CN101764783B - Method for eliminating inter-carrier interference in orthogonal frequency division multiplexing system - Google Patents

Abstract

The invention discloses a method for eliminating inter-carrier interference (ICI) in an orthogonal frequency division multiplexing (OFDM) system in the technical field of wireless communication. The method comprises the following steps of: removing cyclic prefix (CP) of a time domain receiving signal at a receiving end of the OFDM system; dividing a linear time-variant (LTI) channel into the cascade of a linear time-invariant (LTI) channel and a time-variant (TV) channel by using a channel division method; carrying out one-dimensional time domain equalization on the time domain receiving signal after removing the CP by utilizing the TV channel to obtain a time domain initial recovery signal sequence; obtaining a frequency domain initial recovery signal sequence by the fast Fourier transformation; and carrying out one-dimensional frequency domain equalization on the frequency domain initial recovery signal by utilizing the LTI channel. The invention avoids the calculation and inversion processes of the channel frequency domain response matrix with high complexity, efficiently reduces the system time complexity when guaranteeing the practical requirements on the performances of the wireless communication system, and is more suitable for a high-frequency wireless communication system with hardware low-complexity requirements.

Description

The method that the low complex degree inter-carrier interference is eliminated in the ofdm system

Technical field

What the present invention relates to is a kind of method of wireless communication technology field, specifically is (Inter-Carrier Interference, the method for ICI) eliminating of low complex degree inter-carrier interference in a kind of OFDM (OFDM) system.

Background technology

OFDM has characteristics such as transmission rate height, capacity are big, availability of frequency spectrum height, and the Gbps level wireless communication system of realizing on than low-frequency range has at present all adopted the OFDM modulation system.In order to adapt to the demand of multimedia service in the next generation wireless communication, OFDM is as one of key technology of following 4G radio communication.

Because the developing rapidly of the communications industry, in existing low-frequency range wireless communication system, the poor problem of frequency spectrum resource is serious day by day, especially at frequency needs hundreds of MHz at full stretch in the 3GHz radio band, the use highly competitive of frequency spectrum resource.This will become the bottleneck of broadband wireless communication technique to the popularization and application of national economic development every field.A feasible solution of considering at present is to be modulated to the abundant more high band of frequency spectrum resource, for example 6GHz-15GHz to signal.Yet in the high-frequency wireless communication system; The Doppler frequency shift that causes is moved at the terminal to OFDM and the frequency shift (FS) between the transmitting-receiving two-end local oscillator is responsive more; Will destroy the orthogonality between the OFDM subcarrier, thereby produce serious ICI, reduce the performance of system.Therefore the elimination work of carrying out ICI at receiving terminal just seems very important and urgent.

Through existing literature search is found; In the article that is entitled as " Transmission techniques for digital terrestrial TV broadcasting (ground digital television broadcast transmission technology) " that people such as H.Sari delivered on nineteen ninety-five " IEEE Communications Magazine "; Proposed a kind of single tap frequency-domain equalizer, this is the simplest equalization methods.Linear time-varying (Linear Time-Variant for reality; LTV) channel, this method in an OFDM symbol to each channel response constantly average (Average, AVE); Obtain average channel response; Be used for channel situation in the equivalent whole OFDM symbol, constant channel when being converted into time varying channel carries out the signal recovery at receiving terminal with single tap filter.For sub-carrier number is the ofdm system of N, and time complexity is O (N).But this method has been ignored the time variation of channel, can not eliminate ICI very effectively.

Find again by retrieval; In the article that is entitled as " On channel estimation and detection for multicarrier Signals in fastand selective Rayleigh fading channels (channel estimating of multi-carrier signal and detection in the fast selective rayleigh fading channel) " that people such as Y.S.Choi delivered on calendar year 2001 " IEEE Transactions on Communications ", proposed two kinds of fast linear balanced device LS (Least square) and MMSE (Minimum Mean Square Error) and eliminated ICI.These two kinds of methods are calculating channel frequency domain response matrix at first, utilizes this matrix to carry out subsequent treatment and inversion process then, carries out frequency domain compensation and interference eliminated at receiving terminal.LS, MMSE equalization methods have utilized all channel informations, so can eliminate ICI well.But should technology channel frequency domain response matrix find the solution and invert and will carry out a large amount of complex multiplication operations, system complexity is higher.For sub-carrier number is the ofdm system of N, and time complexity is O (N ³), therefore will be very high when concrete the realization to the requirement of hardware.

Also find by retrieval; Schniter P has proposed a kind of low complex degree MMSE (LCMMSE) method and has eliminated ICI in the article of delivering on " IEEE Transactions on Signal Processing " in 2004 that is entitled as " Low-complexity equalization of OFDM in doubly selective channels (low complex degree equalization method of ofdm system under the dual-selection channel condition) ".This method is removed approximated channel frequency domain response matrix with banded structure, and the channel frequency domain response matrix is divided into a series of parton matrixes according to the adjacent carrier distribution character of ICI when receiving terminal is done the MMSE equilibrium.For sub-carrier number is the ofdm system of N, and time complexity is O (N ²Log N+D ²N), D is the width factor of banded structure.But should technology realize the compromise of performance and complexity through the parameter D that selects different sizes.Can know that by expression formula even if sacrifice certain systematic function, required time complexity is still very big.

In a word, present most ofdm system ICI removing method all can only be taken into account in systematic function (leading indicator is output Signal to Interference plus Noise Ratio SNIR), these two indexs of time complexity.Yet these two technical indicators have crucial meaning to the practical application of OFDM technology, especially for the high band wireless system that the requirement of hardware low complex degree is arranged.Therefore, under the prerequisite that guarantees certain performance gain, how to design a kind of ofdm system low complex degree ICI removing method and just have practical significance.

Summary of the invention

The objective of the invention is to overcome the deficiency of above-mentioned prior art, propose the method that the low complex degree inter-carrier interference is eliminated in a kind of ofdm system.The present invention carries out singular value decomposition (Singular Value Decomposition to head and the tail moment channel response in the OFDM symbol; SVD); The method of using channel to cut apart according to SVD result then; The cascade of constant channel and no time delay time varying channel when being with the linear time-variant channel Approximate Equivalent, thus ICI eliminated at receiving terminal through one dimension time domain equalization and one dimension frequency domain equalization.The present invention has avoided the calculating and the inversion process thereof of the channel frequency domain response matrix of high complexity, can under the situation that satisfies the actual system behavior demand, reduce the system time complexity significantly.

The present invention realizes through following technical scheme, the present invention includes following steps:

Step 1:, remove time domain and receive signal { y at the receiving terminal of ofdm system _1nCyclic Prefix (Cyclic Prefix CP), obtains burst { y _n.

Described Cyclic Prefix is to be that the information bit of Cyclic Prefix realizes by removing in the received signal segment length before the symbol.

Step 2: based on the SVD result of head and the tail moment channel response in the OFDM symbol; The method that the utilization channel is cut apart; Linear time-variant channel is divided into LTI (Linear Time-Invariant, LTI) channel and no time time-delay change (Time-Variant, TV) cascade of channel.

The method that described channel is cut apart, concrete steps are:

1), ties up matrix H=(h (1 :) through structure L * 2 according to first channel response h constantly in the OFDM symbol (1 :) and tail channel response h (N :) constantly ^TH (N :) ^T), H is carried out the SVD decomposition obtains:

$H=(h_1.h_2)\left(\begin{array}{cc}\sqrt{{\mathrm{\λ}}_1}& 0\\ 0& \sqrt{{\mathrm{\λ}}_2}\end{array}\right)\left(\begin{array}{cc}{u}_{11}^{*}& u_{21}^{*}\\ u_{12}^{*}& u_{22}^{*}\end{array}\right)$

$=(\sqrt{{\mathrm{\&lambda;}}_1}{u}_{11}^{*}{h}_1+\sqrt{{\mathrm{\&lambda;}}_2}{u}_{12}^{*}{\mathrm{<figure-callout\; id="2"\; label="h"\; filenames="H201010301000720100201E000022.png"\; state="\{\{state\}\}">h</figure-callout>}}_2,\sqrt{{\mathrm{\&lambda;}}_1}{u}_{21}^{*}{h}_1+\sqrt{{\mathrm{\&lambda;}}_2}{u}_{22}^{*}{h}_2)$

That is: $\left\{\begin{array}{c}h{(1,:)}^T=\sqrt{{\mathrm{\&lambda;}}_1}{u}_{11}^{*}{h}_1+\sqrt{{\mathrm{\&lambda;}}_2}{u}_{12}^{*}{h}_2\\ h{(N,:)}^T=\sqrt{{\mathrm{\&lambda;}}_1}{u}_{21}^{*}{h}_1+\sqrt{{\mathrm{\&lambda;}}_2}{u}_{22}^{*}{\mathrm{<figure-callout\; id="2"\; label="h"\; filenames="H201010301000720100201E000022.png"\; state="\{\{state\}\}">h</figure-callout>}}_2\end{array}\right.$

Wherein:

With

The singular value that expression SVD obtains, h ₁And h ₂The unit character vector that expression SVD obtains, h ₁And h ₂Length be L, u ₁₁ ^*, u ₁₂ ^*, u ₂₁ ^*And u ₂₂ ^*The characteristic element that expression SVD obtains, N is the number of subcarrier, L is the number of multipath.

2) to OFDM symbol head and the tail constantly channel response h (1 :), h (N :) carry out approximate processing:

$\left\{\begin{array}{c}h(1,:)\&ap;\sqrt{{\mathrm{\&lambda;}}_1}{u}_{11}^{*}{h}_1^T={\mathrm{\&beta;}}_1{h}_1^T\\ h(N,:)\&ap;\sqrt{{\mathrm{\&lambda;}}_1}{u}_{21}^{*}{h}_1^T={\mathrm{\&beta;}}_N{h}_1^T\end{array}\right.$

3) obtaining in the OFDM symbol constantly through linear interpolation, the channel response h (n. :) of n does

$\left\{\begin{array}{c}h(n,:)={\mathrm{\&beta;}}_n{h}_1^T\\ {\mathrm{\&beta;}}_n={\mathrm{\&beta;}}_1+\frac{{\mathrm{\&beta;}}_N-{\mathrm{\&beta;}}_1}{N-1}(n-1)\end{array}\right.$

4) linear time-variant channel h (n. :) is split into LTI h ₁The cascade of (n. :) and no time delay time varying channel h, that is:

h(n：1)＝h(n)*h ₁(n，1)，

Wherein: $\left\{\begin{array}{c}{h}_1(n,:)=\{h_1(n,l),1\&le;l\&le;L\}=h_1^T\\ h=\{h\left(n\right),1\&le;n\&le;N\}=({\mathrm{\&beta;}}_1,{\mathrm{\&beta;}}_2,L,{\mathrm{\&beta;}}_N)\end{array}\right.$

Step 3: utilize the no time delay time varying channel response in the step 2, the time domain that removes behind the CP is received signal { y _nCarry out the one dimension time domain equalization, obtain the initial restoring signal sequence of time domain y ' _n.

Described one dimension time domain equalization, concrete formula is:

$y_n^{\&prime;}=\frac{{\mathrm{\&beta;}}_n^{*}{y}_n{{\mathrm{\&sigma;}}_x}^2}{{\left|{\mathrm{\&beta;}}_n\right|}^2{{\mathrm{\&sigma;}}_x}^2+{{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H201010301000720100201E000022.png"\; state="\{\{state\}\}">N</figure-callout>}}}^2}$

Wherein: y _nBe to remove behind the CP time-domain signal of n constantly, y ' _nExpression is the initial restoring signal of time domain of n constantly, σ _X ²The variance of burst, σ are sent in expression _N ²The variance of expression interchannel noise.

Step 4: to the initial restoring signal sequence of time domain y ' _nDo fast Fourier transform (Fast Fourier Transform), obtain frequency domain initial restoring signal sequence Y ' _k.

Step 5: utilize the LTI channel response in the step 2, to frequency domain initial restoring signal sequence Y ' _kCarry out the one dimension frequency domain equalization, accomplish ICI and eliminate, obtain final frequency domain send signal recovery value sequence X ' _k.

Described one dimension frequency domain equalization, concrete formula is:

X′ _k＝Y′ _k/H′ _k，

Wherein: $h_1^{\&prime;}(n,k)=\left\{\begin{array}{cc}{h}_1(n,k)& 1\&le;k\&le;L\\ 0& L+1\&le;k\&le;N\end{array}\right.$

X ' _kThe transmission signal recovery value that finally obtains at receiving terminal behind the expression elimination ICI, Y ' _kExpression frequency domain initial restoring signal value, H ' _kExpression LTI channel { h ₁(n.1) sequence spreading of .1≤1≤L} h ' ₁(n.k) N of .1≤k≤N} point DFT sequence, N is the number of subcarrier, L is the number of multipath.

Compared with prior art; The invention has the beneficial effects as follows: calculating and the inversion process thereof of having avoided the channel frequency domain response matrix of high complexity; When guaranteeing the practical wireless communication systems performance requirement; Reduced the system time complexity effectively, time complexity of the present invention is O (N), more is applicable to the wireless communication in high-frequency band system of hardware low complex degree requirement.

Description of drawings

Fig. 1 is a flow chart of the present invention;

Fig. 2 is the Signal to Interference plus Noise Ratio performance sketch map among the embodiment;

Fig. 3 is the Signal to Interference plus Noise Ratio performance sketch map under the embodiment different sub carrier number;

Signal to Interference plus Noise Ratio performance sketch map when Fig. 4 is the embodiment different channels under the change condition;

Fig. 5 is the Signal to Interference plus Noise Ratio performance sketch map under the embodiment different channels decline degree.

Embodiment

Below in conjunction with accompanying drawing embodiments of the invention are elaborated: present embodiment provided detailed execution mode and concrete operating process, but protection scope of the present invention is not limited to following embodiment being to implement under the prerequisite with technical scheme of the present invention.

Embodiment

In the present embodiment: the sub-carrier number N of ofdm system is 256; The length M of Cyclic Prefix is 1/16 of sub-carrier number, i.e. M=16; It is 8 that the multipath of linear time-variant channel is counted L; Interchannel noise is additive white Gaussian noise (AWGN); The input signal-to-noise ratio (SNR) that sends signal is 0dB-30dB; Test result is for receiving the output Signal to Interference plus Noise Ratio (SINR) of signal.

As shown in Figure 1, present embodiment comprises the steps:

Step 1:, be that the information bit of Cyclic Prefix removes the Cyclic Prefix that time domain receives signal through remove receiving in the signal segment length before the symbol at the receiving terminal of ofdm system.

Receive signal y for time domain ₁=[y ₁₁.y ₁₂. ... .y _1M.y _{1 (M+1)}. ... Y _{1 (M+N)}], the time-domain signal that removes behind the CP is:

y＝[y ₁.….y _n.….y _N]＝[y _1(M+1).….y _1(M+N)]

Wherein: M=16, N=256.

Step 2: based on the SVD result of channel response constantly of head and the tail in the OFDM symbol, the method that the utilization channel is cut apart is divided into linear time-variant channel the cascade of LTI channel and no time delay time varying channel.

The method that described channel is cut apart, concrete steps are:

1), ties up matrix H=(h (1 :) through structure L * 2 according to first channel response h constantly in the OFDM symbol (1 :) and tail channel response h (256. :) constantly ^TH (256 :) ^T), H is carried out the SVD decomposition obtains:

$H=(h_1.h_2)\left(\begin{array}{cc}\sqrt{{\mathrm{\&lambda;}}_1}& 0\\ 0& \sqrt{{\mathrm{\&lambda;}}_2}\end{array}\right)\left(\begin{array}{cc}{u}_{11}^{*}& u_{21}^{*}\\ {\mathrm{<figure-callout\; id="12"\; label="u"\; filenames="H201010301000720100201E000022.png"\; state="\{\{state\}\}">u</figure-callout>}}_{12}^{*}& u_{22}^{*}\end{array}\right)$

$=(\sqrt{{\mathrm{\&lambda;}}_1}{u}_{11}^{*}{h}_1+\sqrt{{\mathrm{\&lambda;}}_2}{\mathrm{<figure-callout\; id="12"\; label="u"\; filenames="H201010301000720100201E000022.png"\; state="\{\{state\}\}">u</figure-callout>}}_{12}^{*}{\mathrm{<figure-callout\; id="2"\; label="h"\; filenames="H201010301000720100201E000022.png"\; state="\{\{state\}\}">h</figure-callout>}}_2,\sqrt{{\mathrm{\&lambda;}}_1}{u}_{21}^{*}{h}_1+\sqrt{{\mathrm{\&lambda;}}_2}{u}_{22}^{*}{h}_2)$

That is: $\left\{\begin{array}{c}h{(1,:)}^T=\sqrt{{\mathrm{\&lambda;}}_1}{u}_{11}^{*}{h}_1+\sqrt{{\mathrm{\&lambda;}}_2}{\mathrm{<figure-callout\; id="12"\; label="u"\; filenames="H201010301000720100201E000022.png"\; state="\{\{state\}\}">u</figure-callout>}}_{12}^{*}{h}_2\\ h{(256,:)}^T=\sqrt{{\mathrm{\&lambda;}}_1}{u}_{21}^{*}{h}_1+\sqrt{{\mathrm{\&lambda;}}_2}{u}_{22}^{*}{\mathrm{<figure-callout\; id="2"\; label="h"\; filenames="H201010301000720100201E000022.png"\; state="\{\{state\}\}">h</figure-callout>}}_2\end{array}\right.$

Wherein: With The singular value that expression SVD obtains, h ₁And h ₂The unit character vector that expression SVD obtains, h ₁And h ₂Length be 8, u ₁₁ ^*, u ₁₂ ^*, u ₂₁ ^*Wide and u ₂₂ ^*The characteristic element that expression SVD obtains.

2) to OFDM symbol head and the tail constantly channel response h (1 :), h (256 :) carry out approximate processing:

$\left\{\begin{array}{c}h(1,:)\&ap;\sqrt{{\mathrm{\&lambda;}}_1}{u}_{11}^{*}{h}_1^T={\mathrm{\&beta;}}_1{h}_1^T\\ h(256,:)\&ap;\sqrt{{\mathrm{\&lambda;}}_1}{u}_{21}^{*}{h}_1^T={\mathrm{\&beta;}}_{256}{h}_1^T\end{array}\right.$

3) obtaining in the OFDM symbol constantly through linear interpolation, the channel response h (n. :) of n is:

$\left\{\begin{array}{c}h(n,:)={\mathrm{\&beta;}}_n{h}_1^T\\ {\mathrm{\&beta;}}_n={\mathrm{\&beta;}}_1+\frac{{\mathrm{\&beta;}}_{256}-{\mathrm{\&beta;}}_1}{255}(n-1)\end{array}\right.$

4) linear time-variant channel h (n. :) is split into LTI channel h ₁The cascade of (n. :) and no time delay time varying channel h, that is:

h(n：1)＝h(n)*h ₁(n：1)，

Wherein: $\left\{\begin{array}{c}{h}_1(n,:)=\{h_1(n,l),1\&le;l\&le;8\}=h_1^T\\ h=\{h\left(n\right),1\&le;n\&le;256\}=({\mathrm{\&beta;}}_1,{\mathrm{\&beta;}}_2,L,{\mathrm{\&beta;}}_{256})\end{array}\right.$

Step 3: to removing the time-domain signal { y behind the CP in the step 1 _n, utilize no time delay time varying channel h=(β ₁. β ₂. .... β ₂₅₆) in element β _n, estimate to realize the one dimension time domain equalization through MMSE, obtain the initial restoring signal sequence of time domain y ' _n}:

$y_n^{\&prime;}=\frac{{\mathrm{\&beta;}}_n^{*}{y}_n{{\mathrm{\&sigma;}}_x}^2}{{\left|{\mathrm{\&beta;}}_n\right|}^2{{\mathrm{\&sigma;}}_x}^2+{{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="H201010301000720100201E000022.png"\; state="\{\{state\}\}">N</figure-callout>}}}^2},1\&le;n\&le;256$

Wherein: y ' _nExpression is the initial restoring signal of time domain of n constantly, σ _X ²The variance of burst, σ are sent in expression _N ²The variance of expression channel additive white Gaussian noise.

Step 4: to the initial restoring signal sequence of time domain y ' _nDo 256 point quick Fourier conversion, obtain frequency domain initial restoring signal sequence Y ' _k, concrete formula is:

$Y_k^{\&prime;}=\mathrm{FFT}\left\{y_n^{\&prime;}\right\}=\underset{n=1}{\overset{256}{\mathrm{\&Sigma;}}}{y}_n\exp (-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="H201010301000720100201E000022.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;k}(n-1)/256),1\&le;k\&le;256$

Step 5: utilize the LTI channel response in the step 2, to frequency domain initial restoring signal sequence Y ' _kCarry out the one dimension frequency domain equalization, accomplish ICI and eliminate, obtain final frequency domain send the signal recovery value X ' _k.

Described one dimension frequency domain equalization, concrete formula is:

X′ _k＝Y′ _k/H′ _k，

Wherein: $h_1^{\&prime;}(n,k)=\left\{\begin{array}{cc}{h}_1(n,k)& 1\&le;k\&le;8\\ 0& 9\&le;k\&le;256\end{array}\right.$

X ' _kThe transmission signal recovery value that finally obtains at receiving terminal behind the expression elimination ICI, Y ' _kExpression frequency domain initial restoring signal value, H ' _kExpression LTI channel { h ₁(n, 1), the sequence spreading of 1≤l≤8} h ' ₁(n, k), 256 DFT sequences of l≤k≤256}.

Present embodiment is respectively with the time complexity that the average single tap equalization method (AVE) of channel, MMSE equalization methods (MMSE), low complex degree MMSE equalization methods (LCMMSE) and present embodiment method (CSC) obtain respectively: O (N), O (N ³), O (N ²LogN+D ²N) and O (N).

Present embodiment is as shown in Figure 2 with the Signal to Interference plus Noise Ratio that above-mentioned four kinds of methods obtain respectively, for the more complicated degree, makes LCMMSE method performance identical with the CSC method performance of present embodiment, at this moment banded structure width factor D=70.Can know that by Fig. 2 along with the increase of input signal-to-noise ratio, the performance of MMSE method is superior to CSC, LCMMSE method gradually, and the performance of AVE method is the poorest.When SNR=30dB, though the CSC method that present embodiment proposes has the loss of 5dB with respect to the MMSE method of best performance, the time complexity that the comparison time complexity can be known the CSC method is well below the MMSE method, than little two one magnitude of the latter; Contrast LCMMSE method, the CSC method can reduce time complexity significantly when not losing performance; The simplest AVE method of contrast, the CSC method can be at the performance gain that obtains under the situation of identical time complexity about 6dB.

Fig. 3 has provided the Signal to Interference plus Noise Ratio performance sketch map of present embodiment under the different sub carrier number; Wherein the number of sub-carrier number N increases to 512; Comparison diagram 2 can be known with Fig. 3; The increase of sub-carrier number is little to the SINR performance impact of AVE, CSC, these 3 kinds of methods of MMSE, and the SINR value of LCMMSE method descends to some extent.When high SNR, the CSC method of present embodiment can obtain the performance gain of 2dB with respect to the LCMMSE method.If the LCMMSE method will keep performance constant, then width factor D need increase to 137, and this has increased time complexity undoubtedly.Therefore, the applicability of the CSC method of present embodiment is stronger than LMMSE.

Fig. 4 has provided present embodiment Signal to Interference plus Noise Ratio performance sketch map under the change condition when different channels; Abscissa coefficient correlation ρ among the figure representes the correlation between interior each moment channel response of OFDM symbol under the linear time-variant channel; It is the time variation of channel; This moment system in SNR=30dB, N=256, D=70.Can know when ρ changes that by Fig. 4 the output SINR of system changes hardly under the MMSE method.The MMSE equalization methods has utilized all channel informations in the frequency domain response matrix, is not similar to omission, recovers fully at receiving terminal, and the correlation between the channel response is the characteristic of internal matrix, so the ρ value is little to the MMSE method affect.AVE, CSC, LCMMSE method come down to utilize partial channel knowledge to remove equivalent whole channel response, so receive each restriction of correlation between the channel response constantly.When correlation is strong more, when promptly the channel time variation was more weak, the approximate channel response of equivalence approached real channel response more.Most typical is ρ=1 o'clock, and channel is a LTI, and what AVE, CSC, LCMMSE method obtained is the precise channels response, and portfolio effect is similar.Corresponding to ρ=l, the SINR value of AVE, CSC, LCMMSE, 4 kinds of methods of MMSE equates, is approximately 29dB among Fig. 4.

Can be known that by Fig. 4 on the one hand, with respect to AVE, ρ is more little, the SINR gain that the CSC method obtains is bigger, is 6dB at least, explains that the CSC method can obtain the more property gain when channel time variation was big more; With respect to LCMMSE, when ρ changed, the CSC method can keep performance consistent when reducing complexity; Because the time variation of channel in an OFDM symbol is less, so coefficient correlation ρ is bigger, but the CSC method that is proposed by Fig. 4 and time complexity knowledge capital embodiment can reduce time complexity significantly when performance moves closer to the MMSE method.

Fig. 5 has provided the Signal to Interference plus Noise Ratio performance sketch map of present embodiment under different channels decline degree; For the channel response power spectrum; The channel internal power that definition-20dB width factor Q is illustrated in time delay expansion drops to-time width of 20dB, and it has characterized channel fading speed degree.This moment system in, signal to noise ratio snr=30dB, sub-carrier number N=256, width factor D=70, coefficient correlation ρ=0.9.In the wireless communication in high-frequency band system, multipath component is abundant not as low frequency signal, and the propagation decline of signal is bigger, causes power attenuation very fast, so the Q value is less.Can know that in conjunction with Fig. 5 and time complexity when Q diminished, the performance of the CSC method that present embodiment proposes surpassed the LCMMSE method, and approached the MMSE method of best performance gradually, and the time complexity of CSC method well below the back both.So in the exigent high-frequency wireless system to hardware complexity, the CSC method that present embodiment proposes has more wide prospect.

Claims (4)

1. the method that the low complex degree inter-carrier interference is eliminated in the ofdm system is characterized in that, comprises the steps:

Step 1:, remove time domain and receive signal { y at the receiving terminal of ofdm system _1nCyclic Prefix, obtain burst { y _n;

Step 2: based on the SVD result of channel response constantly of head and the tail in the OFDM symbol, the method that the utilization channel is cut apart is divided into the cascade of LTI channel and no time delay time varying channel with linear time-variant channel, and concrete steps are:

1), ties up matrix H=(h (1 :) through structure L * 2 according to first channel response h constantly in the OFDM symbol (1 :) and tail channel response h (N :) constantly ^TH (N :) ^T), H is carried out the SVD decomposition obtains:

$H=(h_1.h_2)\left(\begin{array}{cc}\sqrt{{\mathrm{\&lambda;}}_1}& 0\\ 0& \sqrt{{\mathrm{\&lambda;}}_2}\end{array}\right)\left(\begin{array}{cc}{u}_{11}^{*}& u_{21}^{*}\\ u_{12}^{*}& u_{22}^{*}\end{array}\right)$

$=(\sqrt{{\mathrm{\&lambda;}}_1}{u}_{11}^{*}{h}_1+\sqrt{{\mathrm{\&lambda;}}_2}{u}_{12}^{*}{h}_2,\sqrt{{\mathrm{\&lambda;}}_1}{u}_{21}^{*}{h}_1+\sqrt{{\mathrm{\&lambda;}}_2}{u}_{22}^{*}{h}_2)$

That is: $\left\{\begin{array}{c}h{(1,:)}^T=\sqrt{{\mathrm{\&lambda;}}_1}{u}_{11}^{*}{h}_1+\sqrt{{\mathrm{\&lambda;}}_2}{u}_{12}^{*}{h}_2\\ h{(N,:)}^T=\sqrt{{\mathrm{\&lambda;}}_1}{u}_{21}^{*}{h}_1+\sqrt{{\mathrm{\&lambda;}}_2}{u}_{22}^{*}{h}_2\end{array}\right.$

Wherein:

With

The singular value that expression SVD obtains, h ₁And h ₂The unit character vector that expression SVD obtains, h ₁And h ₂Length be L, u ₁₁ ^*, u ₁₂ ^*, u ₂₁ ^*And u ₂₂ ^*The characteristic element that expression SVD obtains, N is the number of subcarrier, L is the number of multipath;

2) to OFDM symbol head and the tail constantly channel response h (1 :), h (N :) carry out approximate processing:

$\left\{\begin{array}{c}h(1,:)\&ap;\sqrt{{\mathrm{\&lambda;}}_1}{u}_{11}^{*}{h}_1^T={\mathrm{\&beta;}}_1{h}_1^T\\ h(N,:)\&ap;\sqrt{{\mathrm{\&lambda;}}_1}{u}_{21}^{*}{h}_1^T={\mathrm{\&beta;}}_N{h}_1^T\end{array}\right.$

3) obtaining in the OFDM symbol constantly the channel response h of n (n :) through linear interpolation is:

$\left\{\begin{array}{c}h(n,:)={\mathrm{\&beta;}}_n{h}_1^T\\ {\mathrm{\&beta;}}_n={\mathrm{\&beta;}}_1+\frac{{\mathrm{\&beta;}}_N-{\mathrm{\&beta;}}_1}{N-1}(n-1)\end{array}\right.$

4) linear time-variant channel h (n :) is split into LTI h ₁The cascade of (n :) and no time delay time varying channel h, that is: h (n, 1)=h (n) * h ₁(n, 1)

Wherein: $\left\{\begin{array}{c}{h}_1(n,:)=\{h_1(n,l),1\&le;l\&le;L\}=h_1^T\\ h=\{h\left(n\right),1\&le;n\&le;N\}=({\mathrm{\&beta;}}_1,{\mathrm{\&beta;}}_2,L,{\mathrm{\&beta;}}_N)\end{array}\right.;$

Step 3: utilize the no time delay time varying channel response in the step 2, the time domain that removes behind the CP is received signal { y _nCarry out the one dimension time domain equalization, obtain the initial restoring signal sequence of time domain y ' _n;

Step 4: to the initial restoring signal sequence of time domain y ' _nDo fast Fourier transform, obtain frequency domain initial restoring signal sequence Y ' _k;

Step 5: utilize the LTI channel response in the step 2, to frequency domain initial restoring signal sequence Y ' _kCarry out the one dimension frequency domain equalization, accomplish ICI and eliminate, obtain final frequency domain send signal recovery value sequence X ' _k.

2. the method that the low complex degree inter-carrier interference is eliminated in the ofdm system according to claim 1 is characterized in that, the Cyclic Prefix described in the step 1 is that to receive in the signal segment length before the symbol be that the information bit of Cyclic Prefix realizes through removing.

3. the method that the low complex degree inter-carrier interference is eliminated in the ofdm system according to claim 1 is characterized in that, the one dimension time domain equalization described in the step 3, and concrete formula is:

$y_n^{\&prime;}=\frac{{\mathrm{\&beta;}}_n^{*}{y}_n{{\mathrm{\&sigma;}}_x}^2}{{\left|{\mathrm{\&beta;}}_n\right|}^2{{\mathrm{\&sigma;}}_x}^2+{{\mathrm{\&sigma;}}_N}^2}$

Wherein: y _nBe to remove behind the CP time-domain signal of n constantly, y ' _nExpression is the initial restoring signal of time domain of n constantly, σ _X ²The variance of burst, σ are sent in expression _N ²The variance of expression interchannel noise.

4. the method that the low complex degree inter-carrier interference is eliminated in the ofdm system according to claim 1 is characterized in that, the one dimension frequency domain equalization described in the step 5, and concrete formula is:

X′ _k＝Y′ _k/H′ _k

Wherein: $h_1^{\&prime;}(n,k)=\left\{\begin{array}{cc}{h}_1(n,k)& 1\&le;k\&le;L\\ 0& L+1\&le;k\&le;N\end{array}\right.,$ X ' _kThe transmission signal recovery value that finally obtains at receiving terminal behind the expression elimination ICI, Y ' _kExpression frequency domain initial restoring signal value, H ' _kExpression LTI channel { h ₁, (n, 1), the sequence spreading of 1≤1≤L} h ' ₁(n, k), the N of 1≤k≤N} point DFT sequence, N is the number of subcarrier, L is the number of multipath.

Images