CN105763490A - Improved in-band noise reduction DFT channel estimation algorithm - Google Patents

Abstract

The invention provides an improved in-band noise reduction DFT channel estimation algorithm. As noise still exists in the traditional DFT algorithm channel estimation value, the signal has a large power change range, the demodulated channel is distorted possibly, transmission of the signal has serious time domain dispersion, and in order to reduce the influences, the algorithm of the invention processes DFT channel estimation-based hDFT(n) with n in a range of 0 to Ng, the remaining remains unchanged, and noise interference is filtered.

Description

The DFT channel estimation method of noise reduction in the band of a kind of improvement

Technical field

The present invention relates to the communications field, more particularly, to the DFT channel estimation method of noise reduction in the band of a kind of improvement.

Background technology

Since nineteen nineties, in global range, wireless communication technology obtains unprecedented development.Wherein WLAN (WLAN) the technical development impetus is swift and violent, finds broad application.WLAN is a kind of Communication Network Technique using wireless channel as transmission medium, is grow up on the basis of cable network, makes terminal have mobility, can solve the connectivity problem that wired mode not easily realizes quickly and easily.Its appearance and development meet people at any time and any place with anyone hope carrying out communication and information sharing and demand.Bandwidth is wider, and speed is faster the inexorable trend of various applied business development from now on.The WLAN planning standard of China relates generally to IEEE802.11 series and HiperLAN1/2, and the actual and product situation according to current China WLAN market, IEEE802.11 series is main standard.Mainly there are IEEE802.11, IEEE802.11b/a/g and IEEE802.11n.And up-to-date 802.11ac the maximum data transfer rate is up to 1Gb/s, thus playing, for universal high definition real-time video in the future, the basis providing strong.

The performance of wireless communication system is limited primarily by the restriction of wireless channel.Propagation path between transmitter and receiver is extremely complex, from simple line-of-sight propagation to the landforms suffering various complexity, such as the propagation that building, mountain range and forest etc. affect, in addition wireless channel unlike wire message way fixing and it is contemplated that, but there is very big randomness, cause the distortion receiving the amplitude of signal, phase and frequency, be difficult to be analyzed.So the design of receiver is proposed very big challenge by these problems, and in receivers, channel estimator is a critically important ingredient.

Along with the fast development of wireless communication system and perfect, channel estimation technique increasingly becomes a critically important technology in radio communication undoubtedly.Emission source signal is carried out data transmission by wireless channel, at receiving terminal in order to detect correct transmission information, it is necessary to know correct physical channel relevant parameter, and thus can compensating property reparation to received signal.Channel estimating has a great impact for the overall performance of MIMO-OFDM system.Because, first if system is the modulation system of difference form, data message is encoded according to the difference value of two continuous symbols, the time domain changing value of channel is only small, now system does not need channel estimator, but just due to needs channel estimation process, system receiver is substantially bigger than the Receiver Complexity aspect having channel estimator.But, in practice it has proved that, difference detector compares the snr loss that can cause that system produces 3dB with coherent detector.Therefore, in order to the coherence detection that serviceability is more excellent, research channel estimation technique is very necessary.

Based in the MIMO-OFDM system of 802.11n standard, under slow fading quasi-static channel opens, in practice, receiver end mainly adopts least square (LS, LesatSquare) channel estimation method.LS algorithm computational complexity is low, and hardware is easily achieved, and is substantially the requirement that can reach estimator under quasi-static channel opens, so being widely used in actual wireless LAN system.But, LS algorithm for estimating owing to have ignored the factor of noise when estimating, so channel estimation value is more sensitive on the impact of noise jamming and inter-sub-carrier interference (ICI) and intersymbol interference (ISI), when interchannel noise is relatively larger, the accuracy of channel estimation value is substantially reduced.

In order to carry out channel estimating more accurately, it is necessary to consider the effect of interchannel noise, so just there being the channel estimation method based on DFT, first this algorithm carries out the channel estimating of LS algorithm, enters time domain then through IDFT.In time domain, channel energy concentrates on relatively small number of sample point, and sample point relatively low for energy is considered as zero, eventually passes DFT and enters frequency domain, namely obtains the result of channel estimation in frequency domain.Make use of a characteristic of ofdm system based on the channel estimating of DFT, the length of Cyclic Prefix is typically greater than the shock response length of channel.Channel energy owing to being obtained by LS channel estimation method is concentrated on some sample points of beginning, and DFT channel estimating eliminates effect of noise simply by the channel response value zero setting more than circulating prefix-length.This algorithm there are about the lifting of 5dB compared with LS algorithm in BER performance.Although having filtered the noise beyond Cyclic Prefix code length based on DFT channel estimation method, then the noise within Cyclic Prefix code length does not process or eliminates, and still the end value of channel estimating can be produced impact.So the estimation performance of this algorithm is it is also possible that further improve.

In quasi-static channel opens environment, channel changes not quite within several OFDM symbol cycles.Channel estimating based on training symbol requires to send known symbol on all subcarriers of some OFDM symbol.Owing to there being frequency pilot sign on each subcarrier, receiving terminal just can estimate the frequency domain characteristic of all sub-channel of training symbol according to the change of these known symbols.

LS channel estimation method can be used in the middle of various wireless communication system, by utilizing pilot tone to carry out channel estimating.When user is in the middle of wireless local area network (WLAN) system, it is utilize high-throughput long training sequence in 802.11n mixed format frame structure and Block-type pilot that channel response value is estimated.This algorithm have ignored the existence of noise, and therefore its end value is highly susceptible to the impact that noise brings, and ICI and ISI is more sensitive.Owing to this algorithm has only to do a division calculation, it is not necessary to the prior information of any channel, hardware realizes simply and substantially meeting being wirelessly transferred requirement under quasi-static channel opens, therefore also extensively uses in MIMO-OFDM system.Traditional channel estimation method based on DFT, by time domain by the channel response value zero setting beyond Cyclic Prefix code length, eliminate the impact on channel estimation value of the noise beyond circulating prefix-length code.

The frequency domain LS that traditional channel estimation method based on DFT obtains channel first with high-throughput long training sequence estimates, then inverse Fourier transform (IDFT) is utilized to estimate to be transformed into time domain by the LS of channel, rational linear transformation is carried out again in time domain, carry out the process of denoising, finally utilize Fourier transformation to be transformed into frequency domain and obtain the frequency domain response estimation of channel.For sampling interval channel, all concentrate on several limited multipaths at time domain major part energy, remaining major part item all only comprises noise and does not comprise channel power, processes so the item beyond Cyclic Prefix can do simple zero setting when time domain carries out linear transformation.The just ratio more than once IDFT of LS algorithm and DFT transform in processing procedure, and DFT has fast algorithm FFT, and therefore the complexity of system does not include too high yet.

Summary of the invention

The present invention provides a kind of effective minimizing noise on the DFT channel estimation method of noise reduction in the band of the improvement that channel estimating affects.

In order to reach above-mentioned technique effect, technical scheme is as follows:

The DFT channel estimation method of noise reduction in the band of a kind of improvement, comprises the following steps:

S1: adopt LS algorithm to calculate the frequency domain response value H of channel_LSK (), to H_LSK () is done N point IDFT conversion and is obtained LS time domain initial estimation h_LSN (), wherein N is the quantity of channel sub-carrier；

S2: to h_LSN () processes, as 0≤n≤N_gTime, retain h_LSN the value of (), works as N_gDuring≤n≤N, to h_LSN () does linear transformation, obtain the channel estimation value h of the time domain of DFT algorithm_DFT(n), wherein N_gLength for cyclic prefix code；

S3: work as N_gDuring≤n≤N, channel response value is all caused by noise, calculates h_LSN () is at N_gThe meansigma methods W of channel time domain response value amplitude during≤n≤N, is noise figure:

$W=\frac{1}{N-N_g}\underset{n=N_g}{\overset{N}{\Σ}}\left|h_{LS}\left(n\right)\right|;$

S4: using W as noise threshold, retains h_DFTN in (), amplitude is more than W/2 channel estimation value, for h within circulating prefix-length_DFTN () channel estimation value less than W/2, directly does zero setting and processes, the DFT channel estimating time-domain value h being improved_DFT-U(n), 0≤n≤N-1；

S5: to h_DFT-UN () does DFT transform must based on the channel estimation in frequency domain value of the innovatory algorithm of DFT:

$H_{DFT-U}\left(k\right)=\frac{1}{N}\underset{n=0}{\overset{N-1}{\Σ}}{h}_{DFT-U}\left(n\right)e^{-j2\mathrm{\π}\frac{nk}{N}},0\≤k\≤N-1.$

Further, to h in described step S2_LSN () makes the detailed process of linear transformation:

Retain the channel response value within Cyclic Prefix code length, and by the channel response value zero setting beyond Cyclic Prefix code length.

Further, N is worked as_gDuring≤n≤N-1, h_LSN () is all noise, without any useful information, N_gChannel estimation value zero setting during≤n≤N-1, its residual value is constant, obtains the DFT algorithm channel estimating expression formula of time domain:

$h_{DFT}\left(n\right)=\left\{\begin{array}{cc}h\left(n\right)+w\left(n\right)& 0\≤n\≤N_g-1\\ 0& N_g\≤n\≤N-1\end{array}\right..$

Further, the process of described step S1 is as follows:

LS algorithm is adopted to calculate the frequency domain response value H of channel_LSK (), to H_LSK () is done N point IDFT conversion and is obtained LS time domain initial estimation h_LS(n):

$h_{LS}\left(n\right)=\underset{k=0}{\overset{K}{\Σ}}{H}_{LS}\left(k\right)e^{j2\mathrm{\π}\frac{nk}{N}},0\≤k\≤K-1,0\≤n\≤N-1.$

Compared with prior art, technical solution of the present invention provides the benefit that:

Due to DFT algorithm channel estimation value in there is also noise, the changed power scope making signal is very big, the channel channel so demodulated likely can distortion, and the propagation of signal has serious time domain dispersivity, propose the innovatory algorithm of DFT channel estimating to reduce this present invention of impact, for based on DFT channel estimating at 0≤n≤N_g(N_gRepresent the length of cyclic prefix code) h_DFTN () carries out processing and other are constant, filters noise jamming therein.

Accompanying drawing explanation

Fig. 1 is traditional channel estimating theory diagram based on DFT；

Fig. 2 is the DFT algorithm flow chart improved；

Fig. 3 is noise reduction DFT algorithm principle block diagram in the band improved.

Detailed description of the invention

Accompanying drawing being merely cited for property explanation, it is impossible to be interpreted as the restriction to this patent；

In order to the present embodiment is better described, some parts of accompanying drawing have omission, zoom in or out, and do not represent the size of actual product；

To those skilled in the art, in accompanying drawing, some known features and explanation thereof are likely to omission and will be understood by.

Below in conjunction with drawings and Examples, technical scheme is described further.

Embodiment 1

In the band of the present invention, the DFT channel estimation method of noise reduction is based on the algorithm for estimating of Block-type pilot, and in 802.11n wireless local area network (WLAN) system, Block-type pilot is present in Frame with the formation of training sequence, so the DFT algorithm improved is applicable to 802.11n system.In general the channel Doppler spread of 802.11n WLAN is less, say, that have bigger coherence time, and channel belongs to slow fading channel and quasi-static channel opens, it is possible to consider to have only to estimate a secondary channel within a Frame.So when sending Frame, first send training sequence, send data symbol.This example utilizes the high-throughput long training sequence in 802.11n WLAN mixed format frame structure to do the channel estimation method of the improvement based on DFT.

An example based on the channel estimating of the MIMO-OFDM system of 802.11n standard given below.Here two streams, the example of two transmission antennas and two reception antennas are considered, it is considered to spatial stream number and transmitting antenna number are the relations directly mapped, so containing two high-throughput long training sequences in each 802.11n mixed format frame sent here.A kind of simple and clear expression is as follows, and first antenna of example, wherein x represents high-throughput long training sequence, Y_t1Represent t₁First code element of the receiving terminal of time, Y_t2Represent t₂Second code element of the receiving terminal of time.w_t1Represent t₁The corresponding white Gaussian noise of time, w_t2Represent t₂The corresponding white Gaussian noise of time, W is white noise.

Y_t1=diag (x) H₁₁+diag(x)H₁₂+w_t1

Y_t2=diag (-x) H₁₁+diag(x)H₁₂+w_t2

Order

$X=\left[\begin{array}{cc}diag\left(x\right)& diag\left(x\right)\\ diag(-x)& diag\left(x\right)\end{array}\right]$

$Y=\left[\begin{array}{c}{Y}_{t1}\\ Y_{t2}\end{array}\right]$

$H=\left[\begin{array}{c}{H}_{11}\\ H_{12}\end{array}\right]$

Then least-squares estimation result can be written as

${\hat{H}}_{LS}=X^{-1}Y$

This is the frequency domain channel shock response value that LS algorithm obtains.The channel estimation in frequency domain that LS algorithm obtains is done IDFT conversion and obtains time domain channel estimated value:

$h_{LS}\left(n\right)=\underset{k=0}{\overset{K}{\Σ}}{H}_{LS}\left(k\right)e^{j2\mathrm{\π}\frac{nk}{N}},0\≤k\≤K-1,0\≤n\≤N-1$

In formula, k represents the sequence number of subcarrier in frequency domain, and maximum is the sampled point that K, n represent time domain, and maximum is N, N and K desirable 64 or 128 in 802.11n system.

In MIMO-OFDM system designs; in order to reduce ISI (intersymbol interference) and the ICI (inter-sub-carrier interference) impact on system; the protection interval (Cyclic Prefix) of OFDM symbol should more than the maximum delay of channel impulse response, and the channel impulse response major part energy that the domain channel response value tried to achieve by LS algorithm is obtained through IDFT conversion all concentrates on several sampled points of relatively few.Therefore the channel impulse response of the multipath component that power is bigger, it should drop within protection interval, and protect the component beyond interval to can be regarded as noise.DFT algorithm make use of the length of the channel impulse response of MIMO-OFDM system under normal circumstances less than the length of cyclic prefix code, is all caused by noise more than the estimated value of the length of cyclic prefix code.

So working as N_gDuring≤n≤N-1, h_LSN () is all noise, without any useful information, therefore N_gChannel estimation value zero setting during≤n≤N-1, its residual value is constant, it is possible to the DFT algorithm channel estimating expression formula obtaining time domain is shown below:

$h_{DFT}\left(n\right)=\left\{\begin{array}{cc}h\left(n\right)+w\left(n\right)& 0\≤n\≤N_g-1\\ 0& N_g\≤n\≤N-1\end{array}\right.$

In ofdm system, maximum multipath time delay is less than the length of cyclic prefix code, it is possible to effectively avoid the impact of intersymbol interference, within the time domain response value of channel concentrates on Cyclic Prefix code length.Therefore N is worked as_gDuring≤n≤N-1, LS algorithm channel time domain response value is all caused by noise, it is possible to try to achieve the meansigma methods W of noise amplitude:

$W=\frac{1}{N-N_g}\underset{n=N_g}{\overset{N-1}{\Σ}}\left|h_{LS}\left(n\right)\right|$

H within Cyclic Prefix_DFTN () channel response value, is mostly the effective information of channel, but still there is partial noise.This partial noise is hidden in the middle of virtual value, it is necessary to be further processed noise reduction.Obtained the noise average W beyond Cyclic Prefix code length by above formula, though and have influence of noise within Cyclic Prefix code length, but main or effective information, consider, in order to reduce the influence of noise within Cyclic Prefix code length, to retain h so compromise_DFTN in (), amplitude is more than W/2 channel estimation value, for h within circulating prefix-length_DFTN () channel estimation value less than W/2, directly does zero setting and processes.Namely the DFT channel estimation value h being improved_DFT-U(n),0≤n≤N-1.Then to h_DFT-UN () is done DFT computing and can be obtained, the DFT channel estimation value h improved in frequency domain_DFT-UK (), is shown below:

H_DFT-U(k)=DFT (h_DFT(n)),0≤k≤K-1

The corresponding same or analogous parts of same or analogous label；

Position relationship described in accompanying drawing be used for the explanation of being merely cited for property, it is impossible to be interpreted as the restriction to this patent；

Obviously, the above embodiment of the present invention is only for clearly demonstrating example of the present invention, and is not the restriction to embodiments of the present invention.For those of ordinary skill in the field, can also make other changes in different forms on the basis of the above description.Here without also cannot all of embodiment be given exhaustive.All any amendment, equivalent replacement and improvement etc. made within the spirit and principles in the present invention, should be included within the protection domain of the claims in the present invention.

Claims (4)

1. the DFT channel estimation method of noise reduction in the band improved, it is characterised in that comprise the following steps:

S1: adopt LS algorithm to calculate the frequency domain response value H of channel_LSK (), to H_LSK () is done N point IDFT conversion and is obtained LS time domain initial estimation h_LSN (), wherein N is the quantity of channel sub-carrier；

S2: to h_LSN () processes, as 0≤n≤N_gTime, retain h_LSN the value of (), works as N_gDuring≤n≤N, to h_LSN () does linear transformation, obtain the channel estimation value h of the time domain of DFT algorithm_DFT(n), wherein N_gLength for cyclic prefix code；

S3: work as N_gDuring≤n≤N, channel response value is all caused by noise, calculates h_LSN () is at N_gThe meansigma methods W of channel time domain response value amplitude during≤n≤N, is noise figure:

$W=\frac{1}{N-N_g}\underset{n=N_g}{\overset{N}{\Σ}}\left|h_{LS}\left(n\right)\right|;$

S4: using W as noise threshold, retains h_DFTN in (), amplitude is more than W/2 channel estimation value, for h within circulating prefix-length_DFTN () channel estimation value less than W/2, directly does zero setting and processes, the DFT channel estimating time-domain value h being improved_DFT-U(n), 0≤n≤N-1；

S5: to h_DFT-UN () does DFT transform must based on the channel estimation in frequency domain value of the innovatory algorithm of DFT:

$H_{DFT-U}\left(k\right)=\frac{1}{N}\underset{n=0}{\overset{N-1}{\Σ}}{h}_{DFT-U}\left(n\right)e^{-j2\mathrm{\π}\frac{nk}{N}},0\≤k\≤N-1.$

2. the DFT channel estimation method of noise reduction in the band of improvement according to claim 1, it is characterised in that to h in described step S2_LSN () makes the detailed process of linear transformation:

Retain the channel response value within Cyclic Prefix code length, and by the channel response value zero setting beyond Cyclic Prefix code length.

3. the DFT channel estimation method of noise reduction in the band of improvement according to claim 2, it is characterised in that work as N_gDuring≤n≤N-1, h_LSN () is all noise, without any useful information, N_gChannel estimation value zero setting during≤n≤N-1, its residual value is constant, obtains the DFT algorithm channel estimating expression formula of time domain:

$h_{DFT}\left(n\right)=\left\{\begin{array}{cc}h\left(n\right)+w\left(n\right)& 0\≤n\≤N_g-1\\ 0& N_g\≤n\≤N-1\end{array}\right..$

4. the DFT channel estimation method of noise reduction in the band of improvement according to claim 1, it is characterised in that the process of described step S1 is as follows:

LS algorithm is adopted to calculate the frequency domain response value H of channel_LSK (), to H_LSK () is done N point IDFT conversion and is obtained LS time domain initial estimation h_LS(n):

$h_{LS}\left(n\right)=\underset{k=0}{\overset{K}{\Σ}}{H}_{LS}\left(k\right)e^{j2\mathrm{\π}\frac{nk}{N}},0\≤k\≤K-1,0\≤n\≤N-1.$