CN101471905B - Multi-path channel estimation method based on all-pole model - Google Patents

Abstract

The invention relates to a multipath channel estimation method based on an all-pole model, which comprises the following steps: converting an impulse response of a multipath channel from time domain to frequency domain to obtain a frequency domain response of the multipath channel; expressing the frequency domain response with an all-pole model to obtain an all-pole expression of the multipath channel model; recurring the all-pole model with the corresponding sending end data and receiving end data to obtain model parameters, and constructing an all-pole model with the model parameters; searching poles in the constructed all-pole model to obtain an initial estimated value of a multipath channel time-lag from the searched poles; and determining the impulse response of the multipath channel model according to the path number of the multipath channel and the initial estimated value of the multipath channel time-lag, and obtaining a multipath channel model. The method roughly estimates the multipath channel time-lag, which can prevent the multipath channel parameters trapping in local optimum in the subsequent refining, so as to reduce the complexity of the available multipath channel model estimation and improve the robustness, the anti-jamming property and accuracy.

Description

A kind of multi-path channel estimation method based on all-pole modeling

Technical field

The present invention relates to field of information transmission, be specifically related to a kind of multi-path channel estimation method based on all-pole modeling.

Background technology

Channel (information channels) is the transmission medium of signal, can be divided into wire message way and wireless channel two classes.Wire message way comprises open-wire line, symmetrical cable, coaxial cable and optical cable etc.; Wireless channel has the propagation of ground wave, shortwave ionospheric reflection, ultrashort wave or microwave sighting distance relaying, artificial satellite relaying and various scatter channels etc.Signal is propagated in wireless channel, and received signal is not only by single direct projection path and is obtained, and also comprises reflection, diffraction and data-signal that different paths arrive, and this phenomenon is called multipath transmisstion, and this channel is called multipath channel (Multi-path channel).

When signal transmits in multipath channel, in the signal analysis multipath channel is analyzed as a multipath channel system as shown in Figure 1, make the input signal x (n) of this system from the signal of transmitting terminal emission, at last the signal that receives at receiving terminal is as the output signal y (n) of this system, x (n) wherein, y (n) is the expression mode of discrete signal in the signal analysis, specific to n is the sequence number of signal data, the 3rd data as x (3) expression input signal, below be this implication with the similar n of this expression formula, according to prior art, this system has impulse response h (n), n is the sequence number of impulse response data, and has such relation between input signal and the output signal:

y(n)＝x(n)*h(n) (1)

Promptly the signal of Jie Shouing is obtained by the impulse response convolution of input signal and this system, in practice, signal that after the multipath channel transmission, receives and the actual signal that sends and inconsistent at receiving terminal, send through multipath channel as information a picture, if the signal that receives is directly shown at receiving terminal, the pictorial information that obtains and the information of emission are inconsistent, have the phenomenon of signal distortion after the multipath channel transmission, and Here it is sets up the necessity that multipath road model carries out channel estimating.

The multipath channel models of in Multiple Channel Analysis, setting up, its form with impulse response is represented the multipath channel models set up, existing multipath channel models analysis, typical method is the impulse response h (n) of the multipath channel models set up with sparse model (sparse representation) expression, this expression formula as the formula (2):

$h\left(n\right)=\underset{\stackrel{i=1}{M\≤20}}{\overset{M-1}{\mathrm{\Σ}}}{A}_i\mathrm{\δ}(n-d_i)---\left(2\right)$

In the formula (2), M is the footpath number of multipath channel, and in existing the analysis, general footpath number is smaller or equal to 20, and i is an integer, and the span of i is 1～M, and i represents the sequence number of multipath channel footpath number, d in the formula _iThe time-delay of i channel in the expression multipath channel, n is the sequence number of impulse response data, and δ is an impulse function, and impulse function is to be the function of 0 o'clock output amplitude at input signal only, A _iFor the amplitude of i channel impulse function in the multipath channel is adjusted coefficient, so the multipath channel models that will obtain setting up, need be model parameter A _iWith d _iEstimate.Generally in addition M also be unknown, but when the actual estimated multipath channel models,, suppose generally that M has been estimated or known for the simplification problem.With model parameter A _iWith d _iAfter estimating, by the relation between input signal in the formula (1) and the output signal, utilize signal analysis can obtain the signal of actual transmission, we are called channel-decoding with this process, so the accuracy of channel model directly has influence on the accuracy by the actual transmission signal that obtains after the decoding of received signal channel.

Multipath channel models parameter A in a lot of estimators (2) is arranged in the existing channel method of estimation _iWith d _iMethod, but the ubiquitous problem of setting up with this method of multipath channel models: the multipath channel models of foundation is accurate inadequately, and estimation procedure complexity and redundancy height, makes the channel-decoding difficulty big, can't accurately obtain original transmission signal.

Summary of the invention

The purpose of this invention is to provide a kind of multi-path channel estimation method based on all-pole modeling, the multipath channel models stable performance that utilizes this method to set up, strong interference immunity, can avoid being absorbed in local optimum in the refinement afterwards of multipath channel parameter, reduce the complexity that existing multipath channel models is estimated, improved the accuracy of received signal.

For achieving the above object, the present invention adopts following technical scheme:

A kind of multi-path channel estimation method based on all-pole modeling, the footpath number of described multipath channel is M, and impulse response is h (n), and the parameter of impulse response h (n) is footpath number M and multipath channel time-delay d _i, i is an integer, the span of i is 1～M, d _iThe time-delay of i channel in the expression multipath channel, this method may further comprise the steps:

(1) the impulse response h (n) with described multipath channel is transformed into frequency domain from time domain, obtains the frequency domain response H (z) of multipath channel;

(2) represent the frequency domain response H (z) of multipath channel with all-pole modeling, obtain the full limit expression formula of multipath channel;

(3) according in frequency domain, the data of receiving terminal are obtained by the frequency domain response product of transmitting terminal data and multipath channel, with corresponding sending terminal and receiving terminal data described all-pole modeling recursion is obtained its model parameter, set up all-pole modeling by described model parameter;

(4) all-pole modeling of setting up is carried out the limit search, obtain multipath channel time-delay d by the limit of searching for _iInitial estimate;

(5) count M and multipath channel time-delay d by the footpath of described multipath channel _iInitial estimate determine the impulse response h (n) of described multipath channel to obtain initial multipath channel models.

In step (1), be that impulse response h (n) is carried out the z conversion from the method that time domain is transformed into frequency domain with the impulse response h (n) of described multipath channel.

In step (3), with row Vincent-Du Bin Levinson-Durbin recurrence method described all-pole modeling recursion is obtained its model parameter according to corresponding sending terminal and receiving terminal data

The impulse response h (n) of the multipath channel sparse model representation of following formula,

$h\left(n\right)=\underset{\stackrel{i=1}{M\≤20}}{\overset{M-1}{\mathrm{\Σ}}}{A}_i\mathrm{\δ}(n-d_i)$

Wherein M is the footpath number of multipath channel, and described footpath is counted M and is less than or equal to 20, and i is an integer, and the span of i is 1～M, d _iThe time-delay of i channel in the expression multipath channel, δ is an impulse function, A _iFor the amplitude of i channel impulse function in the multipath channel is adjusted coefficient, in this method, obtain multipath channel time-delay d in step (4) _iInitial estimate after, further comprising the steps of:

(4A) produce the input signal p (n) of multipath channel with pseudo-random noise sequence generator, the signal that receives from actual multipath channel is t (n), by the input signal p (n) of multipath channel and impulse response h (n) convolution of multipath channel, obtain output signal y (n);

(4B) utilize the output signal y (n) of multipath channel and the mean square deviation E between the actual reception signal t (n), by expression formula:

$\frac{\∂E}{\∂A_i}=0$

Draw A _iWith d _iBetween relational expression A _i=g (d _i), the multipath channel time-delay d that step (4) is obtained _iInitial estimate substitution relational expression A _i=g (d _i), the amplitude that obtains the multipath channel impulse function is adjusted coefficient A _i, wherein i is an integer, the span of i is 1～M;

In step (5), count M, multipath channel time-delay d by the footpath of described multipath channel _iThe initial estimate and the amplitude of multipath channel impulse function adjust coefficient A _iDetermine the impulse response h (n) of described multipath channel, obtain initial multipath channel models.

After step (5) obtains initial multipath road model, also comprise step (6): with match tracing method Matching Pursuits refinement multipath channel time-delay d _iThe initial estimate and the amplitude of multipath channel impulse function adjust coefficient A _i

With match tracing method Matching Pursuits refinement multipath channel time-delay d _iThe initial estimate and the amplitude of multipath channel impulse function adjust coefficient A _iMethod may further comprise the steps:

(6A) estimate d based on the multipath channel time-delay that obtains in the step (4) _iInitial estimate make up the time-delay code table, some groups of unequal multipath channels time-delay { d are arranged in the described time-delay code table _i;

(6B) by the relational expression A in the step (4B) _i=g (d _i) draw with step (6A) in { d _iSome groups of corresponding unequal amplitudes adjustment coefficients, some groups of multipath channel parameter { A obtained _i, d _i, count the some groups of impulse responses { h (n) } that M determines described multipath channel in conjunction with the footpath of multipath channel;

(6C) produce the input signal p (n) of multipath channel with pseudo-random noise sequence generator, some groups of impulse responses { h (n) } convolution by input signal p (n) and multipath channel obtains several output signals { y (n) }, and the signal that receives from the actual reception multipath channel is that t (n) is the refinement object vector;

(6D) multipath channel parameter { A of the mean square deviation E minimum of one group of output signal that makes multipath channel of search and refinement object vector from several output signals { y (n) } _I0, d _I0;

(6E) count M and multipath channel parameter { A by the footpath of described multipath channel _I0, d _I0Determine described multipath channel impulse response h ' (n);

(6F) according to multipath channel parameter { A _I0, d _I0The refinement object vector is upgraded, the refinement object vector after the renewal is t ¹(n)=t (n)-y _I0(n), y wherein _I0Obtain with the middle impulse response h ' convolution (n) of step (6E) by input signal p (n);

(6G) circulation execution in step (6D)～(6F) limited number of time is by the multipath channel models after the multipath channel parameter of search and the footpath of multipath channel are counted M and determined refinement in step (6D) for the last time.

In step (6A) by multipath channel time-delay d _iThe initial estimate method that makes up the time-delay code table be: on number axis with multipath channel time-delay d _iInitial estimate be the center, at d _iThe left and right sides is chosen several values symmetrically, obtains some groups of unequal multipath channel time-delays and estimates { d _i.

The advantage of utilizing the present invention to carry out the multipath channel estimation has:

1. utilize the channel initial delay of all-pole modeling to estimate, make next step refinement parameter algorithm that a good initial value be arranged, thereby miss global optimum and make the inaccurate defective of calculating by multipath channel models of received signal owing to the initial condition of difference may be absorbed in a local optimum when having avoided refinement;

2. estimate based on the refinement of the channel parameter of matching pursuit algorithm Matching Pursuits, because matching pursuit algorithm is a kind of greedy algorithm, it is easy to realize and have good approximate characteristic usually, each circulation all can guarantee to reduce mean square error, therefore the required multipath channel parameter refinement of matching pursuit algorithm is estimated to optimize the multipath channel parameter, make the multipath channel models stable performance of foundation, strong interference immunity;

3. when the initial channel time-delay is estimated, set up a rational dictionary or code table, make the complexity that has reduced the matching pursuit algorithm inquiry in the thinning process like this, this multipath channel models method of estimation complexity is reduced.

Description of drawings

Fig. 1 is an input/output signal graph of a relation in the multipath channel system;

Fig. 2 is based on the initial time delay rough estimate flow chart of all-pole modeling in the multi-path channel estimation method;

Fig. 3 carefully estimates flow chart in the multi-path channel estimation method of the present invention based on the channel parameter of matching pursuit algorithm.

Embodiment

Following examples are used to illustrate the present invention, but are not used for limiting the scope of the invention.

In the signal analysis multipath channel is analyzed as a multipath channel system as shown in Figure 1, make the input signal x (n) of this system from the signal of transmitting terminal emission, at last the signal that receives at receiving terminal is as the output signal y (n) of this system, x (n) wherein, y (n) is the expression mode of discrete signal in the signal analysis, specific to n is the sequence number of signal data, the 3rd data that expression transmits as x (3), below be this implication with the similar n of this expression formula, according to prior art, this system has impulse response h (n), and has such relation between input signal and the output signal:

y(n)＝x(n)*h(n) (1)

Promptly the signal of Jie Shouing is obtained by the impulse response convolution of input signal and this system, in practice, signal that after the multipath channel transmission, receives and the actual signal that sends and inconsistent at receiving terminal, in practice, for return to the signal of original transmission from the signal that receives, the erasure signal transmission distortion, when channel-decoding, need set up multipath channel models, the impulse response of this model is the h (n) in the following formula, if can determine accurate h (n) expression formula, just equal to have set up the precise channels model, obtain more accurate original transmitted signal.

Embodiment 1

Multipath channel models in the present embodiment, typical method is represented in its impulse response h (n) employing prior art, promptly uses the impulse response of the multipath channel models of sparse model (sparse representation) expression foundation, as the formula (2):

$h\left(n\right)=\underset{\stackrel{i=1}{M\≤20}}{\overset{M-1}{\mathrm{\Σ}}}{A}_i\mathrm{\δ}(n-d_i)---\left(2\right)$

In the formula (2), M is the footpath number of multipath channel, and in existing the analysis, general footpath number is less than or equal to 20, and i is an integer, and the span of i is 1～M, and i represents the sequence number of multipath channel footpath number, d in the formula _iBe the time-delay of i channel in the multipath channel, δ is an impulse function, and impulse function is to be the function of 0 o'clock output amplitude at input signal only, A _iFor the amplitude of i channel impulse function in the multipath channel is adjusted coefficient, so the multipath channel models that will obtain setting up, need be the multipath channel models parameter A _iWith d _iEstimate.Generally in addition M also be unknown, but when the actual estimated multipath channel models,, suppose generally that M has been estimated or known for the simplification problem.M is known in the present embodiment, the estimation model parameter A _iWith d _iMethod be described in detail as follows.

Be illustrated in figure 2 as in the present embodiment based on the initial delay rough estimate flow chart of all-pole modeling, this method mainly may further comprise the steps:

(1) the impulse response h (n) (formula (2)) with described multipath channel is transformed into frequency domain from time domain, obtain the frequency domain response H (z) of multipath channel, it is this area signal analysis means commonly used that time-domain signal is transformed into frequency-domain analysis, signal in the present embodiment is discrete signal, therefore utilize signal analysis means commonly used, the impulse response of multipath channel is carried out the z conversion obtain its frequency domain response, the z conversion is a kind of signal analysis means that in the signal analysis time domain are transformed in the frequency domain, no longer describes in detail here.For some signals, be transformed into the easier analysis of frequency domain with time domain, to present embodiment also is like this, present embodiment is that the impulse response h (n) (formula (2)) to multipath channel utilizes fast fourier transform FFT to obtain the frequency domain response H (z) of multipath channel, also can Fourier transform obtains the frequency domain response of multipath channel;

(2) represent the frequency domain response H (z) of multipath channel with all-pole modeling, obtain the full limit expression formula of multipath channel models, this step is the full limit form with frequency domain response H (z) expression (3) of multipath channel:

$H\left(z\right)=g/\underset{\stackrel{i=1}{M\≤20}}{\overset{M}{\mathrm{\Σ}}}{a}_i{z}^{-i}---\left(3\right)$

Wherein, M is the footpath number of multipath channel, and in existing the analysis, i is an integer, and the span of i is 1～M, and i represents the sequence number of multipath channel footpath number in the formula, and g is the molecule of full limit expression formula, a _iBe the coefficient of z conversion in the full limit expression formula, z ^-iBe the expression way of z conversion, to the frequency domain response H (z) of expression in the formula (3), the form that can also be expressed as:

$H\left(z\right)=\frac{1}{\frac{a_1}{g}{Z}^{-1}+\frac{a_2}{g}{Z}^{-2}+......\frac{a_M}{g}{Z}^{-M}}---\left(4\right)$

Wherein M, i, g, a _i, Z ^-iWith formula (3) in implication identical, in the formula (3) on year-on-year basis g obtain formula (4), can learn that by formula in fact having only one group of parameter is a _i/ g;

Representing the frequency domain response H (z) of multipath channel in the present embodiment with all-pole modeling, is the conversion to the expression-form of frequency domain response H (z).

(3) according in frequency domain, the relation between the signal of transmitting terminal and the signal of receiving terminal: the signal X (z) of transmitting terminal and the product of frequency domain response H (z) obtain the signal Y (z) of receiving terminal, i.e. Y (z)=X (z) * H (z) is because have only one group of parameter a _i/ g, utilize the linear prediction analysis method can obtain this group parameter, in the present embodiment, be with row Vincent-Du Bin Levinson-durbin recursive algorithm (ASIC Implementation of Levinson-Durbin Arithmetic in LPC, computer and digital engineering, 2005 33 the 9th phases of volume, 118-119,156 pages) recursion obtains the parameter a of all-pole modeling respectively _iAnd g, by model parameter a _iSetting up all-pole modeling with g, also is the signal analysis means that those skilled in the art use always, no longer describes in detail here;

(4) set up all-pole modeling after, can carry out limit search to it, the method for carrying out the limit search has a lot, generally is the model parameter a according to all-pole modeling _iMaking up unit circle with g, is exactly the channel time delay d of multipath channel in the present embodiment by the position of limit in the unit circle _iInitial estimate;

The position of limit is exactly multipath channel time-delay d in this step _iThis relation of initial estimate, be by the decision of the characteristics of formula (2) expression formula, be the characteristic of signal, here to its principle specific explanations no longer.

(5) multipath channel time-delay d _iInitial estimate be the parameter of comparison key, by multipath channel time-delay d _iInitial estimate ask coefficient A in the model parameter _iMethod as follows:

Produce the input signal p (n) of multipath channel with pseudo-random noise sequence generator, the signal that receives from actual channel is t (n), obtains output signal by the input signal p (n) and impulse response h (n) (formula (the 2)) convolution of multipath channel models

$y\left(n\right)=p\left(n\right)*h\left(n\right)=\underset{\stackrel{i=1}{M\≤20}}{\overset{M-1}{\mathrm{\Σ}}}{A}_{i}p(n-d_i);---\left(5\right)$

M, i, d in the formula (5) _i, A _iIdentical with implication in the formula (2);

Utilize the output signal y (n) of multipath channel models in the formula (5) and the mean square deviation E between the actual reception signal t (n),

$E=\underset{n}{\mathrm{\&Sigma;}}{[t\left(n\right)-\underset{\stackrel{i=0}{M<20}}{\overset{M-1}{\mathrm{\&Sigma;}}}{A}_{i}p(n-d_i)]}^2---\left(6\right)$

Wherein, n is the sequence number of signal data, marks n in the formula (6) below the summation symbol and represents square summation after all inputoutput data differences of correspondence.By expression formula

$\frac{\&PartialD;E}{\&PartialD;A_i}=0$

Promptly to A in the formula (6) _iDifferentiate draws A _iWith d _iBetween relational expression A _i=g (d _i), d is estimated in the multipath channel time-delay _iThis expression formula of initial estimate substitution obtain described coefficient A _i

Present embodiment is at definite channel model impulse response A _iWith d _iAfter, count the impulse response h (n) that M determines described channel model in conjunction with the footpath of described multipath channel, obtain the initial channel model.

Utilize the method for present embodiment, with all-pole modeling to the rough estimate of channel initial delay, make next step refinement parameter algorithm that a good initial value be arranged, thereby miss global optimum and make the inaccurate defective of calculating by multipath channel models of received signal owing to the initial condition of difference may be absorbed in a local optimum when having avoided refinement; This method estimated channel model accuracy is also than higher in addition, but estimation procedure is simple, greatly reduces the complexity of calculating.

Embodiment 2

Estimation model parameter A in the present embodiment _iWith d _iThe method key step be divided into for two steps:

(1) based on the initial channel of all-pole modeling time-delay rough estimate;

(2) based on matching pursuit algorithm (Matching Pursuits) to the multipath channel parameter A _iWith d _iRefinement estimate.

Step in the present embodiment (1) obtains the rough estimate { A of channel for carrying out the process of embodiment 1 method in embodiment 1 _i, d _iAfter, for obtaining more accurate multipath channel models, also comprise with matching pursuit algorithm model parameter { A _i, d _iThe process of refinement.

Matching pursuit algorithm MP (Matching Pursuits) is a kind of superpose out algorithm of a specific signal of some base vector of selecting from an extremely redundant dictionary, this algorithm is a kind of specific implementation method of signal decomposition, but it then is a kind of method of estimated signal model parameter in essence, thereby can be directly used in input and parameter estimation, present this algorithm has been successfully used in video compression and other field, but also is not applied in the channel model method for parameter estimation.

The MP algorithm still is a kind of greedy algorithm, and it is easy to realize and have good approximate characteristic usually, and common algorithm based on MP needs a criterion and selects from a plurality of possible candidate's estimated parameters circularly.Here our criterion introduced be in PN (pseudo noise) sequence PN training sequence between the mean square error (MSE) between the signal that receives from actual channel and the signal of from channel model, exporting, circulation all can guarantee to reduce mean square error at every turn.Process below in conjunction with its refinement parameter of description of drawings.

1. make up the time-delay code table

Set up the complexity that a rational dictionary or code table can reduce the MP inquiry before with MP method refinement channel model parameter, present embodiment makes up code table and is based on the one group of initial channel time-delay that obtains noted earlier and estimates { d _i(1≤i≤M, M are the footpath number of multipath channel, and the value of M is less than or equal to 20), concrete construction method is:

On number axis with channel time delay d _iInitial estimate be the center, at d _iThe left and right sides is chosen several values symmetrically, obtains some groups of unequal multipath channel time-delays and estimates { d _i.

In the present embodiment promptly at each d _iCenter on d in the scope of closing on of value _iGet several values symmetrically, the size of scope is set as required, if d _iBetween difference can correspondingly scope be got when big more greatly, the number of institute's value is corresponding more, if d _iBetween difference hour can correspondingly scope be got a little bit smaller, the number of institute's value is corresponding few.Around d _iMake up code table and guarantee that the interior value of code table is all at time-delay estimation d _iAround, be more rational.

Be illustrated in figure 3 as in the present embodiment channel parameter and carefully estimate flow chart, obtain initial time delay based on all-pole modeling among Fig. 3 and be estimated as { n based on matching pursuit algorithm ₁, n ₂, n ₃... n _M, wherein M is the footpath number of multipath channel, the value of M is less than or equal to 20, n _MRepresent the time-delay of M channel, get two values symmetrically when making up code table around each time-delay estimated value, constructed code table is { n ₁-1, n ₁, n ₁+ 1, n ₂-1, n ₂, n ₂+ 1, n ₃-1, n ₃, n ₃+ 1 ... n _M-1, n _M, n _MTherefore+1} obtains three groups of unequal time-delays from described code table and estimates { n ₁-1, n ₂-1, n ₃-1 ... n _M-1}, { n ₁, n ₂, n ₃N _M, { n ₁+ 1, n ₂+ 1, n ₃+ 1 ... n _M+ 1}.

Obtain some groups of unequal multipath channel time-delays and estimate { d _iBack by foregoing A _iAnd d _iBetween concern A _i=g (d _i), some groups of unequal multipath channel time-delay { d in the time-delay code table that can draw and make up _iCorresponding coefficient, some groups of multipath channel models parameter { A obtained _i, d _i, count the some groups of different impulse responses { h (n) } that M determines described channel model in conjunction with the footpath of multipath channel.

2. based on the thin estimation of the channel parameter of matching pursuit algorithm MP

Before thin estimation, produce the input signal p (n) of multipath channel as shown in Figure 3 with pseudo-random noise sequence generator, obtain several output signals { y (n) } (formula (5)) by input signal p (n) with some groups of channel model different impulse responses { h (n) } convolution, the signal that receives from actual channel is that t (n) is the refinement object vector.

Search makes the output signal y (n) of multipath channel models and the minimum model parameter { A of the mean square deviation E of refinement object vector t (n) (formula (6)) for one group from several output signals { y (n) } _I0, d _I0, be specially:

$d_{i0}=\underset{d_i\&Element;\mathrm{dictionary}}{\arg \min }\underset{n}{\mathrm{\&Sigma;}}{[t\left(n\right)-\underset{\stackrel{i=0}{M<20}}{\overset{M-1}{\mathrm{\&Sigma;}}}{A}_{i}p(n-d_i)]}^2---\left(7\right)$

A _i0＝g(d _i0) (8)

Wherein, d _i∈ dictionary represents multipath channel time-delay d _iBelong to the multipath channel time-delay in the time-delay code table that time-delay makes up, n is the sequence number of signal data, in the formula (8) below the summation symbol mark n represent square summation after corresponding all data difference;

Count M and multipath channel parameter { A by the footpath of described multipath channel _I0, d _I0Determine described multipath channel impulse response h ' (n), promptly

$h^{\&prime;}\left(n\right)=\underset{\stackrel{i=1}{M\&le;20}}{\overset{M-1}{\mathrm{\&Sigma;}}}{A}_{i0}\mathrm{\&delta;}(n-d_{i0})---\left(9\right)$

M, i are identical with implication in the formula (2) in the formula (9);

According to model parameter { A _I0, d _I0Refinement object vector t (n) is upgraded, the refinement object vector after the renewal is

t ¹(n)＝t(n)-y _i0(n) (10)

$y_{i0}\left(n\right)\underset{\stackrel{i0=0}{M<20}}{\overset{M-1}{\mathrm{\&Sigma;}}}{A}_{i0}p(n-d_{i0})---\left(11\right)$

Be y _I0(n) by input signal p (n) and channel parameter { A _I0, d _I0Determine described multipath channel impulse response h ' (n) convolution of (formula (9)) obtain.

With the refinement object vector t after upgrading ¹(n) search for one group of output signal y (n) and refinement object vector t that makes multipath channel models again ¹Model parameter { the A of mean square deviation E minimum (n) _I1, d _I1; upgrade the refinement object vector by this group parameter by formula (9), (10) and (11) accordingly; again search for one group with the refinement object vector after upgrading once more and make the output signal y (n) of multipath channel models and the model parameter of the mean square deviation E minimum of refinement object vector; several times are carried out in circulation like this, count channel model after M determines refinement by the footpath of the model parameter of last search and multipath channel.

The present invention introduces the rough estimate of multipath channel models parameter with all-pole modeling, and after the initial channel based on all-pole modeling postponed rough estimate, thin estimation can obtain the channel model of global optimum to the channel model parameter to utilize existing algorithms most in use; After utilization the present invention is based on the initial channel delay rough estimate of all-pole modeling, can the better optimize channel parameter based on the thin estimation of channel parameter of matching pursuit algorithm Matching Pursuits (MP), set up more precise channels, improve the anti-interference and the robustness of multipath channel.

Though the present invention specifically illustrates and illustrates in conjunction with a preferred embodiment; but the personnel that are familiar with this technical field are appreciated that; wherein no matter still can make various changes in detail in form, this does not deviate from spirit of the present invention and scope of patent protection.

Claims (5)

1. multi-path channel estimation method based on all-pole modeling, the footpath number of described multipath channel is M, and impulse response is h (n), and the parameter of impulse response h (n) is footpath number M, multipath channel time-delay d _iAdjust coefficient A with the amplitude of multipath channel impulse function _i, i is an integer, the span of i is 1～M, d _iThe time-delay of i channel in the expression multipath channel, A _iAdjust coefficient for the amplitude of i channel impulse function in the multipath channel, it is characterized in that this method may further comprise the steps:

(1) the impulse response h (n) with described multipath channel is transformed into frequency domain from time domain, obtains the frequency domain response H (z) of multipath channel;

(2) represent the frequency domain response H (z) of multipath channel with all-pole modeling, obtain the full limit expression formula of multipath channel;

(3) according in frequency domain, the data of receiving terminal are obtained by the frequency domain response product of transmitting terminal data and multipath channel, with corresponding sending terminal and receiving terminal data described all-pole modeling is obtained its model parameter by the linear prediction analysis method, set up all-pole modeling by described model parameter;

(4) all-pole modeling of setting up is carried out the limit search, according to the model parameter a of all-pole modeling _iWith g component unit circle, and the position of limit is exactly the channel time delay d of multipath channel in the unit circle _iInitial estimate, g is the molecule of all-pole modeling expression formula here, a _iBe z in the all-pole modeling expression formula ^-iThe coefficient of variable; Then, determine the amplitude adjustment coefficient A of impulse function by following steps _i:

(4A) produce the input signal p (n) of multipath channel with pseudo-random noise sequence generator, the signal that receives from actual multipath channel is t (n), by the input signal p (n) of multipath channel and impulse response h (n) convolution of multipath channel, obtain output signal y (n);

(4B) utilize the output signal y (n) of multipath channel and the mean square deviation E between the actual reception signal t (n), by expression formula:

$\frac{\&PartialD;E}{{\&PartialD;A}_i}=0$

Draw A _iWith d _iBetween relational expression A _i=g (d _i), with the multipath channel time-delay d that obtains previously _iInitial estimate substitution relational expression A _i=g (d _i), the amplitude that obtains the multipath channel impulse function is adjusted coefficient A _i

(5) determine the impulse response h (n) of described multipath channel by following formula, obtain initial multipath channel models:

$h\left(n\right)=\underset{i=1}{\overset{M-1}{\mathrm{\&Sigma;}}}{A}_i\mathrm{\&delta;}(n-d_i)$

Wherein δ is an impulse function.

2. the multi-path channel estimation method based on all-pole modeling as claimed in claim 1 is characterized in that, in step (1), is that impulse response h (n) is carried out the z conversion from the method that time domain is transformed into frequency domain with the impulse response h (n) of described multipath channel.

3. the multi-path channel estimation method based on all-pole modeling as claimed in claim 1, it is characterized in that, in step (3), with row Vincent-Du Bin Levinson-Durbin recurrence method described all-pole modeling recursion is obtained its model parameter according to corresponding sending terminal and receiving terminal data.

4. the multi-path channel estimation method based on all-pole modeling as claimed in claim 1 is characterized in that, described footpath is counted M and is less than or equal to 20.

5. the multi-path channel estimation method based on all-pole modeling as claimed in claim 1 is characterized in that, after step (5) obtains initial multipath road model, also comprises step (6): with match tracing method Matching Pursuits refinement multipath channel time-delay d _iThe initial estimate and the amplitude of multipath channel impulse function adjust coefficient A _i, may further comprise the steps:

(6A) estimate d based on the multipath channel time-delay that obtains in the step (4) _iInitial estimate make up the time-delay code table, wherein, the method that makes up the time-delay code table is that the initial estimate with multipath channel time-delay di on number axis is the center, at d _iThe left and right sides is chosen several values symmetrically, obtains some groups of unequal multipath channel time-delays and estimates { d _i;

(6B) by the relational expression A in the step (4B) _i=g (d _i) draw with step (6A) in { d _iSome groups of corresponding unequal amplitudes adjustment coefficients, some groups of multipath channel parameter { A obtained _i, d _i, count the some groups of impulse responses { h (n) } that M determines described multipath channel in conjunction with the footpath of multipath channel;

(6C) produce the input signal p (n) of multipath channel with pseudo-random noise sequence generator, some groups of impulse responses { h (n) } convolution by input signal p (n) and multipath channel obtains several output signals { y (n) }, and the signal that receives from the actual reception multipath channel is that t (n) is the refinement object vector;

(6D) multipath channel parameter { A of the mean square deviation E minimum of one group of output signal that makes multipath channel of search and refinement object vector from several output signals { y (n) } _I0, d _I0;

(6E) count M and multipath channel parameter { A by the footpath of described multipath channel _I0, d _I0Determine described multipath channel impulse response h ' (n);

(6F) according to multipath channel parameter { A _I0, d _I0The refinement object vector is upgraded, the refinement object vector after the renewal is t ¹(n)=t (n)-y _I0(n), y wherein _I0Obtain with the middle impulse response h ' convolution (n) of step (6E) by input signal p (n);

(6G) circulation execution in step (6D)～(6F) limited number of time is by the multipath channel models after the multipath channel parameter of search and the footpath of multipath channel are counted M and determined refinement in step (6D) for the last time.

Images