Ofdm Channel Estimation and Tracking for Multiple Transmit Antennas

Abstract

Multiple Transmit Multiple Receive Orthogonal Frequency Division Multiplexing (‘OFDM’) comprising generating bit streams and corresponding sets of N frequency domain carrier amplitudes ({tilde over (s)}(kN+j), 0≦j≦N−1) modulated as OFDM symbols subsequently to be transmitted from a transmitter, where k is the OFDM symbol number and j indicates the corresponding OFDM carrier number. Affix information is inserted at the transmitter into guard intervals between consecutive time domain OFDM symbols and are used at the receiver to estimate the Channel Impulse Response (Hlm) of the transmission channels, the estimated Channel Impulse Response (Ĥlm) being used to demodulate the bit streams in the signals received at the receiver. The affix information is known to the receiver, as well as to the transmitter, and is mathematically equivalent to a vector (cD) that is common to the time domain OFDM symbols multiplied by at least first weighting factors (αk) that are different for one time domain OFDM symbol (k) than for another and second weighting factors (wi(k)) that enable one of the transmit antenna means (i) to be distinguished from another.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a schematic block diagram of a transmitter in a communication system in accordance with one embodiment of the invention, given by way of example,

FIG. 2 is a schematic block diagram of a receiver in the communication system whose transmitter is shown in FIG. 1,

FIG. 3 is a diagram of signals appearing in operation of the modulator of FIG. 2, and

FIG. 4 is a diagram of a moving average Doppler module for the demodulator of FIG. 1,

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

FIG. 1 and FIG. 2 show an OFDM communication system in accordance with one embodiment of the invention comprising a transmitter including an OFDM modulator 1 and a receiver including an OFDM demodulator 2, the transmitter and the receiver communicating over a communication channel 3.

The OFDM communication method of this embodiment of the present invention enables estimation and tracking of the Multiple Input Multiple Output (‘MIMO’) channels in coherent multiple transmit antenna/multiple receive antenna (‘MTMR’) systems, without any specific limit to the number of transmit (TX) and receive (RX) antennas and without imposing any particular Space-Time Code (STC). Data and affix vectors are independently encoded by two STC and enable a semi-blind estimation of all the MIMO channels exploiting only the order-one statistics of the received signal.

In the following description, lower (upper) boldface symbols will be used for column vectors (matrices) sometimes with subscripts N or P emphasizing their sizes (for square matrices only); tilde (‘{tilde over ( )}’) will denote frequency domain quantities; argument i will be used to index blocks of symbols; H(T) will denote Hermitian (transpose) operations.

FIG. 1 depicts the baseband discrete-time block equivalent model of an N-carrier PRP-OFDM MTMR transceiver with N_t transmit and N_r receive antennae. The communication system is described with reference to Space-Time (ST) block codes (‘STBCs’) but it will be appreciated that the invention is also applicable to other code systems. In the transmitter, the Initial serial bit stream of constellation symbols {tilde over (s)}(jN), . . . , {tilde over (s)}(jN+N−1) is converted to a set of vectors in a serial-to-parallel converter (not shown); the jth N×1 input digital vector {tilde over (s)}(j) is then modulated by an Inverse Fast Fourier Transform (‘IFFT’) matrix F_N^H in a transformer 4, where

${\left[F_N\right]}_{k,l}=\frac{1}{\sqrt{N}}W_N^{\mathrm{kl}},0\le k<N,0\le l<N\mathrm{and}W_N={}^{-\mathrm{j2\pi }/N}.$

The resulting N−1 time domain vector s(j) is processed by a suitable ST encoder matrix in an encoder 10, as shown in FIG. 1, creating outputs S(t)=(s(iN_t), . . . , s(iN_t+N_t−1))={s(iM+k), 1≦l≦N_t, 0≦k<M} where i is the block number l is the number of the TX antenna and n=iM+k indexes the outputs in FIG. 1. It will be appreciated that, at least in the context of STBCs, M can differ from N_t; In particular, the STBC may lead to rectangular S(i), that is to say that M>N_t. For the sake of simplicity, we assume in the following description that operates on N_t; inputs (s(iN_t), . . . , s(iN_t+N_t−1)). However, the invention can be straightforwardly applied to with other numbers of inputs. In this embodiment of the invention, the s(iM+k) are linearly precoded in a precoder 11 by a zero-padded OFDM (‘ZP-OFDM’) precoding matrix T_ZP, where

$T_{\mathrm{ZP}}={\left[\begin{array}{c}{I}_N\\ 0_{D,N}\end{array}\right]}_{P\times N}$

and u(n)=T_ZPs(n), 1≦l≦N_t.

The affix contents v(n) are added to the data symbols u(n)=T_ZPs(n), 1≦l≦N, resulting in the output vectors q(n). The output vectors q(n) are converted to a series signal by a parallel-to-series converter 6, a pseudo random postfix being inserted in the signal into guard intervals between each consecutive OFDM symbol to produce a series digital signal s(n) on the lth TX atenna. The series digital signal s(n) is then converted to an analogue signal s(t) by a digital-to-analogue converter 7 and transmitted over the channel 3.

More particularly, in a preferred embodiment of the invention, the postfix that is added in the guard interval comprises a pre-calculated suitable vector that is independent of the data and that is weighted by a first factor α_k and a second factor w_i(k). In one embodiment of the invention, the first factor α_k is different from one time-domain OFDM symbol to another and is known both to the transmitter 1 and to the receiver 2, so that any time domain (cyclo-)stationarity (leading to strong undesired frequency contributions at the repetition frequency) is avoided. In another embodiment of the invention, in which the symbols are coded in blocks, the first factor α_k is different from one time-domain OFDM symbol block to another but is the same for each symbol of the same block. The second factor w_i(k) enables one of the transmit antennas to be distinguished from another.

With an OFDM modulator in the transmitter functioning in this way, semi-blind channel estimation in the receiver can be done simply and at low arithmetical complexity. In particular, the receiver can constantly estimate and track the channel Impulse response without any loss of data bandwidth compared to CP-OFDM, other than the transmission of PR-calculation parameters. Moreover, the demodulator at the receiver can have advantageous characteristics, ranging from very low arithmetical cost (at medium performance) to les low arithmetical cost (with very good system performance).

As described In our copending European Patent Application referred to above for the single antenna case, several choices for the first factor α_k are possible. It is possible to choose α_k of any complex value. However, any α_k with |α_k|≠1 leads to performance degradation compared to preferred embodiments of the invention.

It is possible to limit the choice of α_k, somewhat less generally to a complex value with |α_k|=1. This choice usually leads to good system performance, but the decoding process risks to be unnecessarily complex. Preferred values of the first and second factors are described in more detail below.

Preferably, the first factor α_k is pseudo-random. In one embodiment of the present invention the first factor α_k is deterministic and is calculated both by the modulator 1 and the demodulator 2 using the same algorithm and parameters that are stored in memory both in the transmitter and in the receiver. In another embodiment of the present invention, initialisation parameters for the algorithm are transmitted between the transmitter and the receiver to avoid systematically using the same sequence for the first factor α_k. In yet another embodiment of the present invention, the first factor α_k is communicated from the transmitter 1 to the receiver 2, which still represents an acceptable overhead in certain circumstances.

In the embodiment of the present invention shown in FIG. 1 and FIG. 2, the affix is deterministic and is a D×1 postfix vector c treated in an encoder 12 by a specific ST encoder matrix which outputs the D×1 vectors c(n), 1≦l≦N_t. The way in which ensures identification of the complete MIMO channel is described in more detail below. The postfix vectors c(n), 1≦l≦N, are then linearly precoded in a precoder 13 by a ZP-OFDM matrix T_P, where

$T_P={\left[\begin{array}{c}{0}_{N,D}\\ I_D\end{array}\right]}_{P\times D},$

to produce zero-padded postfix vectors as shown at 14 in FIG. 3.

The resulting vectors v_l(n) are finally added to the data symbols u_l(n) by adders 15: q_l(n)=u_l(n)+v_l(n), 1≦l≦N, to produce signals 16 for transmission.

The signals received at the receive antennas are the transmitted signals multiplied by the Channel Impulse Response (‘CIR’) H_lm and with the addition of noise and interference n_l(n). Let H_lm be a P×P circulant matrix whose first row is given by [h_l(0), 0, . . . 0, h_l(L−1), . . . , h_l(1)], where h_l=[h_l(0), . . . , h_l(L−1), 0, . . . 0]^T is the P×1 channel impulse response between the lth transmit and the mth receive antennae; D is chosen such that D≧L−1. Define H_lm^ISI as the lower triangular part of H_lm including the main diagonal which represents the Intra-Symbol-Interference (ISI); H_lm^ISI shall contain the upper triangular part of H_lm representing the Inter-Block-Interference (IBI), such that H_lm=H_lm^ISI+H_lm^IBI. Therefore, the received signal vector on the mth antenna, 1≦m≦N, is, given by:

$r_m\left(n\right)=\sum _{l=1}^{N_t}\left[H_{lm}^{\mathrm{ISI}}q_l\left(n\right)+H_{lm}^{\mathrm{IBI}}q_l\left(n-1\right)\right]+n_m\left(n\right)$

where n_m(n) is an zero-mean additive white independent identically distributed gaussian noise term.

As shown in FIG. 2, The demodulator 2 at the receiver comprises an analogue-to-digital converter 7 that converts the signals r_m(t) received at the receive antennas to digital signals, a serial-to-parallel converter 8, which converts the received digital signals to received vectors r_m(n), and a demodulator and equaliser 9 that uses decoding matrices corresponding to the encoding matrices and to estimate the Channel Impulse Response CIR and demodulate the OFDM signals.

In the following description of the operation of the receiver, an order-one channel estimation algorithm is described first, assuming the channel to be static. Then, the effect of Doppler is introduced for the mobility case and the corresponding channel estimator in the Minimum Mean Square Error (MMSE) sense described.

First the received vector r_m(n) is expressed in an exploitable form for channel estimation. For that purpose, let H_lm^D be the D×D circulant matrix of first row [h_lm(0), 0, . . . 0, h_lm(L−1), . . . , h_lm(1)]. We define H_lm^ISLD and H_lm^IBLD such that H_lm^D=H_lm^ISLD+H_lm^IBLD. The signal r_m(n), received during the nth OFDM symbol on the mth antenna, 1≦m≦N_t, is equal to:

$\begin{array}{cc}{r}_m\left(n\right)=\sum _{l=1}^{N_t}\left[\begin{array}{c}{H}_{lm}^{\mathrm{ISI},D}{\stackrel{\_}{s}}_{l,0}\left(n\right)+H_{lm}^{\mathrm{IBI},D}{\stackrel{\_}{c}}_l\left(n-1\right)\\ ⋮\\ H_{lm}^{\mathrm{IBI},D}{\stackrel{\_}{s}}_{l,1}\left(n\right)+H_{lm}^{\mathrm{ISI},D}{\stackrel{\_}{c}}_l\left(n-1\right)\end{array}\right]+\left[\begin{array}{c}{n}_{m,0}\left(n\right)\\ ⋮\\ n_{m,1}\left(n\right)\end{array}\right]& \mathrm{Equation}1\end{array}$

where s_l,0(n), s_l,1(n), n_m,0(n) are respectively the first D and last D samples of s_l(n) and n_m(n).

Equation 1 indicates that a superimposition of the various postfixes convolved by the corresponding channels is interfering with the useful data. An easy independent retrieval of each of the channels based on the sole observation of the postfix contributions is obtained through isolation of each postfix convolved by its related channel. As detailed below, a way to achieve that condition is to perform a Fast Fourier Transform on the postfixes in the demodulator and equaliser 9 using a weighting ST block coding scheme of the postfix c according to the following postfix generation process:

$\begin{array}{cc}\left[\begin{array}{ccc}{\stackrel{\_}{c}}_l\left(\mathrm{iM}\right)& \dots & {\stackrel{\_}{c}}_l\left(\mathrm{iM}+M-1\right)\\ ⋮& ⋰& ⋮\\ {\stackrel{\_}{c}}_{N_t}\left(\mathrm{iM}\right)& \dots & {\stackrel{\_}{c}}_{N_t}\left(\mathrm{iM}+M-1\right)\end{array}\right]=\hspace{1em}\underset{\underset{W}{}}{\left[\begin{array}{ccc}{w}_l\left(0\right)\alpha \left(\mathrm{iM}\right)& \dots & w_l\left(M-1\right)\alpha \left(\mathrm{iM}+M-1\right)\\ ⋮& ⋰& ⋮\\ w_{N_t}\left(0\right)\alpha \left(\mathrm{iM}\right)& \dots & w_{N_t}\left(M-1\right)\alpha \left(\mathrm{iM}+M-1\right)\end{array}\right]}\otimes c& \mathrm{Equation}2\end{array}$

where is the Kronecker product and c, α(n) are respectively the deterministic postfix and the pseudo-random weighting factors introduced in our co-pending European Patent Application EP 02292730.5 for the single antenna case. The pseudo-random weighting factors α(n) are used to convert the deterministic postfix c into a pseudo-random one. Note that a new set of deterministic weighting factors is introduced, and gathered In the M×N, matrix W corresponding to the matrix W used for encoding the postfixes in the transmitter encoder, with [W]_k,l−1=w_l(k), 0≦k<M, 1≦l≦N_t. W is used to remove the interference between all transmitted postfixes and thus is Invertible in this embodiment of the present invention: full column rank (rank(W)=n_t). In the following description, we choose W orthogonal for this embodiment of the present invention, such that W^HW=I_Nt.

With the assumption of a static channel, an order-one channel estimator in the demodulator and equaliser 9 functions as follows. The first and last D samples of r_m(n) are denoted respectively by r_m,0(n) and r_m,1(n). By setting n=iM+k and assuming the transmitted time domain signal s_l(n) to be zero mean for all l, we use Equations 1 and 2 to compute for each k, 0≦k<M, the following D×1 vector:

$\begin{array}{cc}{d}_m^k\left(i\right)=\frac{r_{m,1}\left(\mathrm{iM}+k\right)+r_{m,0}\left(\mathrm{iM}+k+1\right)}{\alpha \left(\mathrm{iM}+k\right)}& \mathrm{Equation}3\end{array}$

Next d_m^k=E[d_m^k(i)] is defined as the expectation of d_m^k(i). Due to the deterministic nature of the postfixes, it can be verified from Equation 1 that:

$\begin{array}{cc}{d}_m^k=\sum _{l=1}^{N_t}\left[H_{lm}^{\mathrm{ISI},D}+H_{lm}^{\mathrm{IBI},D}\right]w_l\left(k\right)c=\sum _{l=1}^{N_t}H_{lm}^Dw_l\left(k\right)c& \mathrm{Equation}4\end{array}$

Thus the MD×1 vector d_m=[(d_m⁰)^T, . . . , (d_m^M−1)^T]^T can be expressed for each receive antenna as:

$\begin{array}{cc}{d}_m=\sum _{l=1}^{N_t}\left[\begin{array}{c}{H}_{lm}^Dw_l\left(0\right)c\\ ⋮\\ H_{lm}^Dw_l\left(M-1\right)c\end{array}\right]=\left(W\otimes I_D\right)\left[\begin{array}{c}{H}_{lm}^Dc\\ ⋮\\ H_{N,m}^Dc\end{array}\right]& \mathrm{Equation}5\end{array}$

Since W is chosen orthogonal, multiplying each d_m, 1≦m≦N, by (WI_D)^H in the demodulator and equaliser 9 removes completely the interference between channel contributions H_lm^D, 1≦l≦N_t.

Once the interference between channel contributions is removed the estimation algorithms of the single-antenna case of our co-pending European Patent Application EP 02 292 730.5 can be applied in the demodulator and equaliser 9:

H_lm^Dc=C_Dh_lm^D=F_D^H{tilde over (C)}_DF_Dh_lm^D  Equation 6

where C_D is a D×D circulant matrix with the first row [c(0), c(D−1), . . . , c(1)], {tilde over (C)}_D=diag{F_Dc}, and h_lm^D represents the D first coefficients of h_lm. Hence, the estimate ĥ_lm^D of the time domain channel impulse response h_lm^D in the demodulator and equaliser 9 is obtained by multiplying H_lm^D by F_D^H{tilde over (C)}_D⁻¹F_D, 1≦l≦N_t, 1≦m≦N_t. Note that {tilde over (C)}_D⁻¹ is a diagonal matrix that is known to both the transmitter and receiver and can thus be precalculated. Subsequently, h_lm^D is preferably transformed to the P×1 frequency domain vector

${\stackrel{\widehat{~}}{h}}_{lm}={F_P\left[I_D^T,0_{N,D}^T\right]}^T{\hat{h}}_{lm}^D.$

This MIMO channel estimation (i.e. estimation of all channels between any transmit and any receive antenna) is used to space-time decode and equalise the received data signals, as described in more detail in examples below, such that the transmitted data signals can be recovered.

The above channel estimator can be extended to further improve reception in mobile environments by using any Doppler model, and by minimizing any performance criterion. An example is now given, in a preferred embodiment of the present invention based on the introduction of a Doppler model; the estimator aims at minimizing the Mean Square Error (MSE).

The Doppler module shown in FIG. 4 is introduced in the demodulator and equaliser 9 to modify the order-one autoregressive model for the Channel Impulse Response (‘CIR’) between transmit antenna l and receive antenna m separately: h_lm^D(n)=J₀(2πƒ_DΔT)h_lm^D(n−1)+ h_lm^D(n) where J₀(·) is the 0th order Bessel function. ƒ_D is the Doppler frequency, ΔT is the MTMR PRP-OFDM block duration and {hacek over (h)}_lm^D(n) is zero-mean complex Gaussian of constant variance. The same CIR correlations as the ones provided by the known Jakes model are obtained even in the presence of large Doppler frequencies. This is achieved by forcing the correlation E└(h_lm^D(n))^Hh_lm^D(k)┘=J₀(2πƒ_D(k−n)ΔT)∀k>n. This way the approximation J₀(2πƒ_D(k−n)ΔT)=(J₀(2πƒ_DΔT))^(k−n), inherent to the order-one autoregressive estimation, is avoided. This modification leads to the following moving average estimation:

$\begin{array}{cc}{h}_{lm}^D\left(n\right)=\sum _{k=0}^nJ_0\left(2\pi f_Dk\Delta T\right){\stackrel{⋓}{h}}_{lm}^D\left(n-k\right)& \mathrm{Equation}7\end{array}$

As for the order-one autoregressive model, so-called process-noise vectors {hacek over (h)}_lm^D(n) are introduced assuming E└ h_lm^D(n)({hacek over (h)}_lm^D(n))^H┘=0 for n≠k and {hacek over (h)}_lm^D(0)=h_lm^D(0) being the CIR to be estimated. Assuming that CIR h_lm^D(n) is estimated based on Z noisy observations d_lm(i)={{hacek over (d)}_lm(k), i≦k<i+Z}, the expression {hacek over (d)}_lm(i)=└(WI_D)^Hd_m(i)┘_k, (l−1)D≦k>lD results from the convolution of the ith block postfix by channel h_lm^D(n) corrupted both by thermal noise and the OFDM data symbols. The OFDM data symbols are assumed zero-mean and independent of same variance as the postfix samples. d_m(i)=[(d_m⁰(i))^T, . . . , (d_m^M−1(i))^T]T contains the received symbols after equalization of the pseudo random weighting factor of the postfixes (Equation 3). Thus, d_lm(i) can be expressed as follows:

$\begin{array}{cc}{\stackrel{\_}{d}}_{lm}\left(i\right)=\left(J\otimes C_D\right)\left[\begin{array}{c}{\stackrel{⋓}{h}}_{lm}^D\left(0\right)\\ {\stackrel{⋓}{h}}_{lm}^D\left(1\right)\\ ⋮\\ {\stackrel{⋓}{h}}_{lm}^D\left(Z-1\right)\end{array}\right]+n,\mathrm{where}J=\left[\begin{array}{cccc}1& 0& \dots & 0\\ J_0\left(2\pi f_D\Delta T\right)& 1& ⋰& 0\\ ⋮& ⋰& ⋰& ⋮\\ J_0\left(2\pi f_D\left(Z-1\right)\Delta T\right)& J_0\left(2\pi f_D\left(Z-2\right)\Delta T\right)& \dots & 1\end{array}\right]& \mathrm{Equation}8\end{array}$

where n gathers the thermal noise and the interference from the OFDM data symbols. In order to guarantee the unit variance of each channel realisation, the norm of {hacek over (h)}_lm^D(n), n=0, . . . , Z−1 is chosen such that:

$\begin{array}{cc}\sum _{n=0}^{Z-1}{\left[J\right]}_{k,n}^2E\left[{\left({\stackrel{⋓}{h}}_{lm}^D\left(n\right)\right)}^H{\stackrel{⋓}{h}}_{lm}^D\left(n\right)\right]=1,k=0,\dots ,Z-1.& \mathrm{Equation}9\end{array}$

It can thus be verified that the optimum estimator of h_lm^D(0)={hacek over (h)}_lm^D(0) in the MMSE sense is given from Equation 8 by:

$\begin{array}{cc}{\hat{h}}_{lm}^D\left(i\right)={M\left[\left(J\otimes C_D\right){D_{R_b}\left(J\otimes C_D\right)}^H+R_n\right]}^{-1}{\stackrel{\_}{d}}_m\left(i\right)\mathrm{with}\text{:}D_{R_b}=\mathrm{diag}\left\{R_{{\stackrel{⋓\hspace{1em}}{h}}_{lm}^0\left(0\right)},\dots ,R_{{\stackrel{⋓}{h}}_{lm}^0\left(Z-1\right)}\right\},M=\left[1,\dots ,J_0\left(2\pi f_D\left(Z-1\right)\Delta T\right)\right]\otimes \left(R_{{\stackrel{⋓}{h}}_{lm}^0\left(0\right)}C_D^H\right)& \mathrm{Equation}10\end{array}$

and R_m=E└aa^H┘ is the auto-correlation matrix of the vector a.

Since in practice the channel power profile is usually not known, in that case the assumption is made that

$R_{{\stackrel{⋓}{h}}_{lm}^D\left(n\right)}\approx g_nI_D$

for all n. The real gain g_n is introduced for respecting the power constraints of Equation 9.

The above description presents generic channel estimation in the demodulator and equaliser 9 in accordance with embodiments of the present invention for both relatively static and high mobility environments. Their use for two STBC systems will now be described.

The first embodiment of STBC is based on ZP-OFDM decoding. The system includes modulators using pseudo-random postfixes at the transmitter and also equalizer structures derived for the MTMR case from those described for the Single Transmit Single Receive (STSR) case in our co-pending European Patent Application EP 02 292 730.5 based on the transformation of the received PRP-OFDM vector to the ZP-OFDM case. The system is described for the case of N_t=2 transit and N_r=1 receive antennas, although it will be appreciated that the system is applicable to other numbers of antennas. The ST encoder operates over N_t×M vectors with N_t=M=2. Since N_r=1, the subscript 1≦m≦N_r is not used in the following analysis. Perfect knowledge of the channels h_l, I≦l≦N_t is assumed but it will be appreciated that the system is capable of working with imperfect channel knowledge.

At the transmitter, a 2×1 ZP-ST encoder is used, which takes two consecutive OFDM symbols s(2i) and s(2i+1) to form the following coded matrix:

$\begin{array}{cc}\left[\begin{array}{cc}{\stackrel{\_}{s}}_1\left(2i\right)& {\stackrel{\_}{s}}_1\left(2i+1\right)\\ {\stackrel{\_}{s}}_2\left(2i\right)& {\stackrel{\_}{s}}_2\left(2i+1\right)\end{array}\right]=\left[\begin{array}{cc}s\left(2i\right)& -P_N^0s^{*}\left(2i+1\right)\\ s\left(2i+1\right)& P_N^0s^{*}\left(2i\right)\end{array}\right]& \mathrm{Equation}11\end{array}$

where the permutation matrices P_Jⁿ are such that, for a J×1 vector a=[α(0) . . . , α(J−1)]^T, we have {P_N⁰a}_p=α((J−p+n)mod J).

Since the channel has been estimated, as for the single antenna case described in our co-pending European Patent Application, it is always possible to retrieve the MTMR ZP-OFDM signals from Equation 1 by subtracting from the received signal the known PRP contribution:

$\begin{array}{cc}{r}^{\mathrm{ZP}}\left(n\right)=r\left(n\right)-\sum _{l=1}^2\left[H_l^{\mathrm{IBI}}v_l\left(n-1\right)+H_l^{\mathrm{ISI}}v_l\left(n\right)\right]& \mathrm{Equation}12\end{array}$

which leads to

$r^{\mathrm{ZP}}\left(n\right)=\sum _{l=1}^2H_lT_{\mathrm{ZP}}{\stackrel{\_}{s}}_l\left(n\right).$

Note that i) no constraint has to be set on W for the symbol recovery, ii) the PRP interference cancellation procedure proposed is generic and can be applied to any suitable ST encoder .

A suitable detection algorithm is applied to the signal described by Equation 12 by the demodulator and equaliser 9. Noticing that P_P^NT_ZP=T_ZPP_N⁰, we denote by {tilde over (D)}₁=diag{{tilde over (h)}₁}, {tilde over (D)}₁={{tilde over (h)}₁}, ñ(2i)=F_pn(2i) and ñ(2i+1)=F_pP_P^Nn·(2i+1); then if we switch to the frequency domain by computing {tilde over (y)}(2i)=F_pr^ZP(2i) and {tilde over (y)}(2i+1)=F_p(P_P^Nr^ZP(2i+1))ⁿ, exploiting the fact that H_l=F_P^H{tilde over (D)}_lF_P, 1≦l≦2, we can write:

$\begin{array}{cc}\left[\begin{array}{c}\tilde{y}\left(2i\right)\\ \tilde{y}\left(2i+1\right)\end{array}\right]=\underset{\underset{=\tilde{D}}{}}{\left[\begin{array}{cc}{\tilde{D}}_1& {\tilde{D}}_2\\ {\tilde{D}}_2^{*}& -{\tilde{D}}_1^{*}\end{array}\right]}\left[\begin{array}{c}{F}_PT_{\mathrm{ZP}}s\left(2i\right)\\ F_PT_{\mathrm{ZP}}s\left(2i+1\right)\end{array}\right]+\left[\begin{array}{c}\tilde{n}\left(2i\right)\\ \tilde{n}\left(2i+1\right)\end{array}\right]& \mathrm{Equation}13\end{array}$

where {tilde over (D)} is an orthogonal channel matrix. Thus multiplying [{tilde over (y)}(2i)^T {tilde over (y)}(2i+1)^T]^T by {tilde over (D)}^H achieves the separation of the transmitted signals s(2i) and s(2i+1), and it can be shown that full transmit diversity is achieved. Note that the separation of signals allows the same equalisation schemes to be used in this embodiment of the present invention as in the single-antenna case described in our co-pending European Patent Application EP 02 292 730.5.

The second embodiment of STBC system is based on equalization of the full received block by diagonalisation of pseudo-circulant channel matrices. The ST data encoder used in the demodulator and equaliser 9 is based on a version of the single antenna system described in our co-pending European Patent Application EP 02 292 730.5 modified to enable the equalization structure that is detailed below and outputs blocks N_l×M vectors with N_l=M=2. and are specified such that they generate the following matrix Q(i)={q_l(2i+k), 1≦l≦2, 0≦k<2} at the antenna outputs:

$\begin{array}{cc}Q\left(i\right)=\left[\begin{array}{cc}\left[\begin{array}{c}s\left(2i\right)\\ \alpha \left(2i\right)c\end{array}\right]& -P_P^0Q_{\beta \left(i\right)}\left[\begin{array}{c}{s}^{*}\left(2i+1\right)\\ \alpha \left(2i+1\right)c^{*}\end{array}\right]\\ \left[\begin{array}{c}s\left(2i+1\right)\\ \alpha \left(2i\right)c\end{array}\right]& P_P^0Q_{\beta \left(i\right)}\left[\begin{array}{c}{s}^{*}\left(2i\right)\\ \alpha \left(2i+1\right)c^{*}\end{array}\right]\end{array}\right]& \mathrm{Equation}14\end{array}$

P_P⁰ being a permutation matrix defined as previously (inversing the order of the vector elements), α(i) is complex with |α(i)|=1 being pseudo-random complex weighting factors as defined in our co-pending European Patent Application for the single antenna case with α(2i+1)=β²(i)α(2i), and β(i)=α(2i=2)/α(2i). Q_β(i) is defined as:

$Q_{\beta \left(i\right)}=\left[\begin{array}{cc}{0}_{D\times N}& \beta \left(i\right)·I_D\\ I_N& 0_{N\times D}\end{array}\right].$

The D×1 postfix c is chosen such that it has Hermitian symmetry, that is to say that the complex conjugate of the vector read backwards is equal to the original c. As in our co-pending European Patent Application, the channels are represented by P×P pseudo-circulated channel matrices H_l^β(i), 1≦l≦2. These are identical to standard circulant convolution matrices with the upper triangular part multiplied by the scalar factor β(i), in other words H_l^β(i)=H_l^ISI+β(i)H_l^IBI.

With R(i)=[r^T(2i) r^T(2i+1)]^T and the noise matrix N(i)=[n^T(2i) n^T(2i+1)]^T, the received signals over M=2 symbol times are given as follows:

$R\left(i\right)=\left[\begin{array}{c}\sum _{l=1}^2\left[H_l^{\mathrm{IBI}}q_l\left(2i-1\right)+H_l^{\mathrm{ISI}}q_l\left(2i\right)\right]\\ \sum _{l=1}^2\left[H_l^{\mathrm{IBI}}q_l\left(2i\right)+H_l^{\mathrm{ISI}}q_l\left(2i+1\right)\right]\end{array}\right]+N\left(i\right)$

With {circumflex over (R)}(i)=[{circumflex over (r)}^T(2i) {circumflex over (r)}^T(2i+1)]^T, the following operations are performed at the demodulator and equaliser 9 on the received vectors:

$\begin{array}{c}\hat{R}\left(i\right)=\left[\begin{array}{c}r\left(2i\right)+2c_1^{\beta \left(i\right)}\left(i\right)\\ {Q_{\beta \left(i\right)}^H\left(P_P^0\left(r\left(2i+1\right)+2c_2^{\beta \left(i\right)}\left(i\right)\right)\right)}^{*}\end{array}\right]\\ =\underset{\underset{=W_H\left(i\right)}{}}{\left[\begin{array}{cc}{H}_1^{\beta \left(i\right)}& H_2^{\beta \left(i\right)}\\ {\left(H_2^{\beta \left(i\right)}\right)}^H& -{\left(H_1^{\beta \left(i\right)}\right)}^H\end{array}\right]}\left[\begin{array}{c}\left[\begin{array}{c}s\left(2i\right)\\ \alpha \left(2i\right)c\end{array}\right]\\ \left[\begin{array}{c}s\left(2i+1\right)\\ \alpha \left(2i\right)c\end{array}\right]\end{array}\right]\end{array}$ $\mathrm{where}$ $\begin{array}{c}{c}_1^{\beta \left(i\right)}\left(i\right)=\alpha \left(2i\right)\beta \left(i\right)H_1^{\mathrm{IBI}}T_Pc,\\ c_2^{\beta \left(i\right)}\left(i\right)=-\left[2\left\{\alpha \left(2i\right)\right\}H_1^{\mathrm{IBI}}+2\left\{\alpha \left(2i\right)\right\}H_2^{\mathrm{IBI}}\right]T_Pc.\end{array}$

In this embodiment of STBC system, the data symbols are separated in the demodulator and equaliser 9 by premultiplication of {circumflex over (R)}(i) by the Hermitian of the upper channel matrix W_H(i):

$W_H\left(i\right)\hat{R}\left(i\right)=\left[\begin{array}{cc}{H}^{\beta \left(i\right)}& 0_{P\times P}\\ 0_{P\times P}& H^{\beta \left(i\right)}\end{array}\right]\left[\begin{array}{c}\left[\begin{array}{c}s\left(2i\right)\\ \alpha \left(2i\right)c\end{array}\right]\\ \left[\begin{array}{c}s\left(2i+1\right)\\ \alpha \left(2i\right)c\end{array}\right]\end{array}\right]$

with H^β(i)=H₁^β(i)(H₁^β(i))^H+H₂^β(i)(H₂^β(i))^H. The equalisation based on pseudo circulant channel matrices is then performed as presented in our co-pending European Patent Application for the single channel case. The PRP-OFDM postfix based blind channel estimation is performed based on R(i) as presented above.

Claims

1-16. (canceled)

17. A method of communication using Orthogonal Frequency Division Multiplexing (‘OFDM’) from a transmitter comprising a plurality of transit antenna means and a receiver comprising at least one receive antenna means, the method comprising: generating bit streams and corresponding sets of N frequency domain carrier amplitudes ({tilde over (s)}(kN+j), 0≦j≦N−1) modulated as OFDM symbols subsequently to be transmitted from a transmitter, where k is the OFDM symbol number and j indicates the corresponding OFDM carrier number;inserting affix information into guard intervals between consecutive time domain OFDM symbols;transmitting said time domain OFDM symbols including said affix information from said transmitter to said receiver;using said affix information at the receiver to estimate the Channel Impulse Responses (Hlm between the lth transmit and mth receive antenna) of the transmission channels between said transmitter and said receiver; andusing the estimated Channel Impulse Response (Ĥlm between the lth transmit and mth receive antenna) to demodulate said bit streams in the signals received at said receiver, wherein said affix information is known to said receiver as well as to said transmitter, and is mathematically equivalent to a vector (c D) that is common to said time domain OFDM symbols multiplied by at least first weighting factors (αk) that are different for one time domain OFDM symbol (k) than for another and second weighting factors (wi(k)) that enable one of said transmit antenna means (i) to be distinguished from another.

18. A method of communication as claimed in claim 07, wherein said first weighting factors (αk) have pseudo-random values.

19. A method of communication as claimed in claim 07, wherein said first weighting factors (αk) have complex values.

20. A method of communication as claimed in claim 07, wherein said first weighting factors (αk) are deterministic and are known to said receiver as well as to said transmitter independently of current communication between said receiver and said transmitter.

21. A method of communication as claimed in claim 07, wherein said first weighting factors (αk) are communicated from said transmitter to said receiver.

22. A method of communication as claimed in claim 07, wherein said transmitter uses Nt transmit antenna means and the receiver uses Nr receive antenna means, M′ consecutive time domain OFDM data symbols are encoded by a specific space-time encoder such that the encoder produces M time domain OFDM data signals outputs for each of the Nt transmit antenna means and said vector (cD) is encoded by a specific space-time encoder such that the encoder produces M affixes for each of the Nt transmit antenna means corresponding to said affix information weighted by said first and second weighting factors (αk) and wi(k), the resulting affixes being inserted between time domain OFDM data symbols for each of the Nt transmit antenna means.

23. A method of communication as claimed in claim 22, wherein the matrix of said second weighting factors (wi(k)) for said transmit antenna means and for a number NT of consecutive symbols equal to the number NT of said transmit antenna means is a non-orthogonal matrix (W) such that when multiplied by its complex conjugate transpose ((W)T)* the result is different from the identity matrix (I), weighted by a gain factor g0 having a non-zero real value (i.e. g0I≠WHW).

24. A method of communication as claimed in claim 22, wherein all transmit antenna outputs over M consecutive OFDM time domain symbols, including time domain OFDM data symbols space-time encoded by and pseudo-random affixes space-time encoded by , are grouped into a block S, for which said first weighting factors (αk) are the same for OFDM symbols of the same block S but are different for OFDM symbols of different block S.

25. A method of communication as claimed in claim 24, wherein said transmitted affixes enable the separation at said receiver of the transmitted guard interval affix information of said block S, and said second weighting factors (wi(k)) enable the separation and estimation at said receiver of the different physical channels between said transmit antenna means and said at least one receive antenna means.

26. A method of communication as claimed in claim 25, wherein demodulating said bit streams includes, for each said receive antenna means, multiplying a signal derived from the received signal dm by the complex conjugate transpose of the Kronecker product of said matrix of said second weighting factors (wi(k)) for said transmit antenna means by the identity matrix ((W×ID)H) and using channel estimates derived form the results in demodulating said bit streams.

27. A method of communication as claimed in claim 26, wherein the matrix of said second weighting factors (wi(k)) for said transmit antenna means and for a number Nt of consecutive symbols equal to the number Nt of said transmit antenna means is a matrix (W) such that (W) alone is non-orthogonal, but (W) combined with the corresponding pseudo-random factors (60k) is orthogonal.

28. A method of claim 22, wherein the matrix W corresponding to M×Nt of said second weighting factors (wi(k)) for a number M of consecutive symbols and for said Nt transmit antenna means is an orthogonal matrix such that when multiplied by its complex conjugate transpose ((W)T)* the result is the identify matrix (I), weighted by a gain factor g0 having a non-zero real value (i.e. g0I=WHW).

29. A method of claim 17, wherein said second weighting factors (wi(k)) take different values for each of said transmit antenna means so as to enable said physical channels to be distinguished.

30. A method of claim 17, wherein estimating the Channel Impulse Response (Hlm) of the transmission channels between said transmitter and said receiver comprises a step of making a moving average estimation over a plurality of symbol periods of channels which are mathematically equivalent to the relationship:

31. A transmitter comprising generating means for generating bit streams modulated as OFDM symbols to be transmitted and for inserting affix information into a guard intervals between said OFDM symbols, said guard interval affix information being deterministic and suitable to be known to a receiver as well as to said transmitter and including vector (cD) that is common to time domain OFDM symbols multiplied by first weighting factors (αk) that are different for one time domain OFDM symbol (k) than for another and said second weighting factors (wi(k)) that enable transmit antenna means (i) to be distinguished from another.

32. The transmitter of claim 31 comprising demodulating means for receiving signals that comprises said bit streams modulated as said OFDM symbols with said guard interval affix information inserted between said OFDM symbols, said demodulating means being arranged to use said affix information from said guard intervals to estimate a Channel Impulse Response of the transmission channels and to use the estimated Channel Impulse Response to demodulate said bit streams in the signals received at said receiver, said guard interval affix information (wi(k)αk·c0 to wi(k)αk·cD-1) being deterministic and being known to said receiver.