CN100493052C - Method for channel evaluating of weak energy parallel PN sequence - Google Patents

Abstract

This invention provides a method and a device for using parallel PN sequence of weak energy to estimate channels, which utilizes matrix Q determined by the channel S/N ratio by the applied training sequence and the measured channel S/N ratio and matrix z determined by a receiving signal matrix y and a mean matrix A to get the pulse response estiomation value characterizing that the pulse response estimation value is hIMMSEE=Qz, which is a channel estimation technology between the biased and the MMSE estimations and can estimate channel pulse response accurately.

Description

A kind of weak energy parallel PN sequence carries out the method for channel estimating

Technical field:

The present invention relates to the method that a kind of weak energy parallel PN sequence carries out channel estimating, belong to the digital mobile communication field of using the electromagnetic wave technology, particularly digital terrestrial television, single carrier ofdm communication system, multi-carrier OFDM communication system, wireless lan (wlan) etc. adopt the channel estimation technique in the OFDM modulated digital communication system.

Background technology:

As everyone knows, the high frequency band radio wave will be through multiple transmission environment reflection such as house, vehicle, high mountain or diffraction in transmission course.That is: the primary signal that receives of antenna amplitude difference not only, and phase place also there are differences.The result of these signal amplitude combinations will cause the violent fluctuation of signal amplitude, promptly so-called multipath fading.

In ofdm communication system, this multipath fading wireless channel be frequency selectivity with time selectivity, so, before the ofdm signal demodulation, dynamic channel is estimated it is very important.

Channel estimating can adopt to be inserted training sequence with certain cycle and realizes on the subcarrier of OFDM symbol, also can adopt the mode of inserting training sequence on each OFDM symbol.The former is called block-type pilot channel estimation, is mainly used in the estimation to slow fading channel.The latter is called the Comb Pilot channel estimating, is mainly used in the estimation of fast fading channel.These two kinds of methods all can be further divided into least square (Least Square, LS) method or least mean-square error (Minimum Mean-Square-Error, MMSE) method etc.Yet these methods have all taken useful bandwidth and have reduced data transmission efficiency.

It then is another kind of thinking that training sequence is superimposed upon the method for carrying out channel estimating on the time-domain signal.The bandwidth resources that this method biggest advantage is exactly a preciousness of no use exchange performance for estimating channel for.

The thought that overlying training sequence is used for channel estimating originates from nineteen sixty-five, referring to document Van Duc Nguyen, ChristianHansen and Hans-Peter Kuchenbecher, " Performance of channel estimation using pilot symbols forcoherent OFDM system ", WPMC ' 00 November12-15 2000, Bangkok, Thailand, pp:842-847.

Nineteen ninety-five, this thought was used to digital communication, see document C.E.Kastenholz and W.P.Birkemeier, A simultaneousinformation transfer ans channel sounding modulation technique for wide-band channels, IEEETrans.on Communication Technology, June 1965, pp:162-165.

Thereafter different channel estimation methods also are suggested gradually, see following document:

[1]B.Farhang-Boroujeny，Pilot-based?channel?identification：proposal?for?semi-blind?identificationof?communication?channels，IEE?Electronics?Letters，vol.31，no.13，June?1995，pp：1044-104.

[2]G.TZhou，M.Viberg，and?T.Mckelvey，Superimposed?periodic?pilots?for?blind?channelestimation，Proc.35 ^th?Asilomar?Conference?on?Signals，Systems，and?Computers，PacificGrove，CA，Nov.2001，pp：653-657.

[3]N.Chen?and?G.T.Zhou，A?superimposed?periodic?pilot?scheme?for?semi-blind?channel?estimationof?OFDM?systems，Proc.10 ^th?IEEE?DSP?Workshop，Pine?Mountain，GA，Oct.2002，pp：362-365.

[4]J.K.Tugnait?and?W.Luo，On?channel?estimation?using?superimposed?training?and?first-orderstatistics，Proc.IEEE?Int.Conf.Acoust，Speech，Signal?Processing，Hong?Kong，China，Apr.2003，pp：624-627.

[5]G.T.Zhong?and?N.Chen，Superimposed?training?for?doubly?selective?channels，Proc.IEEEStatistical?Signal?Processing?Workshop，St.Louis，MO，Sept.2003，pp：73-76.

[6]X.Meng?and?J.K.Tugnait，Semi-blind?channel?estimation?and?detection?using?superimposedtraining，ICASSP，2004，pp：417-420.

[7]Ning?Chen，G.Tong?Zhou，Superimposed?training?for?OFDM：a?peak-to-average?power?ratioanalysis，IEEE?Trans.on?Signal?Processing，submitted，2004.

Though the method performance that these employing parallel PN sequences carry out channel estimating is pretty good,, its computation complexity is still bigger.

Summary of the invention:

The objective of the invention is: propose the method that a kind of weak energy parallel PN sequence carries out channel estimating, can estimate channel parameter effectively, and reduced computation complexity with respect to the channel estimation methods of prior art.

For achieving the above object, according to an aspect of the present invention, a kind of method of utilizing weak energy parallel PN sequence to carry out channel estimating has been proposed.Its technical scheme is: at transmitting terminal, and the PN sequence of the weak energy of stack on a symbol of time domain OFDM modulation signal; With the matrix Q that the channel signal to noise ratio that can survey determines channel estimating is reduced to linear operation by this PN sequence in the receiving terminal utilization, directly obtains channel state parameter.The impulse response estimated value that it is characterized in that channel ${\hat{h}}_{\mathrm{LMMSEE}}\≈\mathrm{Qz},$ Wherein

z＝Ay

Y is the received signal matrix.

According to another aspect of the present invention, a kind of device that utilizes weak energy parallel PN sequence to carry out channel estimating has been proposed.Utilization obtains the impulse response estimated value of channel by the matrix Q of this PN sequence and the channel signal to noise ratio decision that can survey, by received signal matrix y and the definite matrix z of average matrix A.The impulse response estimated value that it is characterized in that channel ${\hat{h}}_{\mathrm{LMMSEE}}\≈\mathrm{Qz}.$

The method and apparatus that utilizes weak energy parallel PN sequence to carry out channel estimating that the present invention proposes is a kind of between the channel estimation technique that does not have between estimation partially and the MMSE estimation, can estimate channel state parameter effectively and be used for demodulation, thereby effectively improve systematic function.

Channel estimating model of the present invention and principle are described in detail as follows:

For a sub-carrier number is N _cThe system of OFDM modulation, X _i(k) frequency-region signal of k symbol of i piece of expression (block).Suppose that in system protection is longer than channel delay at interval, then do not have inter-block-interference (inter-block-interference, IBI).In order to simplify, can omit piece mark i.Time domain OFDM signal { x (n) } can be expressed as:

$x\left(n\right)=\mathrm{IDFT}\left\{x\left(k\right)\right\}=\frac{1}{\sqrt{N_c}}\underset{k=0}{\overset{N_c-1}{\mathrm{\&Sigma;}}}X\left[k\right]{\mathrm{<figure-callout\; id="2"\; label="e\; j"\; filenames="C200510035892E00102.png,C200510035892E00111.png"\; state="\{\{state\}\}">e</figure-callout>}}^j---\left(1\right)$

The training sequence p (n) of the weak energy of stack on signal, its cycle is N _p, have:

$\tilde{x}\left(n\right)=x\left(n\right)+p\left(n\right)---\left(2\right)$

Then, length is N _gProtection be filled in the information symbol front at interval, have:

$s\left(n\right)=\left\{\begin{array}{cc}G\tilde{x}(N_c+n)& n=-N_g,-N_g+1,...,-1\\ G\tilde{x}\left(n\right)& n=\mathrm{0,1},...,N_c-1\end{array}\right.$

(3)

In the formula, G is the gain of power amplifier.General hypothesis G=1.

For frequency selective fading channels, can represent by constant finite impulse response filter h (n) of time spent.So received signal can be expressed as:

$\begin{array}{cc}\tilde{y}\left(n\right)=s\left(n\right)*h\left(n\right)+w\left(n\right),& -N_g\&le;n\&le;N_c-1\end{array}---\left(4\right)$

Wherein,

$\begin{array}{cc}h\left(n\right)=\underset{i=0}{\overset{r-1}{\mathrm{\&Sigma;}}}{h}_i{\mathrm{<figure-callout\; id="2"\; label="e\; j"\; filenames="C200510035892E00102.png,C200510035892E00111.png"\; state="\{\{state\}\}">e</figure-callout>}}^j\mathrm{\&delta;}(\mathrm{\&lambda;}-{\mathrm{\&tau;}}_i),& 0\&le;n\&le;N_c-1\end{array}---\left(5\right)$

In the formula, * represents linear convolution, and h (n) is the impulse response of transmitter filter, frequency-selective channel, filter for receiver, and w (n) is the zero-mean additive noise, and r is the total number in channel fading footpath, h _iBe the multiple impulse response in i bar footpath,

Be the Doppler frequency shift in i bar footpath, λ postpones the expansion label, and T is the sampling period, τ _iIt is the time delay of carrying out normalized i bar footpath by the sampling time.

At receiving terminal, received signal by A/D and low pass filter after, protection is at interval deleted, has:

$\begin{array}{cc}y\left(n\right)=\tilde{y}(n+N_g)& 0\&le;n\&le;N_c-1\end{array}---\left(6\right)$

After eliminating Cyclic Prefix, linear convolution just becomes circular convolution and (uses

Expression), at this moment, have:

$y\left(n\right)=\tilde{x}\left(n\right)\&CircleTimes;h\left(n\right)+w\left(n\right)=p\left(n\right)\&CircleTimes;h\left(n\right)+u\left(n\right)=q\left(n\right)+u\left(n\right),0\&le;n\&le;N_c-1$

(7)

Wherein,

$q\left(n\right)=p\left(n\right)\&CircleTimes;h\left(n\right)---\left(8\right)$

$u\left(n\right)=s\left(n\right)\&CircleTimes;h\left(n\right)+w\left(n\right)---\left(9\right)$

With matrix (7) formula of rewriting be:

y＝Ph+u (10)

Wherein, P is a N _c* L circular matrix, it first classifies [p[0], p[1 as] ..., p[N _c-1]] ^T, y=[y[0], y[1] ..., y[N _c-1]] ^T, h=[h[0], h[1] ..., h[L-1]] ^T, u=[u[0], u[1] ..., u[N _c-1]] ^T, T represents transposition herein.

For the time frequency-selective channel that becomes, selection cycle is _N, PN sequence p[n] carry out channel estimating.In order to estimate the channel that length is L, require N _p〉=L.Simultaneously, for further simplification, suppose $M=\frac{N_c}{N_p}$ It is an integer.At this moment, size of structure is N _p* N _cMean matrix I wherein _PBe a N _p* N _pUnit matrix.Can get by (10):

Ay＝APh+Au (11)

At this moment, suppose z=Ay, then z is a N _p* 1 to y[n] matrix that averages, its i element is:

$z\left[i\right]=\frac{1}{M}\underset{m=0}{\overset{M-1}{\mathrm{\&Sigma;}}}y[m{N}_p+i],0\&le;i\&le;N_p-1---\left(12\right)$

In like manner, establish u=Au, u is a N _p* 1 to y[n] matrix that averages, its i element is:

$\stackrel{\&OverBar;}{u}\left[i\right]=\frac{1}{M}\underset{m=0}{\overset{M-1}{\mathrm{\&Sigma;}}}u[m{N}_p+i]---\left(13\right)$

Because p[n] be that one-period is N _pSequence, suppose that first column element of the matrix P correspondence in (11) formula is:

Then have:

${\mathrm{<figure-callout\; id="0"\; label="P"\; filenames="C200510035892E00101.png,C200510035892E00102.png"\; state="\{\{state\}\}">P</figure-callout>}}_0=\mathrm{AP}=\left[\begin{array}{cccc}p\left[0\right]& p[N_p-1]& ...& p\left[1\right]\\ p\left[1\right]& p\left[0\right]& ...& p\left[2\right]\\ ...& ...& ...& ...\\ p[N_p-1]& p[N_p-2]& ...& p\left[0\right]\end{array}\right]---\left(14\right)$

So (11) formula of can rewriting is:

z＝Ay＝P ₀h+u (15)

Least mean-square error estimates that (MMSEE) is based on a kind of statistical method that Bayesan estimates, supposes that wherein h is a random vector.Suppose that h obeys the Gaussian Profile of zero-mean, and uncorrelated with u, at this moment, have:

${\hat{h}}_{\mathrm{MMSEE}}=V^{-1}{P}_0^{H}z---\left(16\right)$

Wherein:

$V={\mathrm{\&sigma;}}^2{C}_h^{-1}+D---\left(17\right)$

$D=P_0^H{P}_0---\left(18\right)$

C _h＝E{hh ^H} (19)

The corresponding channel frequency response of MMSEE is:

${\hat{H}}_{\mathrm{MMSEE}}=G{\hat{h}}_{\mathrm{MMSEE}}---\left(20\right)$

Wherein G is a matrix, and its element can be expressed as:

[G] _n，k＝e ^-j2πnk/N |n|≤N _α，0≤k≤L-1 (21)

That is:

${\hat{H}}_{\mathrm{MMSEE}}\left(n\right)=\underset{m=0}{\overset{N_p-1}{\mathrm{\&Sigma;}}}z\left(m\right)Q_{\mathrm{MMSEE}}(n,m)---\left(22\right)$

Wherein:

$Q_{\mathrm{MMSEE}}(n,m)=\underset{k=0}{\overset{L-1}{\mathrm{\&Sigma;}}}{\left[V^{-1}{P_0}^H\right]}_{k,m}{e}^{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="C200510035892E00102.png,C200510035892E00111.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;nk}/N}---\left(23\right)$

Know that by formula (16) the MMSEE channel estimation methods needs the prior information C of channel _h, the accurate acquisition of this information is difficult for; But also must be to C _hInvert, when channel impulse response length was longer, this amount of calculation also was higher.

Propose a kind of linear minimum mean-squared error algorithm below and estimate channel, reduce computation complexity.

Suppose E{ (hh ^H) ^-1}=E{|1/h _k| ²I, wherein I is a unit matrix.Definition average signal-to-noise ratio SNR is E{|x _k| ²/ σ ²], can obtain by (16) formula:

$={(\frac{B}{\mathrm{SNR}}I+P_0^H{P}_0)}^{-1}\&CenterDot;P_0^H\&CenterDot;z$

(24)

Wherein,

B＝E{|x _k| ²}·E{|1/h _k| ²} (25)

In (25) formula, B is a constant relevant with signal constellation (in digital modulation) and channel conditions.

In practice, in order to simplify, B can directly be defined as a constant c.At this moment, because X no longer is a factor of matrix computations, so

Contrary need not recomputate along with the variation of transmission data X.In addition, because

Known in advance, SNR can survey in receiving end, at this moment

Be the function of SNR, in real system,

Can directly obtain by look-up table.So formula (24) just is reduced to a multiplication of matrices computing, has:

${\hat{h}}_{\mathrm{LMMSEE}}\&ap;\mathrm{Qz}---\left(26\right)$

Wherein:

$Q={(\frac{c}{\mathrm{SNR}}I+P_0^H{P}_0)}^{-1}\&CenterDot;P_0^H---\left(27\right)$

At this moment, amount of calculation is all littler than LS method.Because the LS method relates to matrix inversion operation, and Q directly obtains by look-up table in (26) formula, does not have computation complexity; (26) calculating of formula just is mainly reflected in the multiplying of matrix Q and matrix z, and this operand also is indispensable for the LS method, so the LMMSEE method computation complexity after the simplification is all lower than LS method.Certainly, comparing with the amount of calculation of MMSEE method, then is to have significantly reduced.

The present invention is applicable to and adopts OFDM to carry out communication system for modulation, especially is applicable to European digital terrestrial television (DVB-T) system.Though technology of the present invention is primarily aimed at the system of frequency division multiple access, but any engineer with knowledge background such as signal processing, communications, can design corresponding channel estimating apparatus at code division multiple access and time division multiple access according to the present invention, it all should be included in thought of the present invention and the scope.

In the present invention,, can adopt the mode of look-up table to realize, so the present invention compares with existing method has advantage on computation complexity because do not need to calculate inverse matrix.

Simultaneously, prove also that through the ofdm communication system link simulation compare with other conventional methods, the present invention can accurately estimate channel parameter values.

Description of drawings:

Fig. 1 is at the stack PN of different-energy sequence LS method and the art of this patent LMMSEE method channel estimating performance comparison diagram.Fig. 1 is N _c=252 o'clock, LS method, LMMSEE method channel estimating performance were relatively.Wherein, the energy factors that overlying training sequence distributed in LMMSEE method and the LS method is fixed as β=0.3 and 0.5, c=1.0.As seen from the figure, LMMSEE method performance when low signal-to-noise ratio is better than the LS method slightly, and when signal to noise ratio was higher, the LS method was more excellent.

Fig. 2 is at stack PN sequence energy fixedly the time, LS method and the art of this patent LMMSEE method channel estimating performance comparison diagram.Fig. 2 is LS method and LMMSEE method channel estimating MSE comparison diagram, β=0.5 wherein, c=1.0.As seen, two kinds of method performances are close.

Fig. 3 is during to stack PN sequence allocation different-energy, the art of this patent LMMSEE method channel estimating performance comparison diagram.Wherein, N _c=252, B=1.As seen from the figure, be lower than 5dB and β=0.3 o'clock in signal to noise ratio, systematic function and channel estimating performance and ideal situation are approaching.

Embodiment:

Below by concrete enforcement technical scheme of the present invention is further described.

Apply the present invention to ofdm system.This system emulation major parameter is: channel model h (l)=0.6912+2.5340i, 1.0603+0.4829i ,-0.0014-0.5519i ,-0.6523-0.0743i, 0,0.2098+0.1348i ,-0.0287-0.0617i; Pilot tone p (n)=σ _pExp{j π/7}exp{-j π (n ²+ 1)/and 7}, cycle N _p=7, its DFT is P (k)=2 σ _pExp{j π (k ²+ 1)/7}

It is a chirp.Our fixed transmission power ${\mathrm{\&sigma;}}_s^2+{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="p"\; filenames="C200510035892E00102.png,C200510035892E00111.png"\; state="\{\{state\}\}">p</figure-callout>}}^2=1.$ The 4QAM modulation.

Concrete steps are:

1, transmitting terminal is sent into the OFDM base band signal modulated.

2, selection cycle is N _pThe PN sequence carry out channel estimating.In order to estimate the channel that length is L, require N _p〉=L.The PN sequence of the weak energy of stack on the unique therein information symbol.

3, be N with each information symbol tail length _gData be replicated in the information symbol front at interval as protection, through handling, generation transmits.

4, at receiving terminal, received signal by A/D and low pass filter after, protection is at interval deleted, obtains the received signal matrix:

y＝Ph+u

5, suppose $M=\frac{N_c}{N_p}$ It is an integer.At this moment, size of structure is N _p* N _cMean matrix

I wherein _PBe a N _p* N _pUnit matrix.Calculate matrix z, wherein z=Ay.

6, establish P ₀=AP calculates P ₀

7, actual measurement receiving end signal to noise ratio snr is set invariant c, and I is a unit matrix, obtains matrix Q by look-up table.

$Q={(\frac{c}{\mathrm{SNR}}I+P_0^H{P}_0)}^{-1}\&CenterDot;P_0^H$

8, obtain the channel characteristics parameter by look-up table or calculating

${\hat{h}}_{\mathrm{LMMSEE}}\&ap;\mathrm{Qz}$

Claims (2)

1, a kind of method of utilizing Weak energy parallel pseudorandom (PN) sequence to carry out channel estimating is characterized in that comprising the steps:

Step 1, transmitting terminal are sent into OFDM (OFDM) base band signal modulated;

Step 2, selection cycle are N _pPN sequence p[n] carry out channel estimating, in order to estimate the channel that length is L, require N _p〉=L, and in the time domain OFDM information symbol PN sequence of the weak energy of stack on information symbol;

Step 3, be N with each information symbol tail length _gData be replicated in the information symbol front at interval as protection, through handling, produce s emission signal s [n];

Step 4, at receiving terminal, received signal is by behind digital-to-analogue conversion (A/D) and the low pass filter, protection is deleted at interval, obtains the received signal matrix:

y＝Ph+u

Wherein, P is a N _c* L circular matrix, it first classifies [p[0], p[1 as] ..., p[N _c-1]] ^T, y=[y[0], y[1] ..., y[N _c-1]] ^T, [] ^TThe transposition of representing matrix, h[n] be the impulse response of transmitter filter, frequency-selective channel and filter for receiver, h=[h[0], h[1] ..., h[L-1]] ^T, w[n] and be the zero-mean additive noise, $u\left(n\right)=s\left(n\right)\&CircleTimes;h\left(n\right)+w\left(n\right),$ U=[u[0], u[1] ..., u[N _c-1]] ^TIt is the matrix representation forms of u (n);

Step 5, suppose $M=\frac{N_c}{N_p}$ Be an integer, at this moment, constructing a size is N _p* N _cMean matrix

I wherein _PBe a N _p* N _pUnit matrix, N _cBe the OFDM sub-carrier number, calculate matrix z, wherein:

z＝Ay

Step 6, establish P ₀=AP calculates P ₀

Step 7, actual measurement receiving end signal to noise ratio snr are set invariant c, obtain matrix Q by look-up table:

$Q={(\frac{c}{\mathrm{SNR}}I+P_0^H{P}_0)}^{-1}\&CenterDot;P_0^H$

In the formula, l is a unit matrix;

Step 8, pass through look-up table or calculating acquisition channel characteristics parameter,

Be estimation, have h

${\hat{h}}_{\mathrm{LMMSEE}}\&ap;\mathrm{Qz}.$

2, the method for utilizing weak energy parallel PN sequence to carry out channel estimating as claimed in claim 1 is characterized in that: c=1.

Images