JP3259339B2 - Viterbi decoding method and device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a Viterbi decoding method as a maximum likelihood decoding for decoding a convolutional code and an apparatus therefor.

[0002]

2. Description of the Related Art Maximum likelihood decoding is to determine the event posterior probability Pui of all signal sequences ui that may be transmitted with respect to a received signal sequence, and decode the signal sequence having the highest probability among them. This is a decoding method. The posterior probability Pui when assuming an unstored communication channel can be obtained by the following equation by averaging the posterior occurrence probability pjui at time j since the posterior occurrence probability can be equally divided into time. . n Pui = 1 / n · Σpj ui (1) j = 0 Actually, the event posterior probability Pui cannot be obtained, and thus the judgment is made based on the quantity (metric) Mui instead of the probability. The metric for the signal sequence is called the branch metric mj ui and is calculated from them.
The determination is made based on the magnitude of the path metric . The path metric is time-divided, and is obtained by the sum of branch metrics as substitutes for the posterior occurrence probabilities of signals that can be transmitted for each signal at each time (step). Pu → Mui pj ui → mj uin Mui = 1 / n · Σmj ui (2) j = 0

[0003] Viterbi decoding is an algorithm for efficiently performing the maximum likelihood decoding using a convolutional code by selecting a path that is the minimum distance from a received sequence from two paths that merge. a high correction capability for random errors occurring in the communication path, when combined with the so-called soft decision demodulation method, it is possible to obtain particularly large coding gain.

FIG. 3 is a diagram showing a configuration example of an encoder for generating a convolutional code. This encoder has a coding rate R = 1 /
2, which generates a convolutional code having a constraint length K = 3, and includes, for example, a shift register including registers 1A and 1B, and modulo-2 adders 2A, 2B and 2C.

The state (b ₁ b ₂ ) of the shift register in such an encoder includes a state (00) and a state (0
There can be four states of 1), state (10) and state (11), and when an input is given, there are always two states that can be transitioned. That is, in the state (00), when the input is “0”, the state transits to the state (00), and the input is “1”.
In the case of, the state transits to the state (01). In the case of the state (01), when the input is “0”, the state transits to the state (10), and when the input is “1”, the state transits to the state (11). In the case of the state (10), when the input is “0”, the state (00)
To the state (01) when the input is “1”. In the case of state (11), the state transits to state (10) when the input is “0”, and changes to state (1) when the input is “1”.
Transition to 1).

FIG. 4 is a trellis diagram showing such a state transition. In the figure, a branch indicated by a solid line represents a transition caused by an input “0”, and a branch indicated by a broken line represents a transition caused by an input “1”. Represents. The number attached along the branch indicates the code (G ₁ G ₂ ) output when the transition of the branch occurs. As can be seen from FIG. 4, two paths always merge in each state.
The Viterbi decoding algorithm selects the maximum likelihood path from the two paths in each state, selects a path that survives to a predetermined length, and selects the maximum likelihood path from the paths selected in each state. This is for decoding the received code.

FIG. 5 is a block diagram showing a configuration example of a conventional Viterbi decoding device configured based on the above algorithm. As shown in FIG. 5, the Viterbi decoding device includes a branch metric calculation circuit 11, an addition / comparison / selection circuit 12, a path memory circuit 13, and a maximum likelihood determination circuit 14.

In the Viterbi decoder having such a configuration, the branch metric calculation circuit 11 calculates a branch metric from the input received code RS. The calculation result BM is input to the addition / comparison / selection circuit 12 and stored in the path memory circuit 13. In the addition / comparison / selection circuit 12, the branch metric is added to the path metric one time before, the respective path metrics are compared, the maximum likelihood path is selected, and the result is sent to the maximum likelihood determination circuit 14. Is output. In the maximum likelihood determination circuit 14, the original data is decoded for the maximum likelihood path, and the decoded output DC
_OUT is obtained.

[0009]

However, in the conventional Viterbi decoding device, since the branch metmec is not corrected by the reliability of the communication path due to the time variation of the communication path characteristic, there is no significant meaning such as fading or noise. Unreliable parts are also subject to determination, and unreliable branch metrics are handled in the same way as when they are reliable. Therefore, when the path metric is calculated, there is a problem that the influence of the unreliable branch metric on the maximum likelihood determination is large, and the error rate increases.

The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a Viterbi decoding method and a device capable of obtaining a highly reliable decoded output with a small error rate.

[0011]

In order to achieve the above object, a Viterbi decoding method according to the present invention employs a method for decoding a received code from an input received code.
Calculate the launch metric and calculate the branch metric
Detect multiple minimum branch metrics at different times,
Detect the fluctuation component of the detected minimum branch metric,
The branch metric calculated according to the fluctuation component
Weighted and corrected, and based on the weighted correction result,
Select a likelihood path and perform decoding on the selected maximum likelihood path.
U.

In the Viterbi decoding method of the present invention, the minimum block
Detection of the fluctuation component of the launch metric is performed at different times.
Calculate the average value of the above minimum branch metrics.
Calculate the variance of the minimum branch metric for the average
Perform by putting out.

In the Viterbi decoding apparatus according to the present invention, the input reception code
Branch metric to calculate the branch metric from the
Calculation means and a plurality of minimum branch metrics at different times.
Minimum branch metric detection means for detecting
Component that detects the component of the minimum branch metric
Minute detecting means, and the flow rate calculated above according to the fluctuation component.
Weighting means for weighting and correcting the launch metric
And the maximum likelihood path based on the weighted correction result.
Decoding means for decoding the selected maximum likelihood path.
You.

In the Viterbi decoding device according to the present invention, the variation component
The minute detecting means includes a plurality of the minimum branch
Means for calculating an average value of tricks;
The minimum branch metrics for the average of the small branch metrics
Means for calculating a variance value of a trick.
And said weighting means, intends row weighting for branch metrics in accordance with the variance value.

[0015]

According to the Viterbi decoding method of the present invention, a fluctuation component of a branch metric of a received code is obtained, and data is decoded according to the fluctuation component.

According to the Viterbi decoder of the present invention, the branch metric is calculated from the input received code by the branch metric calculation circuit, and the result is output to the branch metric correction circuit. In the branch metric correction circuit, a fluctuation component of the branch metric is extracted, and weight correction is performed in accordance with the fluctuation. Then, the maximum likelihood path is selected based on the weighting result of the branch metric correction circuit, and data decoding is performed on the selected maximum likelihood path.

According to the Viterbi decoding apparatus of the present invention, in the branch metric correction circuit, the minimum value of the branch metric calculated by the branch metric calculation circuit is detected by the minimum branch metric detection means, and the result is used as the fluctuation component extraction means. Is output to In the fluctuation component extracting means, the fluctuation component of the received code is extracted from the detected minimum branch metric and output to the weighting means. The weighting means weights the branch metric according to the extracted fluctuation component.

According to the Viterbi decoding apparatus of the present invention, in the fluctuation component extracting means, the average value of the minimum branch metric is calculated by the average value calculating means, and further, the variance value calculating means calculates the average value and the minimum value of the minimum branch metric. A variance value is calculated from the branch metric. The weighting means weights the branch metric according to the calculated variance value.

[0019]

FIG. 1 is a block diagram showing an embodiment of a Viterbi decoding apparatus according to the present invention. The same components as those in FIG. That is, 11 is a branch metric calculation circuit that calculates a branch metric from a received code, and 12 is an ACS (Adder Comparator Se
lector) operation, that is, an addition / comparison / selection circuit for performing addition of the previous path metric and the branch metric, comparing the metrics, and selecting the maximum likelihood path metric.
Reference numeral 4 denotes a maximum likelihood determination circuit that decodes original data using the maximum likelihood path, and reference numeral 15 denotes a branch metric correction circuit that corrects the obtained branch metric based on the reliability of the transmission path.

The branch metric correction circuit 15 weights the branch metric output from the branch metric calculation circuit 11 in accordance with the reliability characteristics of the communication channel, when the reliability is low, the weight is low, and when the reliability is high, the weight is large. And control the ratio of the branch metric at the time when it affects the path metric.

The weight correction of the branch metric correction circuit 15 will be described in more detail. In a communication channel without memory, that is, a generalized communication channel,
It is necessary to set a random variable such that the sum of the product qj · mj ui of the branch metric and the weighting factor for all the signals becomes 1 by the weighting factor qj at the time j, and this can be calculated by the following equation. As described above, the path metric can be a variable closer to the posterior occurrence probability by the weight coefficient qj at the time j.

Here, it is considered that the weight coefficient qj is approximately obtained. Since the weighting factor qj is determined by the time characteristic of the reliability Tj of the communication channel, the reliability of the communication channel is detected by the variance of the minimum branch metric, and this is determined by the weighting factor qj
Instead of Specifically, the following procedure is performed. 1) Find the minimum branch metric over several hours. 2) Find the average of those branch metrics. 3) Find the variance. Accordingly, the calculation of the path metric is based on the reliability T due to the branch metric at time j and the variance near the time at time j.
This can be performed based on the following equation using j.

The probability distribution in an actual communication channel is not constant but is constantly changing. Therefore, ideally, it is sufficient to obtain the probability that the signal is considered to have been transmitted from the received signal received at that time according to the probability distribution at each time. However, it is not actually possible to know a correct probability distribution that changes with time. Therefore, generally, a posterior probability at that time is obtained by fixing a substitute for a probability or a probability distribution.
But that may not be the case. Therefore, here, the variance in the vicinity of time at that time is used as the reliability characteristic of the communication channel using the memory property of the communication channel, and is used as the correction value of the probability distribution at that time. Then, the branch metric is corrected by this value to approximate the posterior occurrence probability. As a result, decoding closer to the maximum likelihood becomes possible, and a reduction in the decoding error rate can be realized.

FIG. 2 shows a branch metric correction circuit 15.
FIG. 3 is a block diagram illustrating a configuration example of FIG. As shown in FIG.
The branch metric correction circuit 15 includes a minimum branch metric detection circuit 151, a branch metric memory circuit 152, an average value calculation circuit 153, and a variance value calculation circuit 154.
And a weighting circuit 155.

Minimum branch metric detection circuit 151
Detects the minimum branch metric from the branch metrics that are the calculation results BM of the branch metric calculation circuit 11. The branch metric memory circuit 152 includes:
The minimum branch metric detected by the minimum branch metric detection circuit 151 is stored. Average value calculation circuit 15
3 calculates a time average value from some minimum branch metrics at different times stored in the branch metric memory circuit 152. Variance calculation circuit 154
Calculates the time variance value of the minimum branch metric stored in the branch metric memory circuit 152 based on the time average value calculated by the average value calculation circuit 153. The weighting circuit 155 weights the branch metric, which is the calculation result of the branch metric calculation circuit 11, based on the time variance value calculated by the variance value calculation circuit 154.

Next, the operation of the above configuration will be described. The branch metric calculation circuit 11 calculates a branch metric from the input received code RS, and outputs the calculation result BM to the branch metric correction circuit 15.

In the branch metric correction circuit 15, first, in the minimum branch metric detection circuit 151,
The smallest branch metric is detected from all the inputted branch metrics, and the detection result is stored in the branch metric memory circuit 152. Next, the average value calculation circuit 153 calculates the average value of the minimum branch metric stored in the branch metric memory circuit 152 over a certain period of time. Then, the variance value calculation circuit 154 calculates the time variance value of the minimum branch metric stored in the branch metric memory circuit 152 based on the time average value calculated by the average value calculation circuit 153, and the result is weighted by the weighting circuit. 155. The weighting circuit 155 uses the input dispersion value as the reliability characteristic of the communication channel at that time, and weights all branch metrics at that time. Specifically, when the variance value is large, a small weight is assigned because the reliability of the branch metric is small, and when the variance value is small, a large weight is assigned because the reliability of the branch metric is large. This result is output to the addition / comparison / selection circuit 12 and stored in the path memory circuit 13.

In the addition / comparison / selection circuit 12, the branch metric is added to the path metric one time before, and then the respective path metrics are compared to select the maximum likelihood path. Output to the circuit 14. As a result of the above-described weight correction, the influence of the branch metric at a highly reliable time increases when the path metric is compared with the maximum likelihood in the addition / comparison / selection circuit 12.
In the maximum likelihood determination circuit 14, the original data is decoded for the maximum likelihood path, and a decoded output DC _OUT is obtained.

As described above, according to the present embodiment,
Detect the minimum value of the branch metric, calculate its time average value by storing the value in the memory for a certain time, further calculate their time dispersion value,
From this, the reliability of the channel is estimated, the branch metric is weighted according to the reliability of the channel, and the Viterbi decoding is performed by the path metric using these branch metrics. Since Viterbi decoding can be performed using a path metric based on a branch metric weighted in accordance with the reliability characteristic of the communication channel, decoding adapted to time variations in the reliability characteristic of the communication channel can be realized, and the decoding error rate can be reduced.

In this embodiment, as an example of a method of extracting a fluctuation component, a case has been described in which an average value of the minimum branch metric is calculated, and a variance value is calculated from the minimum branch metric and the calculated average value. However, it is needless to say that the present invention is not limited to this. Further, the method and apparatus of the present invention can be applied not only to the communication field but also to other fields, for example, a reproduction system of a magnetic disk device.

[0031]

As described above, according to the present invention,
Since the correction of the branch metric is performed according to the change of the value of the branch metric, decoding adapted to the time change of the transmission path characteristic can be realized, and the decoding error rate can be reduced.

[Brief description of the drawings]

FIG. 1 is a block diagram showing one embodiment of a Viterbi decoding device according to the present invention.

FIG. 2 is a block diagram illustrating a configuration example of a branch metric correction circuit according to the present invention.

FIG. 3 is a block diagram illustrating an example of an encoder for a convolutional code.

FIG. 4 is a trellis diagram showing a state transition in Viterbi decoding.

FIG. 5 is a block diagram illustrating a configuration example of a conventional Viterbi decoding device.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 11 ... Branch metric calculation circuit 12 ... Addition and comparison circuit 13 ... Path memory circuit 14 ... Maximum likelihood judgment circuit 15 ... Branch metric correction circuit 151 ... Minimum branch metric detection circuit 152 ... Branch metric memory circuit 153 ... Average value calculation circuit 154 ... Variance calculation circuit 155 ... weighting circuit

Continuation of front page (58) Field surveyed (Int. Cl. ⁷ , DB name) H03M 13/00

Claims (4)

(57) [Claims] In a Viterbi decoding method for decoding a convolutional code, a branch metric is calculated from an input received code, a plurality of minimum branch metrics having different times of the branch metric are detected, and a fluctuation component of the detected minimum branch metric is detected. Wherein the calculated branch metric is weighted and corrected according to the fluctuation component, a maximum likelihood path is selected based on the weighted correction result, and decoding is performed on the selected maximum likelihood path. Method. 2. The detection of the fluctuation component of the minimum branch metric is performed by calculating an average value of the plurality of minimum branch metrics at different times and calculating a variance of the minimum branch metric with respect to the average value. 2. The Viterbi decoding method according to 1. 3. A Viterbi decoding device for decoding a convolutional code, comprising: a branch metric calculating means for calculating a branch metric from an input received code; a minimum branch metric detecting means for detecting a plurality of minimum branch metrics at different times; A variation component detecting means for detecting a variation component of a minimum branch metric; a weighting means for weighting and correcting the calculated branch metric according to the variation component; and selecting a maximum likelihood path based on the weighted correction result. A Viterbi decoding device having decoding means for decoding the selected maximum likelihood path. 4. The variable component detecting means calculates an average value of a plurality of minimum branch metrics at different times, and a variance calculating a variance value of the minimum branch metric with respect to the average value of the minimum branch metric. The Viterbi decoding device according to claim 3, further comprising: a value calculating unit, wherein the weighting unit weights the branch metric according to the variance value.