CN1731772A - High-performance OFDM channel estimation method based on combined pilot - Google Patents

Abstract

This invention relates to a high-performance OFDM channel estimating method based on grouped polit, belonging to communication technology fields, which comprises the following steps: inserting polits in severel special frequency as frequency domain polit, superposing periodical pseudo-random sequence in line form on the DFDM time domain signal as time domain superposing polit; through multi-path channel, averaging first order statistics on received signal in the time domain to attain signal parameter, and using sent polit frequency and periodical pseudo-random sequence frequency characteristics in frequency domain to assess the channel, finally doing weighted mean on the parameter attained by the two methods as the final access result, or only adopting channel access attained by frequency domain. T he invention has lower calculating complexity and higher utility ratio of frequency spectrum, and improves channel access performance.

Description

A kind of high-performance OFDM channel estimation method based on combined pilot

Technical field

The invention belongs to communication technical field, be specifically related to a kind of high-performance OFDM channel estimation methods based on combined pilot.

Background technology

OFDM (OFDM) is divided into many low speed datas to high-speed serial data and respectively a plurality of orthogonal sub-carriers is modulated, thereby the character rate on the subcarrier is reduced greatly, symbol duration extends greatly, thereby multidiameter delay expansion had stronger resistance, reduced the influence of intersymbol interference to system ^[1]

Yet in new generation of wireless communication,,, need to carry out channel estimating if use coherent detection (coherent detection) to obtain higher performance because transmission rate is higher based on OFDM.The OFDM channel estimation methods mainly is based on the non-blind estimating method of pilot tone at present ^[2]The blind algorithm for estimating of some OFDM is also arranged certainly but still research among ^[3]Divide from the pilot tone insertion position, mainly contain bulk (block type) and pectination (comb type) based on pilot-based channel estimation method ^[4]Block some OFDM symbol all as pilot signal, and pectination is chosen some subchannels of each OFDM symbol as pilot tone.In time not under the Change channel condition, two kinds of method performances are just the same at AWGN, but under the fast situation of change of channel, pectination is better than bulk ^{[4] [5]}Be to have reduced the availability of frequency spectrum based on the pilot-based channel estimation method major defect.

Document [6] has proposed a kind of OFDM channel estimation methods of superimposed pilot.The periodic pulse train that on the OFDM time-domain signal, superposes, and utilize the first-order statistics information of received signal to estimate channel information.This method does not take frequency spectrum resource, and amount of calculation is low, but will improve the transmitting power of signal.Die Jia pulse train has increased the weight of OFDM peak averaging energy than (PAPR) simultaneously ^[6], because superimposed pilot has the character of similar " noise " when channel equalization, make equilibrium have noise gate (error floor), decreased performance under the high s/n ratio situation on the other hand ^[7]

Summary of the invention

It is low to the objective of the invention is to propose a kind of computation complexity, the OFDM channel estimation methods that the availability of frequency spectrum is high,

The OFDM channel estimation methods that the present invention proposes is a kind of channel estimation methods based on the combined pilot technology.At first system is divided into four modules: constellation mapping 1, adding pilot tone 2, multipath channel 3, channel estimating 4, enter adding pilot tone 2 modules through the signal after constellation mapping 1 module, pass through multipath channel 3 modules again, enter channel estimating 4 modules at last and carry out calculating such as channel estimating.

System block diagram is seen shown in Figure 1.The concrete function of each module is as follows:

1, constellation mapping.Binary data stream becomes frequency-region signal through constellation mapping.Constellation mapping can adopt QPSK (four phase place phase shift keyings), 16QAM (quadrature amplitude modulation) or 64QAM.

2, add pilot tone.The concrete steps that add pilot tone are by shown in Figure 2.As can be seen from Figure 2, its step is exactly to go up stack periodicity pseudo random sequence s (n) as the time domain superimposed pilot at OFDM time-domain signal x (n), inserts pilot tone at the non-vanishing Frequency point of periodicity pseudo random sequence frequency domain response simultaneously, as pilot tone.The method is called " combined pilot technology ".

3, multipath channel.Channel model is described by multipath channel, and each channel is obeyed Rayleigh (Rayleigh) respectively and distributed.

4, channel estimating.The concrete steps of channel estimating are by shown in Figure 3.As can be seen from Figure 3, the step of channel estimating is, earlier time domain to received signal y (n) carry out first-order statistics and on average try to achieve channel parameter, try to achieve channel parameter at frequency domain by the method for pilot tone being carried out least square (LS) or least mean-square error (MMSE) then; Channel parameter with time domain and frequency domain is weighted on average as channel estimation results at last.

Description of drawings

Fig. 1 system block diagram.

Fig. 2 adds the pilot module flow chart.

Fig. 3 channel estimation module flow chart.

Peak averaging energy under the different stack of Fig. 4 sequence energy is than (PAPR).

Fig. 5 channel estimation errors relatively.

Fig. 6 different channels method of estimation portfolio effect relatively.

Label among Fig. 1: 1, constellation mapping, 2, add pilot tone, 3, multipath channel, 4, channel estimating.

Embodiment

In native system, establish the OFDM symbol and constitute, and to make the channel impulse response of multipath channel be h=[h (0) by N subcarrier, h (1) ... h (M-1)] ^T, T is the matrix transpose symbol in the formula, M is a channel exponent number.

The concrete steps of channel estimation methods of the present invention are as follows:

(1) chooses pseudo random sequence

The time domain pilot of order stack, promptly periodically pseudo random sequence s (n) is to be the signal of P in the cycle, i.e. s (n)=s (n+P).Make c=[s (0), s (1) ... s (P-1)] ^T, be the one-period of stack sequence, P 〉=M.

The one-period of now choosing pseudo random sequence is as follows:

$c\left(n\right)={\mathrm{\σ}}_c{e}^{j\frac{\mathrm{\π}}{P}n(n+v)}$ V=2, P are idol; V=1, P are strange n=0,1 ... P-1 (1)

Wherein ${{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="c"\; filenames="A20051002845800091.png,A20051002845800111.png"\; state="\{\{state\}\}">c</figure-callout>}}}^2=\frac{1}{P}\underset{n=0}{\overset{P-1}{\mathrm{\&Sigma;}}}{\left|c\left(n\right)\right|}^2$ It is the average energy of stack sequence.

(2) time domain channel is estimated

Make C=circ (c), circ (c) is converted into circular matrix C to vectorial c ^[9]

Order $z\left(i\right)=\frac{1}{\mathrm{Np}}\underset{j=0}{\overset{\mathrm{Np}-1}{\mathrm{\&Sigma;}}}y(i+\mathrm{jP})$ i＝0，1，…P-1，z＝[z(P-1)，z(P-2)，…z(0)] ^T。

Np=N/P in the formula, establishing Np for the sake of simplicity is integer, then has time domain to estimate

For:

${\hat{h}}_t=C^{-1}z---\left(2\right)$

Following formula promptly is the algorithm that time domain channel is estimated.

(3) the characteristic frequency point of selection pilot tone

Definition is S (k) to the Fourier transform of s (n), then has

$S\left(k\right)=\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}s\left(n\right)\exp (-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="A20051002845800091.png,A20051002845800111.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;nk}/N)=\underset{m=0}{\overset{P-1}{\mathrm{\&Sigma;}}}\mathrm{\&delta;}(k-m\&CenterDot;\mathrm{Np})\underset{i=0}{\overset{P-1}{\mathrm{\&Sigma;}}}\mathrm{Np}\&CenterDot;c\left(i\right)\exp (-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="A20051002845800091.png,A20051002845800111.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;im}/P)$

When P is even number, (1) substitution following formula is obtained:

$S\left(k\right)=\underset{m=0}{\overset{P-1}{\mathrm{\&Sigma;}}}\mathrm{\&delta;}(k-m\&CenterDot;\mathrm{Np})\underset{i=0}{\overset{P-1}{\mathrm{\&Sigma;}}}\mathrm{Np}\&CenterDot;c\left(i\right)\exp (-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="A20051002845800091.png,A20051002845800111.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;im}/P)$

$=\mathrm{Np}\&CenterDot;{\mathrm{\&sigma;}}_c\underset{m=0}{\overset{P-1}{\mathrm{\&Sigma;}}}\mathrm{\&delta;}(k-m\&CenterDot;\mathrm{Np})\underset{i=0}{\overset{P-1}{\mathrm{\&Sigma;}}}\exp [\mathrm{j\&pi;}(i+2-2m)i/P]$

Then have ^[10]:

Same conclusions when being odd number, P is also arranged.

The specific Frequency point of pilot tone that the present invention adopts promptly is periodicity pseudo random sequence s (n) frequency domain response S (k) non-vanishing Frequency point: k=mNp, m=0, and 1 ... P-1 is P point altogether.Choose this P point as the pilot tone point, being equivalent to has increased the signal energy of pilot tone $\mathrm{Np}\&CenterDot;\sqrt{P}\&CenterDot;{\mathrm{\&sigma;}}_c$ Times, thereby equivalence has improved the signal to noise ratio of pilot frequency information.Thereby can under each signal to noise ratio, obtain the performance for estimating channel that other method can not obtain.

(4) frequency domain channel is estimated

Adopt least square (LS) or Minimum Mean Square Error (MMSE) at P pilot tone point ^[11]Channel estimating is promptly carried out frequency domain channel and is estimated, establishes the channel parameter that obtains and is

(5) final channel estimation results

Calculate

With

Weighted average be

$\hat{h}=\mathrm{\&alpha;}{\hat{h}}_t+{\widehat{\mathrm{\&beta;h}}}_f,$ Alpha+beta=1 β ≠ 0 (4) is as final channel estimation results.

Especially, when α then=0, $\hat{h}={\hat{h}}_f,$ When $\mathrm{\&alpha;}=\mathrm{\&beta;}=\frac{1}{2}$ The time, $\hat{h}=({\hat{h}}_t+{\hat{h}}_f)/2.$

The technology of the present invention effect

This trifle is by simulating, verifying performance of the present invention.In emulation, realized system shown in Figure 1,, after trying to achieve channel parameter, carried out the single step frequency domain equalization of channel for the accuracy that authenticated channel is estimated.

Establish OFDM sub-carrier number N=1024 in the emulation, quarternary phase-shift keying (QPSK) (QPSK) modulation.Do not adopt chnnel coding, Cyclic Prefix gets 1/32, then circulating prefix-length CP=1024/32=32.Channel is a multidiameter fading channel, and its impulse response is: $h\left(\mathrm{\&tau;}\right)=\underset{k=0}{\overset{M-1}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_k\mathrm{\&delta;}(\mathrm{\&tau;}-{\mathrm{\&tau;}}_k).$ M is a channel exponent number in the formula, sets M=8 in the emulation.α _kBe the amplitude of k paths, it is the multiple Gaussian random variable of zero-mean, and its power time delay profile is exponential distribution θ (τ _k)=exp (τ _k/ τ _Rms), order $\underset{k=0}{\overset{M-1}{\mathrm{\&Sigma;}}}{\left|{\mathrm{\&alpha;}}_k\right|}^2=1,$ The normalization of expression channel energy.τ _kBe the even distribution on the CP/2, promptly with normalized channel delay of systematic sampling cycle, the channel maximum delay is 1024/32/2=16, i.e. P=16.τ _RmsFor time delay expansion root-mean-square value, be made as CP/8.Make τ ₀=1, represent the synchronous fully of time-domain symbol.Produce 1000 groups of channels with Monte Carlo method (Monte-Carlo) and carry out emulation.

1) the peak averaging energy is than (PAPR)

The PAPR of the whole bag of tricks when being SNR=15dB among Fig. 4.The PAPR of original OFDM time-domain signal when not superposeing any sequence, delta function is the PAPR of signal after the periodic pulse train that superposeed, and the PAPR after the stack periodicity pseudo random sequence that proposed proposes for the present invention.Can make on scheme PAPR increase along with the increase of pulse train energy, and the pseudo random sequence shown in (1) formula is owing to have constant amplitude, i.e. its PAPR=1, thus not only can not increase the weight of PAPR, can reduce PAPR on the contrary.

2) channel estimation errors relatively

Channel estimation errors is used ${{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="e"\; filenames="A20051002845800091.png,A20051002845800111.png"\; state="\{\{state\}\}">e</figure-callout>}}}^2=E\left[{|\hat{h}-h|}^2\right]$ Weigh.

Get α=0.4 (stack sequence and burst amplitude ratio) in the method.Because Np=N/P=64, according to (3) formula can get the pilot tone insertion position for 1,65,129,257 ..., 961}, 16 pilot tones altogether.

As a comparison, introduce the method for estimation of tradition based on pilot tone.Adopt Comb Pilot, pilot tone insertion position be 1,33,65 ..., 993}, 32 pilot tones altogether, this is the pilot frequency locations that MMSE goes up optimization ^[4]Channel estimation methods adopts LS and MMSE algorithm ^[11]

LS among Fig. 5, MMSE is based on the channel estimation errors that traditional transmission pilot frequency system obtains, and pro time, proaverage, pro LS, pro MMSE utilizes among the present invention pilot frequency system that time domain is estimated (seeing formula (2)), and average is estimated in time-frequency two territories, and frequency domain LS and frequency domain MMSE method are estimated the error that obtains.As seen weighted average method low signal-to-noise ratio (＜12dB) under, more effective than time domain MMSE algorithm, and only compare then about the error drop by half with the superimposed pilot estimation approach.Adopt the frequency domain LS of pilot tone of the present invention close, but they are more much better than the method for estimation performance of conventional method when the low signal-to-noise ratio with MMSE method performance.It is emphasized that the LS method has low computation complexity.

3) systematic function relatively

Pkh represents portfolio effect under the perfect knowledge of h situation among Fig. 6.As seen approaching based on the time domain LS and the MMSE method performance of pilot tone, with equilibrium under the known channel information 2dB gap is arranged, the 0.5dB gap and this method (pro proposed) and ideal situation are only had an appointment all has good behaviour under each state of signal-to-noise.Only adopt superimposed pilot estimation approach (pro time) decreased performance behind SNR＞6dB, 0.3% BER still has an appointment under high s/n ratio.This is because S (k) is not that 0 Frequency point produces balancing error and causes in amplitude.

If consider owing to introduce the reduction 0.6dB that the stack sequence causes SNR (the past 0.6dB that moves to right of curve that this method is obtained), then can get this method and still be better than the LS method of estimation from Fig. 4, it is emphasized that simultaneously, the amount of calculation of the inventive method is less than the method for estimation of traditional LS, be that the present invention has just obtained the performance that is better than 32 pilot tones of conventional method with 16 pilot tones, also improved the availability of frequency spectrum when improving systematic function.

List of references

[1]F?B?Frede?ricksen，R?Prasad.An?overview?of?OFDM?and?related?techniques?towards?development?of?futurewireless?multimedia?communications[C].Radio?and?Wireless?Conference，IEEE，Aug.2002.19-22.

[2]S?Coleri，M?Ergen.A?study?ofchannel?estimation?in?OFDM?systems[C].VTC?2002，IEEE?56th，vol.2，Sep.2002.894-898.

[3] Zhang Jidong, Zheng Baoyu. based on the OFDM channel estimating and the progress [J] thereof of pilot tone. communication journal, 2003,24 (11): 116-124.

[4]R?Negi，J?Cioffe.Pilot?tone?selection?for?channel?estimation?in?mobile?OFDM?system[J].IEEETrans.Consum.Electron，vol.44(3)，Aug.1998.1122-1128.

[5]M?Dong，L?Tong.Optimal?design?and?placement?ofpilot?symbols?for?channel?estimation[A].IEEE?Trans.Signal?Processing，vol.50，Dec.2002.3055-3069.

[6]N?Chen，G?T?Zhou.A?superimposed?periodic?pilot?scheme?for?semi-blind?channel?estimation?of?OFDMsystems[A].Digital?Signal?Processing?Workshop，Proceedings?ofIEEE，Oct.2002.362-365.

[7]S?Ohno，G?B?Giannakis.Optimal?training?and?redundant?precoding?for?block?transmissions?with?applicationto?wireless?OFDM[J].IEEE?Trans.Commun，vol.50，Dec.2002.2113-2123.

[8]Orozco-Lugo，G?Aldo.Channel?estimation?using?implicit?training[A].IEEE?Trans.Signal?Processing，vol.52.Jan.2004.240-254.

[9]P.J.Davis，Circulant?Matrix[M]，2nd?ed，New?York：Chelsea，1994.

[10] Xiao Guo town. " pseudo random sequence and application thereof " [M], National Defense Industry Press, 1985.

[11]van?de?Beek，O?Edfors.On?Channel?Estimation?in?OFDM?Systems[C].VTC?1995，IEEE?45th，vol.2，July.1995.815-819.

Claims (3)

1, a kind of high-performance OFDM channel estimation method based on combined pilot is characterized in that at first system being divided into four

Individual module: constellation mapping (1), adding pilot tone (2), multipath channel (3), channel estimating (4), enter adding pilot tone (2) module through the signal after constellation mapping (1) module, pass through multipath channel (3) module again, enter channel estimating (4) module at last and carry out calculating such as channel estimating; Wherein,

Constellation mapping is a frequency-region signal with the binary data stream signal map;

Adding pilot tone is to go up stack periodicity pseudo random sequence s (n) as the time domain superimposed pilot at OFDM time-domain signal x (n), inserts pilot tone at the non-vanishing Frequency point of periodicity pseudo random sequence frequency domain response simultaneously, as pilot tone;

The step of channel estimating is, earlier time domain to received signal y (n) carry out first-order statistics and on average try to achieve channel parameter, try to achieve channel parameter at frequency domain by the method for pilot tone being carried out least square or least mean-square error then; Channel parameter with time domain and frequency domain is weighted on average as channel estimation results at last.

2, signal estimation method according to claim 1 is characterized in that the concrete operations step is as follows:

(1) chooses pseudo random sequence

The time domain pilot of order stack, promptly periodically pseudo random sequence s (n) is to be the signal of P: s (n)=s (n+P) in the cycle; Make c=[s (0), s (1) ... s (P-1)] ^T, being the one-period of stack sequence, P 〉=M, M are channel exponent number;

The one-period of choosing pseudo random sequence is as follows:

$c\left(n\right)={\mathrm{\&sigma;}}_c{e}^{j\frac{\mathrm{\&pi;}}{P}n(n+v)}$ V=2, P are idol; V=1, P are strange n=0,1 ... P-1 (1)

Wherein ${{\mathrm{\&sigma;}}_c}^2=\frac{1}{P}\underset{n=0}{\overset{P-1}{\mathrm{\&Sigma;}}}{\left|c\left(n\right)\right|}^2$ It is the average energy of stack sequence;

(2) time domain channel is estimated

Make C=circ (c), circ (c) is converted into circular matrix C to vectorial c ^[9]Order $z\left(i\right)=\frac{1}{\mathrm{Np}}\underset{j=0}{\overset{\mathrm{Np}-1}{\mathrm{\&Sigma;}}}y(i+\mathrm{jP})$ I=0,1 ... P-1, z=[z (P-1), z (P-2) ... z (0)] ^TNp=N/P in the formula, Np are integer, then have time domain to estimate

For:

${\hat{h}}_t=C^{-1}z---\left(2\right)$

(3) the characteristic frequency point of selection pilot tone

Adopt periodically pseudo random sequence s (n) frequency domain response S (k) non-vanishing Frequency point: k=mNp, m=0,1 ... P-1 is a P point, as the pilot tone point altogether;

(4) frequency domain channel is estimated

Adopt least square or Minimum Mean Square Error to carry out channel estimating at P pilot tone point, obtain the frequency domain channel parameter and be

(5) final channel estimation results calculates With

Weighted average be

$\hat{h}=\mathrm{\&alpha;}{\hat{h}}_t+\mathrm{\&beta;}{\hat{h}}_f,\mathrm{\&alpha;}+\mathrm{\&beta;}=1--\mathrm{\&beta;}\&NotEqual;0---\left(4\right)$

As final channel estimation results.

3, signal estimation method according to claim 2 is characterized in that getting α=0, $\hat{h}={\hat{h}}_f,$ Perhaps $\mathrm{\&alpha;}=\mathrm{\&beta;}=\frac{1}{2},$

$\hat{h}=({\hat{h}}_t+{\hat{h}}_f)/2.$

Images