JPH04183143A - Transmission line train estamating unit - Google Patents

Abstract

PURPOSE:To make it possible to follow up a transmission line that fluctuates at high speed while carrying out transmission line estimation for each train of transmission lines stably by obtaining an estimated value for the new present point of time from an estimated value of transmission line response at a past point of time and an estimated value of the present point of time obtained by solving a transmission line equation. CONSTITUTION:A transmission line response vector ht.l3 at the present point of time obtained using a transmission line response calculating circuit 103 by the first weight coefficient for all states using the first multiplier 109. On the other hand, the second multiplier 111 is used to multiplying an estimated value of transmission line response vector ht1' rearranged for a preceding point of time determined using a matrix switch 106 for all states by the second weight coefficient. Further, outputs from multipliers 109 and Ill are added to each other by an adder 110 to obtain an estimated value of transmission line response vector ht' at the present point of time for respective states, and a result is stored in a transmission line response storage circuit 105 and supplied to a branch metric calculating circuit 107. Next, vector ht-1' at the preceding point of time stored in the transmission line response storage circuit 105 is supplied to the matrix switch 106 and rearranged in accordance with connection information of a survival transmission line train prepared by a viterbi processor.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a sequence estimating device that estimates a transmission signal sequence by following temporal fluctuations in characteristics of a transmission path.

(Prior art) Maximum likelihood sequence estimator (MLSE) is known as an equalization method with the best equalization ability (for example, Document 1: G,
D. Forney, “Maximum Likel”
ihood Sequence Estimate
of Digital 5equences in t
he presence of intersymbol
1nterference,”IEEE Tran
saction onInformationTheo
ry, vol, IT-18, no, 3. May 197
2). A maximum likelihood sequence estimator generally includes a single transmission channel response estimator, and transmission channel estimation is performed using this transmission channel response estimator when receiving a known sequence.

In addition, when the characteristics of the transmission path change over time, an adaptive maximum likelihood sequence estimator has been proposed that follows the change in the characteristics of the transmission path over time (for example, see Reference 2:
G. Ungerboeck, “Adaptive
Maximum Likelihood Receiving
er for Carrier-Modulated
Data Transmission Systems
,” IEEE Transaction on Comm
unications, vol, C0M-22, n
o, 5° May 1974). The adaptive maximum likelihood sequence estimator first calculates the transmission path response when receiving a known sequence, and then operates the transmission path estimator using an adaptive algorithm when receiving an information data sequence to sequentially calculate the transmission path response. It is characterized by being able to follow temporal changes in transmission path characteristics by updating. However, in the adaptive maximum likelihood sequence estimator, when the channel characteristics change rapidly, the adaptive operation of the channel estimator cannot keep up.

In contrast, Japanese Patent Application No. 203436/1999 proposes a new type of sequence estimating device that can follow even rapidly changing transmission paths. This device is characterized in that not only the transmission signal sequence but also the characteristics of the transmission path are unknown, and the transmission path response corresponding to each sequence is estimated and the Viterbi algorithm is applied. The transmission path response is estimated by solving the transmission path equation for each sequence, which is determined by the transmission signal sequence candidate, the transmission path response, and the received signal.

Since this corresponds to successively finding the optimal solution for the transmission path response, the sequence estimating device disclosed in Japanese Patent Application No. 2-203436 can follow even high-speed transmission path fluctuations.

(Problem to be Solved by the Invention) However, in the sequence estimation device of Patent Application No. 2-203436, depending on the pattern of the transmission signal sequence candidate, the transmission path equation is determined by the transmission signal sequence candidate, the transmission path response, and the received signal. There is a drawback that there are cases where the transmission path response becomes undefined because it is essentially unsolvable.

Furthermore, since the sequence estimating apparatus disclosed in Japanese Patent Application No. 203436 obtains solutions one after another, it has the disadvantage that the estimated transmission channel response varies too sensitively from time to time due to noise, etc., resulting in errors in sequence estimation.

SUMMARY OF THE INVENTION Therefore, an object of the present invention is to provide a sequence estimating device that can follow a rapidly changing transmission path while stably performing transmission path estimation for each sequence.

(Means for Solving the Problems) A sequence estimating device according to the present invention includes a register that stores a plurality of sample values of a received signal, inputs the plurality of sample values from the register, and inputs the plurality of sample values from the register, and calculates a plurality of sequences. a transmission path response calculation circuit that estimates the transmission path response at the current time;
a weighting factor control circuit that calculates a first weighting factor and a second group of weighting factors for each sequence according to the transmission path response for the plurality of sequences obtained by the transmission path response calculation circuit; a first multiplier that multiplies a weighting coefficient of 1 by the output of the transmission path response calculation circuit; and a second multiplier that multiplies the second weighting coefficient group and the past rearranged transmission path response estimation value sequence. an adder that adds the output of the first multiplier and the output of the second multiplier, and stores the output of the adder for each of the plurality of sequences as a transmission path response estimated value at the current time. a transmission path response storage circuit that calculates a transmission path response, and a correspondence relationship between a past transmission path response estimated value sequence given as an output of the transmission path response storage circuit and the transmission path response at the current time given as an output of the transmission path response calculation circuit. is checked from the connection information of the surviving series for each of the plurality of series, the past transmission path response estimates are sorted according to the correspondence relationship, and the past transmission path response estimates are output as the rearranged transmission path response estimates. Branch metric calculation that calculates a virtual reception signal point for each series based on matrix sinch and the estimated transmission channel response value at the current time given by the adder, and calculates the distance from the sample value of the reception signal. and a Viterbi processor that receives the output of the branch metric calculation circuit, determines the received signal by a Viterbi algorithm, and outputs the connection information of the surviving sequence to the matrix switch.

(Function) In the following, a transmission line in which there are a plurality of delayed wave responses to the main wave as shown in FIG. 4 will be considered. The transmission line impulse response is expressed as a vector htT=(h,.,htl,...,h
tL), transmit signal sequence as vector 't''=(St, 5t
-1, . . . , 5t-L), and let vt be the additive transmission line noise including the observation process independent of the transmitted signal. At this time, the receiver input r at time t is given by the sum of the vector salt, the convolution of the vector 8-body, and the noise, as shown in equation (1).

zuT→ r(=81−ht+Vj , , , , , , , , ,
,,,-,--1-(1) This situation is shown in FIG. Hereinafter, equation (1) will be referred to as the transmission path equation at time t.

Next, a method of least squares estimation of the transmission line impulse response vector salt from N received signals from time t-N+1 to time t will be described. First, for this purpose, from time t-N+1 to time t, N transmission signal sequence vector salts (t-N+1
≦τ≦t), the transmission signal matrix S is defined as follows.

Further, the received signal vector rt and the noise vector ⊖ are defined below.

? tT=(rt-rt-x-=-rt-N+1)
(3)? ,”=(y, Vt-1,..., Vt-
N+1) (4) From the above, the transmission path equation over N time points can be written as equation (5).

(5) At this time, transmission line impulse response vector slope + ls by least squares estimation
ht+1s=(StT'5t)1. BtT, rt(6)
(For example, Reference 3: Nakamizo [Signal Analysis and System Identification] Corona Publishing, 1988). In particular, the number of received signals (N) used for impulse response estimation is the number of transmission channel responses (L
+1), the transmitted signal matrix S becomes a square matrix, and therefore, by simply multiplying the received signal by the inverse of the transmitted signal matrix 5tI7), a channel response estimated value can be obtained by least squares estimation.

(7) - The sequence estimating device of patent application No. 2-203436 shown in FIG. ,...,5t-L-N+1
), respectively, the transmission path response estimated value vector Kt,
Find the solutions for Is, and based on them, calculate each combination of transmission signals (St, 5t-1, ..., at-L-N+1) at each time.
The likelihood (branch metric) shown in equation (8) is calculated for .

lrt "S wind, ts12 (8) Then, use the Viterbi algorithm to find the transmission signal sequence over all times that minimizes the value (path metric) determined by the sum of these values over all times t. Here, the Viterbi algorithm is The state of the trellis diagram to be operated is the combination of transmission signals (St, 5t
-1,..., 5t-L-N+1) is determined. In this sense, below we will discuss the combination of transmitted signals (St, 5t-150
01,5t-L-N+1) will simply be called a state. The series of maximum likelihood states obtained by the Viterbi algorithm over all times becomes the transmission signal sequence estimated value.

Now, in the sequence estimation device of this patent application No. 2-203436, the transmission path response vector bt,
When calculating Is, the matrix (StSt) and the matrix S
, is being calculated. Therefore, for a state in which the matrices S, S, or matrix St become singular, the transmission path response cannot be determined as is. If the channel response becomes unstable, branch metric calculation cannot be performed and the Viterbi algorithm cannot be operated. On the other hand, generally the transmission path fluctuation is the transmission signal matrix 5t
(7) Time interval (L+N-1)T(
T has the property that it is negligibly small in the transmission symbol time interval). Using this property, matrix 5
As the transmission path response estimation value vector h,,is for a state where tTSt or matrix S is singular, the transmission path response estimation value vector ht-1,Is in the previous time state of the surviving sequence that transitions to that state is substituted. be able to. That is, basically, the sequence estimating device of the present invention has a matrix 5tT
For a state in which S or matrix S is singular, use the transmission channel response estimate for the state that the surviving sequence took at the previous time, and create a state in which matrix StT'S or matrix S is non-singular. For this case, the transmission channel response obtained by multiplying the received signal by the inverse matrix of the matrix S is set to be used. As a result, it is possible to always stably supply a transmission line response estimate to the branch metric calculation circuit, and by operating the conventional Viterbi algorithm based on this, it can stably track rapidly changing transmission lines. A sequence estimating device that can perform the following steps can be realized.

Furthermore, the sequence estimating device disclosed in Japanese Patent Application No. 2-203436 only calculates the channel response estimated values hourly and does not utilize the correlation between the estimated values at each time. Therefore, the transmission path response estimate for each state changes greatly at each time due to noise, which may lead to errors in sequence estimation. However, if noise is ignored, the transmission path response originally has the property of continuously changing according to the Doppler frequency. That is,
Although it is time-varying, the change in channel response is generally gentler than the symbol transmission interval, and there is a high correlation between the channel responses at each time. Therefore, in the second sequence estimation device according to the present invention, a conversion operation that utilizes the high correlation between transmission path responses is newly introduced. That is, for each state, a new current estimated value is obtained by this conversion operation from the transmission path response estimated value at a past point in time and the current estimated value obtained by solving the transmission path equation. This results in
It is possible to prevent the phenomenon in which the estimated transmission path response value changes greatly from time to time, and to obtain a stable and reliable estimated transmission path response value with respect to the true state. Further, by smoothing the estimated transmission channel response, noise in the transmission channel estimation process can be effectively removed. the result,
A sequence estimation device that can stably follow a rapidly changing transmission path can be realized.

In the following, a case will be explained in which a simple average operation is selected as the conversion operation. When calculating the estimated value at the current time by averaging operation from multiple past transmission channel response estimates, the weights will change as shown in Fig. It is often more accurate to select a weighted average that follows a negative exponential function. FIG. 3 is a diagram showing changes in the weights for the transmission path response estimated value vectors h, , Is at time t in the weighted averaging operation after the time. This weighted average, in which the weights follow a negative exponential function, is based on, for example, the RLS algorithm suitable for time-varying systems (for example, Reference 3: Cowan, Grand, Adaptive Filters, Prentice-Hall, 1985), the judgment errors at each time related to control of filter coefficients are weighted by the power of the forgetting coefficient and added.

Additionally, when performing weighted averaging where the weights are negative and exponentially attenuated, it is known that a filter with an IIR configuration can achieve its characteristics with a smaller number of taps than a filter with an FIR configuration (for example, as mentioned above). Reference 3). In the series device of the present invention, a transmission path response storage circuit 105, a matrix switch 106, a first multiplier 109, an adder 1
10. The second multiplier 111 corresponds to configuring an IIR type filter, and weighted averaging can be realized with a small hardware scale. Also, matrix switch 1
06 receives the connection information of the surviving series from the Viterbi processor and controls the correspondence between the transmission path response estimates at multiple times in the past and the transmission path response estimates at the current time. If the number of controllers is small, the scale of this control device is also small.

(Example) Next, the present invention will be described with reference to the drawings. FIG. 1 shows an embodiment of a sequence estimation device according to the present invention.

In the following, an example will be described in which only the transmission path response estimated value at the previous time is stored as the past transmission path response estimated value string stored in the transmission path response storage circuit 105.

Receiver input r at time t provided to input terminal 101.

is stored in the register 102 and sent to the branch metric calculation circuit 107. N received signals stored in the register 102 from time t-N+1 to time t are input to the transmission path response calculation circuit 103. The transmission path response calculation circuit 103 calculates vectors h, , Is for each state according to equations (6) to (7). In the transmission path response calculation in equations (6) to (7), matrix (S, TSt)-1, matrix 3. -1 are determined only by the transmitted signal, so all states, i.e. (St, 5t
-11011, 8l-L-N+1) may be calculated in advance for all possible combinations of transmission signals, and the results may be stored. For combinations of transmitted signals in which the matrix 5tTSt or the matrix S is singular, the transmission path response calculation circuit 103 sets a predetermined value (for example, 0
), and informs the weighting coefficient control circuit 104 that the vectors h, , Is are indefinite. Here, a transmission channel response estimated value obtained when the response is not indefinite for each state is called a valid estimated value. The weighting coefficient control circuit 104 first checks whether the vectors 't and Is are indefinite for all states. Next, for an undefined state, the first weighting coefficient output to the first multiplier is set to 0, and the second weighting coefficient is set to 0.
The second weighting coefficient output to the multiplier is set to 1. Here the second weighting factor is 1 for each state.
The reason why only the transmission path response estimated value at the previous time is handled is as the past transmission path response estimated value string stored in the transmission path response storage circuit. For non-indeterminate states, two predetermined weighting coefficients, for example 0.6 and 0.4, are set as the first and second weighting coefficients, respectively. Note that even for a non-indeterminate state, the first and second Each weighting coefficient is set to 1.0, and the transmission path response calculated by the transmission path response calculation circuit 103 at the current time is used as is.

Now, the first multiplier 109 performs the first multiplier 109 for all states.
The weighting coefficient of is multiplied by the transmission path response vector ht, Is at the current time calculated by the transmission path response calculation circuit 103.

The second multiplier 111 multiplies the rearranged transmission path response estimate vector ht-1' given by the matrix switch 106 for all states by the second weighting coefficient. Then, the output of the first multiplier 109 and the output of the second multiplier 111 are added by the adder 110, and for each state, the transmission path response estimated value vector h at the current time is
, ′ is obtained. Since the first weighting coefficient calculated by the transmission path response calculation circuit 103 is set as O1 and the second weighting coefficient is 1, the transmission path response estimated value vector h at the current time,
′, the transmission line response estimated value vector h at the previous time is automatically
t=i'' is taken. Also, in order to prevent the transmission response estimated value vector salt' at the current time from divergence, the first
The sum of the weighting coefficient and the second weighting coefficient must always be set to be 1 or less.

The transmission path response estimated value vector ht' at the current time obtained for all states is stored in the transmission path response storage circuit 105 and supplied to the branch metric calculation circuit 107. Here, the transmission path response stored in the transmission path response storage circuit 105 is an estimated value after undergoing a conversion operation. The transmission path response estimated value vector ht-i' at the previous time stored in the transmission path response storage circuit 105 is the matrix switch 1
06. In the adder 110, the matrix switch 106 calculates the transmission line response estimated value vector ht at the previous time regarding the previous time state taken by the sequence that transitions to that state for the transmission line response estimated value vector ht at the current time of each state.
Viterbi processor 10 so that -1' corresponds correctly.
The transmission path response estimated value vector ht-1' at the previous time is rearranged according to the connection information of the surviving sequence given by No.8.

Further, the branch metric calculation circuit 107 calculates the branch metric determined by equation (8) for all possible states based on the current transmission path response estimated value vector h,' given by the transmission path response storage circuit 105. Calculate separately.

The branch metric calculation circuit 107 outputs the branch metric calculated for each state to the Viterbi processor 108. The Viterbi processor 108 outputs to the output terminal 112 the transmission signal estimated value at a certain time of the series for which the sum of all the times of the metric of equation (8) is the minimum, and also outputs the connection information of the surviving series to the matrix switch 106. supply The operation of the Viterbi processor 108 is similar to that of the sequence estimating device in Reference 1.2, so the details will be omitted.

In the above embodiment, an example has been described in which only the transmission path response estimated value vector ht-1' at the previous time is stored as the past transmission path response estimated value sequence stored in the transmission path response storage circuit 105. Vector ht-1' as response estimate
, vector fence, etc. are stored at multiple times in the past, and second weighting coefficients corresponding to them are stored as a coefficient group, and used for conversion operations of the transmission path response. Good too. Furthermore, in the above embodiment, the weight control circuit 1
In 04, the weighting coefficient is controlled only depending on whether or not the transmission path response estimate value is indefinite, but in addition to the case where the transmission path response estimate is uncertain, there is also a case where a transmission path response that is undesirable for the branch metric calculation circuit 107 is supplied from the transmission path response calculation circuit 103. The weighting coefficient may also be set adaptively. Furthermore, in the above embodiments,
The explanation was based on a transmission line where there are multiple delayed waves relative to the main wave.A transmission line where there are multiple preceding wave responses to the main wave, where preceding waves and delayed waves coexist. It is also clear that the sequence estimating device of the present invention is effective for such transmission paths.

(Effects of the Invention) As described in detail above, the present invention uses a small amount of memory and a small device scale, and can always stably estimate a transmission path for each sequence even in the presence of noise, and can handle rapidly changing transmission paths. A sequence estimation device that follows can be provided.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an embodiment of a sequence estimating device according to the present invention, FIG. 2 is a block diagram showing a conventional sequence estimating device, and FIG. 3 is a block diagram showing a conventional sequence estimating device.
FIG. 4 is a diagram for explaining the transmission channel response, and FIG. 5 is a diagram for explaining the transmission channel model. 101.201.501...Input terminal, 102.20
2. ...Register, 103.203...Transmission path response calculation circuit, 104...Weighting coefficient control circuit, 105...
Transmission path response storage circuit, 106-1. Matrix switch, 107.204... Branch metric calculation circuit,
lO8,205... Viterbi processor, 109...
First multiplier, 110.504.505... adder,
111...Second multiplier, 112.206.506.
... Output terminal, 401... Main wave response, 402... Delayed wave response, 502... Delay element, 503... Multiplier.

Claims (1)

[Claims] a register that stores a plurality of sample values of a received signal; a transmission path response calculation circuit that inputs the plurality of sample values from the register and estimates a transmission path response at the current time for each of the plurality of sequences; a weighting factor control circuit that calculates a first weighting factor and a second group of weighting factors for each sequence in accordance with the transmission path response for the plurality of sequences obtained by the transmission path response calculation circuit; a first multiplier that multiplies the weighting coefficient of and the output of the transmission path response calculation circuit;
a second multiplier that multiplies the second group of weighting coefficients by the past sorted transmission channel response estimation value sequence; and adds the output of the first multiplier and the output of the second multiplier. a transmission path response storage circuit that stores each of the adder outputs as a transmission path response estimate value at the current time for a plurality of sequences; and a transmission path response storage circuit that stores past transmission path response values as outputs of the transmission path response storage circuit. The correspondence relationship between the response estimation value sequence and the transmission line response at the current time given as the output of the transmission line response calculation circuit is checked from the connection information of the surviving series for each of the plurality of sequences, and the correspondence relationship is determined according to the correspondence relationship. a matrix switch that rearranges the past transmission path response estimated values and outputs the past rearranged transmission path response estimated values;
a branch metric calculation circuit that calculates a virtual reception signal point for each series based on the transmission path response estimate at the current time given by the adder and calculates a distance from the sample value of the reception signal; A sequence estimating device comprising: a Viterbi processor that receives an output from a metric calculation circuit, determines a received signal using a Viterbi algorithm, and outputs connection information of the surviving sequence to the matrix switch.