CN102263713A - Two-dimensional OFDM (orthogonal frequency division multiplexing) channel estimation method based on filtering in transform domain - Google Patents

Abstract

The invention belongs to the technical field of mobile communication and relates to a two-dimensional OFDM (orthogonal frequency division multiplexing) channel estimation method based on filtering in a transform domain. The method comprises the following steps: inserting pilot frequency into input data and sending; after a receiving end receives the data, extracting data at a pilot frequency position so as to obtain a channel response at the pilot frequency position, namely, a pilot frequency response; and carrying out interpolation and filtering on OFDM data containing the pilot frequency in a frequency direction, and applying a linear interpolation principle to carry out interpolation for each subcarrier in a time direction so as to realize the two-dimensional OFDM channel estimation. The method can be used to filter out the most of AWGNs (additive white Gaussian noises) in channels and receivers, thus improving the accuracy of the channel estimation; and the method is realized simply, thus reducing the complexity of an algorithm greatly.

Description

A kind of two-dimentional OFDM channel estimation methods based on transform domain filtering

Technical field

The present invention relates to a kind of OFDM channel estimation methods, be specifically related to a kind of two dimensional channel algorithm for estimating of wireless fading channel.

Background technology

Further develop along with informationalized, future mobile communication system need provide more communication service, higher traffic rate and better communication quality with limited frequency spectrum resources more, and the raising of transmission rate has brought the problem of intersymbol interference (ISI) and frequency selective fading for conventional single-carrier system.OFDM (Orthogonal Frequency Division Multiplexing, OFDM) technology is one of effective method that addresses these problems.It is a kind of special multi-carrier communication scheme, and it can realize high speed data transfer in wireless channel.It passes through serial/parallel a plurality of low rate code streams that are transformed to the two-forty information flow of unique user, and each code stream all uses a subcarrier to send, and is mutually orthogonal between each subcarrier.

OFDM compares with common frequency division multiplexing (FDM) technology, has the higher availability of frequency spectrum, can avoid the intersymbol interference (ISI) on each subcarrier effectively,, make the realization of OFDM receiver become very simple by using fast fourier algorithm (FFT).The OFDM technology is as a kind of modulation technique efficiently, and its clear superiority aspect the antagonism multipath fading, all considers to select the OFDM technology as its air interface technologies in many 4G cellular system schemes.The OFDM technology has been acknowledged as the core technology of cellular communication system of future generation.

For ofdm system, under the requirement of same bit error rate, adopt differential ference spiral than the high 3-4dB of signal to noise ratio that adopts coherent demodulation; In addition, differential ference spiral relatively is suitable for use in the system of low data rate, and the ofdm system that requires for more high data rate transfers rate and high spectrum, coherent demodulation is more suitable for.But coherent demodulation must be carried out real-time tracking to the wireless multipath channel of frequency selectivity, estimates channel parameter.

Channel estimating based on pilot tone is divided into two steps usually: at first utilize various mathematical operation criterions to obtain the channel response at pilot frequency locations place, obtain the channel response of all data then by various interpolation arithmetics.Optimum filter is Weiner filter (Wiener filter) in the interpolation method, yet the Wiener filtering channel estimation methods need use complicated matrix operation, is unfavorable for the hardware real-time implementation.Linear interpolation method is realized simple, and performance is good inadequately, for this reason, a kind of channel estimation method balanced between system complexity and performance occurred, promptly based on the channel estimation methods of FFT.Consider The noise.Document 1 (Y Zhao, A Huang.A novel channel estimation methods for OFDM mobile communication systems based on pilot signals and transform domain processing[C] //Proc.IEEE 47th Vehicular Technology Conference.Phoenix, USA, May 1997,3:2089～2093.) a kind of one dimension channel estimation methods of and transform domain auxiliary based on pilot tone proposed.It at first transforms to transform domain with the response at the pilot tone place under the LS criterion through discrete Fourier transform (DFT), then after transform-domain signals being analyzed its characteristic of emulation, signal is carried out filtering, again to signal zero padding after the filtering, at last the transform-domain signals after this zero padding is made inverse discrete Fourier transform (IDFT) and be transformed into frequency domain, the result who obtains has been exactly the transfer function of whole channel.The flow process of the described channel estimation methods of the document as shown in Figure 1, M is a number of pilots, N is a number of sub carrier wave, as we can see from the figure, this method has adopted a kind of filtering technique of crying transform domain filtering, can improve the algorithm channel estimating performance with being positioned at the HFS filtering of the additive white Gaussian noise of full range band like this.This filtering technique is to use transform domain interior corresponding " high frequency " to be changed to 0, and " low frequency " part is changed to 1 and disturbs.But it is how to realize this filtering technique that the document does not specify.

Because the signal distributions of multicarrier system is in time domain and frequency domain, for Channel Track, it is essential that two dimensional channel is estimated.For the channel estimating that the two-dimension pilot frequency symbol is assisted, optimum filter is two-dimentional Weiner filter (Wiener filter), and complexity is higher.The normal one dimension interpolation method of two cascades that adopts realized two-dimentional channel estimating during common engineering was used.In a large amount of Chinese and English documents the two dimensional cascade channel estimation method was proposed all.Document 2 (M.Julia.DFT-based Channel Estimation in 2D-Pilot-Symbol-Aided OFDM Wireless Systems[C] //Proc.IEEE 53rd Vehicular Technology Conference, Rhodes, May, 2001,2:810～814.) propose on frequency direction, to carry out the FFT time domain interpolation, in the two dimensional channel method of estimation of the enterprising line linearity interpolation of time orientation.But in frequency direction owing to do not consider The noise, performance is still good inadequately, document 3 (Amir H.Khanshan, HamidrezaAmindavar.Performance evaluation of two-dimensional interpolations on OFDM channel estimation[C], //Proc.IEEE Telecommunication Networks and Applications Conference, Australasian, 2007, pp:460～464) propose linear interpolation, interpolation methods such as Cubic interpolation are used in the method on frequency and the time orientation respectively, but these two dimensional channel methods of estimation have been ignored interference of noise, and performance is good inadequately.

Summary of the invention

Based on above consideration,, the invention provides a kind of two dimensional channel method of estimation based on transform domain in order to dwindle the difference between channel estimation value and real channel frequency response.The two dimensional channel method of estimation that the present invention proposes, situation at the white Gaussian noise existence, the link of transform domain filtering is incorporated in the interpolation algorithm on the frequency direction, on time orientation, use simultaneously linear interpolation, finish the estimation of the channel response in whole moment, this two dimensional channel method of estimation can make channel estimating more accurate the most of filtering of white Gaussian noise in the receiver, and is more approaching with actual channel.For this reason, the present invention adopts following technical scheme:

A kind of two-dimentional OFDM channel estimation methods based on transform domain filtering comprises the following steps:

(1) the input data is carried out the planisphere mapping;

(2) insert pilot tone

1) adopts rectangle pilot tone mode, on frequency direction and time orientation, insert pilot tone according to the interval of setting respectively;

2), insert pilot tone according to the frequency interval identical with step 1) to last OFDM symbol;

(3) utilize the invert fast fourier transformation method that the data of inserting after the pilot tone are carried out the OFDM modulation;

(4) adding protection interval or Cyclic Prefix are gone here and there data by also becoming; And send by transmitting antenna;

(5) after receiving terminal receives data, at first become also, remove protection interval or Cyclic Prefix, realize the rectification of OFDM then by fast fourier transform by string;

(6) extract the data of pilot frequency locations, obtain the channel response at pilot frequency locations place, be referred to as the pilot tone response;

(7) receiving terminal carries out the interpolation and the filtering of frequency direction to the OFDM data that contain pilot tone according to following step:

1) fast fourier transform is carried out in response to pilot tone, obtains transform domain G;

2) calculate all sampling point total energy value EH among the G;

3) initialization: establish p1=3, p2=M-GI, wherein, M is the quantity of pilot tone in each OFDM symbol, GI is protection length at interval;

4) the ENERGY E noise of the sampling point of sequence number from p1 to p2 among the calculating G;

5) establish ratio R=0.3 * snr that noise accounts for gross energy, wherein, snr is a jamtosignal, and establishes Q=Enoise/EH;

7) size of comparison Q and R, if Q＜R, then the value of p2 adds 1, gets back to the 4th) step;

8) magnitude relationship of Q and R relatively repeatedly up to Q＞R, stops to circulate;

9) element value of sequence number among the G from p1 to p2 is changed to 0, obtains vectorial H_FFT;

10) zero padding in the middle of H FFT makes its length equate with total sub-carrier number;

11) carry out invert fast fourier transformation, obtain frequency domain value, and add penalty coefficient;

(8) to each subcarrier, the principle of utilization linear interpolation, interpolation on time orientation realizes two-dimentional OFDM channel estimating.

Useful result of the present invention is, filtering additive white Gaussian noise in most of channel and the receiver, improved channel estimated accuracy, simultaneously, the interpolation algorithm of two one dimensions of utilization is finished the estimation of whole moment channel responses, and the matrix operation of this method is considerably less, realize simply greatly reducing the complexity of algorithm.

Description of drawings

Fig. 1 is the realization block diagram of the channel estimation method of document 1.

Fig. 2 is the realization block diagram of channel estimation method of the present invention.

Fig. 3 is a pilot design pattern of the present invention.

Fig. 4 is two-dimensional interpolation implementation procedure figure.

Fig. 5 is the transform domain characteristic of pilot tone response under the 5dB signal to noise ratio.

Fig. 6 is conversion and the characteristic through filtered pilot tone response.

Fig. 7 is that the inventive method and other two-dimensional interpolation channel estimation methods RMSE compare.

Embodiment

Below in conjunction with the drawings and specific embodiments the channel estimation methods that the present invention proposes is carried out concrete being described below.

The two dimensional channel method of estimation that is a kind of based on transform domain filtering of the present invention, this is a kind of pilot tone assisted channel estimation method, what Fig. 2 represented is the block diagram of channel estimation method of the present invention.This schematic diagram has provided the signal of being accepted by receiver and has passed through after a series of conversion in frequency domain and in the time domain, has estimated the frequency response of channel accurately.

In order to realize this method, need in sending signal, insert pilot tone.What Fig. 3 represented is pilot design scheme of the present invention, and pilot tone adopts the rectangle inserted mode.According to shown in Figure 3, the pilot-frequency expense of the scheme of this pilot design is 25%,, is spaced apart 2 on the frequency direction that is, is spaced apart 2 on the time orientation, and pilot interval can be other value, as long as satisfy the pilot interval on the time orientation

Fd wherein _MaxBe maximum doppler frequency, Tf is the symbol period of OFDM, the pilot interval on the frequency direction

τ _MaxBe the maximum delay of channel, Δ df is the intercarrier frequency interval.Pilot tone adopts at random

Pilot tone.It should be noted that and to insert pilot tone for last symbol with same frequency interval.

1. receiver utilizes the least square (LS) of pilot frequency sequence calculating channel frequency domain response to estimate.In ofdm system, call algoritic module, set up matrix P, have only the pilot bit value of being equipped with among the P, be the pilot tone that sends, the value of other positions is 0, obtains this inverse of a matrix matrix P ^-1, this matrix and receiving matrix Y _p(having only the pilot bit value of being equipped with, the data that its value receives for receiver) multiplied each other and estimated H at the least square LS of the frequency domain response of pilot frequency locations for channel _N=P ^-1Y _p, Y wherein _pBe the reception frequency-region signal matrix of N*NL, P is the matrix of being made up of pilot tone of N * NL of sending, (*) ^-1Representing matrix is inverted, and NL is the OFDM symbolic number that sends, and N is the number of sub carrier wave in the ofdm system, also is that the computing of discrete Fourier transform and inverse discrete Fourier transform is counted.

2. the pilot tone response matrix is done the two-dimensional interpolation conversion, as shown in Figure 4.Concrete interpolation process is divided into two one dimension interpolations, is respectively FFT interpolation on the frequency direction and the linear interpolation on the time orientation, and step is as follows:

◆ filtering on the frequency direction and interpolation.Make all conversion that contain the OFDM symbol of pilot tone through following step (1) (2) (3) (4) (5), until all channel responses that contain the position on whole frequency direction of symbol of pilot tone value of estimating all.

(1) response is N * 1 vector in the matrix to p OFDM symbol frequency, extracts pilot tone place frequency response estimated value and is placed in the vector, is designated as H _{M, p}, H _{M, p}Middle element is for removing the pilot tone response matrix of unknown position, and the adjacent position, front and back is constant, and the number of pilot tone is M=N/interval_f, and wherein N is the subcarrier number, and interval_f is the interval on the pilot carrier frequency direction, so, and H _{M, p}Vector for M * 1.

(2) to H _{M, p}Do the DFT conversion, the DFT here can utilize the fast algorithm FFT of existing various inverse discrete Fourier transforms to realize, promptly utilizes the IFFT conversion that the frequency response of pilot tone is converted to the transform domain response G that transform domain obtains the pilot tone place _{M, p}=FH _{M, p}DFT transformation matrix F=[r wherein _{M, n}], $r_{m,n}=\frac{1}{\sqrt{N}}\exp [-j2\mathrm{\π}(m-1)(n-1)/N],$ (1≤m, n≤N), (*) ^HThe conjugate transpose of representing matrix.

The transform domain characteristic of pilot tone response when Fig. 5 represents that signal to noise ratio is 5dB.

(3) result to (2) step carries out filtering.Filtering is as follows:

(i) calculate G _{M, p}In all sampling point total energy value EH.

(ii) initialization: establish p1=3, p2=M-GI, GI is protection length at interval.

(iii) calculate G _{M, p}The ENERGY E noise of the middle sampling point of sequence number from p1 to p2.

(iv) establish ratio R=0.3 * snr that noise accounts for gross energy, wherein, snr is a jamtosignal, and establishes M=Enoise/EH;

(the size that v) compares M and R.If M＜R, then the value of p2 adds 1, gets back to for the (iii) step.

(magnitude relationship that vi) compares M and R repeatedly up to M＞R, stops circulation.

(vii) with G _{M, p}The middle element value of sequence number from p1 to p2 is changed to 0, obtains vectorial H_FFT; The characteristic of H_FFT as shown in Figure 6

(4) H_FFT is carried out centre zero padding operation, obtain vector

The consistent vector of quantity N of one length and subcarrier is set

The M/2 element replaces with M/2 element among the H_FFT before this vector, and back M/2 element replaces with back M/2 element among the H_FFT.Available following matrix is represented the process of zero padding: ${\hat{G}}_{N,p}=\left[\begin{array}{ccc}H\_\mathrm{FFT}(1:M/2)& 0& H\_\mathrm{FFT}(M/2+1:M)\end{array}\right],$ Wherein 0 is the full null vector of (N-M) * 1.

(5) right

Make discrete Fourier transform IDFT, this IDFT also is to use the fast algorithm IFFT of inverse discrete Fourier transform, promptly utilizes IFFT with vector Become frequency domain again,

Obtain vector

With vector

Be placed in the receiv_h matrix as column vector, wherein receiv_h is the matrix of a N * NL, is the matrix of depositing last estimated result, it is the matrix of channel frequency response, a is the conversion penalty coefficient, changes according to the value of N and M, and NL is the OFDM symbolic number.

◆ on the time orientation, all subcarriers are carried out following interpolation conversion.To each subcarrier is the interpolation method of the row vector utilization linear interpolation of receiv_h, and according to the channel response of known location on the time orientation, linear interpolation can obtain being in the channel response of two positions between the given data.

Through above-mentioned conversion and calculating, just obtained the channel estimation value of OFDM, adopted this method that ofdm system is carried out channel estimating, filtering most of white Gaussian noise, and eliminated part because the error noise that interpolation caused has effectively been estimated the frequency response of channel.For the performance of the inventive method and traditional two-dimensional linear interpolation algorithm is carried out concrete comparative descriptions, compare in the normalization mean square error (RMSE) of following test platform to the channel response of two kinds of channel estimation methods and real channel:

1. build the ofdm system platform.

2. test channel is COST207 typical case city (TU) channel, adopts 6 footpath channels, and maximum doppler frequency is 100Hz, and corresponding translational speed is 108km/h.

3. the sub-carrier number of transmitting terminal Modulation OFDM signal is 128, and the OFDM symbolic number is any, gets 80 in order to test here, and the length of Cyclic Prefix is 32.

4. the planisphere mapping mode is QPSK.

Pilot tone to be designed on frequency direction be 2, be 2 on time orientation, last symbol is dealt with the back pilot-frequency expense is roughly 25%.

6. the sampling interval is 1 μ s.

7. the channel signal to noise ratio is got 5dB, 10dB, 15dB, 20dB, 25dB, 30dB, 35dB, these several numbers of 40dB.

Fig. 7 is the comparison diagram of the normalization mean square error (RMSE) of the inventive method and other two kinds of two-dimensional interpolation channel estimation methods on above-mentioned test platform.The channel estimating method of linear-linear representative all adopts the mode of linear interpolation on time and frequency direction, the channel estimating method of DFT-linear representative adopts the time domain interpolation method based on DFT on frequency direction, adopt linear interpolation method on time orientation.As can be seen from Figure, the inventive method is compared with linear-linear, and the performance gain of about 3dB is arranged when signal to noise ratio is low, and along with the increase of signal to noise ratio, the advantage of algorithm of the present invention is increasing; Compare with DFT-linear, the performance of algorithm of the present invention under low signal-to-noise ratio is better than the DFT-linear algorithm, the gain of nearly 4-5dB.

The present invention utilizes FFT transform domain interpolation and filtering, and linear interpolation combines, and has formed a kind of two dimensional channel method of estimation based on transform domain filtering.Use this method, can reduce effectively that noise obtains good estimation effect to the influence of channel estimating in the receiver.The inventive method is simple, matrix operation is few, be easy to software and realize, and the use of fast Fourier transform (FFT) and inverse fast Fourier transform (IFFT) also makes and has reduced much on the complexity, easy operating, so the present invention has good application prospects.

The above only is a preferred forms of the present invention; but protection scope of the present invention is not limited thereto; any those skilled in the art are in the disclosed technical scope of the present invention; the conversion that can expect easily or replacement; all should be included within the protection category of the present invention; therefore, protection scope of the present invention should be as the criterion with the scope of claim.

Claims (1)

1. the two-dimentional OFDM channel estimation methods based on transform domain filtering comprises the following steps:

(1) the input data is carried out the planisphere mapping;

(2) insert pilot tone

1) adopts rectangle pilot tone mode, on frequency direction and time orientation, insert pilot tone according to the interval of setting respectively;

2), insert pilot tone according to the frequency interval identical with step 1) to last OFDM symbol;

(3) utilize the invert fast fourier transformation method that the data of inserting after the pilot tone are carried out the OFDM modulation;

(4) adding protection interval or Cyclic Prefix are gone here and there data by also becoming; And send by transmitting antenna;

(5) after receiving terminal receives data, at first become also, remove protection interval or Cyclic Prefix, realize the rectification of OFDM then by fast fourier transform by string;

(6) extract the data of pilot frequency locations, obtain the channel response at pilot frequency locations place, be referred to as the pilot tone response;

(7) receiving terminal carries out the interpolation and the filtering of frequency direction to the OFDM data that contain pilot tone according to following step:

1) fast fourier transform is carried out in response to pilot tone, obtains transform domain response G;

2) calculate all sampling point total energy value EH among the G;

3) initialization: establish p1=3, p2=M-GI, wherein, M is the quantity of pilot tone in each OFDM symbol, GI is protection length at interval;

4) the ENERGY E noise of the sampling point of sequence number from p1 to p2 among the calculating G;

5) establish ratio R=0.3 * snr that noise accounts for gross energy, wherein, snr is a jamtosignal, and establishes Q=Enoise/EH;

7) size of comparison Q and R, if Q＜R, then the value of p2 adds 1, gets back to the 4th) step;

8) magnitude relationship of M and R relatively repeatedly up to Q＞R, stops to circulate;

9) element value of sequence number among the G from p1 to p2 is changed to 0, obtains vectorial H FFT;

10) zero padding in the middle of H_FFT makes its length equate with total sub-carrier number;

11) carry out invert fast fourier transformation, obtain frequency domain value, and add penalty coefficient;

(8) to each subcarrier, the principle of utilization linear interpolation, interpolation on time orientation realizes two-dimentional OFDM channel estimating.

Images