CN101355541A - Blocking equalizing method for OFDM system under Quick-Change channel condition - Google Patents

Abstract

The invention provides a block balancing method for an orthogonal frequency division multiplexing (OFDM) system on the condition of a fast time variant channel. The method comprises the following: at a sending terminal, a bit signal subjected to cyclic redundancy codification is subjected to channel codification, interlacement, symbol map, serial/parallel conversion and IFFT conversion to be converted to a time domain signal; the time domain signal which is subjected to serial/parallel conversion and is inserted with a cyclic prefix is sent to a radio channel; at a receiving terminal, the signal is received by a traditional first-order frequency domain equalizer, and the signal guess value is checked after the signal is subject to serial/parallel conversion, demap, deinterlacement and signal path decoding; if the check is correct, the signal is outputted; if the check is not correct, the block balancing process is performed, during which, firstly, a time domain OFDM sending signal is restructured; secondly, the received time domain OFDM symbol is divided into a plurality of equilong signal receiving subblocks, the loop restructuring and the interference removing processes are performed by a time domain OFDM restructured signal; and finally, each subblock is subjected to frequency domain balance, the balanced guess values of sub blocks are merged to obtain the block balance output serving as the input of the serial/parallel conversion.

Description

Blocking equalizing method in the OFDM system under Quick-Change channel condition

Technical field

The invention belongs to OFDM (OFDM) mobile communication technology field.

Background technology

Reason more and more is subjected to people's attention to OFDM (OFDM, Orthogonal Frequency Division Multiplexing) technology owing to its availability of frequency spectrum height, implementation complexity be low etc.Since the eighties in 20th century, the OFDM technology not only is widely used in broadcast type digital audio and video field, and has become the part of WLAN standard.Along with the enhancing of people to communication dataization, broadband, individualized and mobile demand, the OFDM technology is applied at many high speed information transmission fields.At present, people are considering to use the OFDM technology in the radio honeycomb mobile communication system of future generation in future.Yet work in OFDM mobile system under high carrier frequency, the high rate travel condition field selectivity and frequency domain selectivity, i.e. double selectivity will cause channel to exist simultaneously the time.In this case, become the orthogonality of having destroyed between subcarrier during channel quick, thereby disturb (ICI) between causing carrying, traditional single order equalizer equalizes performance is affected, even can't work, so the balancing technique of ofdm system becomes the focus of current research under the Quick-Change channel.

In order to improve the robustness that frequency-domain equalizer is worked in the time varying channel environment, the interference cancellation techniques of cdma system Multiuser Detection can be combined with frequency-domain equalization technology, utilize the means of iterative detection to reduce the influence of ICI to frequency-domain equalizer.As people such as Wesheng Hou at " IEEE Trans.On WirelessCommunications " vol.4, no.5, pp.2100-2110, September 2005 has delivered " ICICancellation for OFDM Communication Systems in Time-Varing MultipathFading Channels " (IEEE radio communication journal in September, 2005, the 4th volume, the 5th phase, 2100 to 2110 pages, the time become carry in the ofdm communication system under the multidiameter fading channel between interference eliminated).To bring bigger processing delay to system but this method is introduced Interference Cancellation, the performance of its performance and the elementary detection of frequency-domain equalizer is closely related simultaneously, and has the error propagation problem.Another kind method be be used to improve when there is carrier wave frequency deviation in transceiver service behaviour the self-interference cancellation technology be used for ofdm system equilibrium under the time varying channel, as people such as Yuping Zhao at " IEEE Trans.On Communications " vol.49, no.7, pp.1185-1191, July 2001 has delivered " Intercarrier Interference Self-CancellationScheme for OFDM Mobile Communication Systems " (ieee communication journal July calendar year 2001, the 49th volume, the 7th phase, 1185 to 1191 pages, disturb self elimination method between carrying in the mobile ofdm communication system).But this method need be carried out special precoding at transmitting terminal, has reduced the band efficiency of system.In addition, Philip Schniter is at " IEEE Trans.On Signal Processing " vol.52, no.4, pp.1002-1011, April 2004 has delivered " Low-Complexity Equalization of OFDM inDoubly Selective Channels " (IEEE signal processing journal in April, 2004, the 52nd volume, the 4th phase, 1002 to 1011 pages, OFDM low complex degree equalization under the dual-selection channel), this method utilizes time-domain windowed that time varying channel is carried out filtering earlier, utilizing iteration MMSE estimated signal, its shortcoming is accurately to know the channel condition information of statistical information He each moment point of channel, and this channel estimator for actual ofdm system is to be difficult for realizing.People such as Li Wei propose to utilize the method antagonism time varying channel of piecemeal frequency domain equalization in single-carrier frequency domain equalization system and ofdm system, but this method is divided into received signal a plurality of sub-pieces simply to be handled, do not consider sub-inter-block-interference problem, cause equalization performance to descend.

Summary of the invention

The objective of the invention is to propose blocking equalizing method and device in the ofdm system under a kind of Quick-Change channel condition, to reduce because the performance loss that the time variation of channel brings.

The present invention is achieved through the following technical solutions:

1, blocking equalizing method in the ofdm system under a kind of Quick-Change channel condition is characterized in that:

At transmitting terminal, through the bit signal behind cyclic redundancy code stream b (n) channel encode, interweave, formation frequency domain orthogonal frequency-division multiplex singal behind sign map and the serial to parallel conversion, i frequency domain orthogonal frequency division multiplex ransmitting feed signals is expressed as s (i)=[s (i, 0), s (i, 1),, s (i, N-1)] ^T, N is a sub-carrier number, s (i) becomes time domain orthogonal frequency division multiplex ransmitting feed signals through contrary fast fourier transform

After, inserting length is the Cyclic Prefix of G, and sends to wireless channel behind parallel serial conversion;

At receiving terminal, the first step: after removing Cyclic Prefix, serial to parallel conversion, i time domain orthogonal frequency division multiplexing receiving symbol $\tilde{r}\left(i\right)={[\tilde{r}(i,0),\tilde{r}(i,1),\·\·\·,\tilde{r}(i,N-1)]}^T$ Can be expressed as:

$\tilde{r}\left(i\right)=H_t^i\tilde{s}\left(i\right)+\tilde{w}\left(i\right)=H_t^i{F}^{H}s\left(i\right)+\tilde{w}\left(i\right)$

Here H _t ⁱBe channel matrix,

Expression power is σ ²The white Gaussian noise vector, F represents fourier transform matrix, () ^HThe operation of expression conjugate transpose, in order to carry out the equilibrium of single-order frequency domain equalization device, it is static earlier the channel hypothesis to be as the criterion, and promptly the variation of channel can be ignored in 1 OFDM symbol transmission time, multipath channel is modeled as the FIR filter of constant finite impulse response when having the L-1 rank, uses h _l(i) expression l rank impulse response.

This moment H _t ⁱBe a convolution matrix, can be expressed as

$H_t^i=F^H{\mathrm{\Λ}}^{i}F$

Here Λ ⁱBe diagonal matrix,

Become frequency domain OFDM receiving symbol through fast fourier transform $r\left(i\right)=F\tilde{r}\left(i\right),$ Utilize the balanced frequency domain OFDM receiving symbol r (i) of single-order frequency domain equalization device, the frequency domain orthogonal frequency-division multiplex singal estimated value of equalizer output

Can be expressed as:

$\hat{s}\left(i\right)={\mathrm{\Λ}}^{\mathrm{iH}}{({\mathrm{\Λ}}^i{\mathrm{\Λ}}^{\mathrm{iH}}+{\mathrm{\σ}}^{2}I)}^{-1}r\left(i\right)$

Through parallel serial conversion, separate mapping, interweave, decoding obtains the estimated value of transmitting terminal bit signal stream

Second step: to the transmitting terminal bit signal estimated value that obtains

Carry out cyclic redundancy check (CRC), if cyclic redundancy check (CRC) is correct, then output; If cyclic redundancy check (CRC) is incorrect, then carry out blocking equalizing, with the output of blocking equalizing input as parallel serial conversion in the first step,

Above-mentioned blocking equalizing method can be expressed as follows:

A. with the estimated value of the transmitting terminal bit signal that obtains in first step stream

Carry out the chnnel coding identical, interweave, after the sign map, serial to parallel conversion, contrary fast fourier transform, obtain time domain orthogonal frequency division multiplexing reconstruction signal with transmitting terminal ${\tilde{s}}^{\′}\left(i\right)={[{\tilde{s}}^{\′}(i,0),{\tilde{s}}^{\′}(i,1),\·\·\·,{\tilde{s}}^{\′}(i,N-1)]}^T,$

B. the time domain orthogonal frequency division multiplexing symbol that each is received

Be divided into M the isometric sub-piece of received signal, sub-piece is counted M and is satisfied:

And be 2 integral number power, here

Expression rounds operation downwards, and the length of each sub-piece is T=N/M, T＞L, and k sub-piece can be expressed as ${\tilde{r}}_k\left(i\right)={[{\tilde{r}}_k(i,0),{\tilde{r}}_k(i,2),\·\·\·,{\tilde{r}}_k(i,T-1)]}^T={[\tilde{r}(i,\mathrm{kT}),\tilde{r}(i,\mathrm{kT}+1),\·\·\·,\tilde{r}(i,(k+1)T-1)]}^T,$ K ∈ [0 ..., M-1], in the time of each sub-piece transmission, think that the time change of wireless channel can be ignored, promptly $h^i(n,l)\≡h_k^i\left(l\right),$ N=kT ..., (k+1) T-1, l ∈ [0 ... L-1] be the channel l rank impulse response of the time of k sub-piece transmission, at this moment

Can be expressed as

${\tilde{r}}_k\left(i\right)=H_k^{\′}{\tilde{s}}_{k-1}\left(i\right)+H_k{\tilde{s}}_k\left(i\right)+{\tilde{w}}_k\left(i\right)$

Here ${\tilde{s}}_k\left(i\right)={[{\tilde{s}}_k(i,0),{\tilde{s}}_k(i,1),\·\·\·,{\tilde{s}}_k(i,T-1)]}^T={[\tilde{s}(i,\mathrm{kT}),\tilde{s}(i,\mathrm{kT}+1),\·\·\·,\tilde{s}(i,(k+1)T-1)]}^T,$

${\tilde{w}}_k\left(i\right)={[{\tilde{w}}_k(i,1),{\tilde{w}}_k(i,2),\·\·\·,{\tilde{w}}_k(i,T)]}^T,$

Having characterized k-1 signal subspace piece crosstalks to the sub-interblock that current k sub-piece causes.

The time domain of reconstruct is sent signal

Being divided into length is the sub-piece of T ${\tilde{s}}_k^{\′}\left(i\right)={[{\tilde{s}}_k^{\′}(i,0),{\tilde{s}}_k^{\′}(i,1),\·\·\·,{\tilde{s}}_k^{\′}(i,T-1)]}^T={[{\tilde{s}}^{\′}(i,\mathrm{kT}),{\tilde{s}}^{\′}(i,\mathrm{kT}+1),\·\·\·,{\tilde{s}}^{\′}(i,(k+1)T-1)]}^T,$ Utilize

Eliminate k-1 signal subspace piece the interblock that current k sub-piece causes crosstalked,

${\tilde{r}}_k^{\′}\left(i\right)={\tilde{r}}_k\left(i\right)-H_k^{\′}{\tilde{s}}_{k-1}^{\′}\left(i\right)=H_k^{\′}({\tilde{s}}_{k-1}\left(i\right)-{\tilde{s}}_{k-1}^{\′}\left(i\right))+H_k{\tilde{s}}_k\left(i\right)+{\tilde{w}}_k\left(i\right)$

And carry out loop restructuring, and guaranteed the cycle characteristics of channel matrix, disturb between suppressing to carry,

${\tilde{r}}_k^{\′\′}\left(i\right)={\tilde{r}}_k^{\′}\left(i\right)+H_k^{\′}{\tilde{s}}_k^{\′}\left(i\right)=H_k^{\′}({\tilde{s}}_{k-1}\left(i\right)-{\tilde{s}}_{k-1}^{\′}\left(i\right))+H_k{\tilde{s}}_k\left(i\right)+H_k^{\′}{\tilde{s}}_k^{\′}\left(i\right)+{\tilde{w}}_k\left(i\right)$

If do not consider the error that reconstruct causes, promptly ${\tilde{s}}_k\left(i\right)\≡{\tilde{s}}_k^{\′}\left(i\right),\∀k,$ Then

${\tilde{r}}_k^{\′\′}\left(i\right)=H_{k,\mathrm{cir}}^{\′}{\tilde{s}}_k\left(i\right)+{\tilde{w}}_k\left(i\right)$

Convolution matrix H ' _{K, cir}=H _k+ H ' _kBe the channel matrix of k sub-piece correspondence,

C. eliminate that sub-interblock is crosstalked and reconstruction cycle prefix basis on each is received sub-piece, carry out fast fourier transform earlier, the length of its fast fourier transform is consistent with sub-block length, each receives sub-piece to utilize the equilibrium of single-order frequency domain equalization device again, the balanced valuation of the sub-piece of each after the equilibrium is through contrary fast fourier transform, obtain the time domain equalization estimated value of each sub-piece, the time domain equalization estimated value of k sub-piece can be expressed as

${\hat{\tilde{s}}}_k\left(i\right)=F^H({\mathrm{\Λ}}_k^i{\mathrm{\Λ}}_k^{\mathrm{iH}}+{\mathrm{\σ}}_w^2{I}_T){\mathrm{\Λ}}_k^{\mathrm{iH}}F{\tilde{r}}_k^{\′\′}\left(i\right)$

Here diagonal matrix ${\mathrm{\Λ}}_k^i={\mathrm{FH}}_{k,\mathrm{cir}}^{\′}{F}^H,$

D. last balanced estimated value to each sub-piece merges, and the time domain equalization estimated value of each sub-piece is the N symbol according to the synthetic length of the der group of piecemeal $\hat{\tilde{s}}\left(i\right)={[{\hat{\tilde{s}}}_0^T\left(i\right),{\hat{\tilde{s}}}_1^T\left(i\right),\·\·\·,{\hat{\tilde{s}}}_{M-1}^T\left(i\right)]}^T,$

After fast fourier transform, obtain the output of piecemeal frequency-domain equilibrium method

It promptly is the estimated value of the frequency domain orthogonal frequency-division multiplex singal of transmission.

Advantage of the present invention:

(1) the present invention proposes blocking equalizing algorithm in a kind of OFDM system under Quick-Change channel condition, become the influence of disturbing between carrying of causing when having reduced owing to channel, method is simple, and highly versatile, can be used for any ofdm system;

The increase of range degree when (2) blocking equalizing algorithm of the present invention is along with channel, advantage is more obvious.

(3) blocking equalizing algorithm of the present invention does not need to know all channel parameters constantly in the whole OFDM symbol transmission time, has reduced the complexity of system channel estimator, and is practical.

(4) the present invention will utilize the sub-piece of reconstruction signal, eliminating the sub-interblock that is caused by previous signal subspace piece in the sub-piece of each received signal crosstalks, and constitute the Cyclic Prefix of the sub-piece of current received signal, and then guarantee the cycle characteristics of its channel matrix and suppress to disturb between remaining carrying, equalization performance improved.

Description of drawings

Fig. 1 ofdm system transmitting terminal block diagram

Fig. 2 is based on the ofdm system receiving terminal block diagram of blocking equalizing

Fig. 3 is that the receiving terminal piecemeal is handled schematic diagram

Fig. 4 is that the normalization Doppler frequency shift is f _NThe bit error rate comparison diagram of=0.0124 o'clock 3 kind of equalization methods

Fig. 5 is that the normalization Doppler frequency shift is f _NThe bit error rate comparison diagram of=0.05 o'clock 3 kind of equalization methods

Fig. 6 is that the normalization Doppler frequency shift is f _NThe bit error rate comparison diagram of=0.1 o'clock 3 kind of equalization methods

Embodiment

1, blocking equalizing method in the ofdm system under a kind of Quick-Change channel condition is characterized in that:

At transmitting terminal, through the bit signal behind cyclic redundancy code stream b (n) channel encode, interweave, formation frequency domain orthogonal frequency-division multiplex singal behind sign map and the serial to parallel conversion, i frequency domain orthogonal frequency division multiplex ransmitting feed signals is expressed as s (i)=[s (i, 0), s (i, 1),, s (i, N-1)] ^T, N is a sub-carrier number, s (i) becomes time domain orthogonal frequency division multiplex ransmitting feed signals through contrary fast fourier transform

After, inserting length is the Cyclic Prefix of G, and sends to wireless channel behind parallel serial conversion;

At receiving terminal, the first step: after removing Cyclic Prefix, serial to parallel conversion, i time domain orthogonal frequency division multiplexing receiving symbol $\tilde{r}\left(i\right)={[\tilde{r}(i,0),\tilde{r}(i,1),\·\·\·,\tilde{r}(i,N-1)]}^T$ Can be expressed as:

$\tilde{r}\left(i\right)=H_t^i\tilde{s}\left(i\right)+\tilde{w}\left(i\right)=H_t^i{F}^{H}s\left(i\right)+\tilde{w}\left(i\right)$

Here H _t ⁱBe channel matrix, Expression power is σ ²The white Gaussian noise vector, F represents fourier transform matrix, () ^HThe operation of expression conjugate transpose, in order to carry out the equilibrium of single-order frequency domain equalization device, it is static earlier the channel hypothesis to be as the criterion, and promptly the variation of channel can be ignored in 1 OFDM symbol transmission time, multipath channel is modeled as the FIR filter of constant finite impulse response when having the L-1 rank, uses h _l(i) expression l rank impulse response.

This moment H _t ⁱBe a convolution matrix, can be expressed as

$H_t^i=F^H{\mathrm{\Λ}}^{i}F$

Here Λ ⁱBe diagonal matrix,

Become frequency domain OFDM receiving symbol through fast fourier transform $r\left(i\right)=F\tilde{r}\left(i\right),$ Utilize the balanced frequency domain OFDM receiving symbol r (i) of single-order frequency domain equalization device, the frequency domain orthogonal frequency-division multiplex singal estimated value of equalizer output

Can be expressed as:

$\hat{s}\left(i\right)={\mathrm{\Λ}}^{\mathrm{iH}}{({\mathrm{\Λ}}^i{\mathrm{\Λ}}^{\mathrm{iH}}+{\mathrm{\σ}}^{2}I)}^{-1}r\left(i\right)$

Through parallel serial conversion, separate mapping, interweave, decoding obtains the estimated value of transmitting terminal bit signal stream

Second step: to the transmitting terminal bit signal estimated value that obtains

Carry out cyclic redundancy check (CRC), if cyclic redundancy check (CRC) is correct, then output; If cyclic redundancy check (CRC) is incorrect, then carry out blocking equalizing, with the output of blocking equalizing input as parallel serial conversion in the first step,

Above-mentioned blocking equalizing method can be expressed as follows:

A. with the estimated value of the transmitting terminal bit signal that obtains in first step stream

Carry out the chnnel coding identical, interweave, after the sign map, serial to parallel conversion, contrary fast fourier transform, obtain time domain orthogonal frequency division multiplexing reconstruction signal with transmitting terminal ${\tilde{s}}^{\′}\left(i\right)={[{\tilde{s}}^{\′}(i,0),{\tilde{s}}^{\′}(i,1),\·\·\·,{\tilde{s}}^{\′}(i,N-1)]}^T,$

B. the time domain orthogonal frequency division multiplexing symbol that each is received

Be divided into M the isometric sub-piece of received signal, sub-piece is counted M and is satisfied: And be 2 integral number power, here

Expression rounds operation downwards, and the length of each sub-piece is T=N/M, T＞L, and k sub-piece can be expressed as ${\tilde{r}}_k\left(i\right)={[{\tilde{r}}_k(i,0),{\tilde{r}}_k(i,2),\·\·\·,{\tilde{r}}_k(i,T-1)]}^T={[\tilde{r}(i,\mathrm{kT}),\tilde{r}(i,\mathrm{kT}+1),\·\·\·,\tilde{r}(i,(k+1)T-1)]}^T,$ K ∈ [0 ..., M-1], in the time of each sub-piece transmission, think that the time change of wireless channel can be ignored, promptly $h^i(n,l)\≡h_k^i\left(l\right),$ N=kT ..., (k+1) T-1, l ∈ [0 ... L-1] be the channel l rank impulse response of the time of k sub-piece transmission, at this moment

Can be expressed as

${\tilde{r}}_k\left(i\right)=H_k^{\′}{\tilde{s}}_{k-1}\left(i\right)+H_k{\tilde{s}}_k\left(i\right)+{\tilde{w}}_k\left(i\right)$

Here ${\tilde{s}}_k\left(i\right)={[{\tilde{s}}_k(i,0),{\tilde{s}}_k(i,1),\·\·\·,{\tilde{s}}_k(i,T-1)]}^T={[\tilde{s}(i,\mathrm{kT}),\tilde{s}(i,\mathrm{kT}+1),\·\·\·,\tilde{s}(i,(k+1)T-1)]}^T,$

${\tilde{w}}_k\left(i\right)={[{\tilde{w}}_k(i,1),{\tilde{w}}_k(i,2),\·\·\·,{\tilde{w}}_k(i,T)]}^T,$

Having characterized k-1 signal subspace piece crosstalks to the sub-interblock that current k sub-piece causes.

The time domain of reconstruct is sent signal

Being divided into length is the sub-piece of T ${\tilde{s}}_k^{\′}\left(i\right)={[{\tilde{s}}_k^{\′}(i,0),{\tilde{s}}_k^{\′}(i,1),\·\·\·,{\tilde{s}}_k^{\′}(i,T-1)]}^T={[{\tilde{s}}^{\′}(i,\mathrm{kT}),{\tilde{s}}^{\′}(i,\mathrm{kT}+1),\·\·\·,{\tilde{s}}^{\′}(i,(k+1)T-1)]}^T,$ Utilize

Eliminate k-1 signal subspace piece the interblock that current k sub-piece causes crosstalked,

${\tilde{r}}_k^{\′}\left(i\right)={\tilde{r}}_k\left(i\right)-H_k^{\′}{\tilde{s}}_{k-1}^{\′}\left(i\right)=H_k^{\′}({\tilde{s}}_{k-1}\left(i\right)-{\tilde{s}}_{k-1}^{\′}\left(i\right))+H_k{\tilde{s}}_k\left(i\right)+{\tilde{w}}_k\left(i\right)$

And carry out loop restructuring, and guaranteed the cycle characteristics of channel matrix, disturb between suppressing to carry,

${\tilde{r}}_k^{\′\′}\left(i\right)={\tilde{r}}_k^{\′}\left(i\right)+H_k^{\′}{\tilde{s}}_k^{\′}\left(i\right)=H_k^{\′}({\tilde{s}}_{k-1}\left(i\right)-{\tilde{s}}_{k-1}^{\′}\left(i\right))+H_k{\tilde{s}}_k\left(i\right)+H_k^{\′}{\tilde{s}}_k^{\′}\left(i\right)+{\tilde{w}}_k\left(i\right)$

If do not consider the error that reconstruct causes, promptly ${\tilde{s}}_k\left(i\right)\≡{\tilde{s}}_k^{\′}\left(i\right),\∀k,$ Then

${\tilde{r}}_k^{\′\′}\left(i\right)=H_{k,\mathrm{cir}}^{\′}{\tilde{s}}_k\left(i\right)+{\tilde{w}}_k\left(i\right)$

Convolution matrix H ' _{K, cir}=H _k+ H ' _kBe the channel matrix of k sub-piece correspondence,

C. eliminate that sub-interblock is crosstalked and reconstruction cycle prefix basis on each is received sub-piece, carry out fast fourier transform earlier, the length of its fast fourier transform is consistent with sub-block length, each receives sub-piece to utilize the equilibrium of single-order frequency domain equalization device again, the balanced valuation of the sub-piece of each after the equilibrium is through contrary fast fourier transform, obtain the time domain equalization estimated value of each sub-piece, the time domain equalization estimated value of k sub-piece can be expressed as

${\hat{\tilde{s}}}_k\left(i\right)=F^H({\mathrm{\Λ}}_k^i{\mathrm{\Λ}}_k^{\mathrm{iH}}+{\mathrm{\σ}}_w^2{I}_T){\mathrm{\Λ}}_k^{\mathrm{iH}}F{\tilde{r}}_k^{\′\′}\left(i\right)$

Here diagonal matrix ${\mathrm{\Λ}}_k^i={\mathrm{FH}}_{k,\mathrm{cir}}^{\′}{F}^H,$

D. last balanced estimated value to each sub-piece merges, and the time domain equalization estimated value of each sub-piece is the N symbol according to the synthetic length of the der group of piecemeal $\hat{\tilde{s}}\left(i\right)={[{\hat{\tilde{s}}}_0^T\left(i\right),{\hat{\tilde{s}}}_1^T\left(i\right),\·\·\·,{\hat{\tilde{s}}}_{M-1}^T\left(i\right)]}^T,$ After fast fourier transform, obtain the output of piecemeal frequency-domain equilibrium method

It promptly is the estimated value of the frequency domain orthogonal frequency-division multiplex singal of transmission.

Below in conjunction with accompanying drawing the present invention is described further, so that purpose of the present invention, feature and advantage are clearer.

The consideration sub-carrier number is N, and cyclic prefix CP (cyclic prefix) is the CP-OFDM system of G.As shown in Figure 1, at transmitting terminal, bit stream b (n) channel behind cyclic redundancy code coding interweaves, sign map, forming size behind the serial to parallel conversion is the grouping of N, i grouping sheet be shown s (i)=[s (i, 0), s (i, 1) ..., s (i, N-1)] ^TVector s (i) is transformed into the time domain vector through N point IFFT modulation

After, insert the cyclic prefix CP that length is G again, behind parallel serial conversion, send into wireless channel.

As shown in Figure 2, at receiving terminal, the time-domain signal that receives is at first removed CP, through serial to parallel conversion, and i the OFDM symbol that receives $\tilde{r}\left(i\right)={[\tilde{r}(i,0),\tilde{r}(i,1),\·\·\·,\tilde{r}(i,N-1)]}^T$ Can be expressed as:

$\tilde{r}\left(i\right)=H_t^i\tilde{s}\left(i\right)+\tilde{w}\left(i\right)=H_t^i{F}^{H}s\left(i\right)+\tilde{w}\left(i\right)---\left(1\right)$

Here

Expression power is σ ²The white Gaussian noise vector, F represents FFT matrix, () ^HThe operation of expression conjugate transpose.

If it is static that the channel hypothesis is as the criterion, promptly the variation of channel can be ignored in 1 OFDM symbol transmission time, and multipath channel is modeled as the FIR filter of constant finite impulse response when having the L-1 rank, uses h _l(i) expression l rank impulse response.

H then _t ⁱBe exactly a convolution matrix, can be expressed as

$H_t^i=F{\mathrm{\Λ}}^i{F}^H$

Here Λ ⁱIt is diagonal matrix.Therefore at receiving terminal, send signal and can utilize simple single order equalizer to detect:

${\tilde{s}}^{\′}\left(i\right)={\mathrm{\Λ}}^{\mathrm{iH}}{({\mathrm{\Λ}}^i{\mathrm{\Λ}}^{\mathrm{iH}}+{\mathrm{\σ}}^{2}I)}^{-1}F\tilde{r}\left(i\right)---\left(2\right)$

Yet under Quick-Change channel condition, the variation of channel be can not ignore in 1 OFDM symbol transmission time, disturbs (ICI) between having caused carrying.Be modeled as the L-1 rank limited FIR filter that when having become impulse response with multipath channel this moment, uses h _{N, l}(i) the expression l rank impulse response of n constantly, therefore

At this moment,

$H_t^i={\mathrm{FH}}_f^i{F}^H$

Here H ⁱ _fNot diagonal matrix, H ⁱ _fIn element on the off-diagonal disturb that (inter-carrierinterference, ICI), the equalization performance of the single order equalizer that this moment is traditional must be affected between having caused year.Some equalization algorithms that propose can alleviate or eliminate the influence of channel ICI, but inevitably bring the change of signal format, the increase of computational complexity, the unfavorable factors such as reduction of efficiency of transmission.

In real system, the significance degree that becomes when channel is is relevant with the length of observing time.Though, for the high-speed mobile ofdm system, in 1 OFDM mark space, become during channel and can not ignore, but carry out equilibrium respectively if 1 OFDM symbol is divided into the experimental process piece, then for each sub-piece experience the time at interval in channel the time to become may be very little, even can ignore.Based on this thought, the present invention proposes blocking equalizing method in the ofdm system under a kind of Quick-Change channel condition

As shown in Figure 2, carry out the process identical earlier at receiving terminal and carry out initial estimation with traditional ofdm system demodulation.

Transform to frequency domain through FFT, the frequency-domain OFDM symbolic representation is r (i).After carrying out the equilibrium of frequency domain single order, obtain taking place the estimation of signal s (i) by formula (2) Through parallel serial conversion, separate mapping, interweave, decoding obtains transmitting terminal bit signal estimated value

After, carry out cyclic redundancy check (CRC).If cyclic redundancy check (CRC) is correct, then output; Otherwise, then carry out blocking equalizing.

Before blocking equalizing, utilize decoding to obtain earlier

Carry out the chnnel coding identical, interweave, after the sign map, serial to parallel conversion, contrary fast fourier transform, obtain time domain orthogonal frequency division multiplexing reconstruction signal with transmitting terminal ${\tilde{s}}^{\′}\left(i\right)={[{\tilde{s}}^{\′}(i,0),{\tilde{s}}^{\′}(i,1),\·\·\·,{\tilde{s}}^{\′}(i,N-1)]}^T.$

As shown in Figure 3, the time domain orthogonal frequency division multiplexing symbol that each is received Be divided into M the isometric sub-piece of received signal, sub-piece is counted M and is satisfied:

And be 2 integral number power, here Expression rounds operation downwards, and the length of each sub-piece is T=N/M, T＞L, the k of i OFDM symbol sub-piece ${\tilde{r}}_k\left(i\right)={[{\tilde{r}}_k(i,0),{\tilde{r}}_k(i,2),\·\·\·,{\tilde{r}}_k(i,T-1)]}^T={[\tilde{r}(i,\mathrm{kT}),\tilde{r}(i,\mathrm{kT}+1),\·\·\·,\tilde{r}(i,(k+1)T-1)]}^T,$ k∈[0，…，M-1]。In the time of each sub-piece transmission, think that the time change of wireless channel can be ignored, promptly $h^i(n,l)\≡h_k^i\left(l\right),$ N=kT ..., (k+1) T-1, l ∈ [0 ... L-1] be the channel l rank impulse response of the time of k sub-piece transmission,

Can be expressed as,

${\tilde{r}}_k\left(i\right)=H_k^{\′}{\tilde{s}}_{k-1}\left(i\right)+H_k{\tilde{s}}_k\left(i\right)+{\tilde{w}}_k\left(i\right)---\left(3\right)$

Here $s_k\left(i\right)={[s_k(i,1),s_k(i,2),\·\·\·,s_k(i,T)]}^T,$

${\tilde{w}}_k\left(i\right)={[{\tilde{w}}_k(i,1),{\tilde{w}}_k(i,2),\·\·\·,{\tilde{w}}_k(i,T)]}^T.$

In the formula (3)

Having characterized previous signal subspace piece crosstalks to the sub-interblock that current sub-block causes.The piecemeal frequency-domain equilibrium method that people such as LiWei propose is not only considered

Influence, and simply with H _kDeng being all convolution matrix, its equalization methods causes between bigger carrying at frequency domain and disturbs, and causes equalization performance to descend.The present invention utilizes

Right Carry out loop restructuring, both solved sub-interblock cross-interference issue, guaranteed the cycle characteristics of channel matrix again, disturb between suppressing to carry, obviously improved equalization performance.

The time domain of reconstruct is sent signal Being divided into length is the sub-piece of T ${\tilde{s}}_k^{\′}\left(i\right)={[{\tilde{s}}_k^{\′}(i,0),{\tilde{s}}_k^{\′}(i,1),\·\·\·,{\tilde{s}}_k^{\′}(i,T-1)]}^T={[{\tilde{s}}^{\′}(i,\mathrm{kT}),{\tilde{s}}^{\′}(i,\mathrm{kT}+1),\·\·\·,{\tilde{s}}^{\′}(i,(k+1)T-1)]}^T,$ Utilize

Eliminate k-1 signal subspace piece the interblock that current k sub-piece causes crosstalked,

${\tilde{r}}_k^{\′}\left(i\right)={\tilde{r}}_k\left(i\right)-H_k^{\′}{\tilde{s}}_{k-1}^{\′}\left(i\right)=H_k^{\′}({\tilde{s}}_{k-1}\left(i\right)-{\tilde{s}}_{k-1}^{\′}\left(i\right))+H_k{\tilde{s}}_k\left(i\right)+{\tilde{w}}_k\left(i\right)---\left(4\right)$

And carry out loop restructuring, and guaranteed the cycle characteristics of channel matrix, disturb between suppressing to carry,

${\tilde{r}}_k^{\′\′}\left(i\right)={\tilde{r}}_k^{\′}\left(i\right)+H_k^{\′}{\tilde{s}}_k^{\′}\left(i\right)=H_k^{\′}({\tilde{s}}_{k-1}\left(i\right)-{\tilde{s}}_{k-1}^{\′}\left(i\right))+H_k{\tilde{s}}_k\left(i\right)+H_k^{\′}{\tilde{s}}_k^{\′}\left(i\right)+{\tilde{w}}_k\left(i\right)---\left(5\right)$

If do not consider the error that reconstruct causes, promptly ${\tilde{s}}_k\left(i\right)\≡{\tilde{s}}_k^{\′}\left(i\right),\∀k,$ Then

${\tilde{r}}_k^{\′\′}\left(i\right)=H_{k,\mathrm{cir}}^{\′}{\tilde{s}}_k\left(i\right)+{\tilde{w}}_k\left(i\right)---\left(6\right)$

Here convolution matrix H ' _{K, cir}=H _k+ H ' _kIt is the channel matrix of k sub-piece correspondence.Eliminate that sub-interblock is crosstalked and reconstruction cycle prefix basis on each is received sub-piece, carry out fast fourier transform earlier, the length of its fast fourier transform is consistent with sub-block length, each receives sub-piece to utilize the equilibrium of single-order frequency domain equalization device again, the balanced valuation of the sub-piece of each after the equilibrium is through contrary fast fourier transform, obtain the time domain equalization estimated value of each sub-piece, the time domain equalization estimated value of k sub-piece can be expressed as

${\hat{\tilde{s}}}_k\left(i\right)=F^H({\mathrm{\Λ}}_k^i{\mathrm{\Λ}}_k^{\mathrm{iH}}+{\mathrm{\σ}}_w^2{I}_T){\mathrm{\Λ}}_k^{\mathrm{iH}}F{\tilde{r}}_k^{\′\′}\left(i\right)$

Here diagonal matrix ${\mathrm{\Λ}}_k^i={\mathrm{FH}}_{k,\mathrm{cir}}^{\′}{F}^H,$

Merging each sub-piece estimates

The then estimation of i OFDM time domain transmission symbol $\hat{\tilde{s}}\left(i\right)={[{\hat{\tilde{s}}}_1^T\left(i\right),{\hat{\tilde{s}}}_2^T\left(i\right),\·\·\·,{\hat{\tilde{s}}}_M^T\left(i\right)]}^T,$

Transform to frequency domain through FFT, obtain the output of the piecemeal frequency-domain equilibrium method of i OFDM symbol Through separating mapping, after deinterleaving and the decoding, carry out cyclic redundancy check (CRC).If cyclic redundancy is correct, then output; Otherwise, carry out the blocking equalizing process once more.

It is can not ignore that starting point of the present invention becomes during channel in 1 OFDM mark space, but, 1 OFDM symbol carries out equilibrium respectively if being divided into the experimental process piece, then for each sub-piece experience the time at interval in channel the time to become may be very little, even can ignore.Therefore to guarantee each sub-piece experience the time at interval in channel the time become and can ignore, the size of selection block count M that must be suitable.M is bigger, and then each sub-piece is more little, thus each sub-piece experience the time at interval in the Shi Bianyue of channel little.Yet, to crosstalk in order to eliminate sub-interblock, the selection of M must guarantee sub-block length T＞L.And owing to each sub-piece all needs to carry out loop restructuring before equilibrium, so the selection of block count M size also will consider because the error propagation that loop restructuring causes and the influence of system's computational complexity.Simultaneously, in order to lower complexity, the time-frequency domain conversation of each sub-piece realizes with fast Fourier transform (FFT) that normally therefore sub-block length T should be 2 integral number power when realizing.Therefore sub-piece is counted M and need be satisfied:

And be 2 integral number power, here

Expression rounds operation downwards, and the length of each sub-piece is T=N/M, and T＞L selects M=8 for the N=1024 system, and it is more satisfactory that M=4 selects in the N=256 system.

Be example with WiMAX (IEEE802.16) system below, further specify the blocking equalizing algorithm of proposition.Its sub-carrier number is M=256.The operating frequency of system is f _c=3.5GHz, sample frequency is f _s=2MHz, CP length is 8; Channel model adopts Stanford University's the third model of intermediate channels model (SUI3) of the modification of IEEE802.16 working group suggestion, and maximum delay expands to 1 μ s, has 3 footpaths, time delay is respectively 0 μ s, 0.5 μ s, 1 μ s, the average power in 3 footpaths is respectively 0dB ,-5dB ,-10dB.Definition normalization Doppler frequency shift is $f_N=N\frac{{\mathrm{vf}}_c}{{\mathrm{cf}}_s},$ Here v is the translational speed of communication terminal, and c is the light velocity.Consider that because the error propagation that causes of loop restructuring and the influence of system's computational complexity, selecting the piecemeal number is M=4.

Fig. 4,5 and 6 has shown portable terminal under different translational speeds, and the piecemeal frequency-domain equilibrium method of traditional single order equilibrium, Li Wei and the performance of BER that the present invention proposes the blocking equalizing algorithm are relatively.

As shown in Figure 4, the portable terminal translational speed is 30km/h, and this moment, corresponding normalization Doppler frequency shift was 0.0124.As can be seen from the figure, under the low speed mobile environment, Doppler frequency shift is less, the time range degree of channel is not obvious, the piecemeal frequency-domain equilibrium method of Li Wei has influenced equalization performance, its performance of BER even poorer slightly than traditional single order equalizer owing to do not consider sub-inter-block-interference.And the blocking equalizing algorithm that the present invention proposes solves sub-interblock cross-interference issue, has guaranteed the cycle characteristics of channel matrix again, disturbs between suppressing to carry, and has obviously improved equalization performance, is 10 in bit error rate ^-2The time, compare with traditional single order equalizer, obtained the gain of about 2dB.

Under the low speed mobile environment, Doppler frequency shift is smaller, and the time range degree of channel is not obvious, and therefore the blocking equalizing algorithm advantage that proposes is not obvious.

As shown in Figure 5, the portable terminal translational speed is 120km/h, and this moment, corresponding normalization Doppler frequency shift was about 0.05.As can be seen from the figure, because translational speed increases, the time range degree of channel is apparent in view, and therefore traditional single order equalizer performance reduces, and has occurred the error code flat bed after signal to noise ratio is higher than 20dB.The equalization performance of the blocking equalizing algorithm that the present invention proposes obviously is better than traditional single order equalizer performance, and is 10 in bit error rate ^-2The time, compare the gain that 5dB is also arranged with the piecemeal frequency-domain equilibrium method of LiWei.

As shown in Figure 6, the portable terminal translational speed is 250km/h, and this moment, corresponding normalization Doppler frequency shift was about 0.1.As can be seen from the figure, because translational speed further increases, the time range degree aggravation of channel, therefore traditional single order equalizer performance further reduces, and the error code flat bed just occurred after signal to noise ratio is higher than 15dB.This moment, the piecemeal frequency-domain equilibrium method equalization performance of Li Wei also decreased, and the equalization performance of the blocking equalizing algorithm that the present invention proposes still can bit error rate reach 10 when signal to noise ratio is 21dB ^-2, obviously be better than the piecemeal frequency-domain equilibrium method of Li Wei.

Claims (1)

1, blocking equalizing method in the ofdm system under a kind of Quick-Change channel condition is characterized in that:

At transmitting terminal, through the bit signal behind cyclic redundancy code stream b (n) channel encode, interweave, the formation frequency domain is handed over positive frequency-division multiplex singal behind sign map and the serial to parallel conversion, i frequency domain orthogonal frequency division multiplex ransmitting feed signals is expressed as s (i)=[s (i, 0), s (i, 1),, s (i, N-1)] ^T, N is a sub-carrier number, s (i) becomes time domain orthogonal frequency division multiplex ransmitting feed signals through contrary fast fourier transform

After, inserting length is the Cyclic Prefix of G, and sends to wireless channel behind parallel serial conversion;

At receiving terminal, the first step: after removing Cyclic Prefix, serial to parallel conversion, i time domain orthogonal frequency division multiplexing receiving symbol $\tilde{r}\left(i\right)={[\tilde{r}(i,0),\tilde{r}(i,1),\·\·\·,\tilde{r}(i,N-4)]}^T$ Can be expressed as:

$\tilde{r}\left(i\right)=H_t^i\tilde{s}\left(i\right)+\tilde{w}\left(i\right)=H_t^i{F}^{H}s\left(i\right)+\tilde{w}\left(i\right)$

Here H _t ⁱBe channel matrix,

Expression power is σ ²The white Gaussian noise vector, F represents fourier transform matrix, () ^HThe operation of expression conjugate transpose, in order to carry out the equilibrium of single-order frequency domain equalization device, it is static earlier the channel hypothesis to be as the criterion, and promptly the variation of channel can be ignored in 1 OFDM symbol transmission time, multipath channel is modeled as the FIR filter of constant finite impulse response when having the L-1 rank, uses h _l(i) expression l rank impulse response.

This moment H _t ⁱBe a convolution matrix, can be expressed as

$H_t^i=F^H{\mathrm{\Λ}}^{i}F$

Here Λ ⁱBe diagonal matrix,

Become frequency domain OFDM receiving symbol through fast fourier transform $r\left(i\right)=F\tilde{r}\left(i\right),$ Utilize the balanced frequency domain OFDM receiving symbol r (i) of single-order frequency domain equalization device, the frequency domain orthogonal frequency-division multiplex singal estimated value of equalizer output Can be expressed as:

$\hat{s}\left(i\right)={\mathrm{\Λ}}^{\mathrm{iH}}{({\mathrm{\Λ}}^i{\mathrm{\Λ}}^{\mathrm{iH}}+{\mathrm{\σ}}^{2}I)}^{-1}r\left(i\right)$

Through parallel serial conversion, separate mapping, interweave, decoding obtains the estimated value of transmitting terminal bit signal stream

Second step: to the transmitting terminal bit signal estimated value that obtains

Carry out cyclic redundancy check (CRC), if cyclic redundancy check (CRC) is correct, then output; If cyclic redundancy check (CRC) is incorrect, then carry out blocking equalizing, with the output of blocking equalizing input as parallel serial conversion in the first step,

Above-mentioned blocking equalizing method can be expressed as follows:

A. with the estimated value of the transmitting terminal bit signal that obtains in first step stream

, carry out the chnnel coding identical, interweave, after the sign map, serial to parallel conversion, contrary fast fourier transform, obtain time domain orthogonal frequency division multiplexing reconstruction signal with transmitting terminal ${\tilde{s}}^{\′}\left(i\right)={[{\tilde{s}}^{\′}(i,0),{\tilde{s}}^{\′}(i,1),\·\·\·,{\tilde{s}}^{\′}(i,N-1)]}^T,$

B. the time domain orthogonal frequency division multiplexing symbol that each is received

Be divided into M the isometric sub-piece of received signal, sub-piece is counted M and is satisfied:

And be 2 integral number power, here

Expression rounds operation downwards, and the length of each sub-piece is T=N/M, T＞L, and k sub-piece can be expressed as ${\tilde{r}}_k\left(i\right)={[{\tilde{r}}_k(i,0),{\tilde{r}}_k(i,2),\·\·\·,{\tilde{r}}_k(i,T-1)]}^T={[\tilde{r}(i,\mathrm{kT}),\tilde{r}(i,\mathrm{kT}+1),\·\·\·,\tilde{r}(i,(k+1)T-1)]}^T,$ K ∈ [0 ..., M-1], in the time of each sub-piece transmission, think that the time change of wireless channel can be ignored, promptly $h^i(n,l)\≡h_k^i\left(l\right),$ N=kT ..., (k+1) T-1, l ∈ [0 ... L-1] be the channel l rank impulse response of the time of k sub-piece transmission, at this moment

Can be expressed as

${\tilde{r}}_k\left(i\right)=H_k^{\′}{\tilde{s}}_{k-1}\left(i\right)+H_k{\tilde{s}}_k\left(i\right)+{\tilde{w}}_k\left(i\right)$

Here ${\tilde{s}}_k\left(i\right)={[{\tilde{s}}_k(i,0),{\tilde{s}}_k(i,1),\·\·\·,{\tilde{s}}_k(i,T-1)]}^T={[\tilde{s}(i,\mathrm{kT}),\tilde{s}(i,\mathrm{kT}+1),\·\·\·,\tilde{s}(i,(k+1)T-1)]}^T,$

${\tilde{w}}_k\left(i\right)={[{\tilde{w}}_k(i,1),{\tilde{w}}_k(i,2),\·\·\·,{\tilde{w}}_k(i,T)]}^T,$

Having characterized k-1 signal subspace piece crosstalks to the sub-interblock that current k sub-piece causes.

The time domain of reconstruct is sent signal

Being divided into length is the sub-piece of T

${\tilde{s}}_k^{\′}\left(i\right)={[{\tilde{s}}_k^{\′}(i,0),{\tilde{s}}_k^{\′}(i,1),\·\·\·,{\tilde{s}}_k^{\′}(i,T-1)]}^T={[{\tilde{s}}^{\′}(i,\mathrm{kT}),{\tilde{s}}^{\′}(i,\mathrm{kT}+1),\·\·\·,{\tilde{s}}^{\′}(i,(k+1)T-1)]}^T,$ Utilize

Eliminate k-1 signal subspace piece the interblock that current k sub-piece causes crosstalked,

${\tilde{r}}_k^{\′}\left(i\right)={\tilde{r}}_k\left(i\right)-H_k^{\′}{\tilde{s}}_{k-1}^{\′}\left(i\right)=H_k^{\′}({\tilde{s}}_{k-1}\left(i\right)-{\tilde{s}}_{k-1}^{\′}\left(i\right))+H_k{\tilde{s}}_k\left(i\right)+{\tilde{w}}_k\left(i\right)$

And carry out loop restructuring, and guaranteed the cycle characteristics of channel matrix, disturb between suppressing to carry,

${\tilde{r}}_k^{\′\′}\left(i\right)={\tilde{r}}_k^{\′}\left(i\right)+H_k^{\′}{\tilde{s}}_k^{\′}\left(i\right)=H_k^{\′}({\tilde{s}}_{k-1}\left(i\right)-{\tilde{s}}_{k-1}^{\′}\left(i\right))+H_k{\tilde{s}}_k\left(i\right)+H_k^{\′}{\tilde{s}}_k^{\′}\left(i\right)+{\tilde{w}}_k\left(i\right)$

If do not consider the error that reconstruct causes, promptly ${\tilde{s}}_k\left(i\right)\≡{\tilde{s}}_k^{\′}\left(i\right),\∀k,$ Then

${\tilde{r}}_k^{\′\′}\left(i\right)=H_{k,\mathrm{cir}}^{\′}{\tilde{s}}_k\left(i\right)+{\tilde{w}}_k\left(i\right)$

Convolution matrix H ' _{K, cir}=H _k+ H ' _kBe the channel matrix of k sub-piece correspondence,

C. eliminate that sub-interblock is crosstalked and reconstruction cycle prefix basis on each is received sub-piece, carry out fast fourier transform earlier, the length of its fast fourier transform is consistent with sub-block length, each receives sub-piece to utilize the equilibrium of single-order frequency domain equalization device again, the balanced valuation of the sub-piece of each after the equilibrium is through contrary fast fourier transform, obtain the time domain equalization estimated value of each sub-piece, the time domain equalization estimated value of k sub-piece can be expressed as

${\hat{\tilde{s}}}_k\left(i\right)=F^H({\mathrm{\Λ}}_k^i{\mathrm{\Λ}}_k^{\mathrm{iH}}+{\mathrm{\σ}}_w^2{I}_T){\mathrm{\Λ}}_k^{\mathrm{iH}}F{\tilde{r}}_k^{\′\′}\left(i\right)$

Here diagonal matrix ${\mathrm{\Λ}}_k^i=F{H}_{k,\mathrm{cir}}^{\′}{F}^H,$

D. last balanced estimated value to each sub-piece merges, and the time domain equalization estimated value of each sub-piece is the N symbol according to the synthetic length of the der group of piecemeal $\hat{\tilde{s}}\left(i\right)={[{\hat{\tilde{s}}}_0^T\left(i\right),{\hat{\tilde{s}}}_1^T\left(i\right),\·\·\·,{\hat{\tilde{s}}}_{M-1}^T\left(i\right)]}^T,$

After fast fourier transform, obtain the output of piecemeal frequency-domain equilibrium method

, promptly be the estimated value of the frequency domain orthogonal frequency-division multiplex singal of transmission.

Images