for reducing the error sensitivity of the second bit group, and for this purpose, an algorithm for directly reducing the error sensitivity of the bits of the second bit group. And a mapping process using an algorithm that increases the error sensitivity of the bits of the first bit group and indirectly lowers the error sensitivity of the bits of the second bit group. There are things to do.

Next, for the sake of simplicity, an embodiment of the present invention will be described under the following conditions. The signal to be quantized is a three-dimensional vector formed by collecting three consecutive samples of the signal white Gaussian signal sequence. The three-dimensional vector quantizer that achieves this has a 6-bit output, and the quantization table (64 = 2 ⁶ three-dimensional vector signals) has L as a teacher using a three-dimensional signal generated from a white Gaussian signal sequence.
Learned with the BG algorithm. In the error protection method, only the more important three bits out of the six bits output from the quantizer are subjected to error correction / detection coding, and the remaining three bits are transmitted as they are.

From the above conditions, the main part of the present embodiment is to construct a mapping table of 6 bits input and 6 bits output and apply this to the 6 bits output of the quantizer. Next, a configuration method of the mapping table will be described.

Step 1: An average distortion when each bit of the quantizer output 6 bits alone causes a bit error is determined for each bit, and the three bits with the larger average distortion are in the first group and the smaller three bits are the first group. Is a second group.

Step 2: A list of error patterns of all the residual errors after error correction, which occur in the bits of the first group, and their occurrence probabilities are prepared. A list of all the transmission line error patterns occurring in the bits of the second group and their occurrence probabilities is created. For an error pattern whose occurrence probability is sufficiently negligible, the occurrence probability is approximated to zero.

Step 3: Using the error pattern prepared in step 2 and its occurrence probability table, a document (K. Zeger an
d A. Gersho, "Psuedo-Gray Coding", IEEE trans on Com
m, vol38, pp2147-2158, Dec.1990)
A mapping table for converting a given quantization table is created by using an ary switching algorithm (BSA) so as to minimize distortion of decoded speech caused by an error in a transmission path.

Next, the application of the obtained mapping will be described. (1) Quantizer output 6 produced by LBG algorithm
A method of mapping the bits to obtain the desired 6-bit pattern by performing the mapping process obtained in step 3 and (2) combining the quantization table created by the LBG algorithm and the mapping table obtained in step 3 to form a new quantization table And a method of obtaining a desired 6-bit pattern by one-time quantization.

Next, another embodiment will be described. In this embodiment, in step 1 of the above-described embodiment, the elements (64
Is divided into eight clusters, and the number of elements included in each cluster is all eight. With this division, each element is represented by 3 bits for the cluster to which it belongs, and is represented using 3 bits to indicate one of the 8 elements in the cluster. Here, the first group is constituted by 3 bits representing a cluster, and the second group is constituted by 3 bits representing 8 elements in the cluster. Except for this, it is the same as the above-described embodiment.

Another embodiment will be described. In this embodiment, a mapping table is created in the same manner as in the first embodiment, assuming that there is no residual error after error correction occurring in the first group of bits in step 2 of the above-described embodiment. In the present embodiment, in step 3, the error sensitivity of the bits of the second group is efficiently reduced by an operation that dares to increase the error sensitivity of the bits of the first group. The operation to dare increase the bit error sensitivity is
This is realized by reversing the update decision in the BSA (in the direction of increasing the distortion).

Further, another embodiment will be described. This embodiment shows an example in which the quantization table itself is also changed on the assumption that three bits out of six bits are protected and the remaining three bits are not protected. This embodiment is characterized in that 3-bit / 3-bit two-stage vector quantization is used instead of quantizing 6 bits at once. The first three bits are subject to error protection (first group), and the second three bits (second group) are not subject to error protection. In general, when L bits are not subject to error protection and M bits are not, two-stage vector quantization of L bits / M bits is used. When L and M are about 10 bits or more, further multi-stage quantization is applied due to the limitation of the processing amount. The above is the main point of the present embodiment, and the rest is the same as the above-described embodiments.

FIG. 2 is a graph showing the result of comparing the characteristics of the error protection processing method of the present invention with that of the conventional method (LBG). It can be seen that there is an effect of suppressing the deterioration of the quality of the decoded voice as compared with

Further, the error protection processing method according to the present invention is similar to that shown in FIG.
As shown in FIG. 7, it can be seen that the error sensitivity is also flattened (the importance of the bits is made uniform).

[0024]

As described above, according to the present invention,
The bits representing the quantization information are divided into a first bit group consisting of bits protected by an error correction / detection code and a second bit group consisting of unprotected bits, and the error sensitivity of the bits in the second bit group is determined. A predetermined number of quantized bits are converted using a smaller mapping process, the converted bits of the second bit group are transmitted as they are, and the converted bits of the first bit group are converted to error bits. Since transmission is performed after correction / detection encoding, even if a bit error in a transmission path is superimposed on a bit, deterioration of the quality of decoded speech can be suppressed, and the quality of decoded speech can be improved.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of an error protection processing device that performs an error protection processing method according to an embodiment of the present invention.

FIG. 2 is a graph showing the results of comparing the characteristics of the error protection processing method of the present invention with those of a conventional method.

FIG. 3 is a graph showing the effect of flattening the error sensitivity by the error protection processing method of the present invention.

FIG. 4 is a block diagram showing a configuration of a conventional error protection processing device.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Quantization part 3 Mapping part 5 Grouping part 7 Error correction / detection coding processing part 9 Multiplexing part 11 Transmission line

──────────────────────────────────────────────────続 き Continuation of the front page (72) Inventor Kazunaga Ikeda 1-6-1, Uchisaiwaicho, Chiyoda-ku, Tokyo Nippon Telegraph and Telephone Corporation Examiner Masaaki Suzuki (56) References JP-A-2-113633 (JP) JP-A-62-117422 (JP, A) JP-A-3-175830 (JP, A) JP-A-1-177225 (JP, A) JP-A-56-25853 (JP, A) (58) Field surveyed (Int. Cl. ⁷ , DB name) H04B 14/00-14/06 H03M 13/00

Claims (2)

(57) [Claims] [Claim 1] by dividing the signal into frames, the analysis information of the signal waveform or the signal waveform of the frame is quantized to a predetermined number of bits, the signal transmission method for transmitting erroneously Ri correction / detection coding, the a second bit group of bits that are not protected with the first bit group of the bits quantized bits protected by error correcting / detecting code, a smaller error sensitivity of bits of the second bit group the mapping process converts the quantization bits of the predetermined number of bits, bits the converted bits of the first bit group
And the bits of the second bit group are transmitted as they are, and the converted bits of the first bit group are subjected to error correction / detection coding and transmitted. Error protection processing method. 2. The mapping process uses a quantization table that operates so that the bit error rate of the first bit group is lower than the bit error rate of the second bit group. 2. The error protection processing method according to claim 1, wherein