JPH06268531A - Viterbi equalizer - Google Patents

Abstract

PURPOSE:To provide a Viterbi equalizer by which an impulse response between a transmitter and a receiver is estimated with high accuracy and less arithmetic operation quantity. CONSTITUTION:In the case of setting a model to the estimation of channel response between a transmitter and a receiver, a correlation rj between a synchronizing signal and a received signal is calculated (step S3), the result of calculation is normalized and the result is given as an initial value of the channel response (step S4), and an impulse response estimate is revised by using a recurrence formula by the LMS method for each symbol (steps S5-S7). the revision arithmetic operation is repeated up to N at maximum (e.g. N=4) for one symbol (steps S7-S10).

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a Viterbi equalizer,
In particular, the present invention relates to a Viterbi equalizer suitable for use in a TDMA digital transmission system in mobile communication such as a car telephone.

[0002]

2. Description of the Related Art In the United States, Europe and Japan, car telephone system digitalization is in progress. In mobile communication such as a car telephone, a high-rise building or the like is interposed between a mobile station mounted on a car and a base station, so that transmission characteristics between the base station and the mobile station are affected by so-called multipath. Is significantly deteriorated, which makes it difficult to perform data transmission with few errors. Moreover, this equivalent transmission characteristic changes from moment to moment.

In such a mobile communication system,
In order to realize reception with few errors, equalization technology for correcting such transmission characteristics is indispensable. Hitherto, as such an equalization technique, a Viterbi equalizer has been proposed which decodes transmission data based on maximum likelihood sequence estimation using transmission characteristics between a base station and a mobile station. The Viterbi equalizer is mainly composed of a Viterbi estimation unit and a transmission line characteristic estimation unit. To summarize the operation, as shown in the flowchart of FIG. 18, first, the transmission path characteristic between the transmitter and the receiver is estimated (step S201), and then the transmission data is decoded based on this estimated data (step S201). That is, step S202).

By the way, in the GSM (Group Special Mobile) system adopted in European car phones, the communication channel from the base station to the mobile station (car) is shown in FIG.
The frame structure is as shown in (A) and (B).
Since a synchronization signal (SYNC) having a known pattern is added to the central portion of each time slot and transmitted, the impulse response of the transmission system (hereinafter, (Referred to as channel response)
Identify.

In the case of this GSM system, a modulation system called GMSK is adopted, but since the high frequency transmission system is converted into a baseband signal by passing through a demodulator, in the following, in order to simplify the explanation. First, let us proceed with the signal processing in the baseband. This G
In the SM system, eight types of data series are designated in advance as the synchronization signal pattern, and one of the series is shown in FIG. A conventional general procedure for modeling a channel response using this synchronization signal pattern will be described.

Now, the case where the channel response is as shown in FIG. 21 is taken as an example (actually, this channel response is unknown). In FIG. 21, the unit in the time axis direction is equal to the symbol transmission interval. The sync signal pattern of FIG. 21 is the sync signal pattern of FIG. The synchronization signal pattern received when passing through the transmission system having such a channel response is expressed by the following equation.

[Equation 1] Here, y _i is the received signal, x _i is the synchronization signal pattern, h
_i represents the channel response and is a value sampled at each symbol time interval T.

When the received signal corresponding to the sync signal pattern portion is calculated according to the equation (1), the received signal y as shown in FIG. 21 is obtained. On the receiver side, the known information is the sync signal pattern x _i and the received signal y _i . In the transmission path characteristic estimating unit 3, in the conventionally used modeling processing procedure, first, the synchronization signal data portion is detected by calculating the correlation between the reception signal and the synchronization signal pattern. Next, the cross-correlation function r _j between the sync signal data portion and the sync signal pattern is calculated using the following equation.

[Equation 2] Next, normalization is performed using the maximum value of this cross-correlation function r _j . The cross-correlation function r _j thus calculated is shown in FIG.

[0008]

In the conventional Viterbi equalizer described above, it is confirmed that the channel response h _i can be estimated with a certain degree of accuracy by comparing the channel response h _{i of} FIG. 21 with the cross-correlation function r _i. I can, but
On the other hand, it is revealed that "fake impulse response", which should not appear in the original case, is also detected. The cause is the autocorrelation function a _j of the synchronization signal pattern.
This is because there is a problem with itself. That is, it is clear that the autocorrelation function a _j of the synchronization signal pattern is calculated by using the following equation.

[Equation 3]

The autocorrelation function a _j calculated based on the above equation
Is shown in FIG. As is clear from FIG. 21, there are some side lobes having a considerably large level other than the main peak, which was a factor that deteriorates the accuracy when estimating the channel response. The present invention has been made in view of the above problems, and an object thereof is to provide a Viterbi equalizer capable of accurately estimating an impulse response between a transmitter and a receiver with a small amount of calculation. It is in.

[0010]

In order to achieve the above object, a Viterbi equalizer according to the present invention detects a sync signal data part from a received signal data sequence and uses this sync signal data part as a reference signal for LMS. Using the Viterbi algorithm based on the transmission model obtained by this transmission line characteristic estimation means, which models the channel response between the transmitter and the receiver using the recurrence formula It is configured to include a decoding unit that decodes the transmission data sequence.

[0011]

In the TDMA digital transmission system, a synchronization signal having a known pattern is transmitted for each time slot. When identifying the channel response between the transmitter and the receiver using this synchronization signal pattern, a recurrence formula by the LMS method is used with the synchronization signal data portion in the reception signal data sequence as a reference signal. Since the calculation is performed using the recurrence formula instead of calculating the matrix, the channel response between the transmitter and the receiver can be accurately estimated with a small amount of calculation.

[0012]

Embodiments of the present invention will now be described in detail with reference to the drawings. FIG. 2 is a block diagram showing the basic configuration of the Viterbi equalizer according to the present invention. In the figure, a received signal input is supplied to a branch metric calculation circuit 11 and a transmission path characteristic estimation unit 2 which form a Viterbi estimation unit 1. The transmission path characteristic estimation unit 2 is a feature of the present invention, detects a synchronization signal data portion from a received signal data sequence, and uses this synchronization signal data portion as a reference signal for the LMS method (minimum 2
The channel response between the transmitter and the receiver is modeled using a recurrence formula by multiplication.

In explaining a concrete example of the transmission path characteristic estimating section 2, the transmission path is modeled as a channel response as shown in FIG. If modeled like this,
Obviously, the signal y _i that is expected to be received is represented by the above equation (1). On the other hand, if the actually received signal is represented by Y _i , the error regarding the i-th symbol is represented by the following equation.

(4) e _i = y _i −Y _i

The sum of squares of this error is obtained by the following equation.

[Equation 5] Impulse response sequence h to minimize this error
It is a translation that determines _n , but as one of the solutions, LM
The S algorithm is often used. In this LMS algorithm, the impulse response sequence h _j is updated according to the procedure shown by the following equation.

[Equation 6] Here, j is the j-th tap of the FIR filter, n is the n-th update when updated, and α is the step gain (0 <
α ≦ 1), K is the number of times of averaging in the process of averaging the estimation error, y is the sample value of the received signal input to the equalization filter, and e is the equalization error.

Further, the equalization error e is expressed by the following equation, where the output of the equalization filter is e _out and the reference signal is r.

[Expression 7] e = e _out −r When using the recurrence formula by the LMS method, it is necessary to first set some initial value as the impulse response sequence. As this initial value, the cross-correlation function r _j shown in FIG. 21 is used.

The sync signal data x shown in FIG. 21 will be described below.
The processing procedure of the first embodiment of the transmission path characteristic estimation unit 2 which estimates the impulse response sequence from the received signal y and the received signal y using the recurrence formula by the LMS method will be described with reference to the flowchart of FIG. First, the received signal data is stored in the memory (step S1), and the sync signal (SYNC) data portion is detected from the received signal data series (step S2). Next, a cross-correlation function r between the synchronization signal and the received signal
_j is calculated (step S3), and the calculation result is normalized and used as the initial value of the channel response (step S3).
4).

Next, the sync signal pattern 2 shown in FIG.
For 6 symbols, the impulse response estimation value is updated for each symbol by the procedure shown in the recurrence formula of the equation 6 (steps S5 to S7). The update calculation is repeated up to N (for example, N = 4) times for one symbol (step S7).
~ S10), the estimation error is calculated based on the equation of Equation 4 in step S8, and if it is determined in step S9 that the absolute value of the estimation error is smaller than a predetermined reference value, the processing is performed for all symbols. It is determined whether or not the process is completed (i ≦ 26) (step S11). If i ≠ 26, the process returns to step S5 and the same calculation is performed for the next symbol that follows. Further, in steps S9 and S10, if the estimation error does not become equal to or less than the predetermined reference value even if the update calculation is repeated up to N times for one symbol, i ≠ 2
Under the condition of 6, the process returns to step S5, and the same calculation is started for the next succeeding symbol.

Here, the step gain α is α = 0.5
000. The reference value for evaluating the level of the estimation error is a value obtained by multiplying the absolute value of the reference signal by a certain coefficient β. For the sake of convenience, this constant coefficient β will be referred to as an error determination coefficient. This error determination coefficient β is β = 0.100.
It was set to 0. FIG. 5A shows the convergence result when the parameters for the arithmetic processing are set as described above. The sum of squares of the error shown in FIG. 5 (A) is an amount calculated based on the above-described equation (5). The maximum error is the maximum absolute value among the errors calculated based on the equation (4) for each symbol.

In order to compare with the calculation result shown in FIG. 5A, a calculation example in the case of setting other initial values is shown in FIG.
Shown in (B) and (C). In this comparison, all parameters except the initial value were the same. Comparing these calculation results, the number of updates can be reduced to about 1/7 and the estimation accuracy can be reduced by using the method shown in FIG. It can be seen that can be greatly improved. It was confirmed that the channel response can be accurately determined by the above procedure.

Next, an outline of the operation of the Viterbi estimation unit 1 will be described. After identifying this channel response, the transmission data sequence is decoded using the Viterbi algorithm.
That is, the example of FIG. 4 in which the general model shown in FIG. 3 is modeled by specifically limiting its channel response length will be described. When modeled as in FIG. 4, it can be viewed as a convolutional encoder with a constraint length k = 4 and a code rate r = 1/1. However, the difference from the normal convolutional encoder is the following two points. That is, 1. The adder ADD performs a linear operation. 2. The symbols input to the shift registers T _{0 to} T ₄ are binary values of <+1> and <−1>, and the respective outputs of the shift register are channel responses h ₋₁ , h ₀ , h ₊₁ and h.
_It is added to the adder ADD after weighting corresponding to ₊₂ and h ₊₃ .

The symbol G sent in the case of modeling in this way is expressed by the following equation.

[Equation 8] Here, <T _j > represents the content stored in the register T _j . FIG. 6 is a trellis diagram showing transitions of internal states of the transmission line in the transmission line model shown in FIG. In FIG. 6, the three-letter alphabet corresponding to each state node S _i represents the internal state of the shift register in each time slot. Here, since the shift register takes the values of <+1> and <-1>, they are represented as H and L for convenience of expression. In addition, in FIG. 6, by modifying the normally used lattice structure diagram, a solid line is used when the information input symbol <-1> is input, and a data input symbol <+1> is input when the information input symbol <+1> is input. This indicates that the transition shown by the broken line occurs.

On the other hand, in the Viterbi estimation unit 1, the received data Y _k is input to the branch metric calculation circuit 11 and the likelihood regarding the transition is calculated. Here, the branch metric is the likelihood calculated for each transition. Although some metric has been proposed for measuring the likelihood, the Hamming distance, which is the most general evaluation measure in the Viterbi decoder, is broadly applied. Now, the branch metric at time slot t (k) is calculated based on the following equation.

B (k, S _i → S _n ) = | Y _k −G _k | Here, Y _k is the received data, and G _k is the symbol sent from the equivalent transmission model, It is a value calculated based on the formula.

The branch metric calculated by the branch metric calculation circuit 11 is ACS (Add Compare S
elect) 12 is supplied. The ACS circuit 12 is composed of an adder, a comparator, and a selector, and selects a surviving path in each state. The basic operation is to add the branch metric and the path metric one time slot before for each of two paths that have the possibility of reaching a certain state node, and select the one with the smaller total value as the likely survival path. To do. Here, the path metric is a value obtained by adding up past branch metrics in the surviving path.

The output signal of the ACS circuit 12 is supplied to the path metric storage circuit 13 via the normalization circuit 14 and directly to the maximum likelihood path detection circuit 15. The maximum likelihood path detection circuit 15 detects the path having the smallest path metric value and updates the contents of the path memory 16 corresponding to the path. The path memory 16 is a memory that estimates and stores the information symbol sequence, and outputs the content corresponding to the detected path as decoded data when updated by the maximum likelihood path detection circuit 15. Next, the processing procedure in the Viterbi estimation unit 1 will be described with reference to the flowchart in FIG.

In FIG. 7, the branch metric calculation circuit 11 first calculates the branch metric (step S21), and then starts the calculation for the Nth state (step S22). Next, the address of State-1 one time slot before is set (step S2
3) The path metric stored in the path metric storage circuit 13 of the set address is read (step S24), the branch metric calculated in step S21 is added by the ACS circuit 12, and the added output is registered. Store in P1 (step S2
5).

Next, state-2 one time slot before
Is set (step S26), the path metric stored in the path metric storage circuit 13 of the set address is read (step S27), and the branch metric calculated in step S21 and the ACS circuit 12 Addition is performed, and the addition output is stored in the register P2 (step S28). Next, the ACS circuit 12 compares and selects the stored values of the registers P1 and P2 (step S
29, S30), and outputs the selected value (step S
31), the path metric storage circuit 13 is updated with this value (step S32), and the path memory 16 is further updated (step S33).

Steps S22 to S3 described above
The processes up to 3 are repeated until it is determined in step S34 that the number of states is 2 ^k-1 . After the above processing is completed, the maximum likelihood path detection circuit 15 detects the path having the minimum path metric value (step S35), and further, the minimum value of the path metric is subtracted from each path metric amount to perform normalization. Perform processing (step S3)
6). Subsequently, the maximum likelihood path detection circuit 15 sets the address of the maximum likelihood path (step S37), and the path memory 1
The contents of 6 are output as decoded data (step S3
8).

FIG. 8 shows a logical unit forming the Viterbi equalizer. In this figure, each metric has the following contents. P (k-1, S _i ): Path metric of the surviving path reaching the state node S _i at time slot t (k-1) P (k-1, S _j ): Time slot t (k-1) Path metric b (k, S _i → S _n ), which the surviving path has reached at the state node S _{j at} , the branch metric b corresponding to the transition from the state node S _i to the state node S _{n at} time slot t (k) (K, S _j → S _n ): Branch metric corresponding to the transition from the state node S _j to the state node S _n in the time slot t (k)

M (k−1, S _i ): Path memory of the surviving path reaching the state node S _i at time slot t (k−1) M (k−1, S _j ): Time slot t (k -1), the path memory <-1>, <+1> of the surviving path that has reached the state node S _j is: Information symbol P (k, S _n ), which is estimated to be transmitted in the time slot t (k): Time slot t (k) path surviving path which reaches the state node S _n has the metric M (k, S _n): a path memory with the survival path which reaches the state node S _n in time slot t (k)

Here, assuming that the constraint length is k, the number of states is 2
Since there are only ^k−1, the number of logical units shown in FIG. 8 is basically required to be 2 ^k−1 . Further, as in the block configuration of the Viterbi equalizer shown in FIG. 2, a normalization circuit 13 is provided to reduce the size of the path metric storage circuit 13 and to prevent overflow at the time of path metric calculation. As a specific process of this normalization, a process of first detecting the minimum value of the path metric and then subtracting the value from each path metric amount is performed. The number of surviving path metrics selected in this way is 2 ^{k−1, which} is the same as the number of states.

In each time slot, the operation of selecting the surviving path metric and the operation of updating the path metric and path memory corresponding to that path are repeated. It is known that when this operation is performed for a sufficiently long time, the same path is merged before a certain time, which is shown in FIG. The length of the path from the latest processing time point to the time when the paths are merged is called the truncated path length. In the maximum likelihood determination, the path having the smallest path metric value is detected and the path memory 1 corresponding to the path is detected.
Cut off the contents of 6 (normally 3 times the constraint length to 4
The information symbol at a time point that is traced back by (about twice as much) is output as a decoded symbol.

Next, a second embodiment of the present invention will be described. In this embodiment, when the impulse response sequence is estimated, the step gain is continuously controlled corresponding to the estimated error level in the process of sequentially updating using the LMS method. In the first embodiment, the value of the step gain α in the equation (6) is always constant, whereas in the second embodiment, the estimation error of the estimation error increases as the impulse response sequence gradually approaches the convergence value. Since the level also gradually decreases, the step gain α is correspondingly set to a gradually smaller value. Figure 5
For the purpose of comparison with the calculation result in the first embodiment shown in (A), the case where the same initial value is given will be described first.

The processing procedure of the second embodiment will be described below.
Description will be given according to the flowchart of FIG. First, the received signal data is stored in the memory (step S41), and the SYNC pattern is detected from the received signal data series (step S42). Next, the cross-correlation function between the synchronization signal and the reception signal is calculated, and the calculation result is normalized and used as the initial value of the channel response (step S43).
Subsequently, two types of step gains α ₁ and α ₂ are set as the step gain α (step S44). In this example, α ₁ = 0.7000 and α ₂ = 0.4000.
Further, a value obtained by multiplying the absolute value of the reference signal by a certain error determination coefficient β is used as a reference value for evaluating the level of the estimation error. As this error determination coefficient β, two types of error determination coefficient β are also used. ₁ and β ₂ are set (step S4)
5). In this example, β ₁ = 0.5000 and β ₂ = 0.10.
00.

Next, the impulse response estimation value is updated for each symbol using the recurrence formula of Equation 6 (step S).
46-S51). In the first update process, the step gain α ₁ and the error determination coefficient β ₁ are used. The update calculation is repeated up to N (for example, N = 4) times for one symbol (steps S48 to S51). Then, in step S49
In, when it is determined that the absolute value of the estimation error calculated based on the equation (4) becomes smaller than a predetermined reference value,
Increment the continuous pass counter (step S5
2) It is determined whether or not the count value M has become a predetermined value (for example, 3) or more (step S53).

If M> 3, that is, if the judgment of error <reference value (step S49) is cleared three times in succession in the above-described update processing, β ₂ is set as the error judgment coefficient β (step S54), Α ₂ is set as the step gain α (step S55), and then the continuous pass counter is reset (step S56). Then, it is determined whether or not the processing has been completed (i ≦ 26) for all symbols (step S57), and if i ≠ 26, the process returns to step S46 and the same calculation is performed for the next subsequent symbol.

In steps S48 to S51, if the estimation error does not become less than the predetermined reference value even if the updating operation is repeated up to N times for one symbol, the process proceeds to step S56 and the continuous pass counter is reset.
In step S57, if i ≠ 26, step S4
Returning to step 6, a similar operation is started for the next subsequent symbol. Further, when it is determined in step S52 that the determination of error <reference value (step S49) is cleared three more times in succession, the step gain α ₂ is set to a gradually smaller value in step S55 each time this determination is repeated. . Specifically, the step gain α calculated by the following equation 17
_{Use 2} .

[Formula 10] α ₂ (n + 1) = 0.80 · α ₂ (n)

FIG. 11A shows the convergence result when the parameters for the arithmetic processing are set as described above. For comparison with the calculation result shown in FIG. 11A, calculation examples when other initial values are set are shown in FIGS. 11B and 11C. In this comparison, all parameters except the initial value were the same. Comparing these calculation results, for example, FIG. 5B and FIG. 11B, it can be seen that the number of updates can be significantly reduced. In the above description of the second embodiment, the processing of the transmission path characteristic estimation unit 2 has been described, but the processing of the Viterbi estimation unit 1 is exactly the same as in the first embodiment.

Next, a third embodiment of the present invention will be described. In this embodiment, when the impulse response sequence is estimated, the magnitude of the above-described error determination coefficient is continuously controlled corresponding to the estimated error level in the process of sequentially updating using the LMS method. In the first example, the value of the error determination coefficient β was always constant.
On the other hand, in the third embodiment, the level of the estimation error gradually decreases as the impulse response sequence gradually approaches the convergence value, and accordingly, the error determination coefficient β is set to a gradually smaller value. To do. Figure 5
For the purpose of comparison with the calculation result in the first embodiment shown in (A), the case where the same initial value is given will be described first.

The processing procedure of the third embodiment will be described below.
It will be described according to the flowchart of FIG. First, the received signal data is stored in the memory (step S61), and the SYNC pattern is detected from the received signal data series (step S62). Next, the cross-correlation function between the synchronization signal and the reception signal is calculated, and the calculation result is normalized and then used as the initial value of the channel response (step S63).
Then, two types of step gains α ₁ and α ₂ are set as the step gain α, and two types of error determination coefficients β ₁ and β ₂ are set as the error determination coefficient β (step S64,
S64). In this example, α ₁ = 0.7000, α ₂ = 0.
4000, β ₁ = 0.5000, β ₂ = 0.300
Set to 0.

Next, the impulse response estimation value is updated for each symbol using the recurrence formula of Equation 6 (step S
66-S71). In the first update process, the step gain α ₁ and the error determination coefficient β ₁ are used. The update calculation is repeated up to N (for example, N = 4) times for one symbol (steps S68 to S71). Then, in step S69
In, when it is determined that the absolute value of the estimation error calculated based on the equation (4) becomes smaller than a predetermined reference value,
The continuous pass counter is incremented (step S7
2) It is determined whether or not the count value M has reached a predetermined value (for example, 3) or more (step S73).

When the judgment of error <reference value (step S69) is cleared three times in succession in the above-mentioned update processing (M> 3), α ₂ is set as the step gain α (step S74), and the error is set. Β ₂ is set as the determination coefficient β (step S75), and then the continuous pass counter is reset (step S76). Then, it is determined whether or not the processing has been completed for all symbols (i ≦ 26) (step S77), and if i ≠ 26, the process returns to step S66 and the same calculation is performed for the next subsequent symbol.

In steps S68 to S71, if the estimation error does not become less than the predetermined reference value even if the update operation is repeated up to N times for one symbol, the process proceeds to step S76 and the continuous pass counter is reset,
In step S77, if i ≠ 26, step S6
Returning to step 6, a similar operation is started for the next subsequent symbol. If it is determined in step S72 that the determination of error <reference value (step S69) has been cleared three more times consecutively, the error determination coefficient β ₂ is set to a gradually smaller value in step S75 each time this determination is repeated. To do. Specifically, the error determination coefficient β ₂ given by the following equation is used.

[Formula 11] β ₂ (n + 1) = 0.70 · β ₂ (n)

FIG. 13A shows the convergence result when the parameters for the arithmetic processing are set as described above. For comparison with the calculation result shown in FIG. 13A, calculation examples when other initial values are set are shown in FIGS. 13B and 13C. In this comparison, all parameters except the initial value were the same. Comparing these calculation results, for example, FIG. 5B and FIG. 13B, it can be seen that the number of updates can be significantly reduced. In the above description of the third embodiment, the processing of the transmission path characteristic estimation unit 2 has been described, but the processing of the Viterbi estimation unit 1 is exactly the same as in the case of the first embodiment.

Next, a fourth embodiment will be described. In this embodiment, when the impulse response sequence is estimated, the value before update is used or the value after update is used when the criterion of the estimation error cannot be passed in the process of sequentially updating using the LMS method. After selecting whether or not to do so, the next step is entered. For the purpose of comparison with the calculation result in the first embodiment shown in FIG. 5A, the case where the same initial value is given will be described first.

The processing procedure of the fourth embodiment will be described below.
Description will be given according to the flowchart of FIG. First, the received signal data is stored in the memory (step S81), and the SYNC pattern is detected from the received signal data series (step S82). Next, the cross-correlation function between the sync signal and the received signal is calculated, and the calculation result is normalized and used as the initial value of the channel response (step S83).
Then, the step gain α and the error determination coefficient β are set (steps S84 and S85). In this example, α
= 0.5000 and β = 0.1000.

Next, the impulse response estimation value is updated for each symbol using the recurrence formula of Equation 6 (step S).
86-S92). At the time of this updating process, the impulse response estimation value before updating is stored in the memory in advance (step S87). Then, the update calculation is repeated up to N (for example, N = 4) times for one symbol (steps S89 to S92). In step S90, the number 4
When it is determined that the absolute value of the estimation error calculated based on the equation (4) is smaller than the predetermined reference value, it is determined whether the processing is completed (i ≦ 26) for all symbols (step S93), and i If ≠ 26, the process returns to step S86 and the same calculation is performed for the subsequent symbol.

In steps S89 to S92, if the estimation error does not become less than the predetermined reference value even if the updating operation is repeated up to N times for one symbol, the impulse response before updating stored in the memory in step S87. The estimation error 1 by the column is calculated (step S9
4) Subsequently, the estimation error 2 based on the updated impulse response sequence is calculated (step S95), and these estimation errors 1 and 2 are compared (step S96). Then, if the estimation error 1 <estimation error 2, the impulse response sequence before updating is selected (step S97), and if the estimation error 1 ≧ estimation error 2, the impulse response sequence after updating is selected ( Step S98). That is, one having a smaller estimation error is selected, and i ≠ 2 in step S93.
Under the condition of 6, the process returns to step S86 and the same calculation is started for the next succeeding symbol.

FIG. 15A shows the convergence result when the parameters for the arithmetic processing are set as described above. For comparison with the calculation result shown in FIG. 15A, calculation examples when other initial values are set are shown in FIGS. 15B and 15C. In this comparison, all parameters except the initial value were the same. Comparing these calculation results, for example, FIG. 5B and FIG. 15B, it can be seen that the number of updates can be significantly reduced. In the above description of the fourth embodiment, the processing of the transmission path characteristic estimation unit 2 has been described, but the processing of the Viterbi estimation unit 1 is exactly the same as in the first embodiment.

Next, a fifth embodiment will be described. In this embodiment, when the impulse response sequence is estimated, the synchronization signal (SYNC) data sequence stored in the memory is repeatedly used in the process of sequentially updating by using the LMS method, so that the estimation accuracy is improved. It is intended to improve. In each of the first to fourth embodiments, the SYNC data sequence is used only once, but in the fifth embodiment, it is used any number of times. For the purpose of comparison with the calculation result in the first embodiment shown in FIG. 5A, the case where the same initial value is given will be described first.

The processing procedure of the fifth embodiment will be described below.
It will be described according to the flowchart of FIG. First, the received signal data sequence is stored in the memory (step S10
1), a SYNC pattern is detected from this received signal data sequence (step S102). Next, the cross-correlation function between the synchronization signal and the received signal is calculated, and the calculation result is normalized and then used as the initial value of the channel response (step S103). Then, the step gain α and the error determination coefficient β are set (steps S104 and S10).
5). In this example, α = 0.5000 and β = 0.10.
00.

Next, the impulse response estimation value is updated for each symbol using the recurrence formula of Equation 6 (step S
106-S111). The update calculation is repeated up to N (for example, N = 4) times for one symbol (step S1).
08-S111). When it is determined in step S109 that the absolute value of the estimation error calculated based on the equation 4 is smaller than the predetermined reference value, it is determined whether or not the processing is completed (i ≦ 26) for all symbols. (Step S112) If i ≠ 26, the process returns to step S106 and the same calculation is performed for the next symbol that follows. In steps S108 to S111, if the estimation error does not become less than or equal to the predetermined reference value even if the update operation is repeated up to N times for one symbol, step S1
In 12, the processing returns to step S106 under the condition that i ≠ 26, and the same calculation is started for the next succeeding symbol.

When it is determined in step S112 that the process for the 26th symbol has been completed, the number J of repeated processes is counted (step S113), and then the number J of processes is a predetermined value (3 in this example). It is determined whether or not this is the case (step S114), and if J ≠ 3, the process returns to step S106 and the above-described series of update processing is repeated until J ≧ 3 is determined in step S114. That is, the impulse response estimation value updated in the above series of processing is set as a new initial value, and the series of processing described above is repeated, for example, three times.

FIG. 17A shows the result of convergence when the parameters for arithmetic processing are set as described above. In order to compare with the calculation result shown in FIG. 17A, calculation examples when other initial values are set are shown in FIGS. 17B and 17C. In this comparison, all parameters except the initial value were the same. Comparing these calculation results, for example, FIG. 5B and FIG. 17B, it can be seen that the designation accuracy can be significantly improved. In the above description of the fifth embodiment, the processing of the transmission path characteristic estimation unit 2 has been described, but the processing of the Viterbi estimation unit 1 is exactly the same as in the first embodiment.

[0054]

As described above, according to the present invention,
Detect the sync signal data part from the received signal data series,
By using the synchronization signal data section as a reference signal and modeling the channel response between the transmitter and the receiver by using the recurrence formula by the LMS method, the recurrence formula is used instead of calculating the matrix. Since it was a calculation
The channel response between the transmitter and receiver can be accurately estimated with a small amount of calculation.

[Brief description of drawings]

FIG. 1 is a flowchart of a first embodiment according to the present invention.

FIG. 2 is a block diagram showing a basic configuration of a Viterbi equalizer according to the present invention.

FIG. 3 is a block diagram showing a generalized transmission line equivalent model.

FIG. 4 is a block diagram showing an embodied transmission line equivalent model.

FIG. 5 is a diagram showing a calculation result of the first exemplary embodiment of the present invention.

FIG. 6 is a trellis diagram showing transitions of internal states of a transmission line in a transmission line model.

FIG. 7 is a flowchart showing a processing procedure of a Viterbi estimation unit.

FIG. 8 is a block diagram showing a logical unit forming a Viterbi equalizer.

FIG. 9 is a diagram showing metric calculation and survivor paths.

FIG. 10 is a flowchart of a second embodiment according to the present invention.

FIG. 11 is a diagram showing a calculation result of the second exemplary embodiment of the present invention.

FIG. 12 is a flowchart of a third embodiment according to the present invention.

FIG. 13 is a diagram showing a calculation result of the third exemplary embodiment of the present invention.

FIG. 14 is a flow chart of a fourth embodiment according to the present invention.

FIG. 15 is a diagram showing a calculation result of the fourth example of the present invention.

FIG. 16 is a flowchart of a fifth embodiment according to the present invention.

FIG. 17 is a diagram showing a calculation result of the fifth example of the present invention.

FIG. 18 is a flowchart showing the basic operation of the entire Viterbi equalizer.

FIG. 19 is a configuration diagram of a communication channel from a base station to a mobile station.

FIG. 20 is a configuration diagram of a synchronization signal pattern.

FIG. 21 is a diagram for explaining a conventional example.

[Explanation of symbols]

 1 Viterbi Estimator 2 Transmission Line Characteristic Estimator 11 Branch Metric Calculation Circuit 12 ACS Circuit 13 Path Metric Storage Circuit 14 Normalization Circuit 15 Maximum Likelihood Path Detection Circuit 16 Path Memory

Claims (6)

[Claims] 1. A sync signal data part is detected from a received signal data sequence, and an impulse response between a transmitter and a receiver is obtained by using a recurrence formula by the least squares method using this sync signal data part as a reference signal. Viterbi equalization, characterized by comprising: a transmission path characteristic estimating means for modeling; and a decoding means for decoding a transmission data sequence using a Viterbi algorithm based on the transmission model obtained by the transmission path characteristic estimating means. vessel. 2. The transmission path estimating means calculates a cross-correlation function between the sync signal data portion and the sync signal, and uses this cross-correlation function as an initial value of the impulse response to obtain a recurrence formula by the least square method. Use of claim 1
Viterbi equalizer as described. 3. The Viterbi equalization according to claim 1, wherein the transmission line estimating means continuously controls a step gain corresponding to an estimation error level in the recurrence formula based on the least squares method. vessel. 4. The transmission line estimating means continuously controls an error evaluation criterion corresponding to an estimated error level in the recurrence formula by the least squares method.
Viterbi equalizer as described. 5. The transmission line estimating means compares an estimation error based on a value before update and an estimation error based on a value after update when the criterion of the estimation error cannot be passed in the recurrence formula based on the least squares method. The Viterbi equalizer according to claim 1, wherein a value before or after updating is used based on the result. 6. The Viterbi equalizer according to claim 1, wherein the transmission path estimation unit repeatedly uses the synchronization signal data section in a recurrence formula based on the least squares method.