WO2007029142A2

WO2007029142A2 - Method and apparatus for estimating channel based on implicit training sequences

Info

Publication number: WO2007029142A2
Application number: PCT/IB2006/052988
Authority: WO
Inventors: Luxi Yang; Yuanjie Li
Original assignee: Koninklijke Philips Electronics N.V.; Southeast University
Priority date: 2005-09-06
Filing date: 2006-08-29
Publication date: 2007-03-15
Also published as: CN101292481B; CN101292481A; WO2007029142A3

Abstract

The present invention provides a channel estimation method and apparatus based on implicit training sequences for use in wireless communication systems, characterized in that stationary training sequences are superimposed with information sequences for transmission at the transmitting side and channel estimation is performed at the receiving side by using the uncorrelated characteristic between the training sequences and the information sequences, that is, channel estimation is done based on the principle that the estimation result of channel parameter converges to a Wiener solution under the condition that the training sequences and information sequences in the transmission signals are uncorrelated. The method of the invention comprises steps of: obtaining a set of training sequences comprising a plurality of subsets of training sequences, each of the plurality of subsets is generated based on a known initial training sequence and a predetermined channel order; calculating the cross-correlation matrix for the subsets of training sequences and the corresponding inverse matrix; calculating the cross-correlation matrix for the set of training sequences and the received signals; and estimating channel parameters based on the inverse cross-correlation matrix for the subsets of training sequences and the cross-correlation matrix for the set of training sequences and the received signals.

Description

METHOD AND APPARATUS FOR ESTIMATING CHANNEL BASED ON IMPLICIT TRAINING SEQUENCES

FIELD OF THE INVENTION The present invention relates generally to wireless communication systems, and more particularly, to a method and apparatus for estimating channel of a wireless communication system.

BACKGROUND OF THE INVENTION Typically, signal transmission over a wireless channel often suffers from signal fading and distortion due to the propagation distance, barrier obstruction and relative movement between the receiver and the transmitter, and accordingly there exists difference between the data received at the receiving side and that transmitted at the transmitting side. To recover data correctly at the receiving side, known training sequences are inserted in particular positions in the data segments to be transmitted, so that the receiving side can estimate channel parameter estimation for data recovery by processing the expected and distorted training sequences. Fig.l illustrates a typical burst structure in the 3G mobile communication systems. The burst comprises data segment 1, data segment 2 and training sequence 3. Data segments 1 and 2 carry user data, and the training sequence 3 is a known Pseudo-random sequence used for estimating channel parameters. For simplicity, control segments or guard gaps possibly inserted at the two ends of the burst or other positions are omitted herein.

It is clear from Fig.1 that the channel parameters for data recovery can be estimated conveniently based on the training sequences. However, the training sequences occupy a certain bandwidth, which decreases the transmission rate, and the case is true especially for a fast varying environment, in which transmission of the training sequences needs to be done periodically to keep up with the channel variation, and this leads to a significant waste of bandwidth resources in transmission and reception of training sequences without carrying any data information. To those skilled in the art, it becomes an attractive technical problem as to find a channel estimation method, which allows efficient channel estimation without waste of bandwidth resources. Among the various technical solutions, the channel estimation method based on implicit training sequences has grown into a competitive one due to its unique transmission manner and its potential in improving performance.

Fig.2 shows a burst containing an implicit training sequence. It can be seen from Fig.2 that the channel estimation based on implicit training sequences is characterized in that the training sequence is superimposed on the information sequence and transmitted together with the information sequence, such that the training sequence for channel estimation will occupy no dedicated timeslots during transmission, thus reducing the loss in transmission rate due to training sequences to 0. A.GOrozco-Lugo, M.M.Lara, and D.C.McLernon, "Channel estimation using implicit training, " IEEE Trans. Signal Processing, vol. 52, No.l, pp. 240-254, Jan. 2004, had given a detailed description of how to use implicit training sequences in a SISO (single-input single- output) system and its performance is analyzed in details.

In the description of this reference, periodic sequences are superimposed on information sequences and transmitted together with data sequences at the transmitting side, and accordingly, channel estimation is performed at the receiving side by using the signal's cycle- stationary characteristic. The reference also proposes an optimal method for training sequences, i.e. how to choose sequences to reduce the MSE (Mean Square Error) of the estimation and overcome the high PAR (Peak Average Ratio) problem caused by superimposing the implicit training sequences with the information sequences.

However, the method proposed in this reference has a disadvantage that the channel order of a practical system should be known before optimizing the training sequences. When there is a change in the channel order or an error in estimation, the cyclic matrix formed by the training sequences changes and the previous sequence subsequently loses its optimal characteristics. Furthermore, when there exists unknown direct circuit (DC) interference at the receiving side, the method of the reference requires introduction of complicated operations to cancel the DC interference.

A method for performing channel estimation on SIMO (single-input multi-output) systems using implicit training sequences is disclosed in J.K.Tugnait and W.Luo, "On channel estimation using superimposed training and first-order statistics, " in Proc. IEEE ICASSP'03, vol. 4, pp.624-627, Apr. 2003, in which a transform similar to DFT is employed in the proposed algorithm. A time-varying channel can be described with an exponent-base extension model used in the algorithm to estimate time-varying and non time-varying channels. But, the method assumes, when estimating time-varying channels, that the phase shift for each arriving path is known, which is unrealistic and there are too many parameters to be estimated for the exponent-base extension model, thereby channel variations can't be tracked in real time.

There is therefore a need for provide a more effective method and apparatus for estimating channel parameters based on implicit training sequences to improve efficiency of estimating channel parameters.

OBJECT AND SUMMARY OF THE INVENTION

An object of the present invention is to provide a channel estimation method for wireless communication systems to improve efficiency of estimating channel parameters.

Accordingly, the invention provides a channel estimation method based on implicit training sequences, comprising steps of: obtaining a set of training sequences comprising a plurality of subsets of training sequences, each of the plurality of subsets being generated based on a known initial training sequence and a predetermined channel order; calculating the cross-correlation matrix for the subsets of training sequences and the corresponding inverse matrix; calculating the cross-correlation matrix for the set of training sequences and the received signals; and estimating channel parameters based on the inverse cross-correlation matrix for the subsets of training sequences and the cross-correlation matrix for the set of training sequences and the received signals.

In a preferred embodiment of the invention, the received signals are transmission signals from fading channel and the transmission signals are superimposed signals of the initial training sequence and an information sequence, where the initial training sequence is a stationary signal and is uncorrelated with the information sequence.

Another object of the invention is to provide a channel estimation apparatus for wireless communication systems, to improve efficiency for estimation of channel parameters.

Accordingly, the invention provides a channel estimation apparatus based on implicit training sequences, comprising: a generation unit, for generating a set of training sequences comprising a plurality of subsets of training sequences, each of the plurality of subsets being generated based on a known initial training sequence and a predetermined channel order; a first calculation unit, for calculating the cross-correlation matrix for the subsets of training sequences and the corresponding inverse matrix; a second calculation unit, for calculating the cross-correlation matrix for the set of training sequences and the received signals; and an estimation unit, for estimating channel parameters based on the inverse cross-correlation matrix for the subsets of training sequences and the cross-correlation matrix for the set of training sequences and the received signals.

The invention further comprises a receiver, which comprises the channel estimation apparatus provided by the invention.

Other objects and attainments together with a fully understanding of the invention will become apparent and appreciated by referring to the following descriptions and claims taken in conjunction with the accompanying drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

Fig.l illustrates a burst with a training sequence inserted into data segments;

Fig.2 shows a burst containing an implicit training sequence;

Fig.3 shows a MIMO communication system in which channel estimation is performed by using implicit training sequences;

Fig.4 is a flow chart showing the channel estimation method according to an embodiment of the invention; and

Fig.5 is a block diagram showing the channel estimation apparatus according to an embodiment of the invention.

Throughout all the above drawings, same reference numerals will be understood to refer to similar or corresponding features or functions.

DETAILED DESCRIPTION OF THE INVENTION

Compared with the existing channel estimation by superimposing periodic training sequences on the information sequences for transmission and using the signal's cyclic- stationary characteristic at the receiving side, the channel estimation method based on implicit training sequences provided in the invention is different in that the stationary training sequences are superimposed on information sequences for transmission and channel estimation is done at the receiving side mainly using the incorrelate characteristic between the training sequences and the information sequences, that is, channel estimation is done based on the principle that the estimation result of a channel parameter converges to a Wiener solution under the condition that the training sequences and data sequences in the transmission signals are uncorrelated. According to the method of the invention, the training sequences formed by pseudo-random signals can easily satisfy the requirement of optimal correlation function and the corresponding estimation performance is unrelated to the channel order.

The above-mentioned channel order may be defined as the number of discernible multi-path components in the received signals from fading channel. Specifically, the signal bandwidth W is much higher than the channel correlation bandwidth Fc in broadband communication such that the channel shows a frequency selective channel characteristic. At this time, the multi-path components of the signal are discernible and the resolution is IAV. Since the multi-path delay spread is T=l/Fc, T*W discernible components may be generated. Thus, the frequency selective channel may be modeled as a tapped delay-line (FIR) filter with an order (i.e. the number of taps) of L=T*W+1.

A detailed description will be made in conjunction with the drawings below to the principle, method and apparatus for channel estimation based on implicit training sequences as provided in the present invention, in which a relatively complicated MIMO (multi-input multi-output) system in communication systems is illustrated.

Fig.3 shows a system diagram in which channel estimation is done by using implicit training sequences in a MIMO wireless communication system.

Referring to Fig.3, let's consider a multi-antenna wireless communication system having Nt transmit antennas (2-1,...,2- M) and Nr receive antennas (4-1, ..., A- Nr ), in which only the equivalent part of the base -band signal is shown for purpose of simplicity. The basic principle for the channel estimation in the communication system in Fig.3 can be summarized as: when the information sequence on the ith transmit antenna (4- / ,

I = I,-", Nt ) is denoted by S₁ («) , the vector s(n) = [s_l (n), - - -, s_Nt (n)]^T is defined to be the information sequence corresponding to each transmit antenna, in which S₁ (n) is assumed to have a mean value of zero and the information sequences are independent identically distributed between the various antennas. The training sequence {t_t (n)}, i.e. the known initial training sequence, are superimposed synchronously on an information sequence [s^n)} in the adder (1- / , i = l,-- -,Nt ), to form a transmission sequence

[X₁ (It) = s _t {ή) + 1 _t {ή)} , i = \,"-,Nt , to be transmitted via the corresponding i^th transmit antenna. The initial training sequences are stationary signals and the number of them is equal to the number of the active transmit antennas, and different transmit antennas transmit different sequences, that is, the training sequences correspond to the transmit antennas one-to-one.

At the receiving side, the channel estimation apparatus 5 processes the signals received via the receive antenna (4-1, ..., A- Nr ) to estimate channel parameters and the signal detector 6 processes the received signals by using the obtained channel parameters to detect and recover the corresponding information sequence.

Fig.4 is a flow chart showing the channel parameter estimation method based on the implicit training sequences according to an embodiment of the invention. The method comprises steps of: obtaining a set of training sequences comprising a plurality of subsets of training sequences, each of the plurality of subsets being generated based on a known initial training sequence and a predetermined channel order (SlO); calculating the cross-correlation matrix for the subsets of training sequences and the corresponding inverse matrix (S20); calculating the cross-correlation matrix for the set of training sequences and the received signals (S30); and estimating channel parameters based on the inverse cross-correlation matrix for the subsets of training sequences and the cross-correlation matrix for the set of training sequences and the received signals (S40). A detailed description will be given below with reference to Fig.3 and Fig.4 to the channel estimation optimization principle on which the invention is based and to the method provided therein.

The initial training sequence and the information sequence are generated through independent physical processes, so it may be assumed that the initial training sequence and the information sequence are uncorrelated. The channel from the i^th transmit antenna to the i^th receive antenna may be modeled as a FIR filter h _β - [h_β (0), • • • , h_β (L - l)]^τ , where L is the maximum channel order, and then the equivalent received baseband signals on the j th receive antenna are:

X₁ (H) x_Nt (n) y_} (n) = h ⁺ - ⁺ K»r + W₁ (H) (1)

X₁ (H - L + !) x_Nt (n - L + ϊ) where W_} (ή) is additive white Gaussian noise. Let's define h^ = \h.^T _Ji,---,h.^T _jNt]^r , then h_j may be determined to be a vector having Lx M dimensions with reference to the definition of h_p . x(n) = [X₁ («), • • • , x_x (n - L + 1), ^{■ ■ ■},x_m (n),---,x_m (n - L + V)\^T is defined as the transmission signal corresponding to channel h_} , then x(n) is a vector of Lx Nt dimensions. Eq.(l) may be simplified in form of vector: y, (n) = h^x(n) + w, (n) (2)

Considered there are Nr receive antennas, the received signals at instant n are:

Eq.(3) may be given in form of matrix as: y(n) = H^rx(n) + w(n) (4) where y(n) = [J₁ («), • • -, y_Nr (n)]^τ , H = \h_l,---,h_Nr] , and the corresponding additive white noise vector is w(«) = [W₁ (n), • • , w_Nr (n)]^τ .

The parameter estimation result for channel h_p is denoted by h_β , and

h_j = Ih¹ _J1 , ---S⁷ _JNtY is defined as the filter coefficients for the channel parameter estimation. The set of initial training sequences acts as the filter input for channel estimation, and the output result is: b_j (n) = h^τ _j t(n) (5) where t(n) is a set of training sequences generated based on a group of known initial training sequences superimposing on the corresponding information sequences and the maximum channel order, and may further be given by: t(n) = [F₁ (Zi), I₂ {n),- - , J_NM)Y

I₁ (n) = [t, («Vi (» - 1), ^{• • •} , h (n - L + l)J h M = [t₂ (n),t₂(n - 1), ^{• • •} , t₂ (n - L + l)J (6)

^t _Nt M = [*N, (» V Nt (^H -¹V"_' *N, (^{H ~ L} + ¹J] where I₁ (n) is obtained by shifting the known initial training sequence t_t {n) circularly with 0,1,..., L - 1 units. By comparing the filter output with the received signals, we may obtain the error information e_} (n) as:

e_J (n) = y_J (n) -b_J (n) = h^T _J x(n) + w_J (n) -h^T _Jt(n) (7A)

According to Eq. (7A), by taking the MSE as the cost function, we may get: ξ_J=E{\e_J(n)\²} = E{\h^τ _jx(n) + w_J(n)-h^τ _jt(n)\²} (7B)

According to the optimization theory, when the cost function ξ_} in Eq.(7B) is at minimum, the estimation result of a channel parameter converges to a Wiener solution, that is, the coefficients of the estimator will converge to be the channel parameters. It may be seen from the reference by Simon Hayhin, "Adaptive Filter Theory, Fourth Edition," Prentice-Hall, Inc.2002 that the necessary and sufficient condition for minimization of ξ_} is that the partial derivative of ξ_} with respect to h^* _t (k) is zero. By substituting Eq. (2), we can obtain:

where [ ]^* represents conjugate operation and i = !,•'-, Nt; k = 0,---,L — 1. From Eq.(8), we may further get:

E{[y_j(k)-b_j(k)]f(n-k)} = 0 i=l,--,Nt; k = O,--,L-l (9)

that is, the error between the received signals and the estimation filter output at instant k must be orthogonal to each initial training sequence inputted to the estimation filter at the instant.

From the above analysis, the task for channel estimation may be redefined as estimation of the coefficient h^ of the channel parameter filter by using the received signal y _}{n) and the set of training sequences t(n) of the initial training sequences {t_t(n)} superimposing on the information sequence, so as to make it converges to the channel parameter h _} .

The cross-correlation matrix for the subsets of training sequences is denoted by R_t , and the cross-correlation matrix for the received signals y _} (n) and the set of training sequences is denoted by R _{v t} , and the corresponding correlation matrixes are written as:

R_f = ; (10A)

R R ^R ,,t = (10B)

R

InEq.(lOA),

is the cross-correlation matrix for subsets of training sequences

-l)."-Λ i(»-£ + l)} and t_l2(n)={t_l2(n),t_l2(n-l),---,t_l2(n-L + l)} , where /1, /2 = 1,2, • • • , Nt . In Eq.(lOB),

is the cross-correlation vector between the received signals [y^n)], j = 1,2,- ^■-, Nr and the subsets of training sequences {t_t (H)J₁ (n - 1), • • • , t_t (n - L + 1)} , where / = 1,2, ---,Nt. By combining Equations (5), and (9)-(12), Eq.(9) may be obtained in form of matrix as:

R_th,=R_v (13)

Similarly, by extending the result to the received signals on the Nr antennas at the receiving side, Equations (10B) and (13) may be generalized as:

R

^R,_t = (14)

R

R,H=R (15) where R_yt =[R_yit,-,R^_tf .

When the cross-correlation matrix for the subsets of training sequences is denoted by the inverse of R_t , we may obtain the convergence coefficients for the MIMO channel estimation or filter as:

H= R₁-¹R _yt (16)

The above description is made to the basic principle of channel parameter estimation based on implicit training sequences provided in the invention, and the tasks for channel estimation are redefined based on this and a corresponding solution is provided. A further introduction will be given below to the estimation method shown in Fig.4, with reference to the above optimization principle for channel parameter estimation.

In step SlO, the set of training sequences t(n) shown in Eq. (6) is generated based on a group of known initial training sequences {t_t (n), i = 1,2, • • • , Nt] superimposing on the corresponding information sequences and the known predetermined channel order L. The set of training sequences t(n) includes Nt subsets of training sequences, i.e. t(n) = [t_l (n),t₂ (n),-- -,t_Nt (n)]^T , each of the subsets I₁ (n) is formed individually by circularly shifting a known initial training sequence corresponding to the information sequence sent via a different antenna with 0,1,...,L - I units, that is,

I₁ {n) = [t_t (n),t , (n - 1), • • • , t_l (n - L + 1)] . The predetermined channel order L is a known maximum channel order which may be determined in advance based on the channel characteristic, or obtained by estimating the received signals in other ways before this step is performed.

In step S20, the cross-correlation matrix R_t for the subsets of training sequences and the corresponding inverse matrix RJ^"1 are calculated, to obtain the correlation matrixes shown in Eq.(lOA) and Eq.(ll). When the training sequences superimposing on the transmission information sequences are fixed and the maximum channel order L is relatively stable, the cross-correlation matrix for the subsets of training sequences and the corresponding inverse matrix may be calculated in advance and stored in the buffer of the channel estimation apparatus in the receiver, to boost channel estimation speed. When one of the initial training sequences superimposing on the transmission information sequence and the maximum channel order L changes, the cross-correlation matrix R_t and the corresponding inverse matrix RJ^"1 have to be calculated in real time, to adapt to the dynamic change in system parameters.

In step S30, the cross-correlation matrix R _yt for the received signals and the set of training sequences is calculated with the method shown in Equations (10B), (12) and (14).

In step S40, channel parameters H are estimated as shown in Eq.(16), the channel parameters being the convergence coefficients for the channel filter when being optimized with MSE as the cost function. The channel estimation method based on implicit training sequences described with reference to Fig.4 may be implemented in software, or hardware, or in combination of both. Fig.5 illustrates an apparatus for implementing the method. A description is given below to an apparatus for channel parameter estimation according to an embodiment of the invention with reference to Fig.5.

The channel estimation apparatus 100 shown in Fig.5 comprises a generation unit 10, for generating a set of training sequences t(n) as shown in Eq.(6), based on a group of known training sequences {?,-(«)} superimposing on corresponding information sequences and a known maximum channel order L ; a first correlation calculation unit 20, for calculating the cross-correlation matrix R_t for the subsets of training sequences and the corresponding inverse matrix R^"1 , to obtain the matrix described in Eq.(lOA) and Eq.(ll); a second correlation calculation unit 30, for calculating the cross-correlation matrix R _fi for the set of training sequences and the received signals, with the method shown in Equations (10B), (12) and (14); and an estimation unit 40, for estimating channel parameters H in the way shown in Eq. (16), the channel parameters being the convergence coefficients of the channel filter.

The invention further provides a receiver for wireless communication systems, comprising a reception unit, for receiving signals; an estimation unit, for estimating channel parameters based on implicit training sequences and the received signals, the estimation unit being one provided in the invention; and a detection unit, for detecting the corresponding data based on the channel parameters.

As noted above, the channel estimation method and apparatus based on implicit training sequences provided in the invention is characterized in that stationary training sequences are superimposed on information sequences for transmission and channel estimation is done at the receiving side mainly using the uncorrelated characteristic between the training sequences and the information sequences. With the proposed method, a training sequence formed by pseudo-random signals may easily satisfy the requirement for an optimal correlation function by knowing the maximum channel order only, and its performance is unrelated with the practical channel order.

It is to be understood by those skilled in the art that the method and apparatus provided in the invention may be used for estimation of non time-varying channels, real-time tracking of time- varying channels, as well as detection of the data from multiple transmission units by integrating with other components of the receiver to form a integral correlation detector.

It is to be understood by those skilled in the art that the channel estimation method and apparatus based on implicit training sequences for use in mobile communication systems as provided in the invention is applicable not only to MIMO wireless communication systems, but also to SISO, SIMO and MISO wireless communication systems.

It is to be understood by those skilled in the art that various improvements and modifications can be made to the channel estimation method and apparatus based on implicit training sequences as disclosed in the present invention without departing from the spirit and scope of the present invention, the scope of the present invention is to be defined by the attached claims herein.

Claims

CLAIMS:

1. A method for estimating channel based on implicit training sequences, comprising steps of:

(a) obtaining a set of training sequences comprising a plurality of subsets of training sequences, each of the plurality of subsets being generated based on a known initial training sequence and a predetermined channel order;

(b) calculating the cross-correlation matrix for the subsets of training sequences and corresponding inverse matrix;

(c) calculating the cross-correlation matrix for the set of training sequences and the received signals; and

(d) estimating channel parameters based on the inverse cross-correlation matrix for the subsets of training sequences and the cross-correlation matrix for the set of training sequences and the received signals.

2. The method as claimed in claim 1, wherein the received signals are transmission signals from fading channel and the transmission signals are superimposed signals of the initial training sequence and an information sequence, wherein the initial training sequence is a stationary signal and is uncorrelated with the information sequence.

3. The method as claimed in claim 2, wherein the predetermined channel order is a known maximum channel order and the channel order is the number of discernible multi-path components in the received signals from fading channel.

4. The method as claimed in claim 2 or 3, when the known initial training sequence is denoted by {t^njj = 1,2,- ••, Nt) , Nt is the number of active transmit antennas, the predetermined channel order is denoted by L , and the set of training sequences is denoted by t(n) , then the set of training sequences generated at step (a) is as: t(n) = [t_l (n),t₂ (n),---,t_Nt (n)]^T h M = h («)>* i (^Λ - 4 ^{• • •} _' h iⁿ - ^L + ^!)1

I₂ (n) = [t₂ (n),t₂ (n-l),---,t₂(n-L + l)J

t_N, M =

L + 1)] wherein t_l(n)= {t_ι(n),t_ι(n-l),---,t_ι(n-L + ϊ}} is a subset of training sequences, each of the subsets is obtained by shifting the known initial training sequence ?₍(n) circularly with 0,1,...,L-I units.

5. The method as claimed in claim 4, wherein the cross-correlation matrix R_t for the subsets of training sequences at step (b) is calculated as:

R R

R. =

R 'mh R 'm'm (nj where R , , =

is the cross-correlation matrix for two subsets of training sequences h_ι{n)={t_ιl{n),t_ιl{n-\),---,t_ιl{n-L + \)} and t_l2(n)= {t_l2(n),t_l2(n-l),---,t_l2(n-L + l)} , where /1, ϊl = 1,2, ---,Nt, and [ ]^* represents conjugate operation.

6. The method as claimed in claim 5, when the number of active receive antennas is Nr and the corresponding received signals are [y^n), j = 1,2,- --,Nr] , the cross-correlation matrix R _yt for the set of training sequences and the received signals at step (c) is given by:

R Vit

R . =

R y_NA

where R_v y Λ_t

where R =

+ l) is the cross-correlation vector between the received signals {;y₇ («)}, ./ = 1,2,- ^■ - , Nr and the subset of training sequences {t_t (It)J₁ (n - 1), • • • , t_l (n - L + 1)} , where / = 1,2, - - -, Nt , and [ J represents conjugate operation.

7. The method as claimed in claim 6, when the channel parameters are represented by channel estimation matrix H , which is a Nt * L * Nr order matrix formed by channel vector h _β from the i^th transmit antenna to the j^th receive antenna, that is:

H = [h₁₅- - -,hJ

where h^ = [h^,- - -,h^_r]^r , then the corresponding channel parameter estimation

matrix H is given by:

H = R"R_yt where RJ^"1 is the inverse matrix for the cross-correlation matrix R_t for the subsets of training sequences, and R _yt is the cross-correlation matrix for the set of training sequences and the received signals.

8. An apparatus for estimating channel based on implicit training sequences, comprising: a generation unit, for generating a set of training sequences comprising a plurality of subsets of training sequences, each of the plurality of subsets being generated based on a known initial training sequence and a predetermined channel order; a first calculation unit, for calculating the cross-correlation matrix for the subsets of training sequences and corresponding inverse matrix; a second calculation unit, for calculating the cross-correlation matrix for the set of training sequences and the received signals; and an estimation unit, for estimating channel parameters based on the inverse cross-correlation matrix for the subsets of training sequences and the cross-correlation matrix for the set of training sequences and the received signals.

9. The apparatus as claimed in claim 8, wherein the received signals are transmission signals from fading channel and the transmission signals are superimposed signals of the initial training sequence and an information sequence, wherein the initial training sequence is a stationary signal and is uncorrelated with the information sequence.

10. The apparatus as claimed in claim 9, wherein the predetermined channel order is a known maximum channel order and the channel order is the number of discernible multi-path components in the received signals from fading channel.

11. The apparatus as claimed in claim 9 or 10, when the known initial training sequence is denoted by {t^njj = 1,2, ^{■ ■}-, Nt] , Nt is the number of active transmit antennas, the predetermined channel order is denoted by L , and the set of training sequences is denoted by t(n) , then the set of training sequences generated in the generation unit is as: t(n) = [t_l (n),t₂ (n),---,t_Nt (n)]^T h M = [V₁ M>^ri (^Λ -I),- ^■ -J₁(H - L + I)J

I₂ (n) = [t₂ (n),t₂(n - l), ^{■ ■ ■} , t₂ (n - L + l)J

t_m M = L (n),t _Nt (n - 1), ^{• • •} , t_Nt (n - L + 1)] wherein

is a subset of training sequences, each of the subsets is obtained by shifting the known initial training sequence t,.(n) circularly with 0,1,...,L-I units.

12. The apparatus as claimed in claim 11, wherein the cross-correlation matrix R_t for the subsets of training sequences calculated by the first calculation unit is as:

R R h'm

R. =

R R

where R =

is the cross-correlation matrix for two subsets of training sequences t_ιι(n)={t_ιι(n),t_ιl(n-^'ή,-,t_ιl(n-L + l)} and t_l2{n)={t_l2{n),t_l2{n-l),--,t_l2{n-L + \)} , where /1, ϊl = 1,2, ---,Nt, and [ ]^* represents conjugate operation.

13. The apparatus as claimed in claim 12, when the number of active receive antennas is Nr and the corresponding received signals are [y^n), j = 1,2,- --,Nr] , the cross-correlation matrix R _yt for the set of training sequences and the received signals calculated by the second calculation unit is as: R ^",Vit

R . =

R VΛfrt

where R_yt

where R

is the cross-correlation vector between the received signal

j = \,2,---,Nr and the subset of training sequences

(H)J₁ (n - 1), • • • , t_t (n - L + 1)} , where / = 1,2, ---,Nt, and [ ]^* represents conjugate operation.

14. The apparatus as claimed in claim 13, when the channel parameters are represented by channel estimation matrix H , which is a Nt* L* Nr order matrix formed by channel vector h _β from the i^th transmit antenna to the j^th receive antenna, that is:

H=[h₁,...,h_Λfr] where h^ =[h^,-",h^_r]^r, then the corresponding channel parameter estimation

matrix H is given by the estimation unit based on the following equation:

H=R"R_yt where R, ¹ is the inverse matrix for the cross-correlation matrix R_t for the subsets of training sequences, and R_yt is the cross-correlation matrix for the set of training sequences and the received signals.

15. A receiver for wireless communication systems, comprising; a reception unit, for receiving signals; an estimation unit, for estimating channel parameters based on implicit training sequences and the received signals, the estimation unit being the estimation unit as claimed in any of claims 8 to 14; and a detection unit, for detecting the corresponding data based on the channel parameters.