JP2006060374A - Viterbi equalizer - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To conduct viterbi equalizing processing suitable for a multi-value QAM modulation system with a realistic operation quantity. <P>SOLUTION: The viterbi equalizer is provided with a delay wave replica generating means and an error calculating means. In the equalizer, a desired wave replica generating means is provided with a replica candidate generating means for generating a desired wave replica candidate as to a representative point of each block generated by equally dividing a distribution region of a signal point candidate; a candidate error calculating means for obtaining a branch metric for each desired wave replica candidate; a sorting means for selecting the most likelihood representative point based upon the branch metric; a region determination means for determining a distribution region of a next signal point candidate of interest based upon the distance between desired wave replica candidates corresponding to the block corresponding the maximum likelihood representative point and its adjacent block and the branch metric corresponding thereto; and a replica adding means for newly generating a desired wave replica additional candidate, corresponding to a signal point candidate adjacent to the most likelihood signal point candidate in which a corresponding desired wave replica is not yet generated. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a Viterbi equalizer that cancels the influence of delay waves other than the desired wave in an environment where the reception level of the delayed wave is sufficiently small with respect to the desired wave, and in particular, has a calculation amount suitable for the multilevel QAM modulation system Regarding reduction technology.

Viterbi equalizers are often used as error correction means in communication systems that are susceptible to interference waves, such as mobile communications. For this reason, many techniques for reducing the amount of computation of the Viterbi equalizer have been proposed on the assumption that they are applied in a Rayleigh fading environment where there is a delayed wave having a level that cannot be ignored with respect to the desired wave.
For example, pay attention to the fact that the number of delayed waves with a high reception level is not so many in reality, and pay attention to two delayed waves with high power and ignore the influence of other delayed waves. A technique for reducing the amount of calculation by reducing the number of states has been proposed (see Non-Patent Document 1).

A general Viterbi equalizer is designed to cope with a case where a delay wave arrives at an arbitrary delay time within a maximum symbol delay time (for example, 3 symbols) to be considered. Accordingly, the modulation scheme applied is in the case of QPSK (quadrature phase shift keying), the number of states to be considered is 4 ³ i.e. 64.
In the technique of Non-Patent Document 1, for example, the number of states to be considered in the Viterbi equalizer is reduced by performing sequence estimation by paying attention to the preceding wave and the wave delayed by 3 symbols. Specifically, first, as shown in FIG. 8, the transmission sequence indicated by the number assigned in symbol units according to the transmission time series is divided into three equal parts by switching the output every symbol interval T. Viterbi equalization is performed for each series. Then, by synthesizing the Viterbi equalization output obtained in each sequence and reproducing the original sequence, sequence estimation focusing on the preceding wave and the wave delayed by 3 symbols is realized. At this time, the number of states of each divided Viterbi equalization is set to 4 by treating the calculation related to the compensation for the preceding wave and the wave delayed by 3 symbols in the same manner as the Viterbi equalization for compensating the delayed wave of 1 symbol delay. It has been reduced to.

In addition to the above-described techniques, various state number reduction techniques suitable for modulation schemes such as the QPSK scheme and the π / 4 QPSK scheme have been proposed for Viterbi equalizers applied to mobile communication environments. On the other hand, a Viterbi equalizer adapted to multi-level QAM (quadrature amplitude modulation) has not been studied so far.
This is because, if the delayed wave compensation processing by the Viterbi equalizer is applied to the reception signal modulated by the multi-level QAM system, as described above, for the compensation of the preceding wave and the wave delayed by 3 symbols. This is because even if a technique that handles the related operations in the same way as Viterbi equalization that compensates for a delayed wave with a one-symbol delay is applied, the number of states becomes so large that it is considered impossible to achieve with a realistic amount of calculation. .

  This is because if the Viterbi algorithm is applied as it is to the multilevel QAM system, the desired wave replica generation unit 401 and the delayed wave replica generation unit 402 shown in FIG. A wave replica and a delayed wave replica are generated, and the branch metric calculation unit 403 calculates a branch metric with the received signal for all possible combinations of the desired wave replica and the delayed wave replica. This is because the selection processing unit 404 needs to add the path metric and the branch metric, compare the path metric as the addition result, and select the path. Note that the information on the path selected by the addition comparison selection processing unit 404 shown in FIG. 9 is held in the path memory 405 and is output as an output signal for which compensation for the delayed wave has been completed. Further, based on the information held in the path memory 405, the transmission path estimation unit 406 estimates the amplitude and phase rotation amount of the desired wave and the amplitude and phase rotation amount of the delayed wave, and based on the information. The desired wave replica generation unit 401 and the delayed wave replica generation unit 402 described above respectively generate a desired wave replica and a delayed wave replica. In addition, the path metric as a result of the addition process of the branch metric and the path metric performed in the addition comparison selection processing unit 404 described above is held in the path metric memory 407, and the addition process with the branch metric in the next symbol is performed. To be served.

Next, as one of the few examples of the Viterbi equalizer state number reduction technology adapted to the multilevel QAM modulation method, the creation of a desired wave replica corresponding to a transmission signal candidate that is far from the reception signal is suppressed. A technique for reducing the amount of calculation (see Patent Document 1) will be described.
In the technique of Patent Document 1, a transmission signal candidate by the 16QAM system is divided into, for example, four blocks corresponding to quadrants in a complex plane, and a distance between each desired wave replica corresponding to a representative point in each block and the received signal. The branch metric is calculated, and only the transmission signal candidates belonging to the block corresponding to the representative point that gives the minimum branch metric are left. This reduces the process of creating a desired wave replica for a transmission signal candidate far from the received signal and the process of calculating a branch metric for possible combinations of these desired wave replicas and delayed wave replicas.

For example, transmission signal candidates s1 to s16 according to the 16QAM scheme are divided into four blocks as shown in FIG. 10, branch metrics are calculated for representative points a1 to a4 of each block, and based on the obtained branch metrics. The representative point with the maximum likelihood (for example, representative point a1) is selected. The desired wave replicas corresponding to the transmission signal candidates s1 to s4 belonging to the block corresponding to the representative point a1 selected in this way are generated and used for the branch metric calculation process. The number can be greatly reduced, and the amount of calculation associated therewith can be greatly reduced.
JP-A-5-335893 (FIG. 1) Takao Kamio, Eimatsu Moriyama, Shuichi Sasaoka, Yuji Sugimoto "Development of 10Mbps p / 4-QPSK equipment for mobile communications-Outline of simplified Viterbi equalizer", IEICE Technical Report RCS96-67, 1996-08

The technique of Non-Patent Document 1 described above is very useful in a modulation scheme with a small number of states corresponding to each symbol as in the QPSK scheme, but has a large number of transmission signal candidates as in the multilevel QAM scheme. With the modulation method, a sufficient effect cannot be expected.
On the other hand, in the technique of Patent Document 1, in the process of narrowing down the transmission signal candidates by dividing into blocks, if a wrong block is selected, there is no true signal point in the selected block. As a result, communication quality may be significantly degraded.

  For example, in the 64QAM signal point arrangement as shown in FIG. 11, when there is a true signal point at the signal point d2, it is natural that the representative points a1 to a4 in the blocks corresponding to the four quadrants are selected. The representative point a1 is selected, and branch metric calculation processing is performed corresponding to the representative points b1 to b4 of the blocks obtained by dividing the block corresponding to the representative point a1 into four. At this time, if the true signal point is the signal point d2, and the position of the received signal is shifted due to the influence of noise and is located between the representative point b1 and the representative point b2, the representative point There is a possibility that b1 is selected as the most likely representative point. In this case, as a matter of course, the desired wave replica generation process is performed only for the signal points c1 to c4 that are signal points included in the block corresponding to the representative point b1, and is used for the branch metric calculation process. . That is, the signal point d2, which is a true signal point, may be eliminated in the above-described narrowing process.

On the other hand, since the conventional Viterbi equalizer calculation amount reduction technique is premised on application to a mobile communication system, the D / U ratio indicating the ratio between the reception level of the desired wave and the reception level of the delayed wave is low. For this reason, branch metrics corresponding to all possible signal points have been demanded in consideration of frequent occurrence of path branching in Viterbi equalization.
By the way, in the environment where both the transmitting node and the receiving node are fixed and there are few obstacles between the transmitting and receiving nodes and the prospect is good, the surrounding obstacles are compared with the Rayleigh fading environment which is treated as a general environment in the mobile communication system. The level of the delayed wave generated by scattering from the object is extremely low compared to the level of the desired wave. It is known that path branching in Viterbi equalization is very rare in such an environment with a high D / U ratio. Therefore, in the Viterbi equalizer used in an environment where the signal level of the delayed wave can be expected to be extremely small compared to the signal level of the desired wave as described above, for example, all the 64 signal points shown in FIG. It is highly likely that the arithmetic processing spent to generate the corresponding delayed wave replica and calculate the branch metric is useless processing.

The applicant of the present application discloses a Viterbi equalizer to which a possible method for solving the above-described problems is applied in Japanese Patent Application No. 2004-056152 “Viterbi equalizer”, and a similar Viterbi equalizer is disclosed in 2004. Published in Proceedings of the Institute of Electronics, Information and Communication Engineers published on March 8, 2000.
In the Viterbi equalizer disclosed in Japanese Patent Application No. 2004-056152, in the same manner as in Patent Document 1 described above, the center of the block obtained by equally dividing the area where the 64 signal points shown in FIG. Based on the branch metric of the desired wave replica candidate generated from the coordinates and the received signal, the region in which a signal point candidate with a high likelihood is expected to be limited is limited. In the Viterbi equalizer disclosed in Japanese Patent Application No. 2004-056152 at the final output stage as a desired wave replica, the hope corresponding to the signal point selected last by the technique disclosed in Patent Document 1 described above is used. A desired wave replica corresponding to a signal point candidate adjacent to this signal point is generated together with the wave replica.

  That is, in the Viterbi equalizer disclosed in Japanese Patent Application No. 2004-056152, in addition to the signal points finally selected by the technique disclosed in Patent Document 1, the signal points surrounding the signal points are used as true signal point candidates. By leaving, the risk of eliminating the true signal point in the process of limiting the blocks is eliminated, and the estimation accuracy for the true signal point is prevented from being lowered. On the other hand, since the number of desired wave replica candidates or desired wave replicas generated as described above is smaller than the number of all signal point candidates, the amount of computation required for generating the desired wave replica is also reduced. Yes.

  The feature of the Viterbi equalizer disclosed in this Japanese Patent Application No. 2004-056152 is that the region where the signal point candidate with high likelihood is expected to be present is limited, and the maximum likelihood signal point is detected. In addition to the signal point, the signal point candidates around the signal point are used as the generation target of the desired wave replica. And in Japanese Patent Application No. 2004-056152, the example which applied the technique disclosed in patent document 1 is disclosed as an example of the method of limiting the area | region where a signal point candidate with a high likelihood exists. Yes. However, there is room for further improvement in the technique itself that limits the region where a signal point candidate with high likelihood is expected to exist.

  An object of the present invention is to provide a Viterbi equalizer capable of realizing equalization processing compatible with the multi-level QAM modulation method with a realistic calculation amount.

  A first Viterbi equalizer according to the present invention includes delayed wave replica generation means, desired wave replica generation means, and error calculation means. The desired wave replica generation means includes replica candidate generation means, candidate candidates, An error calculation unit, a selection unit, a determination unit, a region determination unit, and a replica addition unit are provided. The region determination unit includes a distance calculation unit, a distance comparison unit, and a region estimation unit.

The principle of the first Viterbi equalizer according to the present invention is as follows.
The delayed wave replica generation means generates a delayed wave replica corresponding to at least a signal point candidate corresponding to the selected survival path among signal point candidates in 4 ^N QAM modulation (N is a natural number). The desired wave replica generation means selects a signal point candidate that can be expected to have a high likelihood from signal point candidates in 4 ^N QAM modulation, and generates a desired wave replica. The error calculation means obtains a branch metric with the input received signal for each possible combination of the delayed wave replica and the desired wave replica. In the desired wave replica generation means, the replica candidate generation means obtains the coordinate value of the centroid position of each predetermined number of equal area blocks obtained by equally dividing the distribution area of the signal point candidate of interest. To generate a desired wave replica candidate. The candidate error calculation means obtains a branch metric with the received signal for each desired wave replica candidate. The selecting means selects a desired wave replica candidate for which the smallest branch metric is obtained by the candidate error calculating means from the desired wave replica candidates. The determining means determines whether or not the barycentric position of each block matches the position of the signal point candidate in 4 ^N QAM modulation. The area determining means determines the distribution area of the signal point candidate to be noticed next when the determination means obtains a determination result that the position of the center of gravity does not match the position of the signal point candidate, and the replica candidate generating means To create a new desired wave replica candidate. The replica adding means is a signal in 4 ^N QAM modulation corresponding to the desired wave replica candidate selected by the selecting means when the determination means obtains a determination result that the position of the center of gravity and the position of the signal point candidate match. A new desired wave replica addition candidate is generated corresponding to a signal point candidate that is adjacent to the point candidate and the corresponding desired wave replica has not yet been generated. The desired wave replica candidate generated by the wave replica candidate generating means is output as a desired wave replica corresponding to a signal point candidate having a high likelihood. The distance calculating means provided in the region determining means includes a desired wave replica candidate selected by the selecting means and a desired wave replica candidate corresponding to two adjacent blocks adjacent to the target block corresponding to the desired wave replica candidate. Each distance is calculated. Similarly, the distance comparison means provided in the area determination means calculates the distances calculated by the distance calculation means for the desired wave replica candidates corresponding to two adjacent blocks, and the candidate error calculation means corresponding to these desired wave replicas. Compare each of the generated branch metrics. Similarly, the region estimation unit included in the region determination unit estimates a region where a signal point candidate with a high likelihood is likely to exist in the target block based on the comparison result obtained by the distance comparison unit. Output as a distribution area of signal point candidates to be noted.

The operation of the first Viterbi equalizer configured as described above is as follows.
In the desired wave replica generation means, for example, the replica candidate generation means generates desired wave replica candidates corresponding to the centroid positions a1 to a4 of each block obtained by equally dividing the signal point distribution in 64QAM shown in FIG. Then, corresponding to these desired wave replica candidates, each branch metric is calculated by the error calculation means. In this case, since the centroid position of each block does not coincide with the 64QAM signal point candidate, the region determination unit determines the region to be focused next, and the desired wave replica candidate generation processing described above is performed for this focused region. Is done.

  At this time, the distance calculating means provided in the area determining means is adjacent to the desired wave replica candidate selected by the selecting means as having the smallest branch metric between the received signal and the block corresponding to this desired wave replica candidate. The distances from the desired wave replica candidates generated based on the centroid positions of the two blocks (for example, centroid positions a2 and a4 shown in FIG. 11) are calculated. For example, when the desired wave replica candidate R1 generated corresponding to the barycentric position a1 of the block corresponding to the first quadrant of the complex plane shown in FIG. 11 is selected, the second quadrant adjacent to this block is selected. The distances between the desired wave replica candidates R2 and R4 generated corresponding to the barycentric position of the block and the barycentric position of the block in the fourth quadrant and the desired wave replica candidate R1 are calculated. Then, these distances are compared with the branch metrics obtained for the above two desired wave replica candidates R2 and R4 by the distance comparison unit, and the comparison result is subjected to the estimation process by the region estimation unit.

  Here, the branch metric represents an error between a received signal replica created based on a certain signal point candidate and an actually received received signal. Therefore, a large branch metric obtained corresponding to a replica of a received signal created based on a certain signal point candidate means that the signal point candidate used for creating the replica is far from the true signal point. Conversely, a small branch metric indicates that the distance between the signal point candidate used to create the replica and the true signal point is short.

  Therefore, the distance between the selected desired wave replica candidates and the two desired wave replica candidates generated by using the centroid positions of the two blocks adjacent to the corresponding desired wave replica candidates, and the corresponding desired wave replica candidates Based on the comparison result with the branch metric calculated as described above, it is possible to further narrow down the area where the true signal point exists in the block corresponding to the selected desired wave replica candidate.

In other words, the region estimation means uses the position of the desired wave replica candidate excluded by the selection means as being far from the reception signal and the branch metric corresponding thereto, so that the reception signal exists. The expected area is roughly estimated, and is used as the next area to be noted for the processing of the replica candidate generation means.
Therefore, the replica candidate generation means sets a region narrower than the block corresponding to the desired wave replica selected in the above processing as the next region of interest, and centroid of each block obtained by further dividing the region of interest into four A new desired wave replica candidate corresponding to the position is generated. For example, if it is determined by the region determining means described above which of the four blocks obtained by dividing the block corresponding to the desired wave replica a1 shown in FIG. The processing of calculating the branch metric for the desired wave replica candidate corresponding to the centroid position (b1, b2, b3, b4) of the block is omitted, and the centroid position of each block obtained by dividing the block to be noticed into four is obtained. It is possible to proceed to the processing relating to the corresponding desired wave replica candidate.

  In this way, the region where the signal point candidates with high likelihood are expected to be limited is limited, and the barycentric position of each block (for example, indicated by reference numerals c1 to c4 in FIG. 11) and the position of the signal point candidate When the two match, the replica adding means determines the signal point candidates selected by the selecting means from the signal point candidates c1 to c4 described above (for example, indicated by reference sign c2 in FIG. 11) according to the determination result by the determining means. The nearest signal point candidate among the adjacent signal points, and the desired wave replica addition candidate corresponding to the signal point candidate d2 and the signal point candidate d4 for which the corresponding desired wave replica has not yet been subjected to the branch metric calculation processing, Along with the desired signal replica candidates generated by the replica candidate generating means and corresponding to the signal point candidates c1 to c4 described above, Output as a desired desired wave replica.

  In this way, when the first Viterbi equalizer limits the region where a signal point candidate with a high likelihood is expected to exist, the desired wave replica candidate excluded as having a low likelihood at that stage is used. By using the corresponding branch metric and the distance between the desired wave replica candidates generated based on the centroid position of each block, fewer operations can be performed by narrowing down the area where signal candidates with high likelihood can be expected to exist. A true signal point can be estimated by quantity.

In addition, at the stage of finally outputting as a desired wave replica, blocks are limited by adding desired wave replicas of adjacent signal point candidates that have not been generated with the desired wave replica closest to the last selected signal point. In the process, the risk of eliminating the true signal point can be eliminated.
As a result, the desired wave replica corresponding to the true signal point can be generated reliably and with a realistic amount of computation, and can be provided to the processing of the error calculating means together with the delayed wave replica generated by the delayed wave replica generating means. it can.

A second Viterbi equalizer according to the present invention is the first Viterbi equalizer described above, wherein the region estimation means includes a first selection means and a second selection means.
The principle of the second Viterbi equalizer according to the present invention is as follows.
In the region estimation means provided in the first Viterbi equalizer described above, the first selection means calculates the distance and branch calculated for the desired wave replica candidate corresponding to the first adjacent block which is one of the two adjacent blocks. Based on the magnitude relationship obtained by the distance comparison means with respect to the metric, the attention area is obtained by dividing the attention area by a straight line connecting the centroid position of the second adjacent block which is the other of the two adjacent blocks and the centroid position of the attention area. One of the two regions to be selected is selected. Similarly, in the region estimation means, the second selection means uses the first relation based on the magnitude relationship obtained by the distance comparison means for the distance and branch metric calculated for the desired wave replica candidate corresponding to the second adjacent block. Select one of the two regions obtained by dividing the region selected by the first selection means by a straight line connecting the center of gravity of the block corresponding to the desired wave replica candidate and the center of gravity of the region of interest, and select the selected region Output as an estimation result.

The operation of the second Viterbi equalizer configured as described above is as follows.
For example, when the desired wave replica candidate R1 generated based on the barycentric position a1 shown in FIG. 11 is selected as a candidate having the smallest branch metric with the received signal, the first selecting unit of the region estimating unit For example, the branch metric _Ba2 corresponding to the desired wave replica candidate R2 generated for one centroid position a2 of the block adjacent to the block corresponding to the centroid position a1, the desired wave replica candidate R2, and the desired wave replica described above comparing the distance a ₁₂ calculated by the distance calculation means for candidates R1. Then, the first selection means determines whether a straight line connecting the centroid position a4 of another adjacent block and the centroid position a1 of the attention area, that is, complex coordinates, depending on whether the branch metric B _a2 is larger or smaller than the distance A _12. Of the two regions obtained by dividing the region of interest by a straight line parallel to the Q axis, the one farther or closer to the center of gravity a2 of the block corresponding to the desired wave replica candidate R2 is selected.

Then, the second selection means then selects the distance A ₁₄ between the desired wave replica candidate R4 generated for the barycentric position a4 of the other adjacent block and the desired wave replica candidate R1 described above, and the desired wave replica candidate described above. The branch metric B _a4 obtained for R4 is compared, and depending on whether the branch metric B _a4 is larger or smaller than the distance A ₁₄ , a straight line connecting the centroid position a2 and the centroid position a1 shown in FIG. Of the two regions obtained by dividing the region selected by the first selection means described above with a straight line parallel to the I-axis of the complex coordinates, from the barycentric position a4 of the block corresponding to the desired wave replica candidate R4 described above. Select the far or near side.

  In this way, the area estimation means is based on the branch metric obtained for the desired wave replica candidate generated corresponding to the area excluded as the possibility that the received signal is low and the position of the excluded desired wave replica candidate, By performing an extremely simple calculation, after narrowing the region where the possibility of the presence of the received signal is extremely high to one of the quadrants delimited by the coordinate axes as described above, this region (for example, the first (Quadrant) can be limited to one of the areas obtained by dividing into four.

  According to the first Viterbi equalizer described above, the information on the desired wave replica candidate generated corresponding to the representative point of the region excluded as the possibility that the true signal point is low is low, By narrowing down the region where signal points are expected to exist, the number of signal point candidates for generating the desired wave replica is narrowed down to a very small number, but the true signal regardless of the position occupied by the received signal in the complex plane. It is possible to reliably generate a desired wave replica having a high likelihood with respect to a point. Therefore, it is possible to achieve both reduction in the amount of calculation required for equalization processing conforming to the multi-level QAM scheme and maintenance of accuracy regarding estimation of the true signal point.

In particular, according to the above-described second Viterbi equalizer, an area where a true signal point is expected to be estimated can be estimated by a very simple calculation, so that an equalization adapted to the multi-level QAM scheme is achieved. The amount of computation required for processing can be greatly reduced.
As described above, according to the present invention, it is possible to realize a Viterbi equalizer that performs equalization processing conforming to the multi-level QAM scheme with a realistic calculation amount.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
FIG. 1 shows an embodiment of a Viterbi equalizer according to the present invention.
1 that are the same as those shown in FIG. 9 are denoted by the same reference numerals as those shown in FIG. 9, and description thereof is omitted.
In the Viterbi equalizer shown in FIG. 1, the training sequence output unit 201 generates a known signal sequence and provides it to the transmission path estimation unit 406 via the selector 202. The selector 202 selects the output of the training sequence output unit 201 and inputs it to the transmission path estimation unit 406 during the training period, and selects and transmits the signal output from the path memory 405 after the training period ends. Input to the route estimation unit 406.

  Further, in the replica generation unit 210 shown in FIG. 1, the multiplication module 211 receives the coordinate value in the complex plane input by the replica generation control unit 213 and the amplitude and phase rotation amount of the delayed wave obtained from the transmission path estimation unit 406. Are respectively input as complex signals, and the resulting delayed wave replica is input to the subtraction module 221 of the branch metric calculation unit 220. On the other hand, the multiplication module 212 shown in FIG. 1 multiplies the coordinate value in the complex plane input by the replica generation control unit 213 and the amplitude and phase rotation amount of the desired wave obtained from the transmission path estimation unit 406 as complex signals. Then, the desired wave replica obtained as a result is input to the subtraction module 222 of the branch metric calculation unit 220.

  The branch metric calculation unit 220 shown in FIG. 1 is basically equivalent to the branch metric calculation unit 403 shown in FIG. In this branch metric calculation unit 220, the subtraction module 221 described above performs a process of subtracting M (corresponding to the number of survival paths) delayed wave replicas generated by the replica generation unit 210 from the received signal in parallel. Has the function to execute. Further, the subtraction module 222 executes in parallel a process of subtracting a plurality of types of desired wave replicas output by the multiplication module 212 described above from each of the M types of subtraction results obtained as outputs of the subtraction module 221 described above. It has a function. Further, the error calculation unit 223 shown in FIG. 1 has a function of executing in parallel the processing of calculating the square of the subtraction result obtained for each combination by the subtraction module 222. The output is output as a branch metric.

  Further, the addition comparison selection processing unit 230 illustrated in FIG. 1 includes a branch metric comparison unit 231, a path metric update unit 233, and a survival path selection unit for performing the same functions as the addition comparison selection processing unit 404 illustrated in FIG. In addition to 234, an arithmetic control unit 232 is provided. The arithmetic control unit 232 has a function of passing the comparison result by the branch metric comparison unit 231 to the region estimation unit 214 and passing information indicating the selection result by the survival path selection unit 234 to the replica generation control unit 213.

  The replica generation control unit 213 illustrated in FIG. 1 receives information indicating signal point candidates corresponding to the M survival paths selected by the survival path selection unit 234 via the arithmetic control unit 232 and receives these signals. Coordinate values corresponding to the points are input to the multiplication module 211 to generate a delayed wave replica. The area estimation unit 214 shown in FIG. 1 receives the desired wave replica generated by the multiplication module 212 and the branch metric calculated by the error calculation unit 223 and at the same time the branch metric comparison unit 231 via the operation control unit 232. The result of the comparison processing is received, and based on these pieces of information, the signal point distribution region where a true signal point is expected to exist is narrowed down as described later. In addition, the region estimation unit 214 passes information on the narrowed region to the replica generation control unit 213 and provides it to a desired wave replica generation process. Here, the multiplication module 211 shown in FIG. 1 has M units based on the M coordinates input as candidates of delayed wave signal points by the replica generation control unit 213 and the input from the transmission path estimation unit 406. It has a function to generate a delayed wave replica. The multiplication module 212 also generates a plurality of desired wave replicas based on a plurality of coordinates input as desired wave signal point candidates by the replica generation control unit 213 and an input from the transmission path estimation unit 406. It has.

Next, a description will be given of a method of performing a compensation process, taking as an example the case where a delayed wave arriving in one symbol is equalized with respect to a received signal modulated by 64QAM by the Viterbi equalizer shown in FIG.
FIG. 2 is a flowchart showing the compensation operation.
In response to an instruction from the arithmetic control unit 232 illustrated in FIG. 1, the replica generation control unit 213 of the replica generation unit 210 receives a minimum from the arithmetic control unit 232 of the addition comparison selection processing unit 230 by, for example, the survival path selection unit 234. 2 is received (step 301 in FIG. 2), and the coordinates of this signal point are input to the multiplication module 211 (see FIG. 1), so that the delayed wave corresponding to the survival path is received. A replica is selectively generated (step 302).

  Next, the replica generation control unit 213 illustrated in FIG. 1 divides the region of interest (initially the entire distribution range of signal points) designated by the region estimation unit 214 into, for example, four equal area blocks, The center of gravity position of the block is obtained as a representative point of the block (step 303 in FIG. 2). For example, as shown in FIG. 3, initially, the distribution range of signal points in 64QAM is divided into four blocks corresponding to quadrants in the complex plane, and in the subsequent processing, the selected block is further divided into four blocks. Divide.

Next, the replica generation control unit 213 inputs the coordinate values of the four representative points obtained in step 303 described above to the multiplication module 212 shown in FIG. Replicas are generated (step 304), and these desired wave replicas are subjected to branch metric calculation processing.
In response to this, each desired wave replica obtained by the multiplication module 212 in step 304 described above from the output of the subtraction module 221 by the subtraction module 222 provided in the branch metric calculation unit 220 shown in FIG. Subtraction is performed, and the result of the subtraction is subjected to processing by the error calculation unit, whereby a branch metric corresponding to the combination of these desired wave replicas and the delayed wave replica generated in step 302 described above is calculated (step 305). .

  The branch metrics obtained in this way are compared by the branch metric comparison unit 231 of the addition comparison selection processing unit 230 shown in FIG. 1, and information indicating the desired wave replica that gives the minimum branch metric is the operation control unit 232. Is passed to the region estimation unit 214. In addition, information indicating the four desired wave replicas generated in the above-described step 304 and the branch metric values obtained corresponding to these are also passed to the region estimation unit 214.

In response to this, the region estimation unit 214 determines whether or not the representative point of each block described above matches the position of the candidate signal point in 64QAM (step 306). ) Is selected as the maximum likelihood block at the current stage (step 307).
At this time, the region estimation unit 214 first selects the desired block corresponding to the maximum likelihood block based on the desired wave replica obtained in step 304 corresponding to the above-described maximum likelihood block and two adjacent blocks. The distance between the wave replica and each desired wave replica corresponding to two adjacent blocks is calculated (step 308).

Next, the region estimation unit 214 compares the branch metric calculated for the desired wave replica corresponding to each adjacent block with the distance calculated in step 308 corresponding to these desired wave replicas (step 309). .
Based on the comparison result, the region estimation unit 214 selects one of the blocks obtained by dividing the maximum likelihood block into four as a new attention region (step 310).

For example, as shown in FIG. 3, when branch metrics Bm1 to Bm4 are calculated for the desired wave replicas generated for the representative points A1 to A4 that are the centroid positions of four blocks corresponding to each quadrant of the complex plane, respectively. As an example, the operation of selecting a region of interest will be described in detail.
When a block corresponding to the representative point A1 is selected based on these branch metrics, the region estimation unit 214 determines the desired wave replica corresponding to the representative points A2 and A4 and the representative point A1 in step 308 described above. The distances D ₁₂ and D ₁₄ with the desired wave replica corresponding to are calculated. Then, in step 309, the region estimating unit 214 compares the branch metrics Bm2 corresponding to each desired wave replica corresponding to the representative points A2, A4, bm4 these distances D _12, D ₁₄ and respectively.

Here, as described above, the magnitude relation between the distance D ₁₂ and the branch metric Bm4 described above branch metric Bm2 the distance D ₁₄ described above, shows the positional relationship between the representative points A2, A4 and the true signal point Yes.
For example, these branch metrics Bm2, bm4 are both distance D _12, if D is greater than _14, the true signal point, than the representative point A1 corresponding to the block of the maximum likelihood, the representative points A2, A4 It is thought that it is in the position far from both. Therefore, in this case, the region estimation unit 214 newly selects a region indicated by reference numeral G1 in FIG.

Conversely, if both of the branch metrics Bm2, bm4 is smaller than the distance D _12, D ₁₄ is, both the true signal point, than the representative point A1 corresponding to the block of the maximum likelihood, the representative points A2, A4 It is thought that it is in the near position. Accordingly, in this case, the region estimation unit 214 newly selects a region indicated by reference numeral G3 in FIG.
On the other hand, the branch metric Bm2 is greater than the distance D _12, when the branch metric Bm4 is smaller than the distance D _14, the true signal point, than the representative point A1 corresponding to the block of the maximum likelihood, from the representative point A2 Is far from the representative point A4. Therefore, in this case, the region estimation unit 214 newly selects a region indicated by reference numeral G4 in FIG.

Conversely, the branch metric Bm2 is smaller than the distance D _12, when the branch metric Bm4 is greater than the distance D _14, the true signal point, than the representative point A1 corresponding to the block of the maximum likelihood, the representative point A2 It is considered that the position is near from and far from the representative point A4. Therefore, in this case, the region estimation unit 214 newly selects a region indicated by reference numeral G2 in FIG.

That is, the region estimation unit 214 uses the position of the desired wave replica generated corresponding to the block excluded as not being the maximum likelihood block and the branch metric calculated corresponding to these desired wave replicas, In the maximum likelihood block, a region where a true signal point is expected to be further narrowed down.
In this way, one of the blocks obtained by further dividing the maximum likelihood block selected in step 307 into four is selected as a region to be noted, and based on the information regarding this region, the replica generation control unit 213 Returning to step 303, a new representative point is obtained for each block obtained by dividing the new region of interest into four, and a new desired wave replica generation process is executed.

  In this way, the range with a high probability that a true signal point exists is narrowed down, and when the position of each representative point obtained in step 303 matches the position of a signal point candidate in 64QAM (affirmative determination in step 306). The region estimation unit 214 selects a signal point candidate that gives the minimum branch metric based on the branch metric corresponding to each signal point candidate obtained in step 305 (step 311). Furthermore, in addition to the selected signal point candidate, the region estimation unit 214 finally selects a signal point that is the closest signal point to this signal point candidate and has not yet been used for generating a desired wave replica. A replica is added to signal point candidates to be created (step 312), and information regarding these signal point candidates is provided to the process of the replica generation control unit 213.

In this way, for example, the true signal point located at the signal point candidate C1 shown in FIG. 4 is moved to an intermediate position between the signal point candidate B3 and the signal point candidate C1 due to the influence of noise. In this case, the signal point candidate C1 can be reliably left as the final signal point candidate.
This is because in such a case, signal point candidates are narrowed down as described below.

  First, as shown in FIG. 4, branch metrics are calculated for the desired wave replicas generated for the representative points A1 to A4 corresponding to the barycentric positions of the four blocks corresponding to the quadrants of the complex plane. Based on the metric, the block corresponding to the representative point A1 is selected. In FIG. 4, the branch metrics calculated corresponding to the signal point candidates and the representative points are shown in parentheses below the reference numerals assigned to the corresponding signal point candidates and the representative points.

At this time, in step 308 described above, the region estimation unit 214 uses the distance d ₁₂ between the desired wave replica corresponding to the representative point A1 and the desired wave replica corresponding to the representative point A2, and the desired wave replica corresponding to the representative point A1. And the distance d ₁₄ between the desired wave replica corresponding to the representative point A4. Here, for simplification of description, assuming that there is no amplitude variation and phase rotation of the desired wave, both the distance d ₁₂ and the distance d ₁₄ are equal to the distance between the corresponding representative points, and therefore the branch corresponding to the representative point A 2. metric Bm2 (= 1.1) and on the basis of comparison of the result and the distance d ₁₂ between the comparison result and the distance d ₁₄ and branch metric bm4 (= 1.25), 4 divides the block corresponding to the representative points A1 Among the blocks obtained in this way, the block located at the upper right (indicated by reference numeral G1 in FIG. 4) is selected.

  In the case of 64QAM, the representative points B1 to B4 of each block obtained by dividing the block thus selected into four coincide with the positions of signal point candidates in 64QAM. Therefore, the signal point B3 is selected as the maximum likelihood candidate based on the branch metrics calculated for the desired wave replicas generated for the signal points B1 to B4 as the representative points. Further, in step 312 shown in FIG. 2, two signal points C1 and C2 closest to the signal point B3 among the signal points adjacent to the signal point B3 are selected as additional candidates (see FIG. 4). .

As described above, if the signal point narrowing method described above is used, it is possible to reliably leave a true signal point as a candidate regardless of the influence of noise.
Next, the replica generation control unit 213 inputs coordinate values corresponding to the additional candidate signal points thus obtained to the multiplication module 212, and generates a desired wave replica corresponding to these candidates ( Step 313). Then, branch metrics corresponding to these additional desired wave replica candidates are calculated by the branch metric calculation unit 220 shown in FIG. 1 (step 314).

  The result obtained by the branch metric calculation unit 220 for the combination of the desired wave replica and the delayed wave replica generated corresponding to the additional candidate obtained in this way, and the maximum likelihood candidate in step 311 described above. The branch metric obtained corresponding to the selected signal point B 3 and the three signal points B 1, B 2 and B 4 with which the desired wave replica is generated is passed to the path metric update unit 233 by the arithmetic control unit 232. , And is subjected to path selection processing based on maximum likelihood estimation of the signal sequence. At this time, the path metric update unit 233 calculates a new path metric by adding the newly obtained branch metric and the path metric stored in the path metric memory 407, and based on this result, the path metric update unit 233 calculates the path metric. The metric memory 407 is updated (step 315).

Here, FIG. 5 is a diagram illustrating sequence estimation.
In FIG. 5, signal points in 64QAM corresponding to time t−1 and time t are indicated by number 0 to number 63. For example, when the path metric corresponding to the signal point of number 1 is the smallest among the path metrics corresponding to each signal point at time t−1, the delay wave replica corresponding to the signal point of number 1 is Combination with desired wave replicas generated for signal point candidates (indicated by symbols B1, B2, B3, B4, C1, and C2 in FIG. 5) selected for the signal at time t by the above-described narrowing-down method For each, a branch metric is calculated. In FIG. 5, the combination whose likelihood is to be estimated is indicated by an arrow connecting the signal point number 1 at time t-1 and the signal point at time t indicated by symbols B1, B2, B3, B4, C1, and C2. Corresponding to these arrows, the values of branch metrics calculated for the respective combinations are shown.

  At this time, the path metric update unit 233 illustrated in FIG. 1 sets the path metric value (for example, 0.2) corresponding to the signal point of number 1 held in the path metric memory 407 to the symbol B1 in FIG. The branch metric values corresponding to the signal points indicated by B2, B3, B4, C1, and C2 are added to obtain the path metric of the series corresponding to each arrow shown in FIG.

  Here, if the true signal point is located at the signal point indicated by C1 in FIG. 4, the signal point and the number 1 at time t indicated by C1 in FIG. It can be expected that the branch metric corresponding to the arrow connecting the signal point at time t-1 shown in FIG. Therefore, of course, it corresponds to this sequence obtained as described above, that is, a sequence including the signal point at time t-1 indicated by number 1 in FIG. 5 and the signal point at time t indicated by reference C1. The path metric to be performed is considered to be the smallest.

  By comparing the path metrics obtained in this way, the survival path selection unit 234 finds a signal point (for example, signal point C1) corresponding to the minimum path metric described above, and calculates information regarding this signal point. The data is transferred to the replica generation control unit 213 via the control unit 232 and used for equalization processing in the next symbol. In addition, the survival path selection unit 234 selects a path including the signal point having the minimum path metric found as described above as a survival path, and adds the end point of the new path to the path memory 405. Accordingly, the path memory 405 outputs a signal corresponding to the oldest time (step 316), whereby a Viterbi equalized output can be obtained.

In the Viterbi equalizer described above, the signal point candidates used for the generation of the delayed wave replica are limited to those corresponding to the survival path, and the signal point candidates for generating the desired wave replica are further narrowed down.
As described above, the signal point candidates used for the generation of the delayed wave replica are limited to those corresponding to one survival path, and the desired wave replica is generated according to the procedure described in steps 303 to 312 described above. By determining candidate signal points, it is possible to greatly limit the number of combinations of delayed wave replicas and desired wave replicas that finally calculate branch metrics. In addition, the number of times of executing the processing for generating the desired wave replica is 10 times in total when the signal points of 64QAM are narrowed down as described above, including the step of narrowing down the signal points for generating the desired wave replica. It is clear that this is far less than when all the 64QAM signal points are candidates. Further, as described above, when the signal points for the desired wave replica are narrowed down by dividing each block into four, the number of desired wave replicas to be generated at the time of narrowing down is four, and the above-described Since the maximum number of signal point candidates to be added in step 312 is 2, the multiplication module 212 of the replica generation unit 210 has a function of generating a desired wave replica according to the coordinates of a maximum of 4 signal points. It is enough to have. Similarly, it is sufficient that the subtraction module 222 of the branch metric calculation unit 220 has a function of subtracting a maximum of four desired wave replicas generated simultaneously by the multiplication module 212 from the output of the subtraction module 221 in parallel. is there.

For these reasons, it is clear that the Viterbi equalization process can be realized with a sufficiently realistic amount of computation even in a communication system that employs a modulation method having an extremely large number of states such as 64QAM.
On the other hand, the Viterbi equalizer shown in FIG. 1 is not limited to the signal point candidates for generating the delayed wave replica and the signal point candidates for generating the desired wave replica. The applicant has confirmed by simulation that extremely high performance can be realized in an environment where the amplitude of the delayed wave is sufficiently small relative to the amplitude of the wave.

FIG. 6 shows the characteristic evaluation results of the Viterbi equalizer according to the present invention.
In FIG. 6, the horizontal axis indicates the ratio Eb / No between the signal power per bit and the noise power, and the vertical axis indicates the so-called bit error rate (BER). Further, in FIG. 6, the BER characteristic of the Viterbi equalizer according to the present invention is shown by a broken line with a circle, while the Viterbi equalization based on the technique disclosed in Patent Document 1 is shown by a broken line with a triangle. The estimated BER characteristics for the vessel were shown.

  In the estimation of the BER characteristics related to the Viterbi equalizer based on the technique disclosed in Patent Document 1, the following conditions were used. For the generation of the desired wave replica, a branch metric for the representative point corresponding to each quadrant is calculated, and all signal points (that is, 16 points) included in the quadrant corresponding to the representative point where the minimum branch metric is obtained. Signal points) are candidates, desired wave replicas corresponding to these candidates are generated, and subjected to branch metric calculation processing considering delayed wave replicas. Further, as delayed wave replicas, they are generated corresponding to all signal points in 64QAM, branch metrics are calculated for all possible combinations of these delayed wave replicas and the desired wave replicas described above, and based on this result. The path selection process was executed.

In addition, the evaluation of the BER characteristics of the Viterbi equalizer disclosed in Patent Document 1 and the Viterbi equalizer according to the present invention is carried out using a delayed wave to be considered as one delayed wave of one symbol delay, and a desired wave. It is assumed that the D / U ratio, which is the power ratio between the delay wave and the delay wave, is 12 dB, the number of observation symbols is 10 ⁶ , ideal transmission path estimation is performed, and the coefficient α of the roll-off filter is 0.5 went.

As can be seen from FIG. 6, the Viterbi equalizer according to the present invention is more than the characteristics expected in the Viterbi equalizer based on the technique disclosed in Patent Document 1, like the Viterbi equalizer according to the prior application. High characteristics can be expected.
Improvement of such characteristics is achieved by adding a signal point adjacent to the signal point in addition to the signal point selected as the maximum likelihood candidate based on the stepwise refinement combined with the region estimation described above. This is considered to be because a desired wave replica corresponding to a true signal point can be generated without omission and used for branch metric calculation processing.

Further, the BER characteristic obtained by the Viterbi equalizer according to the present invention is less deteriorated than the BER characteristic at the time of synchronous detection in AWGN (additive white Gaussian noise) indicated by a broken line in FIG. Can be confirmed.
The applicant has confirmed that the Viterbi equalizer according to the present invention can exhibit high performance even in an environment where the D / U ratio is lower than the above-described conditions.

  It has been confirmed that even when the D / U ratio is reduced to 6 dB, sufficient characteristics can be obtained by increasing the number of survival paths to two in the Viterbi equalizer according to the present invention. When a plurality of survival paths are left in this way, branch metrics are calculated for combinations of desired wave replicas generated using the selected signal point candidates and delayed wave replicas corresponding to the respective survival paths. Therefore, it is considered that the increase in the amount of computation due to the increase in the number of survival paths is proportional to the number of survival paths.

Therefore, in the Viterbi equalizer according to the present invention, even if the D / U ratio is about 6 dB, it is possible to expect a sufficient calculation amount reduction effect while maintaining the characteristics.
Next, the calculation amount reduction effect realized by the Viterbi equalizer according to the present invention will be described.
FIG. 7 shows the calculation amount of the main part of the Viterbi equalizer according to the present invention.

  The numerical values in each column shown in FIG. 7 are the same as those obtained when normal Viterbi equalization is performed for the QPSK method, when Viterbi equalization according to the present invention is performed for the 64QAM method, and for the 64QAM method. 1 symbol required for delayed wave replica generation processing, delayed wave replica subtraction processing, desired wave replica generation processing, desired wave replica subtraction processing, and addition comparison selection processing when Viterbi equalization is performed in accordance with the technique disclosed in FIG. The number of multiplications per win and the number of additions are shown. The last row of the table shown in FIG. 7 shows the sum of the number of multiplications and the number of additions required in each method.

  As is apparent from FIG. 7, the amount of computation required to realize the delayed wave compensation by the Viterbi equalizer according to the present invention is much larger than the amount of computation required in the technique disclosed in Patent Document 1. Very few. Further, the amount of computation (number of multiplications) required in the Viterbi equalizer according to the present invention is about 1.3 times the amount of computation (number of multiplications) in the Viterbi equalizer related to the QPSK scheme that has already been put into practical use. Thus, it has been proved that the Viterbi equalizer according to the present invention provides a delayed wave compensation technique adapted to the 64QAM system with a practical amount of calculation.

  Note that in a communication environment in which the amplitude of the delayed wave is sufficiently smaller than the amplitude of the desired wave, the survival path can be limited to one as described above, so that it is added in step 312 shown in FIG. The branch metric calculated corresponding to the desired wave replica and the branch metric corresponding to the desired wave replica generated for the signal point selected as the most likely signal point are used for path metric calculation and update processing in step 315. A configuration is also conceivable. In this case, a path metric for a signal point (for example, the signal points B1, B2, and B4 shown in FIG. 4) that is included in the signal point distribution area of the last attention but is not selected as the most likely signal point. This calculation process can be omitted. In this way, the signal point that was included in the signal point distribution area of the last attention but was not selected as the most likely signal point was excluded from the final signal point candidates, so that the estimation for the true signal point was performed. The accuracy is not expected to decrease. Because, if the survival path is limited to one, the desired wave replica generated for these signal points is at least more than the desired wave replica generated for the signal point selected as the most likely signal point. It is certain that a large branch metric will be given, and the path metric obtained by adding these branch metrics to the path metric one hour before will naturally correspond to the signal point selected as the most likely signal point. This is because it is clear that it is larger than the calculated path metric.

Viterbi equalizers that use a method of narrowing down signal points for generating a desired wave replica in stages and adding signal points adjacent to the finally narrowed candidates, regardless of the amplitude ratio between the desired wave and the delayed wave Since the amount of calculation can be reduced while maintaining the BER characteristics, it is useful in a communication system using a multilevel QAM modulation system in various environments.
Furthermore, by applying a technique that narrows down the signal points that generate the desired wave replica and a technique that limits the signal points that generate the delayed wave replica to the survival path, the delayed wave is compensated by Viterbi equalization for the 64QAM system. It is possible to reduce the amount of computation required to achieve the same amount as the amount of computation required for Viterbi equalization in a general QPSK system, and maintain communication quality. Thereby, it is possible to realize a Viterbi equalizer that performs delay wave compensation for the 64QAM system with a sufficiently realistic calculation amount.

It is a figure which shows embodiment of the Viterbi equalizer in connection with this invention. It is a flowchart showing compensation operation. It is a figure explaining narrowing down of an attention area. It is a figure explaining narrowing down of a signal point candidate. It is a figure explaining sequence estimation. It is a figure which shows the characteristic evaluation result of the Viterbi equalizer in connection with this invention. It is a figure which shows the operation amount of the principal part of the Viterbi equalizer concerning the present invention. It is a figure explaining the technique which reduces the calculation amount of a Viterbi equalizer paying attention to two waves with big electric power. It is a figure which shows the structural example of a general Viterbi equalizer. It is a figure explaining the calculation amount reduction technique of a Viterbi equalizer of 16QAM system. It is a figure explaining the subject in a prior art.

Explanation of symbols

201 training sequence output unit 202 selector 210 replica generation unit 211, 212 multiplication module 213 replica generation control unit 214 region estimation unit 220, 403 branch metric calculation unit 221, 222 subtraction module 223 error calculation unit 230, 404 addition comparison selection processing unit 231 Branch metric comparison unit 232 Operation control unit 233 Path metric update unit 234 Survival path selection unit 401 Desired wave replica generation unit 402 Delayed wave replica generation unit 405 Path memory 406 Transmission path estimation unit 407 Path metric memory

Claims (2)

Delay wave replica generation means for generating a delayed wave replica corresponding to a signal point candidate corresponding to at least a selected survival path among signal point candidates in 4 ^N QAM modulation (N is a natural number);
A desired wave replica generation means for generating a desired wave replica by selecting a signal point candidate that can be expected to have a high likelihood from the signal point candidates in the 4 ^N QAM modulation;
For each of the possible combinations of the delayed wave replica and the desired wave replica, an error calculating means for obtaining a branch metric with the input received signal,
The desired wave replica generation means includes:
Replica candidate generation means for generating a desired wave replica candidate for each of a predetermined number of equal area blocks obtained by equally dividing a distribution area of signal point candidates of interest using coordinate values of the center of gravity positions of the blocks When,
Candidate error calculation means for obtaining a branch metric with the received signal for each desired wave replica candidate;
Selecting means for selecting a desired wave replica candidate from which the smallest branch metric is obtained by the candidate error calculating means from among the desired wave replica candidates;
Determining means for determining whether the barycentric position of each block coincides with the position of a signal point candidate in the 4 ^N QAM modulation;
When the determination means obtains a determination result indicating that the position of the center of gravity and the position of the signal point candidate do not match, the distribution area of the signal point candidate to be noticed next is determined, and the replica candidate generation means A region determination means for creating a new desired wave replica candidate;
The signal in the 4 ^N QAM modulation corresponding to the desired wave replica candidate selected by the selection unit when the determination unit obtains a determination result that the position of the center of gravity and the position of the signal point candidate match. A new desired wave replica addition candidate is generated corresponding to a signal point candidate adjacent to the point candidate and the corresponding desired wave replica has not been generated, and the selected desired wave replica candidate and simultaneously with this Along with a desired wave replica candidate generated by the desired wave replica candidate generating means, a replica adding means for outputting as a desired wave replica corresponding to a signal point candidate with a high likelihood,
The region determining means includes
Distance calculating means for calculating the distance between the desired wave replica candidate selected by the selecting means and the desired wave replica candidate corresponding to two adjacent blocks adjacent to the target block corresponding to the desired wave replica candidate;
The distances calculated by the distance calculation means for the desired wave replica candidates corresponding to the two adjacent blocks are compared with the branch metrics calculated by the candidate error calculation means corresponding to these desired wave replicas. Distance comparison means;
Based on the comparison result obtained by the distance comparison means, an area where there is a high possibility that a signal point candidate with a high likelihood exists in the target block, and the distribution area of the signal point candidate to be focused next is estimated. And a region estimation means for outputting as a Viterbi equalizer. The Viterbi equalizer according to claim 1,
The area estimation means is
Based on the magnitude relationship obtained by the distance comparison means for the distance and the branch metric calculated for the desired wave replica candidate corresponding to the first adjacent block which is one of the two adjacent blocks, the two adjacent blocks First selection means for selecting one of the two regions obtained by dividing the region of interest by a straight line connecting the center of gravity of the second adjacent block that is the other of the two and the center of gravity of the region of interest;
Based on the magnitude relationship obtained by the distance comparison means for the distance and branch metric calculated for the desired wave replica candidate corresponding to the second adjacent block, the block corresponding to the first desired wave replica candidate A second region obtained by dividing the region selected by the first selection means by a straight line connecting the center of gravity and the center of gravity of the region of interest is selected, and the selected region is output as an estimation result. A Viterbi equalizer comprising a selection means.

Images