CN101192846B - A self-adapted method for setting protection interval of overlapping cutting method - Google Patents

Abstract

The invention discloses a method for protecting interval by adopting an adaptive set overlapping shearing approach. A transmitting end sends given data of a receiving end. The receiving end receives the data and calculates the data error to determine the location and length of frequency domain equalization protection interval. The invention further discloses a method for protecting interval by adopting an adaptive set overlapping shearing approach. If the SNR is smaller than the preset SNR threshold value, the protection interval is set as 0; otherwise, the transmitting end sends given data of the receiving end and the receiving end receives the data to calculate the data error and determine frequency domain equalization protection interval. The invention can effectively reduce bit error rate and further reduce the operation complexity of the frequency domain equalizer while the system performance is not reduced.

Description

A kind of self adaptation is provided with the protection method at interval of overlapping shearing method

Technical field

The present invention relates to moving communicating field, relate in particular to the protection method at interval that self adaptation in a kind of frequency-domain equalization technology is provided with overlapping shearing method.

Background technology

The RAKE reception technique is the most frequently used reception technique of cdma system, and it detects all distinguishable rakes, multipath component is carried out high specific merge.Under the situation that the orthogonal property of spreading code can be maintained, the performance of RAKE receiver will be optimum.But because the multipath transmisstion of wireless channel has destroyed the orthogonal property of spreading code, the performance of RAKE receiver has been subjected to multiple access interference (Multiple Access Interference, MAI) influence inserts in the more sub-district of user at one, and the performance of RAKE receiver will significantly reduce.

Relevant improved technology comprises multiuser detection, and this technology is united the detection performance that the information of using a plurality of users removes to improve unique user by multi-user detector.Though this can eliminate MAI, for the travelling carriage in descending reception, the multiuser detection complexity is too high.

Therefore, people have proposed the detection technique of the many CDMA of being suitable for down links, comprising frequency-domain equalization technology.Frequency-domain equalization technology is proposed in 1973 by Walzman and Schwartz the earliest.It also is well-known conclusion that balanced reception technique can be obtained than RAKE reception technique more performance in cdma system.Theoretical foundation in CDMA downlink system applying frequency domain balancing technique is based on such fact, all users' signal has passed through identical multipath channel simultaneously and has arrived travelling carriage from the base station, therefore applying frequency domain balancing technique (Frequency Domain Equalization) in the down link of CDMA, can eliminate the influence of multipath, keep the orthogonality between the spreading code, finally reach the purpose of eliminating MAI and intersymbol interference (ISI, Inter Symbol Interference) simultaneously.The frequency domain equalization algorithm can carry out in frequency domain based on fast Fourier transform (FFT), and computation complexity is far smaller than multi-user detector.

In order to finish frequency domain equalization to user's chip based on fast Fourier transform/quick inversefouriertransform (FFT/IFFT) module, the effect that at first should satisfy between the signal that is transmitted in channel and the channel is this hypothesis of circular convolution, in order to realize this condition, three kinds of processing methods are arranged: 1, Cyclic Prefix method (Cyclic Prefix Method, CP Method), this method is similar to OFDM (OFDM, orthogonal frequency division multiplexing) in the system to OFDM time domain sample Signal Processing mode, add CP in each segment data front, thereby realize the conversion of channel matrix to the circular matrix form.2, zero padding method (Zero-Padding Method, ZP Method) in each segment data back zero padding, is finished the conversion of channel matrix to the circular matrix form.3, overlapping shearing method (Overlap-Cut Method; OC Method); Frame is adopted the mode of sliding window piecemeal processing; the beginning and end of the chip to be detected of sliding window overlay area part as protection at interval each time; described protection is the overlapping part of data in the sliding window at interval, abandons before spread spectrum and symbol judgement.CP method and ZP method will be adjusted transmitting terminal, and the OC method is just passable as long as increase corresponding processing unit at receiving terminal, has embodied existing system better compatible.In order to reduce the time delay of frequency domain equalization algorithm, inevitable requirement reduces the length of FFT/IFFT, and the effective data rate of CP method and ZP method is reduced; And the OC method can not be subjected to the influence of this factor, and just amount of calculation increases in the frequency domain equalization module.

CP method and ZP method can realize the transformation of the loop structure of channel matrix fully, so bit error rate performance is better; And the OC method is the computational methods of a kind of approximate realization linear convolution to circular convolution, because this approximation has caused bit error rate performance poor slightly; And protection is set at interval is exactly in order to overcome the poor slightly problem of bit error rate performance, thereby reduces the error rate; But in the prior art; protection length at interval generally is the artificial empirical value of setting, and can not be provided with according to the situation that systematic error distributes, so in actual applications; can not reduce the error rate well, also can under situation about there is no need fully, increase the computation complexity of system sometimes.

Summary of the invention

The technical problem to be solved in the present invention just provides the protection method at interval that a kind of self adaptation is provided with overlapping shearing method; to improve existing OC method protection position and length at interval can not be set according to the situation that systematic error distributes, can not reduce the problem of the error rate well.

In order to solve the problems of the technologies described above; the invention provides the protection method at interval that a kind of self adaptation is provided with overlapping shearing method, transmitting terminal transmitting and receiving terminal given data, receiving terminal receives this data; the calculated data error is determined the protection position and the length at interval of frequency domain equalization.

Further, described method comprises the steps:

(1) transmitting terminal is that the base station sends the continuous pilot sequence of length greater than channel delay, and this pilot frequency sequence is the receiving terminal given data;

(2) receiving terminal is after frequency-domain equalizer is received this pilot frequency sequence, calculates the systematic error sequence e of frequency domain equalization;

(3) from the two ends of e to the third side to, segmentation detects respectively, if a certain section error greater than assign thresholds, just needs protection at interval on the then corresponding section, and continues to detect its inboard adjacent segment; Otherwise stop to detect; Each needs the protection interval of much length before and after determining respectively.

Further,

Described base station transmission continuous pilot sequence i.e. all users' piecemeal is closed signal x=[x ₀, x ₁..., x _N-1], wherein N is the length of piecemeal; It is y=[y that frequency-domain equalizer receives corresponding sequence ₀, y ₁..., y _N-1], wherein, the impulse response of the channel of data process is h=[h ₀, h ₁..., h _L-1], L is the length of discrete multipath channel, i.e. channel delay; The circular convolution of x and h is r=[r ₀, r ₁..., r _N-1];

The systematic error sequence e that calculates frequency domain equalization in the described step (2) is:

Calculate e=IFFT[FFT (Δ, N) ./FFT (h, N)], Δ=[Δ wherein ₀, Δ ₁... Δ _L-2], be the difference of preceding L-1 the data of y and r, and FFT (, N) fast Fourier transform that length is N, [] are carried out in representative. and the sequence of the identical size of/[] representative is carried out correspondence and is divided by.

Further, described step (3) comprising:

Get the amplitude equalizing value λ of middle L the value of e, from the two ends of e to the third side to, detect respectively, the error amount sequence that respectively to get one section L length before and after each time be L, ask the amplitude equalizing value of front and rear sections respectively, if the amplitude equalizing value of which section is greater than assign thresholds ρ λ, ρ is set to integer as required, just needs protection at interval on the then corresponding section; Otherwise just do not need; If one section of the outside needs protection, just then the adjacent segment of inboard is carried out the detection of same procedure, to determine whether to need to strengthen protection length; Otherwise just stop detection to its inboard adjacent segment; Each needs the protection interval of much length before and after determining so respectively.

Further, the maximum protection length of front and back all is 4L.

Further, before transmitting terminal transmitting and receiving terminal given data, whether judge signal to noise ratio less than the snr threshold of setting, if less than, it is 0 that protection length at interval then is set; Otherwise, transmitting terminal transmitting and receiving terminal given data.

Further, described snr threshold is 10dB.

Further, described method comprises the steps:

(1) transmitting terminal is that the base station sends the continuous pilot sequence of length greater than channel delay, and this pilot frequency sequence is the receiving terminal given data;

(2) receiving terminal is after frequency-domain equalizer is received this pilot frequency sequence, calculates the systematic error sequence e of frequency domain equalization;

(3) from the two ends of e to the third side to, segmentation detects respectively, if a certain section error greater than assign thresholds, just needs protection at interval on the then corresponding section, and continues to detect its inboard adjacent segment; Otherwise stop to detect; Each needs the protection interval of much length before and after determining respectively.

Further,

Described base station transmission continuous pilot sequence i.e. all users' piecemeal is closed signal x=[x ₀, x ₁..., x _N-1], wherein N is the length of piecemeal; It is y=[y that frequency-domain equalizer receives corresponding sequence ₀, y ₁..., y _N-1], wherein, the impulse response of the channel of data process is h=[h ₀, h ₁..., h _L-1], L is the length of discrete multipath channel, i.e. channel delay; The circular convolution of x and h is r=[r ₀, r ₁..., r _N-1];

The systematic error sequence e that calculates frequency domain equalization in the described step (2) is:

Calculate e=IFFT[FFT (Δ, N) ./FFT (h, N)], Δ=[Δ wherein ₀, Δ ₁... Δ _L-2], be the difference of preceding L-1 the data of y and r, and FFT (, N) fast Fourier transform that length is N, [] are carried out in representative. and the sequence of the identical size of/[] representative is carried out correspondence and is divided by.

Further, described step (3) comprising:

Get the amplitude equalizing value λ of middle L the value of e, from the two ends of e to the third side to, detect respectively, the error amount sequence that respectively to get one section L length before and after each time be L, ask the amplitude equalizing value of front and rear sections respectively, if the amplitude equalizing value of which section is greater than assign thresholds ρ λ, ρ is set to integer as required, just needs protection at interval on the then corresponding section; Otherwise just do not need; If one section of the outside needs protection, just then the adjacent segment of inboard is carried out the detection of same procedure, to determine whether to need to strengthen protection length; Otherwise just stop detection to its inboard adjacent segment; Each needs the protection interval of much length before and after determining so respectively.

Further, the maximum protection length of front and back all is 4L.

The present invention proposes and effectively to reduce the error rate, and under the prerequisite that does not reduce systematic function, further to have reduced the computational complexity of frequency-domain equalizer according to the distribution self adaptation adjustment protection AOC method at interval of systematic error.

The complexity that the AOC method is calculated is low, can effectively suppress multi-user interference in the multi-user CDMA downlink system or the interference between the code channel, do not need to change the air interface standard of existing system, suitably might obtain better detection performance, have a good application prospect with combining of other technologies such as multi-user interference deleting technique.

Description of drawings

Fig. 1 is the amplitude distribution schematic diagram of modified model walking A channel errors sequence;

Fig. 2 is the amplitude distribution schematic diagram of vehicle-mounted A channel errors sequence;

Fig. 3 is the selecting sequence schematic diagram that detects the sequence protection zone;

Fig. 4 is 8 users, 16 rank Walsh spread spectrums, the bit error rate performance of various receivers;

Fig. 5 is 16 users, 16 rank Walsh spread spectrums, the bit error rate performance of various receivers.

Embodiment

By analyzing the systematic error characteristic distributions of OC method, method of the present invention is applicable to the frequency-domain equalizer at the CDMA down channel, the performance of frequency-domain equalizer is improved and keeps low complex degree.

The frequency domain equalization of cdma system is the channel equalization of chip-level.Discrete multipath channel can characterize with a tapped delay line model.The impulse response of channel can be expressed as: h=[h ₀, h ₁..., h _L-2, h _L-1] ^T, the length of the discrete multipath channel of L representative, the transposition of T representation vector.In multi-user's synchronous CDMA system, all users' that send the base station piecemeal closes signal and can be expressed as: x=[x ₀, x ₁..., x _N-2, x _N-1] ^T, N is the length of piecemeal, i.e. sample sequence length.The signal that receives of travelling carriage can be expressed as so:

y＝Hx+n (1)

Just

Wherein H is the channel matrix of the channel impulse response formation of (N+L-1) * N dimension, and n is a white Gaussian noise.

If can allow channel H that user's data passes through as in OFDM by certain mode, equivalence becomes circular matrix

Form, at this moment the time domain expression formula of the discrete vectorial y that receives in the system is:

$y=\tilde{H}x+n---\left(3\right)$

Application class is like the method for OFDM frequency domain equalization, and the output result that can obtain frequency-domain equalizer is:

Be the estimated value that sends signal frequency domain, two identical vectorial corresponding elements of length of [] ⊙ [] expression multiply each other C=[C ₀, C ₁..., C _N-1] be the equalizing coefficient of frequency-domain equalizer.C can (Zero-Forcing, ZF) (Minimum Mean Square Error, MMSE) criterion obtains for criterion or least mean-square error according to ZF.Y=[Y ₀, Y ₁..., Y _N-1] be the Fourier transform of received signal.By the corresponding relation of circular convolution and FFT, the frequency domain sample of receiving sequence multiply by equalizing coefficient has respectively just finished frequency domain equalization calculating, and time domain is returned in the sequence transformation after using IFFT with equilibrium at last, carries out despreading and judgement.

The systematic error characteristic distributions of following surface analysis OC method frequency domain equalization proposes adaptive overlapping cutting method according to these error profile characteristics.

For two sequence x=[x ₀, x ₁..., x _N-1] and h=[h ₀, h ₁..., h _L-1], N＞L wherein, as mentioned above, on behalf of all users' that the base station sends piecemeal, the former close signal, and the latter represents the impulse response of discrete multipath channel.

Use the frequency domain equalization algorithm that the reception data are carried out deconvolution and suppose that exactly x is to have carried out the circular convolution process with h through channel.First value that can obtain circular convolution is:

${\mathrm{<figure-callout\; id="0"\; label="r"\; filenames="S071C9755120070801E000011.png,S071C9755120070801E000022.png"\; state="\{\{state\}\}">r</figure-callout>}}_0=x_0{\mathrm{<figure-callout\; id="0"\; label="h"\; filenames="S071C9755120070801E000011.png,S071C9755120070801E000022.png"\; state="\{\{state\}\}">h</figure-callout>}}_0+\underset{k=1}{\overset{L-1}{\mathrm{\&Sigma;}}}{x}_{N-k}{h}_k---\left(5\right)$

At the h ring shift right after the L-2 position, obtain

$r_{L-2}=\underset{k=0}{\overset{L-2}{\mathrm{\&Sigma;}}}{x}_k{h}_{L-k-2}+x_{N-1}{h}_{L-1}---\left(6\right)$

At the h ring shift right after the L-1 position, obtain

$r_{L-1}=\underset{k=0}{\overset{L-1}{\mathrm{\&Sigma;}}}{x}_k{h}_{L-k-1}---\left(7\right)$

Can obtain circular convolution thus from r ₀To r _L-2L-1 value, and preceding L-1 value of the linear convolution between x and the h is inequality.And the r in the circular convolution _L-1To r _N-1To be worth N value correspondent equal with L of linear convolution.

Can learn that according to top analysis in frequency domain equalization is handled, the top n result of data and channel linearity convolution is handled as the result of circular convolution, front L-1 individual (L is the length of channel) will cause the error of OC method.Further consider and receive data have been carried out segment processing by sliding window effect: because channel and data have been carried out the effect of linear convolution, so preceding L-1 the data of each section can be subjected to the pollution of the multidiameter of earlier data, this also can cause the error of OC method.

The front has obtained the top L-1 of each a section data and will not meet that channel and data are prerequisites of the hypothesis of circular convolution effect in the frequency domain equalization.Below further analyze with data and channel linearity convolution N and sample and treat as circular convolution and carry out frequency domain equalization, to the influence that precision caused of frequency-domain equalizer.

The chip sequence of supposing frequency-domain equalizer processing each time is y=[y ₀, y ₁..., y _N-1], it is that all users' of being sent by the base station piecemeal closes signal x=[x ₀, x ₁..., x _N-1], through h=[h ₀, h ₁..., h _L-1] the top n chip-level sampling (do not consider Gaussian noise in the discussion, can lead to the same conclusion when considering Gaussian noise) that obtains.If sending the circular convolution of sequence x and h is r=[r ₀, r ₁..., r _N-1], preceding L-1 the data of y and r are inequality, difference is Δ=[Δ ₀, Δ ₁... Δ _L-2].The ZF criterion of use frequency domain equalization is come the characteristics of analytical system error, and MMSE can obtain similar conclusion.Frequency domain equalization computational process based on the ZF criterion is:

${\hat{x}}^T=\mathrm{IFFT}[\mathrm{FFT}\left(y\right)./\mathrm{FFT}(h,N)]$

$=F^{-1}\left\{\right[F{({\mathrm{<figure-callout\; id="0"\; label="r"\; filenames="S071C9755120070801E000011.png,S071C9755120070801E000022.png"\; state="\{\{state\}\}">r</figure-callout>}}_0+{\mathrm{\&Delta;}}_0,r_1+{\mathrm{\&Delta;}}_1,...,r_{L-2}+{\mathrm{\&Delta;}}_{L-2},r_{L-1},...,r_{N-1})}^T]./[F{({\mathrm{<figure-callout\; id="0"\; label="h"\; filenames="S071C9755120070801E000011.png,S071C9755120070801E000022.png"\; state="\{\{state\}\}">h</figure-callout>}}_0,h_1,...,h_{L-1},0,...0)}^T\left]\right\}$

$=F^{-1}\left\{\right[F\left(r^T\right)]./[F{({\mathrm{<figure-callout\; id="0"\; label="h"\; filenames="S071C9755120070801E000011.png,S071C9755120070801E000022.png"\; state="\{\{state\}\}">h</figure-callout>}}_0,h_1,...,h_{L-1},0,...0)}^T\left]\right\}+F^{-1}\left\{\right[F{({\mathrm{\&Delta;}}_0,...,{\mathrm{\&Delta;}}_{L-1},0,...0)}^T]./[F{({\mathrm{<figure-callout\; id="0"\; label="h"\; filenames="S071C9755120070801E000011.png,S071C9755120070801E000022.png"\; state="\{\{state\}\}">h</figure-callout>}}_0,...,h_{L-1},0,...0)}^T\left]\right\}$

$=x^T+F^{-1}\left\{\right[F{({\mathrm{\&Delta;}}_0,{\mathrm{\&Delta;}}_1,...,{\mathrm{\&Delta;}}_{L-1},0,...0)}^T]./[F{({\mathrm{<figure-callout\; id="0"\; label="h"\; filenames="S071C9755120070801E000011.png,S071C9755120070801E000022.png"\; state="\{\{state\}\}">h</figure-callout>}}_0,h_1,...,h_{L-1},0,...0)}^T\left]\right\}$

$=x^T+\mathrm{IFFT}[\mathrm{FFT}(\mathrm{\&Delta;},N)./\mathrm{FFT}(h,N)]---\left(8\right)$

Wherein F and F ^-1Be respectively Fourier transform matrix and Fourier inversion matrix, wherein [] ^-1Inverse of a matrix, [] are got in expression ^TThe expression transposition.FFT (, N) fast Fourier transform that length is N, [] are carried out in representative. and the sequence of the identical size of/[] representative is carried out the correspondence position and is divided by.From formula (8) as can be seen, the systematic error of OC method comes from the latter half of (8) formula, below the error part is further analyzed, and makes error partly be e=[e ₀, e ₁..., e _N-1] ^T,

$e=\mathrm{IFFT}[\mathrm{FFT}(\mathrm{\&Delta;},N)./\mathrm{FFT}(h,N)]$

$=(1/N)\&CenterDot;F^H\&CenterDot;{[\underset{i=0}{\overset{L-2}{\mathrm{\&Sigma;}}}{\mathrm{\&Delta;}}_i/\underset{i=0}{\overset{L-1}{\mathrm{\&Sigma;}}}{h}_i,\underset{i=0}{\overset{L-2}{\mathrm{\&Sigma;}}}{\mathrm{\&Phi;}}^{1\&CenterDot;i}{\mathrm{\&Delta;}}_i/\underset{i=0}{\overset{L-1}{\mathrm{\&Sigma;}}}{\mathrm{\&Phi;}}^{1\&CenterDot;i}{h}_i,...,\underset{i=0}{\overset{L-2}{\mathrm{\&Sigma;}}}{\mathrm{\&Phi;}}^{(N-1)\&CenterDot;i}{\mathrm{\&Delta;}}_i/\underset{i=0}{\overset{L-1}{\mathrm{\&Sigma;}}}{\mathrm{\&Phi;}}^{(N-1)\&CenterDot;i}{h}_i]}^T$

$=(1/N)\&CenterDot;{[{\mathrm{\&alpha;}}_0,{\mathrm{\&alpha;}}_1,...,{\mathrm{\&alpha;}}_{N-1}]}^T---\left(9\right)$

Wherein ${\mathrm{\&alpha;}}_m=\underset{k=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{\mathrm{\&Phi;}}^{-k\&CenterDot;m}{[\underset{i=0}{\overset{L-2}{\mathrm{\&Sigma;}}}{\mathrm{\&Phi;}}^{k\&CenterDot;i}{\mathrm{\&Delta;}}_i/\underset{i=0}{\overset{L-1}{\mathrm{\&Sigma;}}}{\mathrm{\&Phi;}}^{k\&CenterDot;i}{h}_i]}^T,m=\mathrm{0,1},...,N-1.---\left(10\right)$

Wherein, the derivation that formula (8), (9), (10) are to use the ZF criterion of frequency domain equalization to carry out, the process of deriving based on the MMSE criterion will be identical with formula (8), (9), (10) of using the ZF criterion to derive under situation about repeatedly simplifying.

This shows that systematic error e will be to detecting sequence

Each chip exert an influence.Make a=[α ₀, α ₁..., α _N-1], apparent, sequence a and error sequence e differ a constant times, below we analyze a sequence, establish

$f_k={[\underset{i=0}{\overset{L-2}{\mathrm{\&Sigma;}}}{\mathrm{\&Phi;}}^{k\&CenterDot;i}{\mathrm{\&Delta;}}_i/\underset{i=0}{\overset{L-1}{\mathrm{\&Sigma;}}}{\mathrm{\&Phi;}}^{k\&CenterDot;i}{h}_i]}^T,k=\mathrm{0,1},...N-1---\left(11\right)$

By formula (10), each α _m(m=0,1 ..., N-1) all be the stack of N item, and wherein each all is element f _k(k=0,1 ... N-1) the phase place deflection Ф corresponding with one ^-kmProduct.For each α _mCalculating, f _kSequence all is identical.We estimate each α _mIn N member f _kWhat phase change remain in the pi/2 continuously.Because the molecule and the denominator of formula (11) have similar forms, so as long as the variation of estimation molecular moiety phase place is just enough.With reference to formula (11), make f _kMolecular moiety

${\mathrm{\&gamma;}}_k=\underset{i=0}{\overset{L-2}{\mathrm{\&Sigma;}}}{\mathrm{\&Phi;}}^{k\&CenterDot;i}{\mathrm{\&Delta;}}_i,k=\mathrm{0,1},...N-1---\left(12\right)$

Each γ in the following formula _kAll be L-1 member's Δ _iMultiply by respective phase deflection Ф respectively ^KiAdd and form.With these Δs _iRegard γ as _kThe initial baseline phase place, suppose from f ₀Continuous z f of beginning _kKeep phase change in pi/2, so by relational expression:

${\mathrm{\&gamma;}}_{z-1}=\underset{i=0}{\overset{L-2}{\mathrm{\&Sigma;}}}{\mathrm{\&Phi;}}^{i\&CenterDot;(z-1)}{\mathrm{\&Delta;}}_i,k=\mathrm{0,1},...N-1---\left(13\right)$

Zui Da phase place deflection is Ф as can be seen ^{(L-2) (z-1)}Cause, so can obtain relational expression:

$\frac{2\mathrm{\&pi;}}{N}(L-2)(z-1)\&le;\frac{\mathrm{\&pi;}}{2}---\left(14\right)$

So obtain:

$z\&le;\frac{N}{4(L-2)}+1---\left(15\right)$

Can see that by formula (15) length that the size of z and the length N of FFT are approximated to direct ratio and channel is approximated to inverse ratio.

Fig. 1 and Fig. 2 are respectively the M.1225 distribution schematic diagrams of the OC method error e of modified model walking A channel and vehicle-mounted A channel of ITU.Wherein abscissa is represented the sequence number of sample sequence, and ordinate is represented the size of error.

Here the size of setting the FFT/IFFT module is 512, produces the transmission data of 512 length at random, utilizes the linear convolution of data and channel and the difference of circular convolution to calculate Δ=[Δ ₁, Δ ₂... Δ _L-1], utilize formula (9) to calculate the distribution of error.

Under the short situation of channel delay (for example modified model walking A channel), according to formula (15), the z value is bigger.Again according to formula (10), when calculating a, when m approaches 0 or during N-1, adjacent f _kPhase weighting coefficient Ф ^-kmPhase difference so just causes continuous z Ф that phase place is approaching in the formula (10) with less ^-kmF _kAddition.N with a is divided into

Section, here

Expression rounds.Ф in each section ^-kmF _kPhase place is close, so add with the mould value greatlyyer in the section, if adding and can not offset between each section, then the concentration of energy of error sequence e is at two ends; Otherwise the two ends error may be less.But whether error sequence e two ends peak value occurs, concrete condition also with the sequence Δ=[Δ of stochastic behaviour ₁, Δ ₂... Δ _L-1] and h=[h ₀, h ₁..., h _L-1] relevant.In the error sequence when m to the N/2 direction near the time, adjacent Ф ^-kmPhase difference increases gradually, when m=N/2, and adjacent Ф ^-kmPhase difference be π; Because have a continuous z f _kPhase difference is in pi/2, so each in the formula (10) adds and offset a front and back, so the mid portion amplitude of error sequence e is less.As depicted in figs. 1 and 2, the error of rear and front end is more likely greater than the error of mid portion.Vehicle-mounted A channel delay is bigger, and according to formula (15), at this moment z is less, has caused the bigger part of error to expand to the centre position.

Length is reasonably protected in estimation below.Calculate in the process of a in formula (10), if continuous z f _kThe phase difference that the front added equals 2 π, unites the result of consideration formula (15) again, all Negative function in individual section strengthens, and this time error diminishes.Among the hypothetical sequence a, begin to become less, can get according to top analysis when m increases to the K time error:

$\frac{2\mathrm{\&pi;}}{N}K[\frac{N}{4(L-2)}+1]=2\mathrm{\&pi;}---\left(16\right)$

So

K≈4(L-2) (17)

The systematic error of OC method is in that to detect the sequence two ends bigger as can be known by the analysis of front, but bigger error does not always appear at two ends simultaneously.Along with the increasing of channel delay, the distribution fluctuates at error sequence two ends also strengthens.Can obtain with reference to formula (10) and formula (11), two sequences of Δ and h also will determine the distribution of estimated sequence front and back error largely.

We propose the improvement of OC method and are referred to as AOC method (Adaptive Overlap-Cut Method) thus: at first utilize pilot tone that the error profile feature of system is estimated, adaptive then protection position and the length at interval that is provided with.This improves algorithm and utilizes the pilot tone of cdma system, and need not change the frame structure of existing system as the CP method largely.

The concrete steps of AOC method are as follows:

(1) at first, iff the down channel receiving algorithm that utilizes OC method frequency domain equalization as cdma system, (such as snr threshold is 10dB under the snr threshold situation of signal to noise ratio less than setting,), can not use protection at interval, can not cause adverse influence in this case, and greatly reduce the complexity of frequency-domain equalizer the performance of system; At this moment, determined protection length is 0, directly enters step (4);

Because different environment channel impulse response differences, and might be changing, according to simulation result, it is proper to think that signal to noise ratio is not provided with protection length below 10dB; This value should guarantee have with two kinds of situations of unguarded interval under the error rate of system performance do not have under the condition of big difference too, obtaining of trying one's best is bigger, because protection frequency domain equalization at interval can reduce the operand of frequency domain equalization; This value also can be provided with according to actual conditions;

(2) under the higher situation of signal to noise ratio (such as greater than 10dB), the length that sends according to the base station is greater than the continuous pilot sequence of channel delay, based on the systematic error sequence e of formula (8) and (9) calculating frequency domain equalization;

(3) get the amplitude equalizing value λ of the individual value of middle L (L is the length of channel) of e, from the two ends of e to the third side to, detect respectively, respectively get the error amount of one section L length each time, ask the amplitude equalizing value of front and rear sections respectively, if the amplitude equalizing value of which section greater than assign thresholds, just needs protection at interval on the then corresponding section; Otherwise just do not need; If one section of the outside needs protection, just then the adjacent segment of inboard is carried out the detection of same procedure, to determine whether to need to strengthen protection length; Otherwise just stop detection to its inboard adjacent segment; Each needs the protection interval of much length before and after determining so respectively;

As shown in Figure 3, from the two ends of e to the third side to, detect respectively.Get the error amount of L length each time, calculate the amplitude average η of first section of front (section 1 in the corresponding diagram 3) at first respectively ₁And the amplitude equalizing value η of back first section (section 1 in the corresponding diagram 3 ') _{1 '}If, η ₁Or η _{1 '}In any one (ρ is the multiple of threshold value about λ, can directly be arranged to integer as required greater than threshold value ρ λ; Can see according to Fig. 1 and Fig. 2, to compare the amplitude of the error of mid portion be very large to the amplitude of the error amount of rear and front end in fact, so that this ρ can be provided with is bigger, in this routine emulation ρ=SF is set, SF is a spread spectrum coefficient), just need protection at interval on the so corresponding section; Otherwise just do not need; If one section of the outside needs protection, just then the adjacent segment of inboard is carried out the detection of same procedure, to determine whether to need to strengthen protection length; Otherwise just stop detection to its inboard adjacent segment; Maximum protection length before and after setting all is 4L, and this determines according to formula (17);

(4) after the protection length that needs before and after having determined, the base station sends data to be sent, and frequency-domain equalizer receives data, according to formula (4) data is carried out frequency domain equalization.

In the present invention; because cdma system generally all has the pilot tone of time-division; just between pilot (being given data) and data segment (being data to be sent) across transmission; so can utilize pilot protection length is at interval estimated and to be revised, the data segment that arrives soon after is carried out AOC method frequency domain equalization.

Annotate: if OC method frequency-domain equalization technology carries out the descending reception of CDMA again in conjunction with the interference delete scheduling algorithm, the protection in the OC method still needs at interval, can avoid the error propagation effect so as far as possible.

Prior art and common ground of the present invention are if protection is arranged at interval, and it necessarily appears at foremost or backmost so, that is to say that position class seemingly.In the existing OC method frequency-domain equalization technology, protection setting and difference of the present invention at interval is:

1) the prior art conclusion is to detect front and back one each time to establish a capital the interval that needs protection, and the present invention has proved that sometimes preceding or back does not need protection at interval; For example under the situation among Fig. 1, the back does not just need protection at interval;

2) the prior art conclusion is all to need protection under any state of signal-to-noise at interval, and the present invention has proved under the low signal-to-noise ratio that simple frequency-domain equalizer can not be provided with protection at interval, and can not cause the decline significantly of performance;

3) in the existing paper, the protection length that is provided with that has is not mention under L-1 even the most of situation how many protection gap lengths is on earth, here L is the length of channel dispersion multipath tapped delay line model, and the protection that L-1 all is set before and after the present invention has proved is unnecessary sometimes at interval; Curtailment again sometimes can reduce the performance of frequency-domain equalizer.

Below by the performance of an example explanation AOC frequency domain equalization algorithm in the CDMA downlink system.

The emulation platform of setting up is as follows: the radio frequency carrier frequency of coded system is not 2GHz, and channel adopts vehicle-mounted A channel model, two-phase PSK (BPSK) modulation.Chip transmits according to time slot, 2560 chips of each time slot, and spreading rate is 3.84Mchip/s.Adopt ideal communication channel to estimate in the emulation.We will contrast traditional OC method, AOC method, the CP method of RAKE receiver, frequency domain equalization and not protect performance under the length situation.In the emulation to traditional OC algorithm, adopt protection length L-1 commonly used, L is the length of channel dispersion multipath tapped delay line model here.We set ρ=SF in the AOC method, and SF is a spread spectrum coefficient.

Shown in Figure 4 is 16 rank Walsh spread spectrums, under 8 users' the situation, and the bit error rate performance of RAKE receiver and different frequency-domain equilibrium methods.Under the situation of high s/n ratio, the algorithm of frequency domain equalization (no matter also being based on the MMSE criterion based on the ZF criterion) all performance than RAKE receiver is good.In the frequency domain equalization algorithm based on ZF and two criterions of MMSE, the performance of AOC algorithm all is better than the algorithm of not protecting length and fixing protection length.Protection does not improve greater than performance after the 12dB in signal to noise ratio (SNR, Signal to Noise Ratio) basically based on the frequency domain equalization of MMSE criterion at interval.

Fig. 5 is 16 rank Walsh spread spectrums, and under 16 users' the situation, the error rate of system performance of various receivers has similar distribution with Fig. 5.Complex chart 5 and Fig. 6; can see; though the frequency-domain equilibrium method of not protecting length is under the situation of high s/n ratio; poor-performing; but under the low signal-to-noise ratio situation; owing to be subjected to the high-energy The noise equally, the performance of its performance and traditional OC method, AOC method and CP method is very approaching.Therefore, can further improve the AOC method, under the situation of low signal-to-noise ratio (below 10dB), the length that can not need protection is fully carried out frequency domain equalization, has further simplified calculating so again, and the performance of frequency-domain equalizer can not be affected yet.

The above is the preferred embodiments of the present invention only, is not limited to the present invention, and for a person skilled in the art, the present invention can have various changes and variation, and protection scope of the present invention is as the criterion with claim.

Claims (6)

1. a self adaptation is provided with the protection method at interval of overlapping shearing method, it is characterized in that,

Transmitting terminal is the continuous pilot sequence of base station transmitting and receiving terminal known data length greater than channel delay, receiving terminal is after frequency-domain equalizer receives this pilot frequency sequence, calculate the systematic error sequence e of frequency domain equalization, determine the protection position and the length at interval of frequency domain equalization, from the two ends of e to the third side to, segmentation detects respectively, if a certain section error is greater than assign thresholds, just need protection at interval on the then corresponding section, and continue to detect its inboard adjacent segment; Otherwise stop to detect; Each needs the protection interval of much length before and after determining respectively.

2. method according to claim 1 is characterized in that, before transmitting terminal transmitting and receiving terminal given data, whether judges signal to noise ratio less than the snr threshold of setting, if less than, it is 0 that protection length at interval then is set; Otherwise, transmitting terminal transmitting and receiving terminal given data.

3. method according to claim 2 is characterized in that, described snr threshold is 10dB.

4. according to the described method of claim 1, it is characterized in that,

Described base station transmission continuous pilot sequence i.e. all users' piecemeal is closed signal x=[x ₀, x ₁..., x _N-1], wherein N is the length of piecemeal; It is y=[y that frequency-domain equalizer receives corresponding sequence ₀, y ₁..., y _N-1], wherein, the impulse response of the channel of data process is h=[h ₀, h ₁..., h _L-1], L is the length of discrete multipath channel, i.e. channel delay; The circular convolution of x and h is r=[r ₀, r ₁..., r _N-1];

The systematic error sequence e that calculates frequency domain equalization is:

Calculate e=IFFT[FFT (Δ, N) ./FFT (h, N)], Δ=[Δ wherein ₀, Δ ₁... Δ _L-2], be the difference of preceding L-1 the data of y and r, and FFT (, N) fast Fourier transform that length is N, [] are carried out in representative. and the sequence of the identical size of/[] representative is carried out correspondence and is divided by.

5. method according to claim 4 is characterized in that:

Get the amplitude equalizing value λ of middle L the value of e, from the two ends of e to the third side to, detect respectively, respectively get the error amount sequence that a segment length is L each time, ask the amplitude equalizing value of front and rear sections respectively, if the amplitude equalizing value of which section is greater than assign thresholds ρ λ, ρ is set to integer as required, just needs protection at interval on the then corresponding section; Otherwise just do not need; If one section of the outside needs protection, just then the adjacent segment of inboard is carried out the detection of same procedure, to determine whether to need to strengthen protection length; Otherwise just stop detection to its inboard adjacent segment; Each needs the protection interval of much length before and after determining so respectively.

6. method according to claim 5 is characterized in that, the maximum protection length of front and back all is 4L.

Images