DE10205742C1 - Transmission channel pulse response estimation device has channel pulse response estimation passed through correction filter - Google Patents

Abstract

The device has an estimation stage (10) for estimation of the channel pulse response (KIA) using received data from the transmission channel, coupled to a device (14) for calculation of the exact autocorrelation function of the estimated channel pulse response, a second device (12) estimating the autocorrelation function of the channel pulse response from the received data. The exact and estimated autocorrelation functions are compared for determination of filter coefficients for a correction filter (16) for the estimated channel pulse response. An Independent claim for a method for estimation of the pulse response of a transmission channel is also included.

Description

The present invention relates to the news transmission and especially on possibilities of treasure tion of a transmission channel, for example a Ent distortion in a receiver, or to direct a equalization.

In transmission systems that vary widely channel is an estimate of the channel's own important to ensure the variable channel properties to be able to take countermeasures. Such a great variety transmission channels occur in particular if the transmission channels free space transmission channels are, for example, how they are available in mobile communications.

The channel estimate is necessary for this, for example any channel equalization at the receiving end using a Undo equalizer.

For this reason, channel receivers are used in modern receivers process used. These channel estimation methods are based insists that a known broadcast sequence from the broadcaster Transmission channel is transmitted to the receiver, and that evaluated the reception sequence with knowledge of the transmission sequence to determine the channel's transfer function. Such known broadcast episodes are also training programs sequences.

For correct equalization it is necessary to impulse the channel word, i.e. the result of an inverse Fourier Transformation of the channel transmission function, if possible too to know at any time. This applies all the more to the  The process of establishing a connection between a transmitter and a receiver over the transmission channel.

However, this means that the channel impulse response is ongoing must be estimated.

As an alternative to the estimation methods based on training sequences, i.e. known reference sequences based on the Channel impulse responses have also been appreciated recently Reference sequence-free estimation procedures for channel identification on, d. H. to estimate the channel impulse response of the channel or to estimate the autocorrelation function of the Channel impulse response has been set out. For this, Robert Klinski, "Blind System Identification and Equalization for OFDM-Systems ", Alcatel Workshop on Equalization, Antwerp, November 1999.

When estimating the usually complex channel can the complex channel impulse response or the complex channel transfer function can be estimated. To estimate the Be carrier frequency response and the phase frequency response of a Certainly complex channels are different devices gene and procedures known. For this purpose, for example Simon Haykin, "Adaptive Filter Theory", Prentice-Hall 1996, directed.

The so-called MMSE approach as a representative of a known channel estimate procedure received (MMSE = Minimum Mean Squared Error Estimation = Estimation using the method of minimal mean quadratic error). Alternatively, you can however also a so-called least squares estimate (LS Estimate).

In the MMSE approach, one is typically used for channel estimation known reference sequence used. This reference sequence will transmitted via an unknown channel h. The recipient  then receives a receive sequence, which in the case of an idea len channel correspond directly to the send reference sequence would. Due to changes in time and space other channels, such as those used in digital Radio transmission occur, however, the channel is very strong distorting, so that correct reception with limited Transmission power can only be guaranteed if the Channel can be estimated to equalize the reception sequence ren what can usually be done with an equalizer is achieved. Through the channel estimation, the Be estimated channel and phase frequency response of the channel, to control the equalizer so that the Ver distortion of the channel can be compensated.

If no channel estimation were used, a could good reception, for example with a bit error rate below half a certain threshold only at a very high hen signal / noise ratio be ensured what either corresponds to a high transmission power, or one relatively low user data rate.

The following equation shows the theoretical relationship between the known input sequence x and the known output sequence y, where h represents the channel impulse response. The output sequence y results from the convolution of the channel impulse response with the input sequence. The channel impulse response can then be calculated as follows by changing the system equation:

y = h⊗x; ⇒ h = r * | yx. [R _xx ] ^-1 ;

The circled cross in the above equation represents the convolution operation. R * | yx is the conjugate cross correlation. The following boundary conditions must be observed when estimating MMSE:

• a) In order to determine the desired channel impulse response h, the input sequence x must be known at the receiving end.
• b) The longer the re., the smaller the estimation error sequence x is.

Condition (b) in particular has the following problems:

• a) As the length of the reference sequence increases, the ef decreases fective data rate of the transmission.
• b) The inversion of the autocorrelation matrix, which is based on is necessary due to the above equation is at long reference sequences problematic.

In other words, each channel estimate produces e on the other hand the question of when the channel impulse response is sufficient is accurately estimated, d. H. the input sequence was x long nug. On the other hand, there is the question of how that already exists existing estimate can be improved without having to a digital transmission can take a very long time te input reference sequence must be sent, which value takes full bits for the user data transmission.

Especially with highly time-variant channels, such as those B. Another problem arises in mobile communications added. Even if the user data bandwidth, i.e. the Ver Ratio of the amount of user data to the amount of known Reference data is not a problem, but can only be very much short reference sequences are used to determine the current Channel properties of the rapidly changing channel to be able to determine. You can see this using a fah imagine cars in which someone has a cellphone  served. The transmission channel between this cellular network advises and a base station changes especially then very strong when the car is moving and at high and drives past low houses or trees and in particular z. B. must cross under a bridge. Here, the Change channel very quickly and very sustainably, see above that the use of long reference sequences is limited It makes sense because the MMSE process is a very precise one Channel estimation allowed, but the channel is in time, in which the exact channel estimate has been calculated, such as who has changed so that little benefit remains.

It can thus be summarized that in particular strongly time-variant channels, but also not so strong time-variant channels in which a maximum transmission short reference sequences must be used to the current channel own be able to determine as accurately and quickly as possible or to ensure a high user data rate and little "Waste" send bits for reference sequences.

On the other hand, however, this also means that the estimate necessarily has an estimation error, wel Usually unknown and based on a short reference follow is relatively large.

The problem with using short reference sequences is that short reference sequences only provide an estimate of the channel impulse response lead the actual channel impulsant word approximates, but has an estimation error is. The longer the re., The smaller the estimation error sequence is, and disappears for infinitely long reference sequence, while, in the opposite case, the estimate error increases, the shorter the reference sequence becomes.  

In the case of the reference data-free estimate, the channel impulsant word z. B. in OFDM systems, can not particularly agile algorithms are used that essentially Chen based on the estimation of the autocorrelation function of the Ka based impulse response. However, their disadvantage is the right relatively slow rate of convergence. To increase this can also use more powerful estimation methods det be such. B. the so-called subspace method. prob However, the lematic of this very complex method is high complexity, that of a practical hardware implementation tion especially for inexpensive recipients when z. B. of digital broadcasting is in the way.

Especially in the field of digital broadcasting, where Emp catchers should be sold as bulk items, play even small price differences play a decisive role, whether there is a product on the market or whether a product is not used. Similar requirements gen exist in the mobile phone area, in which a hard Price competition takes place in that expensive and complex hardware implementations in mobile phones to be considered skeptical.

The object of the present invention is a less complex and efficient concept for estimating the channel impulse response of a transmission channel or To equalize a signal.

This task is accomplished by a device for estimating egg ner channel impulse response according to claim 1, by a Method for estimating a transmission channel according to Pa claim 13, a receiver device according to patent claim 14, a method for receiving according to claim 15, a device for equalization according to claim 16 or a method for equalization according to claim 27 ge solves.  

The present invention is based on the finding that that the Wiener-Hopf equation, both in estimating the channel impulse response of a transmission channel z. B. too Equalizer purposes as well as when directly equalizing one Received signal, d. H. when calculating the equalizer coefficient cient occurs when z. B. based on the MMSE criterion is placed around an error between a defective tenn output size and an error-free or less error Minimize the target size using egg Classic prediction filter with delay element as an auxiliary filter to the actual correction or equalizer computation filter, which is called a "prediction filter" is considered without delay element to solve.

With the problem of estimating a channel impulse response transmission channel, as for example through any Known estimation method that can be free of reference data or not, the Wiener Hopf equation is used to to "afterwards", ie according to the estimate, using a Filter correction filter to estimate the channel im improve pulse response, the correction filter coefficient according to the invention by solving the Wiener-Hopf Equation can be calculated. Here is an estimate of the Channel impulse response of the transmission channel on the one hand and an estimate of the channel correlation function Pulse response, on the other hand, is necessary.

According to the invention, the estimated channel impulse response the autocorrelation function of the estimated channel pulse response calculated as an exact autocorrelation function the estimated channel impulse response is called as calculated exactly and deterministically from the estimated values becomes.

The filter coefficients of the correction filter are then calculated using auxiliary filter coefficients, the auxiliary filter coefficients are chosen so that a product of a vector of the auxiliary filter coefficients  and an auto correlation matrix that is based on the exact auto correlation function of the estimated channel impulse response is directed to the exact autocorrelation function is the estimated channel impulse response. This auxiliary Filter coefficients are obtained by solving the known How ner-Hopf equation or by solving the well-known Jule Walker's equation calculates and describes a classic Prediction filter in that the prediction error between the actual signal and the predicted sig nal according to the MSE or MMSE criterion, d. H. according to the criterion rium of least squares, is minimal.

After filtering the estimate of the Channel impulse response not returned to previous values this prediction approach alone is sufficient not from. The correction filter coefficients for the dimension The correction filter is therefore used of the auxiliary filter coefficients, the correction Door filter coefficients are chosen so that a product egg nes vector of correction filter coefficients and auto correlation matrix based on the exact autocorrelation function of the estimated channel impulse response is derived, equal to the channel's estimated autocorrelation function impulse response is.

In a preferred embodiment of the present Invention that does not require matrix inversion is the Be calculation of the auxiliary filter coefficients iteratively under Ver application of the well-known Levinson-Durbin recursion leads. This means that the elaborate inversion of the autocorrela tion matrix based on the exact autocorrelation function the estimated channel impulse response is bypassed become.

In a further preferred embodiment of the above lying invention will also calculate the corrective door filter coefficients also performed iteratively, and using a new iteration rule, the  on the auxiliary filter coefficient, the exact autocorrela tion function and the estimated autocorrelation function based.

In addition, it is preferred to both the channel pulse answer yourself as well as their auto-correlated reference valued free. Because of the subsequent filtering by the correction filter according to the invention, the estimate channel impulse response and autocorrelation function on the channel impulse response to be relatively rough, since the estimate errors due to the subsequent correction filtering according to the present invention can be significantly reduced.

The concept according to the invention is particularly in this regard advantageous that the rate of convergence by a subsequent filtering of the estimates is increased. In an expressed in words, the estimation error is subsequently added Lich, d. H. through the correction filtering, minimized.

The concept according to the invention is particularly in this regard advantageous that the requirements for the estimation method themselves are not particularly high. So if Estimation methods used with reference data are relative short reference sequences are used, which on the one hand Speed of the whole estimate and secondly the usable bandwidth of the transmission channel favorably influenced influ-.

On the other hand, reference data-free estimation methods are used det, relatively simple methods can also be used here be because the requirements for the estimation quality are not significantly high, especially since a later one anyway che filtering the channel impulse response by the fiction appropriate correction filter is carried out.

Another advantage of the present invention is in that in general the security of convergence, i. H. that an estimation method converges at all, with coarser ones  Estimation method is higher than with more sophisticated estimators drive. The two-stage channel estimation concept according to the before lying invention therefore allows the application of Estimation method for the channel impulse response and for the Au tocorrelation function of the channel impulse response, which is a si convergence, but typically one slow speed of convergence, d. H. how fast one Convergence is achieved. By the later che filtering by the correction filter is however at ei convergence process speed increases, so that a channel estimation concept before lies, on the one hand, a secure convergence in itself and on the other hand has a high rate of convergence.

The concept according to the invention is based on the theory of linear prediction to derive a solution to the determ measurement of the filter coefficients of the correction filter Using the Wiener Hopf equation. The Wiener Hopf However, equation is based on knowledge of the (unknown) Channel impulse response. Since this is unknown - this value should be estimated - provides the present Erfin an approximation, where as statistical information the autocorrelation function of the channel impulse response goes. This is preferably estimated without reference data, low-cost algorithms exist for this.

The concept according to the invention is advantageous in that that there is an increase in the rate of convergence by one Can reach factor 2.

The concept according to the invention can, however, the inversion ner autocorrelation matrix in each iteration step demand. However, it is preferred to use one on the Levinson Durbin recursion based new iteration rule use so that the correction filter coefficients without Matrix inversion can be calculated iteratively. The new recourse sion can be used to iteratively solve the general Viennese Hopf equation, such as those related to  a reference data-based determination of the equalizers coefficients based on the minimum mean criterion Error square occurs. After the Effort to determine the filter coefficients quadratic with the order of the correction filter, d. H. with the number of correction filter coefficients, it will rise before moves to keep the order of the correction filter in the frame. In particular, it has been found that the correction file ter only half as far in terms of its order large as the order of the transmission impulse response channel must be dimensioned in order to accept table Improve the estimate of the channel impulse response through the correction filter.

Preferred embodiments of the present invention are referred to below with reference to the enclosed Drawings explained in detail. Show it:

Fig. 1 is a block diagram of an inventive device for estimating a channel impulse response;

Fig. 2 is a more detailed illustration of the means for calculating the correction filter coefficients;

Fig. 3 is a more detailed illustration of the calculation of auxiliary filter coefficients of a prediction filter;

Figure 4 is a detailed illustration of the calculation of the correction filter coefficient using the auxiliary filter coefficients.

Fig. 5 is an overview of the sizes used in the presen- tation of Figures 3 and 4.

FIG. 6a, a classical prediction filter;

FIG. 6b is a "prediction" without delay;

FIG. 7 is a list of the explanation of the terms Au tokorrelationsmatrix and Autokorrelationsfunkti one;

Fig. 8 is an illustration of a receiver circuit having equalizing function, in the estimation approach according to the invention is implemented;

Fig. 9 is a schematic representation of a distorting transmission channel with a downstream OF INVENTION dungsgemäßem equalizer; and

Fig. 10 is a block diagram of an inventive device for equalizing a second signal, which is a distorted version of a first signal.

Fig. 1 shows a basic block diagram of the inventive device for estimating a channel impulse response of a transmission channel. The device according to the invention comprises a first estimating device for estimating the channel impulse response (KIA) of the channel, which is designated by 10 in FIG. 1. The device according to the invention further comprises a second estimating device for estimating the autocorrelation function of the channel impulse response of the transmission channel, which is denoted by 12 in FIG. 1. The first estimator delivers an estimated channel impulse response on the output side, while the second estimator has the estimated autocorrelation function of the channel impulse response on the output side. The estimated channel impulse response provided by device 10 is fed to means 14 for calculating the autocorrelation function of the estimated channel impulse response. On the output side, the device 14 provides the exact autocorrelation function of the estimated channel impulse response. The estimated channel impulse response, which is provided by the device 10 , is further fed into a correction filter 16 , which provides an improved estimated channel impulse response of the transmission channel on the output side. The correction filter coefficients of the correction filter 16 are calculated by means of a device 18 for calculating the correction filter coefficients and fed to the correction filter 16 .

The transmission channel, from which the data fed into the first treasure device 10 and the second treasure device 12 originate, can, for. B. be a free space transmission channel, as occurs for mobile phones or radio applications. Such a channel is particularly variable, ie it has a strongly varying channel impulse response or, in other words, a strongly varying transfer function which is known to be the Fourier transform of the channel impulse response. The improved estimated channel impulse response of the transmission channel provided by the correction filter 16 can be used to perform receiver-side equalization, as will be explained later with reference to FIG. 8.

The device 18 for calculating the correction filter coefficients is designed to use auxiliary filter coefficients which are selected such that a product of a vector of the auxiliary filter coefficients and an autocorrelation matrix which is derived from the exact autocorrelation function of the estimated channel impulse response, which is provided by the device 14 is equal to the exact autocorrelation function of the estimated channel impulse response, which is also provided by the device 14 .

The device 18 for calculating the correction filter coefficients is also designed to calculate, using the auxiliary filter coefficients, the correction filter coefficients, which in turn are chosen such that a product of a vector of the correction filter coefficients and the auto-correlation matrix is derived from the exact autocorrelation function of the estimated ones Channel impulse response derived from device 14 is equal to the estimated auto-correlation function of the channel impulse response provided from device 12 .

A detailed description of the device 18 for calculating the correction filter coefficients is discussed below with reference to FIG. 2. The device 18 comprises a first sub-device 18 a for calculating auxiliary filter coefficients of a prediction filter, which supplies the auxiliary filter coefficients p _{m of} the prediction filter. The device 18 of FIG. 1 further comprises a second sub-device 18 b, which he averages using the auxiliary filter coefficients p _m, the correction filter coefficients according to the equation shown in block 18 b.

The vector p _m is a vector of the auxiliary filter coefficients, the auxiliary filter having the order m, ie a number m of auxiliary filter coefficients. As will be explained below, the auxiliary filter is a classic prediction filter. The device 18 a also receives on the aisle side the exact autocorrelation function r _{x, m} calculated by the device 14 of FIG. 1 of the estimated channel impulse response. The vector r _{x, m} comprises a discrete-time representation of the autocorrelation function of the estimated channel impulse response, where m indicates the order of the autocorrelation function, ie the number of the most significant element of the autocorrelation function, which is greater than or equal to a predetermined lower threshold. As is known, the auto-correlation function of the channel impulse response of the transmission channel has a length, ie a number of elements, which can vary from transmission channel to transmission channel. In an estimate of the channel impulse response, in which the length of the auto-correlation function of the channel impulse response is not known from the outset, who usually estimates more elements of the channel impulse response or the auto-correlation function of the channel impulse response, which then reveals that from a certain element number the elements are very much accept small values and can therefore be set to 0.

The device 18 a for calculating auxiliary filter coefficients of FIG. 2 is also based in principle on the autocorrelation matrix R _{x, m} , this autocorrelation matrix, as will be shown later, can be constructed from the autocorrelation function r _{x, m} .

The determination equation in block 18 a of FIG. 2 is referred to in the art as the Wiener-Hopf equation, and is described, for example, in the specialist book "Digital signal processing, filtering and spectral analysis with MATLAB exercises" by KD Kammeyer, K. Kroschel, Teubner Study books electrical engineering, Teubner Verlag, treated. The Wiener-Hopf equation results from the solution of the prediction problem if the aim is to minimize the performance of the prediction error. These relationships are known in the art and presented in the specialist book "Nachrichtenübertragung", KD Kammeyer, Verlag BG Teubner Stuttgart, Information Technology series.

The device 18 b for calculating the correction filter coefficients is also based on a Wiener-Hopf equation, the correlation function r _m , which is the estimated autocorrelation function of the channel impulse response provided by the device 12 , being different from the exact autocorrelation function r _m is. In particular, the autocorrelation function r _m includes a coefficient r (0) that does not occur in r _{x, m} . This is due to the fact that, as will be explained with reference to Fig. 6a and Fig. 6b, the correction filter used for filtering the estimated channel impulse response, having no delay element, which contrast in Ge is for classic prediction ,

For a more detailed explanation of this fact, reference is now made to FIGS. 6a and 6b. FIG. 6a shows in principle a prediction filter P (z) 60 and is well known, comprises a delay member 62. It is further assumed that x _{k is} a steady-state and mean-free process, the estimated value of which is denoted by a queue above x _k , and whose estimated value is known to be sought based on the past value. This problem arises for example in parametric spectral estimation. This estimation problem is solved in technology with the aid of a linear prediction filter, the coefficients p of which are to be determined. For this purpose, the input values x _k and the estimated output values are compared with one another using the summer 64 . The prediction filter, which is to be minimized according to the MSE criterion, results on the output side of the summer. A set of filter coefficients p is therefore sought which result in the performance of the error being minimal. The solution to this problem is provided by the Wiener-Hopf equation, which is given in block 18 a in FIG. 2.

Fig. 6b shows a "prediction filter", which corresponds to the correction filter 16 , and whose coefficients are marked with f (z) be, the prediction filter 66 is now, as it were, alone, namely without a time delay element as in Fig. 6a. In contrast to the prediction filter structure shown in FIG. 6a, no continuously received coefficients of a white power process are processed in the filter structure shown in FIG. 6b. However, the estimated values of the channel impulse response and the channel impulse response itself can be treated as dimensionless energy signals, ie signals with limited energy.

Provided that the channel impulse response has been excited with a white, uncorrelated and mean-free process when determining the estimated value for the channel impulse response, the filter coefficients f can also be determined from the minimization of the error at the output of the sum element 68 . Here expectation values occur, which are represented by an operator E {. , .} in technology. There are several ways of calculating an expected value, for example simply by averaging over time or other measures known to eliminate noise processes. The expected values can be understood as system correlation functions. Although it is not always stated in the following, it should be pointed out that a correlation with the channel impulse response, which is not a signal sequence, is a so-called energy correlation, which, for. B. is set out in a block 50 of FIG. 5 by the superscript capital letter E.

The approach for calculating the prediction filter f 66 from FIG. 6b again results in a Wiener-Hopf equation, which, however, as has been explained above, in contrast to the Wiener-Hopf equation of the prediction filter 60 from FIG. 6a has a coefficient r (0) has the autocorrelation function on the right side of the equation, which does not occur in the Wiener-Hopf equation for the classic prediction filter in FIG. 6a due to the delay element 62 .

Before going into more detail on FIGS. 3 and 4, in order to present an iterative solution to the Wiener-Hopf equations listed in blocks 18 a and 18 b, which does not require an explicit matrix inversion, i.e. inversion of the autocorrelation matrix R _{x, m} , whether with reference to FIG. 7 again, the relationship between autocorrelation matrix and autocorrelation function set.

The definition equation of the Autokor relations matrix is shown in a block 70 , where y z. B. is an output signal of a transmission channel. The operator E {. , .} denotes, as has been carried out, the expected value which, for. B. can be calculated by time averaging or other measures to eliminate noise processes. The expression (...) ^H represents the Hermitian operator, which is known to include a transposition and a conjugation. A block 72 gives a definition for a discrete representation of the autocorrelation function r _{yy of} the signal y, the autocorrelation function in the digital signal processing typically being specified as a sequence of discrete values of the autocorrelation function. For this reason, the autocorrelation function is often referred to as the autocorrelation sequence. The autocorrelation function, as shown in block 72 , can be thought of as a sequence of samples of the continuous auto-correlation function, which is plotted as a function of τ and has been sampled at certain τ values.

It is also known in the art that an autocorrelation matrix can be constructed from the autocorrelation function, as shown in block 74 of FIG. 7. It can be seen here that the first line of the autocorrection matrix R _{yy is} identical to the autocorrelation function of block 72 . The remaining lines of the auto correlation matrix are calculated by cyclically interchanging the elements of the autocorrelation function and by entering the conjugate complex of the cyclically interchanged value below the main diagonal. So it is evident that each row on the one hand and each column on the other hand includes the complete autocorrelation function, but depending on the row and column number in a different order and correspondingly conjugated complex.

In the following, Fig. 3 is discussed in order to illustrate a solution of the Wiener-Hopf equation shown in the device 18 a, which is iterative and does not require an explicit inversion of the autocorrelation matrix R _xm , which is derived from the exact autocorrelation function of the estimated channel impulse response is. As has already been explained, this Wiener-Hopf equation, which is shown in a block 30 of FIG. 3, provides the filter coefficients p _{m of} a classic prediction filter according to the structure of FIG. 6a, that is to say a prediction filter with a delay element. A Levinson-Durbin recursion can be used for such a vector-matrix equation, as it is shown in block 30 , as described in Appendix A5 in KD Kammeyer, "News Transmission", published by BG Teubner Stuttgart.

The Levinson-Durbin recursion comprises a function block 31 for calculating the mth auxiliary filter coefficient of an auxiliary filter according to FIG. 6a of the order m + 1, which is referred to in the functional block 31 as p _{m + 1} (m). The rule given in function block 31 for calculating this auxiliary filter coefficient is the main iteration rule. On the basis of the main iteration rule given in function block 31 , the auxiliary filter coefficients of the auxiliary filter of the order m + 1 with a lower number, ie with the numbers 0,. , ., m - 1, be calculated. This calculation rule is shown in a function block 32 .

The iteration calculation is started with an iteration start value for the order index or iteration index m = 0, which, as shown in FIG. 3 in a function block 33 , is the quotient from the first element of the exact auto-correlation function of the estimated channel impulse response, which is provided by the device 14 of FIG. 1 is determined (r _x (1)) and its 0th element (r _x (0)) is determined. p ₁ be thus describes the "set" of auxiliary filter coefficients. Since p ₁ a filter for the order m = 1, this filter has only coefficients of the single Koeffi shown in the functional block 33rd

Now the order index is incremented to m = 1. In accordance with the iteration rule given in function block 31 , the first auxiliary filter coefficient of an auxiliary filter of order 2 is now calculated. The 0th auxiliary filter coefficient of the auxiliary filter of order 2 is then calculated using the calculation rule given in function block 32 . A decision block 34 then examines whether m is to be incremented further. Furthermore, in one exemplary embodiment, the auxiliary filter coefficient vectors can be stored, as shown in a function block 35 , in order then, as shown in a function block 36 , to increment the iteration index m by 1. In the example dealt with here, m now has the number 2. The function block 31 now gives more the specification for calculating the second auxiliary filter coefficient of an auxiliary filter of order 3. The first and the 0th auxiliary filter coefficient of the auxiliary filter of order 3 are then calculated in function block 32 . It should be noted that, as shown in the lower half of function block 32 , the vector p * | m, r is the conjugate coefficient vector in reverse order, as represented by the index r representing reverse stands.

This iterative procedure is repeated until the auxiliary filter coefficients for the auxiliary filter of the order sought m = max have been calculated. The order of the auxiliary filter corresponds to the order of the correction filter 16 of FIG. 1 that is ultimately to be calculated . If the calculation of the different auxiliary filters of different orders is carried out independently of the calculation of the actual correction filter coefficients, the auxiliary filter of the order m = max as well as the compute th subsidiary filter of lower orders are stored, since this is represented as reference to FIG. 4, are needed th in an iterative calculation of the Korrekturfilterkoeffizien, wherein the dahinge in Fig. 4 shown ite rative calculation of the correction filter coefficients based It is advantageous that no explicit inversion of the autocorrelation matrix R _{x, m is} required here either.

When the Levinson-Durbin recursion shown in FIG. 3 has ended, ie when all auxiliary filters of the order m sought have been calculated, the iteration process is ended (block 37 ).

In the following, with reference to FIG. 4 in an iterative solution of the Wiener-Hopf equation 40 in FIG. 4 is received, which has the advantage that no explicit matrix inversion of R _{x, m} is required.

The iterative solution according to the invention of the Wiener-Hopf equation 40 from FIG. 4 in turn comprises a function block 41 for calculating the mth filter coefficient for a correction filter of the order m + 1. The correction filter coefficients with a lower number can be determined by a function block 42 and be determined in particular by the calculation rule given therein. The starting value of the iteration, ie the vector of filter coefficients for a filter of order 1, ie if the iteration index m = 0, is given by block 43 , in particular by the quotient of the 0th element of the estimated autocorporation function of the channel impulse response , which is provided by the device 12 , and the 0th element of the exact autocorrelation function of the estimated channel impulse response, which is provided by the device 14 .

The iteration rule shown in FIG. 4 further includes a decision block 44 for determining whether the iteration has ended, in order then to output the vector of the correction filter coefficients and to supply the correction filter 16, the correction filter having the desired dimension m = max. If max has not yet been reached, m is incremented in an increment block 46 in order to run through the iteration loop.

It can be seen from the equations shown in blocks 41 and 42 that the auxiliary filter coefficients in conjugate complex form are required at corresponding points. If the iteration calculation shown in FIG. 3 is carried out separately from the iteration calculation shown in FIG. 4, the auxiliary filter coefficient vectors p _{m are present} in the spoke 35 of FIG. 3 and can be supplied to the two blocks 41 and 42 from this.

From a comparison with FIG. 3 and FIG. 4 it can be seen that the iteration rules in blocks 41 and 31 and 42 and 32 have a similar structure, however, when calculating the correction filter coefficients on the auxiliary filter coefficients in the blocks 41 and 42 must be used.

For the sake of clarity, an overview of the essential variables shown in FIG. 2 will now be given with reference to FIG . The determination equation relevant to the problem of the channel impulse response is given in a block 50 . A snake above a variable indicates an estimate. The superscript index E indicates an energy correlation. Uppercase letters represent matrices as long as they are bold, while lowercase letters as long as they are bolded represent vectors.

The letter h stands for the impulse response of the transmission channel in the time domain, while, as can be seen on the left side of the equation in block 50 , the roof above the h stands for the estimation of the channel impulse response. The queue above the vector f indicates that the correction filter coefficients are also estimated, that is, they do not correspond to the ideal correction filter coefficients that would be needed to filter the first estimate of the channel impulse response so that the ge filtered estimate exactly matches the exact value ,

However, the improved estimated channel impulse response of the transfer function provided by the correction filter 16 of FIG. 1 is closer to the value of the actual channel impulse response h than the estimate of the channel impulse response associated with the roof over h in block 50 of FIG. 5 is referred to.

In a block 51 of FIG. 5, the correspondences with the variables shown for example in FIGS . 1, 2, 3 and 4 are shown. In block 52 , the estimated autocorrelation function of the channel impulse response is shown as it occurs in block 40 of FIG. 4 in the Wiener-Hopf equation. In block 53 , the autocorrelation function of the estimated channel impulse response is performed, as occurs in block 30 of the Wiener-Hopf equation of FIG. 3. Finally, the autocorrelation matrix derived from the exact autocorrelation function of the estimated channel impulse response is represented in a block 54 , as occurs in blocks 30 and 40 of FIGS. 3 and 4. The autocorrelation matrix is constructed as usual from the autocorelation function of the estimated channel impulse response, and in exactly the same way as is shown in block 74 .

From the determination equation shown in block 50 of FIG. 5 it can be seen that an unknown scaling constant still exists. The determination of the correction filter coefficients according to the invention provides these correction filter coefficients scaled with an unknown scaling constant. This scaling constant, which affects all correction filter coefficients equally, represents an ambiguity that can be resolved, for example, by the scaling constant also being estimated or being set to a predetermined value.

At this point, it should be noted that the estimation of the channel impulse response of the channel by the device 10 of FIG. 1 and the estimation of the autocorrelation function of the channel impulse response of the transmission channel by the second estimator 12 of FIG. 1 are performed independently of each other by two obtain statistical information via the transmission channel. As a reference data-free method for estimating the autocorrelation function of the channel impulse response of the transmission channel, as can be carried out by the second estimating device 12 , a multiplication-free and therefore easy to implement estimation algorithm can be used, which in the German patent application has the official file number 100 56 824.6 -35 entitled "Device and Method for Estimating a Transmission Channel and Receiver" filed on November 16, 2000. The description of the multiplicationless algorithm from this patent application is hereby incorporated by reference. This algorithm has the advantage of the unlimited estimation on the one hand and the possibility of a low-wall implementation on the other. However, it should be noted that any other known algorithm can be used to estimate the autocorrelation function, as well as to estimate the channel impulse response of the channel.

8 is hereinafter referred to Fig. Addressed to represent a dung OF INVENTION proper receiver circuit. The receiver circuit comprises a receiver front end 80 which receives a message signal from a transmission channel 81 after. The receiver front end is any known circuit that is designed to convert the data signal from the transmission channel 81 to the baseband. The baseband signal is then fed to an A / D converter 82 , which on the output side supplies a sequence of digital samples of the baseband signal, which can be stored in a memory 83 for the sampled digital samples 83 . The receiver circuit further comprises a device 84 according to the invention for estimating the channel impulse response of the channel, which estimates the current channel impulse response of the channel from the output signal of the A / D converter. The estimated channel impulse response of the transmission channel, which is output by the correction filter 16 of FIG. 1, is then fed to an equalizer circuit 85 to equalize the digital baseband samples cached in the memory 83 . Using the equalized sample values, demodulation can then be carried out, for example in the form of a Fourier transform, when OFDM systems are being considered. The device for estimating the channel impulse response of the transmission channel, which is denoted by 84 in FIG. 8, can also be connected to a Viterbi decoder which is designed to directly calculate the channel impulse response of the transmission channel when calculating the path metrics to use.

While Fig. 8 shows an embodiment of the present is invention, in which the channel impulse response is determined by the means 84 to calculate equalizer coefficients for the equalizer 85, after following with reference to FIGS. 9 and 10 to a wide Res embodiment of the present invention was received, in which the equalizer coefficients directly, ie without the detour via the channel impulse response, who calculated. In contrast to the above embodiment, in which no reference data are required, for the embodiment described with reference to FIGS. 9 and 10, the signal fed into the transmission channel is required, in order to be carried out later, a cross-correlation function between the channel output signal and the To be able to calculate the channel input signal.

Fig. 9 shows a general example of a distorting transmission channel 90, which is not distorted by the transmission channel 90 in which a first signal is fed, is to output a second signal TEN to preserver, which is a distorted version of the first signal , The distorted second signal is fed into an equalizer 92 according to the invention, which is constructed internally as shown below with reference to FIG. 10 and which supplies an equalized second signal on the output side. The equalization shown in FIG. 9 is thus an equalization without calculation of the channel impulse response. For this, the first signal, ie the signal fed into the transmission channel, must be provided to the equalizer in some way. The provision of the undistorted first signal can consist, for example, in the use of training sequences or reference sequences or reference sequences. Exemplary methods for this are that, for. B. at the beginning of each frame, as agreed, a defined training sequence is transmitted, which is stored in the equalizer 92, that is to say is firmly agreed. The undistorted first signal thus does not necessarily have to be supplied from the channel input to the equalizer 92 , but can also be stored in the equalizer 92 after a corresponding agreement. For this reason, a "transmission line" 93 is shown dotted for the first signal in FIG. 9.

The direct equalization using reference data is described in the specialist publication "News Transmission" by KD Kammeyer, Series Information Technology, BG Teubner Verlag Stuttgart, pages 508 ff. An equalizer filter with filter coefficients f is calculated using the MSE criterion (MSE = Mean Square Error). The so-called MSE target function is that the error between the equalized second signal and the first signal should be minimal. Here again the Wie-Hopf equation is obtained, which is shown in block 40 of FIG. 4, but now, in contrast to the previous explanations, the vector f _{m represents} an equalization filter of the m-th order. The autocorrelation matrix R _{x, m} is the autocorrelation matrix of the received values, ie the second signal, while the vector r _{m represents} the cross-correlation function between the received signal and the reference signal, ie between the second signal and the first signal.

From the above it can be seen that the same Wiener-Hopf equation offers itself for the direct calculation of the equalizer coefficients as for the calculation of the improved estimated channel impulse response, which has been described with reference to FIGS . 1 to 8.

If a corresponding variable substitution is carried out, the same solution concept that has been shown in FIGS. 3 and 4 for the problem of calculating an improved channel impulse response can also be used for the direct calculation of the equalizer coefficients.

10, a corresponding device or a corresponding method for equalizing a second signal, which is a distorted version of a first signal, is discussed in detail below with reference to FIG. 10. In order to show the correspondence to FIG. 1, the corresponding blocks in FIG. 10 have reference numerals which correspond to the reference characters of FIG. 1 apart from a multiplication by "10".

The device according to the invention first comprises a device 100 for obtaining the distorted second signal. The device according to the invention further comprises a device 120 for determining the cross-correlation function of the second signal and the first signal. For this purpose, the device 120 requires not only the data from the transmission channel, ie the second signal, but also the "undistorted" first signal, especially since known reference data are available for the direct calculation of the equalizer coefficients.

The device according to the invention further comprises a device 140 for determining the autocorrelation function of the signal received by the device 100 and a device 180 for calculating the equalization filter coefficients f. The device 180 supplies the calculated equalization filter coefficients to the equalization filter 160 to equalize the distorted second signal received by the device 100 to obtain the (directly) equalized second signal.

Again, as has been explained with reference to FIGS. 1 to 8, the two-stage recursion method is used, so that no inversion of the autocorrection matrix R _{x, m is} necessary to calculate the equalization filter coefficients f. In any case, the process is again carried out in two stages using auxiliary filter coefficients, the auxiliary filter being an equalization filter using a classic prediction for which the known Levinson-Durbin recursion can be used, while then using the auxiliary filter coefficients of Fig. 4 shown he invention modern "extended" Levinson-Durbin recursion can be set if the corresponding variable is performed substitution.

Claims (27)

1. Device for estimating a channel impulse response of a transmission channel, with the following features:
first means ( 10 ) for estimating the channel impulse response of the transmission channel using a signal received from the transmission channel to obtain an estimated channel impulse response of the transmission channel;
means ( 14 ) for calculating an autocorrelation function of the estimated channel impulse response to obtain an exact autocorrelation function of the estimated channel impulse response;
second means ( 12 ) for estimating an auto-correlation function of the channel impulse response of the transmission channel using a signal received from the transmission channel to obtain an estimated auto-correlation function of the channel impulse response;
a correction filter ( 16 ) with correction filter coefficients for filtering the estimated channel impulse response of the transmission channel to obtain an improved estimated channel impulse response of the transmission channel; and
means ( 18 ) for calculating the correction filter coefficients, the means for calculating being designed to use auxiliary filter coefficients ( 18 a) which are selected such that a product of a vector of the auxiliary filter coefficients and an autocorrelation matrix, which is derived from the exact autocorrelation function of the estimated channel impulse response, is equal to the exact autocorrelation function of the estimated channel impulse response ( 30 ), and to calculate, using the auxiliary filter coefficients, the correction filter coefficients chosen so that a product of a vector of Correction turf filter coefficients and the autocorrelation matrix, which is derived from the exact autocorrelation function of the estimated channel impulse response, is equal to the estimated autocorrelation function of the channel impulse response ( 40 ). 2. Device according to claim 1, in which the device ( 18 ) for calculating the correction filter coefficients is designed to iteratively calculate the auxiliary filter coefficients without inversion of the autocorrelation matrix. 3. Apparatus according to claim 1 or 2, wherein the means ( 18 ) for calculating the correction filter coefficients is designed to first calculate all auxiliary filter coefficients before the correction filter coefficients are calculated. 4. Device according to one of the preceding claims, wherein the means ( 18 ) is designed to calculate iteratively to calculate the correction filter coefficients without Inversi on the autocorrelation matrix, which is derived from the exact autocorrelation function of the estimated channel impulse response. 5. Device according to one of the preceding claims, in which the device ( 18 ) is designed to calculate to calculate an m-th filter coefficient for a correction filter of an order m + 1 according to the following rule ( 41 ):
where m is an iteration index that starts at m = 0,
where k is a summation index that runs from 0 to (m - 1),
where r (m) is an mth element of the estimated auto-correlation function of the channel impulse response,
where r _x (0) is a 0th element of the exact autocorrelation function of the estimated channel impulse response,
where p _m (k) is a kth auxiliary filter coefficient of an auxiliary filter of order m,
where r _x (m - k) is an (m - k) th element of the exact auto correlation function of the estimated channel impulse response,
where r _x (k + 1) is a (k + 1) th element of the exact auto correlation function of the estimated channel impulse response, and
where f _{m + 1} (m) is the mth filter coefficient of the correction filter of the order m + 1. 6. The device as claimed in claim 5, in which the device ( 18 ) for calculating the correction filter coefficients is designed to calculate the correction filter coefficients from a number 0 to a number m − 1 of the correction filter of the order m + 1 according to the following regulation ( 42 ):
[f _{m + 1} (0),. , ., f _{m + 1} (m - 1)] = f _m - f _{m + 1} (m) .p * _{m, r}
where f _{m + 1} (i) is a filter coefficient of the correction filter of order m + 1 with a number i;
where f _{m is} a vector of the filter coefficients of the correction filter of order m, and
where p * _{mr is} a vector of the conjugate complex auxiliary filter coefficients in reverse order. 7. The device as claimed in claim 5 or 6, in which the device ( 18 ) is designed to calculate in order to use the following value ( 33 ) as the starting value for the iteration specification for a value of the iteration index m = 0:
f ₁ = r (0): r _x (0),
where f _{1 is} the only filter coefficient of the correction filter of order 0,
where r (0) is the 0th element of the estimated auto-correlation function of the channel impulse response; and
where r _x (0) is the 0th element of the exact autocorrelation function of the estimated channel impulse response. 8. Device according to one of the preceding claims,
in which the device ( 18 ) is designed to calculate in order to calculate the auxiliary filter coefficients using the following iteration rule:
where p _{m + 1} (m) is the mth auxiliary filter coefficient of an auxiliary filter of the order m + 1,
where m is an iteration index,
where k is a summation index that runs from 0 to (m - 1),
where p _m (k) is the kth auxiliary filter coefficient of an auxiliary filter of order m,
where r _x (0) is a 0th element of the exact autocorrelation function of the estimated channel impulse response,
where r _x (m + 1) is the (m + 1) th element of the exact auto correlation function of the estimated channel impulse response,
where r _x (m - k) is the (m - k) th element of the exact auto correlation function of the estimated channel impulse response, and
where r _x (k + 1) is a (k + 1) th element of the exact auto-correlation function of the estimated channel impulse response. 9. The device according to claim 8,
in which the means ( 18 ) for calculating is designed to calculate ( 32 ) the auxiliary filter coefficients from a number 0 to a number m-1 of the correction filter of the order (m + 1):
[p _{m + 1} (0),. , ., p _{m + 1} (m - 1)] = p _m - p _{m + 1} (m) .p * | m, r
where p _{m + 1} (m) is the m-th auxiliary filter coefficient of the auxiliary filter of order m + 1,
where p _{m is} a vector of auxiliary filter coefficients of the auxiliary filter of order m,
where p _{m + 1} (i) is an auxiliary filter coefficient of the auxiliary filter of order m + 1 with a number i; and
where p * | m, r is given as follows:
p * | m, r = [p * | m (m - 1),. , ., p * | m (0)]. 10. Device according to one of the preceding claims, in which the second device ( 12 ) is designed to estimate the autocorrelation function of the channel impulse response of the transmission channel in order to carry out a reference data-free estimation of the autocorrelation function of the channel impulse response. 11. Device according to one of the preceding claims, in which the first device ( 10 ) is designed to estimate the channel impulse response of the transmission channel in order to carry out a reference data-free estimation of the channel impulse response of the transmission channel. 12. The device according to any one of the preceding claims, wherein the correction filter ( 16 ) is designed to have an order that is less than a highest number mer of a coefficient of the estimated channel impulse response, which is greater in magnitude than a predetermined threshold. 13. A method for estimating a channel impulse response of a transmission channel, comprising the following steps:
Estimating ( 10 ) the channel impulse response of the transmission channel using a signal received from the transmission channel to obtain an estimated channel impulse response of the transmission channel;
Calculating ( 14 ) an autocorrelation function of the estimated channel impulse response to obtain an exact autocorrelation function of the estimated channel impulse response;
Estimating ( 12 ) an autocorrelation function of the channel impulse response of the transmission channel using a signal received from the transmission channel to obtain an estimated autocorrelation function of the channel impulse response;
Filtering ( 16 ) the estimated channel impulse response of the transmission channel using correction filter coefficients to obtain an improved estimated channel impulse response of the transmission channel; and
Calculate ( 18 ) the correction filter coefficients using auxiliary filter coefficients ( 18 a), which are chosen so that a product of a vector of the auxiliary filter coefficients and an autocorrelation matrix derived from the exact autocorrelation function of the estimated channel impulse response is equal to that is the exact autocorrelation function of the estimated channel impulse response ( 30 ), the correction filter coefficients being calculated so that a product of a vector of the correction filter coefficients and the autocorrelation matrix derived from the exact autocorrelation function of the estimated channel impulse response is equal to the estimated auto correlation function of the channel impulse response ( 40 ). 14. Receiver circuit with the following features:
equalization means ( 85 ) configured to equalize a signal received from a transmission channel, wherein dimensioning of equalizer coefficients is based on a channel impulse response of the transmission channel;
a device ( 84 ) for estimating the channel impulse response of the transmission channel according to one of the claims 1 to 12;
a sequence control device which is designed
to cause a number of samples of the signal to be stored in a memory ( 83 ),
to cause the device ( 84 ) to estimate the channel impulse response of the transmission channel, which was relevant to the number of samples,
to cause the equalization coefficients to be dimensioned on the basis of the estimated channel impulse response by the equalization device ( 85 ), and
to cause the equalizer ( 85 ) to equalize the plurality of samples stored in the memory ( 83 ) using the dimensioned equalizer coefficients. 15. A method of receiving a signal, comprising the following steps:
Equalizing a signal received from a transmission channel, wherein dimensioning of equalizer coefficients is based on a channel impulse response of the transmission channel;
Estimating the channel impulse response by a method according to claim 13;
Storing a number of samples of the signal;
Estimating the channel impulse response of the transmission channel that is relevant to the number of samples;
Dimensioning the equalizer coefficients based on the estimated channel impulse response; and
Equalize the plurality of samples using the dimensioned equalizer coefficients. 16. Device for equalizing a second signal, which is a distorted version of a first signal, having the following features:
means ( 100 ) for obtaining the second signal;
means ( 140 ) for determining an auto-correlation function of the second signal;
means ( 120 ) for determining a cross correlation function between the second signal and the first signal;
an equalization filter ( 160 ) having equalization filter coefficients for filtering the second signal to obtain an equalized second signal; and
means ( 180 ) for calculating the equalization filter coefficients, the means for calculating being designed to use auxiliary filter coefficients ( 18 a) which are selected such that a product of a vector of the auxiliary filter coefficients and an autocorrelation matrix, which is derived from the autocorrelation function of the second signal is equal to the autocorrelation function of the second signal ( 30 ), and to calculate, using the auxiliary filter coefficients, the equalization filter coefficients chosen to be a product of a vector of the equalization filter coefficients and Autocorrelation matrix, which is derived from the autocorrelation function of the second signal, is equal to the cross-correlation function ( 40 ). 17. The apparatus of claim 16, wherein the means ( 180 ) for calculating the equalization filter coefficients is designed to iteratively calculate the auxiliary filter coefficients without inversion of the Autokor relationsmatrix. 18. The apparatus of claim 16 or 17, wherein the means ( 180 ) for calculating the equalization filter coefficients is designed to first calculate all auxiliary filter coefficients before the equalization filter coefficients are calculated. 19. Device according to one of claims 16 to 18, wherein the means ( 18 ) for calculating is designed to iteratively calculate the equalization filter coefficients without inversion of the autocorrelation matrix derived from the autocorrelation function of the second signal. 20. Device according to one of claims 16 to 19, wherein the means ( 180 ) for calculating is designed to calculate an m-th filter coefficient for an equalization filter of an order m + 1 according to the following rule ( 41 ):
where m is an iteration index that starts at m = 0,
where k is a summation index that runs from 0 to (m - 1),
where r (m) is an mth element of the cross-correlation function between the first and the second signal,
where r _x (0) is a 0th element of the autocorrelation function of the second signal,
where p _m (k) is a kth auxiliary filter coefficient of an auxiliary filter of order m,
where r _x (m - k) is an (m - k) th element of the autocorrelation function of the second signal,
where r _x (k + 1) is a (k + 1) th element of the autocorrelation function of the second signal, and
where f _{m + 1} (m) is the mth filter coefficient of the equalization filter of order m + 1. 21. The apparatus of claim 20, wherein the means ( 180 ) for calculating the equalization filter coefficients is designed to calculate the equalization filter coefficients from a number 0 to a number m - 1 of the equalization filter of order m + 1 according to the following provision ( 42 ):
[f _{m + 1} (0),. , ., f _{m + 1} (m - 1)] = f _m - f _{m + 1} (m) .p * | m, r
where f _{m + 1} (i) is a filter coefficient of the equalization filter of order m + 1 with a number i;
where f _{m is} a vector of the filter coefficients of the equalization filter of order m, and
where p * _{mr is} a vector of the conjugate complex auxiliary filter coefficients in reverse order. 22. The apparatus of claim 20 or 21, wherein the device ( 180 ) is designed for calculation in order to use ( 33 ) as the starting value for the iteration specification for a value of the iteration index m = 0:
f ₁ = r (0): r _x (0),
where f _{1 is} the only filter coefficient of the equalization filter of order 0,
wherein r (0) is the 0th element of the cross-correlation function between the first and the second signal; and
where r _x (0) is the 0th element of the autocorrelation function of the second signal. 23. The device according to one of claims 16 to 22,
in which the means ( 180 ) for calculating is designed to calculate the auxiliary filter coefficients by the following iteration rule:
where p _{m + 1} (m) is the mth auxiliary filter coefficient of an auxiliary filter of the order m + 1,
where m is an iteration index,
where k is a summation index that runs from 0 to (m - 1),
where p _m (k) is the kth auxiliary filter coefficient of an auxiliary filter of order m,
where r _x (0) is a 0th element of the autocorrelation function of the second signal,
where r _x (m + 1) is the (m + 1) th element of the autocorrelation function of the second signal,
where r _x (m - k) is the (m - k) th element of the autocorrelation function of the second signal, and
where r _x (k + 1) is a (k + 1) th element of the autocorrelation function of the second signal. 24. The device according to claim 23,
in which the means ( 18 ) for calculating is designed to calculate ( 32 ) the auxiliary filter coefficients from a number 0 to a number m-1 of the correction filter of the order (m + 1):
[p _{m + 1} (0),. , ., p _{m + 1} (m - 1)] = p _m - p _{m + 1} (m) .p * | m, r
where p _{m + 1} (m) is the m-th auxiliary filter coefficient of the auxiliary filter of order m + 1,
where p _{m is} a vector of auxiliary filter coefficients of the auxiliary filter of order m,
where p _{m + 1} (i) is an auxiliary filter coefficient of the auxiliary filter of order m + 1 with a number i; and
where p * | m, r is given as follows:
p * | m, r = [p * | m (m - 1),. , ., p * | m (0)]. 25. The device according to one of claims 16 to 24,
in which the first signal is a transmission signal that can be fed into a transmission channel,
wherein the second signal is a receive signal that is extractable from the transmission channel and that is a version of the first signal distorted by the transmission channel. 26. Device according to one of claims 16 to 25, where the first signal is a predetermined reference si gnal is. 27. A method of equalizing a second signal, which is a distorted version of a first signal, comprising the following steps:
Receiving ( 100 ) the second signal;
Determining ( 140 ) an autocorrelation function of the second signal;
Determining ( 120 ) a cross-correlation function between the second signal and the first signal;
Filtering ( 160 ) the second signal with equalization filter coefficients to obtain an equalized second signal; and
Calculate ( 180 ) the equalization filter coefficients, using auxiliary filter coefficients ( 18 a), which are chosen so that a product of a vector of the auxiliary filter coefficients and an autocorrelation matrix derived from the autocorrelation function of the second signal is equal to that Autocorrelation function of the second signal is ( 30 ), and where the equalization filter coefficients are further calculated so that a product of a vector of the equalization filter coefficients and the autocorrelation matrix derived from the autocorrelation function of the second signal is equal to the cross correlation function ( 40 ).