JPS62219833A - Adaptive decoding type code transmission system - Google Patents

Abstract

PURPOSE:To attain the effective decoding of a cycle code by giving a marker to a lost bit detected by the decoding of a transmission line code and utilizing the marker so as to apply burst correction if burst correction is enabled for an error caused in the cyclic code and to apply random correction if disabled. CONSTITUTION:Supposing the probability sending a mark or a Space and receiving the Space or the Mark respectively is P and the probability sending the Mark or Space and receiving a bit with marker is P', the probability P is far smaller than the probability P' and the noise fed in the transmission line is due to rule destruction of the transmission line code only and it is detected as the bit with marker 'M' at the reception side. The input signal is decoded into '0', '1' or 'M' by a decoder 2 of the transmission line code, the result is inputted to a decoding circuit 4 and the cyclic block code is decided whether burst enable, random correction enable or decoding disable eased on the pattern of 'M' generated in one frame. Thus, a switch 7 is controlled and the signal is inputted to a decoder. The result of decoding is outputted to an output 12 by a burst correction decoding circuit 9 and a random correction decoding circuit 11 and if decoding is disabled, the signal is outputted to a signal throw-away output 13.

Description

Detailed Description of the Invention (Industrial Application Field) The present invention relates to code/decoding when transmitting block-coded digital signals on a transmission path that undergoes high-speed fading, such as digital data transmission in a mobile radio system. It is related to the method.

(Prior art) Conventionally, when transmitting digital data over a transmission line that is subject to high-speed fading, the transmitting side performs error correction block encoding, and the receiving side corrects errors added on the transmission line.
However, the method relied solely on the algebraic properties of the code. In other words, due to fading, the S/N ratio of the transmission path differs for each bit that makes up one frame when that bit is being received, and even though errors sometimes occur in bursts, the reception In the decoding of the error correction block code on the side, the decoding operation was performed assuming that each bit had the same reliability. For this reason, deep drops in the reception level due to fading occur at multiple locations within one frame, and this causes an error pattern that results in erroneous reception in algebraic decoding (limited distance decoding method), which corresponds to the position of the high reception level. However, even if a bit is determined to be incorrect, it cannot be detected.

On the other hand, in digital data transmission systems, in order to adapt the encoded data to the physical properties of the transmission path, second encoding is often performed to control the spectrum of the code. For example, in a wireless transmission path having low-frequency cutoff characteristics, a data sequence may be encoded into a code that does not have a DC component, such as a split-phase code, and sent to the transmission path.

In this case, the receiving side considers the transmission line code to be an erasure detection code, resulting in an error in the rules of the transmission line code (referred to as bit erasure).

In the previous example, it is possible to detect a split phase (split phase corruption) and place a marker on the corresponding bit in the frame, and use the position of this marker to decode the block code corresponding to the first encoding. . Conventionally, this marker has been used to detect whether the number of markers in one frame is greater than or equal to a predetermined value, and when the number is greater than or equal to the predetermined value, the received frame is rejected. Was.

(Problem to be solved by the invention) In this method, marking and decoding of the block code are performed independently, so if the block code is decoded using the position of the marker, the error correction ability can be expanded. Despite this, there was a drawback that this was not utilized. That is, conventionally, when a bit error occurs during transmission, a marker indicating that the bit is lost is attached, and the transmission line code is decoded to 0 or 1 even if it is detected that an error has occurred in that bit.

In other words, the above marker was not used when decoding . In particular, the block code corresponding to the first encoding is a cyclic code (most commonly used block codes such as Hamming codes and BCH codes are cyclic codes).
By using , in addition to random error correction using its algebraic properties, it is also possible to decode it as a burst error correction code by using the marker position. Conventional decoding methods do not take advantage of this property, so they have the drawback of not being able to improve transmission reliability when a bit error occurs during transmission.

The present invention regards the transmission line code corresponding to the second encoding as an erasure detection code, and uses the marker obtained by decoding the transmission line code for decoding the block code, thereby generating the cyclic code corresponding to the first encoding. Determine whether the error that occurred in the error can be burst corrected or not, perform burst correction if it is possible, and perform random correction if it is not possible, enabling more effective decoding of the cyclic code. The purpose is to

(Means for Solving the Problem) The present invention attaches a marker to the lost bit detected by decoding a transmission line code, and uses this marker to convert the cyclic code into a cyclic block code when decoding a cyclic block code corresponding to the first code. The main feature is that it is determined whether the error that has occurred can be burst corrected or not, and if it is possible, burst correction is performed, and if it is not possible, random correction is performed. conventionally,
Decoding of the first code and decoding of the second code are performed independently, and as a result, even if the reliabilities of multiple bits within one frame differ in decoding of the first code, this will not affect the decoding result. This differs from conventional techniques because an algebraic decoding method was used that did not reflect the results.

(Example) An example of the present invention is shown in FIG. In Fig. 1, 1 is a signal input, for example, a digital signal received by a receiver that is bit-shaped, and 2 is a transmission line code decoder that converts the signal input from input terminal 1 into 0.1 bits or bits with markers. (The bit in which the transmission line code error was detected) is identified and reproduced. 3 is parallel data indicating a pattern of marker bits, 4 is a burst length that is less than or equal to a length that can be corrected by burst correction when decoding a cyclic block code from the pattern of marker bits indicated in data 3; a determination circuit that determines whether it is more than that or whether it cannot be decoded by either burst correction or random correction;
controls the switch 7 depending on whether burst correction is possible, random correction is possible, or correction is not possible using judgment data.

6 is the decoded output of the transmission path code 2, which is a data sequence of 0, 1, and marker bits. 7 is switch, 8 is 5
9 is a cyclic block code burst correction decoding circuit; 1 is a signal input terminal when the judgment data is burst correctable;
0 is a signal input terminal when the judgment data of 5 can be randomly corrected, 11 is a random correction circuit for a cyclic block code, 12 is a decoding output, and 13 is a signal rejection output.

Next, the operation of this embodiment will be explained. For the purpose of explanation, a split phase code will be taken as an example of a transmission line code, and a (7,4) Hamming code will be taken as an example of a cyclic block code. The (7,4) Hamming code is a code constructed for the purpose of correcting random errors, and since its inter-code distance is 3, 1-bit correction is possible. It is also possible to perform 3-bit burst error correction on this code, in which case it is necessary not only to be 4-bit error free but also to be able to estimate the burst position.

Now, the split phase code is bit 1 of the original data (
This is a code that converts Mark) to "01" and bit O (Space) to "10'". Since the split-phase encoded data sequence does not have a DC component, it is often used when transmitting digital signals through a DC cut-off transmission line. On the other hand, there is a drawback that the electric scum vector expands to a component of 2 fb relative to the signal transmission speed fb. Now, assuming that the split phase encoded data is a digital signal sequence with a signal transmission rate of 2 fb, and the bit error rate at this time is P, then Mark = 1'01 is transmitted and '
The probability that 110 It =S pace is received is Pb2 assuming that the occurrences of errors are independent. Furthermore, in normal receivers, noise is removed using a baseband filter after detection, so the characteristics of the filter can be determined so that the autocorrelation characteristics of the noise in the filter output have the same polarity within one bit of the split phase code. . In this case, when transmitting Mark = "01", if noise occurs in the first half bits so that O is mistaken for 1,
In the latter half of the bits, noise that causes 1 to be mistaken for 0 is unlikely to occur. That is, the probability of transmitting the aforementioned Mark="01" and receiving ti 10 u is even lower than P-.Therefore, when transmitting Mark="01" and an error is added on the transmission path.

The received pattern is It OOj due to half bit error
In most cases, it will be either l or "11". That is, if a bit error occurs in the transmission path, the coding rule of the sublit phase is almost destroyed. Here, the split phase destruction rule means that for one bit of the original data, the same bits are consecutive like "II Q Q jj, it l l j". In this case, "
Even if Mark=: “00” is received, Mark=: “01” is sent and the latter half bits are incorrect, because Space=“
The receiving side cannot determine whether ``10'' was sent and the first half bits were incorrect.In other words, even if the transmission line code is decoded as it is, accurate decoding cannot be expected. The same applies to the transmission of ``.

Now, if the receiving side receives 00" or 11", the encoding rules of the split phase code are broken for this bit, so this bit is set as a bit with a marker, and when the block code is decoded later, the marker is used. can be used.

In other words, in the present invention, when the split phase encoding rule is broken, the code is not forcibly decoded and is retained as a bit erasure code. Therefore, when the transmission line code is decoded, ′
In addition to 0'' tt 1 pt, there is a bit erasure code called RM u.To summarize the above, the transmission path model can be expressed as shown in FIG.

In Figure 2, P sends Mark and Space
The probability of receiving M a
The probability of receiving r k is Mark, or s p
is the probability of transmitting an ace and receiving a marked bit. When using a split phase code as a transmission line code, P and P' are respectively.

P≦P− P' = 2Pb (1-Pb) It is.

Next, a method for decoding a cyclic block code using markers will be described. As mentioned above, the probability P is considered to be extremely small compared to P', so if this probability is ignored, the only noise added in the transmission path is the breaking of the rules of the transmission path code, and this error is detected by the marker bit on the receiving side. “Can be detected as Mu.

Now, as an example, the transmitting side transmits X□= (1000101), which is one of the code words of the (7,4) Hamming code,
Consider the case where Y = (IOM M Mo1) is received on the receiving side. Here, M represents a bit with a marker.

This means that a 3-bit error has occurred on the transmission path. As mentioned above, the (7,4) Hamming code is capable of correcting 3-bit burst errors. The method is shown below. Y
If we perform a 2-bit shift σ2(Y) on the is also a codeword. The first four bits of this code are the information part, and the next three bits are the check bits. Therefore, the first 4 bits of σ2(Y) created from the received word are σ! If it is regarded as the information part for (x-work), the remaining 3 bits with markers can be obtained in the same way as encoding. In other words, (7
．． 4) The check pit of the Hamming code is 001 (
In reality, it is found by generating a generator matrix), so by replacing the marker bits with this pattern, σ
2(Y)=0110001. By performing a 2-bit shift in the opposite direction to σ2(Y), the result of burst correction for Y can be obtained. Namely.

σ−2(σ”(Y)) = 1000101, which is consistent with the transmitted word Find the burst length of the data decoded from the transmission line code, detect whether it is longer or shorter than the check pit length of the cyclic code, and if it is short, the information part can be considered to be correct, so check it using the same method as cyclic encoding. If a pit is generated, it can be considered as burst-corrected data.As is clear from this example, in order to perform burst error correction on a cyclic block code, it is necessary to (i) identify the position of the burst error; ii) It is necessary that the burst length does not exceed the check pit length.In the present invention, this condition is made possible by determining the length between the marker bits detected during decoding of the transmission path code.

Next, a case where this condition is not satisfied will be described. If there are 4 or more bits with markers, or 3
Even if the length of a burst including bits with a marker is 4 bits or more, it cannot be decoded by any method. That is, Y=(dMMMddM): M is 4
Since bits and the like are not decoded, the signal is rejected. Here, d
represents O or 1. On the other hand, if the number of marker bits is 2 bits and the burst length is 4 bits or more, it can be corrected as a random error using the following method. As an example of such a case, consider a case where xL=(1000101) is transmitted and y=(IMGOMOI) is received. A bit with a marker is either O or 1, so it is 1
Substituting all possible patterns into M, the transmission layer has Y, = (1000001), Y2 = (100010
1).

y,=(ttooooog, Y,=(1100101)
b. The number of error bits for these transmission layers xL is as follows: Yl: 1 bit error, Y2: 20 bit errors.

Y3 = 2 bit error, Y4 = 1 bit error, so Y□
,Y,,Y, is correctly decoded by random error correction using the algebraic decoding method, and Y, is received incorrectly (as mentioned above, the (7,4) Hamming code can correct 1-bit error), so , if the results of algebraic decoding for these are judged by majority vote, the correct code word will be decoded with a ratio of 3:1. (1oootol)
Therefore, the number of bits with a marker is 2 bits due to 0 or more equal to the transmitted word X1.

It has been shown that even when the burst length is 4 bits or more, correct decoding can be achieved by random error correction.

Next, the embodiment of the present invention shown in FIG. 1 and the above decoding operation will be explained in association with each other. The input signal is decoded by the transmission line code decoder 2 into 0.1 or marker bits. 1
The transmission line code decoding result of the frame is input to the determination circuit 4, which determines whether the cyclic block code can be burst corrected, can be randomly corrected, or cannot be decoded based on the pattern of marker bits generated within one frame. do.

For example, in a cyclic block code consisting of bits for information and bits for check n-, when the inter-code distance is 2d+1, n
Burst correction is possible up to a burst of marker bits of - bits, and random correction is possible up to 2d marker bits. In this case, the burst length of the marked bits is the shorter of the lengths between the marked bits including the non-marked bits, and Min represents the minimum when the received word is next. The output of the decision circuit 4 is a decision result as to whether to perform burst correction, random correction on the received word, or to reject the signal since it cannot be decoded by either method.The output of the decision circuit 4 is the decision result whether to perform burst correction, random correction, or reject the signal because it cannot be decoded by either method. Input the decoding result of the transmission line code. The burst correction decoding circuit 9 and the random correction decoding circuit 11 perform decoding for each case using the method described above, and output the decoding results to the decoding output 12. Furthermore, if the signal cannot be decoded, the signal is output to the signal rejection output 13.

In the above description, a Hamming code is used as an example of the first code, and a split phase code is used as an example of the second code. This can be easily expanded when applying a bit erasure detection code such as a bipolar code.

(Effect of the invention) As described above, in the decoding method according to the present invention, the second
According to the pattern within one frame of marker bits obtained when decoding the transmission line code corresponding to the encoding of
The decoding method for the cyclic block code, which corresponds to the encoding of , is switched between burst correction and random correction. This makes it possible to perform decoding appropriate to the error pattern occurring on the transmission path, so the present invention has the advantage of expanding error correction capability. In general, there are bits for information, check bits n-
In a cyclic block code with an inter-code distance of 2d+1 bits, the burst length n that allows burst correction of marker bits is n.
-, the number of bits with markers that can be randomly corrected is 2d, and n- is ≧2d (for example, in the Golay code, n = 23.
k = 12°2d+1 = 7, and n - k = 11.
2d=6), the improvement effect of the decoding method of the present invention is significant when digital data is transmitted over a transmission line in which marker bits are generated in bursts.

[Brief explanation of drawings]

FIG. 1 is a diagram showing an embodiment of the present invention, and FIG. 2 is a diagram showing a model of a transmission path. 1 is a signal input, 2 is a transmission line code decoder, and 3 is a mark h-
4 is a judgment circuit, 5 is judgment data, 6 is a decoded output of a transmission line code, 7 is parallel data showing a pattern of attached bits,
is a switch, 8 is a burst correction signal input terminal, 9 is a burst correction circuit, 10 is a random correction signal input terminal,
11 is a random correction circuit, 12 is a decoding output, 13 is a signal rejection output, P is the probability of transmitting a mark and receiving a space (probability of transmitting a space and receiving a mark), P'
is the probability of transmitting a mark and receiving a marked bit (the probability of transmitting a space and receiving a marked bit).

Claims (1)

[Claims] On the transmitting side, the input signal is encoded using a first error-correcting cyclic block code, and is encoded into a second code that allows positioning within the frame of bit loss added on the transmission path, and then transmitted. On the receiving side, the bit loss on the transmission path is detected by decoding the second code, and the position of the detected bit loss within the frame of the first code is specified. Determine whether the burst length including the erasure bits of the code is less than or equal to the length that can correctly decode the error correction cyclic block code as a burst error correction code, and if it is less than the length that can be burst corrected, the received frame is An adaptive decoding type code transmission system characterized by decoding the received frame as a burst error correction code, and decoding the received frame as a random error correction code when the length exceeds the length that can be burst corrected.