CN101079863A - Frequency domain balancer design method in orthogonal frequency division multiplexing system - Google Patents

Abstract

The invention discloses a designing method of frequency domain equalizer in the orthogonal frequency division multiplexing system, which comprises the following steps: utilizing MMSE criterion to get optimum expression formula of tapping power factor of frequency domain equalizer; possessing front system and compensating system of the frequency domain equalizer; making the residual interference power to eliminate ISI and ICI completely; inhibiting the noise power of residuals; obtaining the optimum frequency domain equalizer with simple structure; improving the usage of frequency band of system without adding the CP length which is less than the length of channel impaction respond length.

Description

Frequency domain balancer design method in the ofdm system

Technical field

The present invention relates to a kind of channel equalization technique of wireless communication field, specifically is the frequency domain balancer design method in a kind of OFDM (OFDM) system.

Background technology

Along with the develop rapidly of radio communication, data communication and Internet with merge day by day, the mobile subscriber is for the ever-increasing while of the demand of multiple business, and the requirement of traffic rate is also improved constantly.Wireless communication technology is experiencing unprecedented opportunity to develop.Any progress of wireless domain research simultaneously all must be tightly round a theme, and that is exactly must solve preferably that bandwidth anxiety, propagation channel that high-speed radiocommunication self faced are abominable, mobile environment is complicated and the quality problems during limited-service.The channel bandwidth frequency selectivity multipath fading that two-forty causes, the continuous variation of scattering object makes channel have Doppler's time selective fading effect around the fast moving of portable terminal or perhaps the communication environments simultaneously.This associating T/F selectivity causes the multipath-Doppler-decline of channel, can have a strong impact on the overall performance of system.And because the intersymbol interference (ISI) that multipath effect causes is the principal element that the restriction wireless communication system improves transmission rate.Therefore, in system design, the problem that people are most interested in is how to design whole communication system, to eliminate Unknown Channel to the distortion of transmission signals and the influence of additive Gaussian noise.As everyone knows, at receiving terminal receiver is optimized design, i.e. the design of equalizer is one of effective way that addresses the above problem.

OFDM (OFDM) is the effective multi-transceiver technology of a kind of frequency spectrum, thereby it utilizes the orthogonal narrow-band sub-carriers of some to transmit the transmission that low-rate data is realized whole high-speed data concurrently.Ofdm system utilizes Cyclic Prefix can effectively suppress intersymbol interference (ISI) and inter-carrier interference (ICI), but because the insertion of Cyclic Prefix causes the band efficiency of system to descend, in case and the impulse response length of channel is greater than the length of CP, CP can't eliminate ISI and ICI fully, and the performance of system will descend rapidly owing to the existence of residual ISI and ICI so.

In order to solve this two problems, various equalization schemes have appearred at present.These schemes mainly are divided into two classes, and a kind of is to shorten channel approach, and another is non-linear disturbance Canceller.Shortening channel approach is to utilize liftering to shorten the effective length of channel impulse response.This scheme can realize in time domain and frequency domain.Time-domain equalizer utilizes the time domain finite impulse response filter to come the compensate for channel impulse response, and frequency-domain equalizer has been realized further improvement by finite impulse response filter is moved to frequency domain from time domain.In order to finish equilibrium, the scheme of this shortening channel still need increase certain redundancy.Non-linear interference offset device is the duplicate of adjudicating the output of symbol and maximum likelihood detector as Canceller with pre-.Though this scheme can obtain good equalization performance, its computation complexity is unacceptable.

Summary of the invention

The objective of the invention is to the deficiency that exists in the existing balancing technique, the frequency domain balancer design method in a kind of ofdm system is provided.The starting point of this method is under error rate minimum or mean square error minimal condition, ask for the equaliser structure model of performance the best, this method has been broken the design philosophy of conventional equalizer in the past, promptly set the structure of equalizer earlier, this is obtained by priori or hypothesis, be conceived to the research of algorithm then, to scheme to improve the performance of equalizer, opposite the present invention is setting structure not fully, promptly equalizer is not done the constrained a priori assumption of any structure, (MMSE) asks for the system model with best portfolio effect by minimum mean square error criterion, carries out structure according to this equalizer system model with optimum efficiency then and realizes.Equalizer among the present invention is divided into primary mold and improved model, because the existence of white Gauss noise makes primary mold can't eliminate ISI and ICI fully, and improved model can be eliminated ISI and ICI in theory fully.

Of the present invention to the effect that according to ofdm system reception model, obtain the frequency-domain equalizer tap parameter expression formula under the MMSE criterion, in order thoroughly to eliminate ISI and the ICI that exists in the ofdm system, on elementary frequency-domain equalizer model based, obtained the parameter expression of bucking-out system in the frequency-domain equalizer improved model.Under above-mentioned theoretical foundation, the present invention proposes concrete frequency-domain equalizer structure.

For achieving the above object, technical conceive of the present invention and principle are as follows:

(1) foundation of primary mold and analysis

The frequency-region signal that the ofdm system receiver receives is that the input of frequency-domain equalizer can be expressed as:

X(k)＝H(k)S(k)+N(k) (1)

Frequency-domain equalizer is output as:

Y(k)＝C(k)X(k) (2)

And the evaluated error of frequency-domain equalizer is:

$E\left(k\right)=\hat{S}\left(k\right)-Y\left(k\right)---\left(3\right)$

Suppose $\hat{S}\left(k\right)=S\left(k\right)$ , and with (1) and (2) substitution (3) formula, the mean square error that can obtain equalizer is:

Definition ε _I(k) and ε _N(k) be respectively residual interference (ISI and ICI) power and residual noise power,, make ε (k) that minimum value be arranged by asking for the value that satisfies following partial derivative C (k) according to the MMSE criterion:

$\frac{\∂\mathrm{\ϵ}\left(k\right)}{\∂C\left(k\right)}={\mathrm{\σ}}_s^2{C}^{*}\left(k\right){\left|H\left(k\right)\right|}^2-{\mathrm{\σ}}_s^{2}H\left(k\right)+{\mathrm{\σ}}_n^2{C}^{*}\left(k\right)=0---\left(5\right)$

Try to achieve the forward direction system tap weights coefficient of frequency-domain equalizer:

$C\left(k\right)={\mathrm{\&sigma;}}_s^2{H}^{*}\left(k\right){[{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="n"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">n</figure-callout>}}^2+{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="s"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">s</figure-callout>}}^2{\left|H\left(k\right)\right|}^2]}^{-1}---\left(6\right)$

To analyze in order further simplifying, C (k) substitution of obtaining can be obtained minimum mean square error:

${\mathrm{\&epsiv;}}_{\min }\left(k\right)={\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="s"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">s</figure-callout>}}^2{\mathrm{\&sigma;}}_n^2{[{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="n"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">n</figure-callout>}}^2+{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="s"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">s</figure-callout>}}^2{\left|H\left(k\right)\right|}^2]}^{-1}---\left(7\right)$

Residual interference power:

${\mathrm{\&epsiv;}}_I\left(k\right)={\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="s"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">s</figure-callout>}}^2{\mathrm{\&sigma;}}_n^4{[{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="n"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">n</figure-callout>}}^2+{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="s"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">s</figure-callout>}}^2{\left|H\left(k\right)\right|}^2]}^{-2}---\left(8\right)$

Residual noise power:

${\mathrm{\&epsiv;}}_N\left(k\right)={\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="n"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">n</figure-callout>}}^2{\mathrm{\&sigma;}}_s^4{\left|H\left(k\right)\right|}^2{[{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="n"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">n</figure-callout>}}^2+{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="s"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">s</figure-callout>}}^2{\left|H\left(k\right)\right|}^2]}^{-2}---\left(9\right)$

To carry out performance evaluation to frequency-domain equalizer below:

When ${\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="n"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">n</figure-callout>}}^2\&RightArrow;0,$ C(k)＝H ^-1(k)，ε _I(k)→0，ε _min(k)＝ε _N(k)→0。Frequency-domain equalizer becomes an inverse filter completely, and ISI and ICI will all be offset, and obviously noise power is more little, and frequency-domain equalizer is unreasonable to be thought, not only exists but actual conditions are noises, but also has suitable intensity.See that below a reverse side situation ought

$C\left(k\right)\&RightArrow;{\mathrm{\&sigma;}}_s^2{H}^{*}\left(k\right),$ ${\mathrm{\&epsiv;}}_{\min }\left(k\right)={\mathrm{\&epsiv;}}_I\left(k\right)\&RightArrow;{\mathrm{\&sigma;}}_s^2,$ ε _N(k)→0。Frequency-domain equalizer deteriorates to a matched filter, and the useful information of output is flooded by ISI and ICI, has correct information hardly, and system can not work normally.It is very big to the primary mold influence of frequency-domain equalizer to draw noise from top analysis.Must be improved it for this reason.

(2) foundation of improved model and analysis

Because the primary mold of frequency-domain equalizer can not all be offset ISI and ICI under noisy situation,, the output of equalizer is become so on the basis of primary mold, introduce a bucking-out system B (k):

Y(k)＝C(k)X(k)+B(k)S(k) (10)

This moment, the mean square error of equalizer was:

$\mathrm{\&epsiv;}\left(k\right)={\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="s"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">s</figure-callout>}}^2{|1-C\left(k\right)H\left(k\right)-B\left(k\right)|}^2+{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="n"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">n</figure-callout>}}^2{\left|C\left(k\right)\right|}^2---\left(11\right)$

Similarly definition ${\mathrm{\&epsiv;}}_I\left(k\right)={\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="s"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">s</figure-callout>}}^2{|1-C\left(k\right)H\left(k\right)-B\left(k\right)|}^2,$ ${\mathrm{\&epsiv;}}_N\left(k\right)={\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="n"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">n</figure-callout>}}^2{\left|C\left(k\right)\right|}^2$ 。According to the MMSE criterion, bucking-out system tap weights coefficient that can be when making mean square error get minimum value:

$B\left(k\right)=1-C\left(k\right)H\left(k\right)={\mathrm{\&sigma;}}_n^2{[{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="n"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">n</figure-callout>}}^2+{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="s"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">s</figure-callout>}}^2{\left|H\left(k\right)\right|}^2]}^{-1}---\left(12\right)$

Utilize and try to achieve in the primary mold $C\left(k\right)={\mathrm{\&sigma;}}_s^2{H}^{*}\left(k\right){[{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="n"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">n</figure-callout>}}^2+{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="s"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">s</figure-callout>}}^2{\left|H\left(k\right)\right|}^2]}^{-1}$ , finally be improved the parameters of model.C (k) and B (k) substitution error function are expressed Shi Kede: ε _I(k)=0, ${\mathrm{\&epsiv;}}_{\min }\left(k\right)={\mathrm{\&epsiv;}}_N\left(k\right)={\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="n"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">n</figure-callout>}}^2{\mathrm{\&sigma;}}_s^4{\left|H\left(k\right)\right|}^2{[{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="n"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">n</figure-callout>}}^2+{\mathrm{\&sigma;}}_s^2{\left|H\left(k\right)\right|}^2]}^{-2}$ 。From The above results as can be known, improved frequency-domain equalizer model is only real best frequency-domain equalizer, it can thoroughly offset ISI and ICI, make that information is zero-decrement to be passed through, the mean square error of output is littler than primary mold, only produced by noise, the performance of improved frequency-domain equalizer model has arrived the limit, and this is by the getable ideal performance of MMSE criterion institute.

To analyze and handle in order further to simplify, the forward direction system and the bucking-out system of equalizer can be changed into:

C(k)＝snrH ^*(k)[1+snr|H(k)| ²] ^-1 (13)

B(k)＝[1+snr|H(k)| ²] ^-1 (14)

Wherein $\mathrm{snr}={\mathrm{\&sigma;}}_s^2/{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="n"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">n</figure-callout>}}^2$ It is system's received signal to noise ratio.Therefore, as long as obtain the frequency domain estimated value of channel impulse response and the estimated value of received signal to noise ratio at the ofdm system receiving terminal, just can the parameters of frequency-domain equalizer is unique definite.This also makes the levels of precision of channel estimating and signal-to-noise ratio (SNR) estimation will directly influence the equalization performance of frequency-domain equalizer.Can adopt the channel estimation method and the signal-to-noise ratio estimation algorithm of suitable different situations according to different system to the difference of performance requirement.

According to above-mentioned invention technical conceive and principle, the present invention adopts following technical proposals:

Frequency domain balancer design method in a kind of ofdm system is characterized in that design procedure is:

1) method for solving of primary mold parameter and result;

2) method for solving and the result of the improved model parameter of primary mold and bucking-out system formation;

3) the final realization expression formula of each model parameter;

4) structural design of improved model;

5) realization of improved model structure in application.

The method for solving of above-mentioned primary mold is: the frequency-region signal that the ofdm system receiver receives be frequency-domain equalizer be input as X (k)=H (k) S (k)+N (k), frequency-domain equalizer is output as Y (k)=C (k) X (k), the mean square error of frequency-domain equalizer $\mathrm{\&epsiv;}\left(k\right)=E\left[{\left|E\left(k\right)\right|}^2\right]={\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="s"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">s</figure-callout>}}^2{|1-C\left(k\right)H\left(k\right)|}^2+{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="n"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">n</figure-callout>}}^2{\left|C\left(k\right)\right|}^2$ , define residual interference power and be ${\mathrm{\&epsiv;}}_I\left(k\right)={\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="s"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">s</figure-callout>}}^2{|1-C\left(k\right)H\left(k\right)|}^2$ , residual noise power is ${\mathrm{\&epsiv;}}_N\left(k\right)={\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="n"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">n</figure-callout>}}^2{\left|C\left(k\right)\right|}^2$ , under the MMSE criterion, the frequency-domain equalizer primary mold tap weights coefficient results of trying to achieve when making mean square error get minimum value is:

$C\left(k\right)={\mathrm{\&sigma;}}_s^2{H}^{*}\left(k\right){[{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="n"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">n</figure-callout>}}^2+{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="s"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">s</figure-callout>}}^2{\left|H\left(k\right)\right|}^2]}^{-1}$

Above-mentioned primary mold with the method for solving of the improved model parameter that bucking-out system constitutes is: identical in the input of frequency-domain equalizer and the primary mold, be output as Y (k)=C (k) X (k)+B (k) S (k), the mean square error of frequency-domain equalizer $\mathrm{\&epsiv;}\left(k\right)=E\left[{\left|E\left(k\right)\right|}^2\right]={\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="s"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">s</figure-callout>}}^2{|1-C\left(k\right)H\left(k\right)-B\left(k\right)|}^2+{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="n"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">n</figure-callout>}}^2{\left|C\left(k\right)\right|}^2$ , under the MMSE criterion, the frequency-domain equalizer bucking-out system tap weights coefficient results of trying to achieve when making mean square error get minimum value is:

$B\left(k\right)=1-C\left(k\right)H\left(k\right)={\mathrm{\&sigma;}}_n^2{[{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="n"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">n</figure-callout>}}^2+{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="s"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">s</figure-callout>}}^2{\left|H\left(k\right)\right|}^2]}^{-1}$

The forward direction system tap weights coefficient results of improved model is identical with primary mold tap weights coefficient C (k).

The final realization expression formula of each above-mentioned model parameter is: with the expression formula further abbreviation of the C of forward direction system (k) with bucking-out system B (k), become the function expression of certain value that can be obtained in the practical application, with the molecule denominator among C (k) and the B (k) simultaneously divided by σ _s ², the expression formula of forward direction system and bucking-out system becomes:

C(k)＝snrH ^*(k)[1+snr|H(k)| ²] ^-1

B(k)＝[1+snr|H(k)| ²] ^-1

Wherein $\mathrm{snr}={\mathrm{\&sigma;}}_s^2/{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="n"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">n</figure-callout>}}^2$ Be system's received signal to noise ratio, can be estimated at the receiving terminal of system that H (k) can be estimated at the receiving terminal of system equally, Shang Mian C (k) just can obtain in reality system with B (k) like this.

The construction design method of above-mentioned improved model is: after FFT handled, through a finite impulse response filter group with several the single taps of FFT point, hard decision device A was sent in the output one tunnel of bank of filters, and one the tunnel sends into one group of adder.The output of hard decision device is again through a finite impulse response filter group with several the single taps of FFT point, the output of this bank of filters is organized the input of adder as above-mentioned this equally, the output of this group adder is sent into hard decision device B, the output of the hard decision device B output of frequency-domain equalizer improved model just; The finite impulse response filter group also can replace with one group of multiplier, and one tunnel input of each multiplier is the circuit-switched data after handling through FFT, and another road input is a tap weights coefficient of former limited impact response filter group; Suppose that counting of FFT is N, the initial model structure of so whole frequency-domain equalizer is made up of finite impulse response filter group and a hard decision device of a N tap, and the improved model structure of whole frequency-domain equalizer is made of jointly finite impulse response filter group, a N adder and two hard decision devices of two N taps.Wherein can replace the inverse filterbank of N tap with N multiplier.

The above-mentioned implementation method of improved model structure in application is: the structure of frequency-domain equalizer is in case determine, so each realization of handling is exactly a tap coefficient of determining the finite impulse response filter group, in fact just estimates the value of channel impulse response value and signal to noise ratio; The algorithm of channel estimating is to adopt least-squares algorithm, obtains channel estimation values of pilot frequency positions by the method for inserting pilot tone in each symbol, and the estimated value of all the other positions utilizes the method for interpolation to obtain; Estimation for signal to noise ratio is to adopt online signal-to-noise ratio (SNR) estimation scheme, this signal-to-noise ratio (SNR) estimation scheme is based on statistics ratio to one group of specific observable reception data and obtains the signal-to-noise ratio (SNR) estimation value, this scheme has been used the SNR estimator of Sa Musi and the inferior design of Weir, one group of specific observable reception data is to have continued to use the pilot data in the channel estimating, and the signal to noise ratio of other position also utilizes the method for interpolation to obtain.

The present invention compared with prior art, having following conspicuous outstanding substantive distinguishing features is remarkable advantage:

The present invention utilizes the MMSE criterion to solve the optimal solution of the forward direction and the bucking-out system tap coefficient of frequency-domain equalizer, determine the parameter of optimal solution and finally obtain a definite numerical value by channel estimating and signal-to-noise ratio estimation algorithm, the numerical expression of optimal solution will be used for upgrading the tap coefficient of frequency-domain equalizer, because the existence of bucking-out system, the final output of equalizer will not have ISI and ICI, noise power also can be inhibited simultaneously, and the performance of system will be improved greatly.Owing to do not need iteration and multi-dimensional matrix computing to make computation complexity become very low, be fit to be applied in the actual communication systems in the frequency-domain equalizer of the present invention.Frequency-domain equalizer among the present invention can solve the situation that CP length causes systematic function to descend rapidly in the ofdm system less than channel impulse response length, therefore, under the channel condition condition of severe or under the nervous situation that needs suitable reduction CP length of frequency spectrum resource, strong point of the present invention will be embodied fully.In other system, compare with other frequency-domain equalizer, the present invention also has the advantage of himself, improves the availability of frequency spectrum thereby for example reduce CP length, can effectively suppress noise, offsets ICI of Doppler frequency shift generation or the like.

Description of drawings

Fig. 1 is the OFDM baseband system block diagram of primary mold of the present invention.

Fig. 2 is the OFDM baseband system block diagram of improved model of the present invention.

Fig. 3 is a structural representation of the present invention.

Fig. 4 is the ofdm system bit error rate performance under the additive white Gaussian channel.

Fig. 5 is the ofdm system bit error rate performance under walking B (Pedestrian-B) channel.

Fig. 6 is the ofdm system bit error rate performance under vehicle-mounted B (Vehicular-B) channel.

Embodiment

Details are as follows in conjunction with the accompanying drawings for a preferred embodiment of the present invention:

As shown in Figure 1, the data that transmit are at first handled through serial, walk abreast/serial process after handling through IFFT then again, and this has just realized the OFDM modulation.Before each the OFDM symbol that has modulated, add cyclic prefix CP, just it can be sent into channel.Data by channel will experience various influences such as channel multi-path, decline, Doppler frequency shift, additive white Gaussian noise.The signal that receives at first is removed CP, handle through serial then and wait for that FFT handles, handle data afterwards through FFT and just realized the OFDM demodulation, the data of pilot frequency locations will be used to carry out channel estimating and signal-to-noise ratio (SNR) estimation in the data after the demodulation, must just obtain in the channel impulse response of pilot frequency locations and the estimated value of signal to noise ratio, the method by interpolation just can obtain other locational channel impulse response and signal-to-noise ratio (SNR) estimation value.Whole data after the demodulation will be as the input of frequency-domain equalizer, and frequency-domain equalizer has two kinds of primary mold and improved models.

As shown in Figure 1, FFT is the primary mold of a frequency-domain equalizer after handling, and FFT shown in Figure 2 is the improved model of a frequency-domain equalizer after handling, and concrete frequency-domain equalizer improved model structure as shown in Figure 3.Parallel data after the OFDM demodulation will be through the forward direction system of frequency-domain equalizer, and the tap weights coefficient that each circuit-switched data will be multiplied by on the frequency-domain equalizer correspondence position is exported then, and the data one tunnel of output enter decision device A, and another road is as one tunnel input of adder.Decision device A is the hard decision device herein, can certainly use the soft-decision device, and its performance is more excellent, but structure is than hard decision device complexity.The output of decision device A is the input of system by way of compensation.What C of forward direction system (k) and decision device A constituted is the primary mold of a frequency-domain equalizer, and the input of bucking-out system B (k) output of primary mold just.The input of adder is formed in the output of the output of bucking-out system and forward direction system jointly, and decision device B is sent in the output of adder, and same decision device B can be that the hard decision device also can be the soft-decision device, and just the thresholding of judgement can change with the signal of input.The output of decision device B is exactly the output of whole frequency-domain equalizer improved model.

The step of the frequency domain balancer design method in this ofdm system is:

1) method for solving of primary mold parameter and result;

2) method for solving and the result of the improved model parameter of primary mold and bucking-out system formation;

3) the final realization expression formula of each model parameter;

4) structural design of improved model;

5) realization of improved model structure in application.

The method for solving of above-mentioned primary mold is: the frequency-region signal that the ofdm system receiver receives be frequency-domain equalizer be input as X (k)=H (k) S (k)+N (k), frequency-domain equalizer is output as Y (k)=C (k) X (k), the mean square error of frequency-domain equalizer $\mathrm{\&epsiv;}\left(k\right)=E\left[{\left|E\left(k\right)\right|}^2\right]={\mathrm{\&sigma;}}_s^2{|1-C\left(k\right)H\left(k\right)|}^2+{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="n"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">n</figure-callout>}}^2{\left|C\left(k\right)\right|}^2$ , define residual interference power and be ${\mathrm{\&epsiv;}}_I\left(k\right)={\mathrm{\&sigma;}}_s^2{|1-C\left(k\right)H\left(k\right)|}^2$ , residual noise power is ${\mathrm{\&epsiv;}}_N\left(k\right)={\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="n"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">n</figure-callout>}}^2{\left|C\left(k\right)\right|}^2$ , under the MMSE criterion, the frequency-domain equalizer primary mold tap weights coefficient results of trying to achieve when making mean square error get minimum value is:

$C\left(k\right)={\mathrm{\&sigma;}}_s^2{H}^{*}\left(k\right){[{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="n"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">n</figure-callout>}}^2+{\mathrm{\&sigma;}}_s^2{\left|H\left(k\right)\right|}^2]}^{-1}$

Above-mentioned primary mold with the method for solving of the improved model parameter that bucking-out system constitutes is: identical in the input of frequency-domain equalizer and the primary mold, be output as Y (k)=C (k) X (k)+B (k) S (k), the mean square error of frequency-domain equalizer $\mathrm{\&epsiv;}\left(k\right)=E\left[{\left|E\left(k\right)\right|}^2\right]={\mathrm{\&sigma;}}_s^2{|1-C\left(k\right)H\left(k\right)-B\left(k\right)|}^2+{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="n"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">n</figure-callout>}}^2{\left|C\left(k\right)\right|}^2$ , under the MMSE criterion, the frequency-domain equalizer bucking-out system tap weights coefficient results of trying to achieve when making mean square error get minimum value is:

$B\left(k\right)=1-C\left(k\right)H\left(k\right)={\mathrm{\&sigma;}}_n^2{[{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="n"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">n</figure-callout>}}^2+{\mathrm{\&sigma;}}_s^2{\left|H\left(k\right)\right|}^2]}^{-1}$

The forward direction system tap weights coefficient results of improved model is identical with primary mold tap weights coefficient C (k).

The final realization expression formula of each above-mentioned model parameter is: with the expression formula further abbreviation of the C of forward direction system (k) with bucking-out system B (k), become the function expression of certain value that can be obtained in the practical application, with the molecule denominator among C (k) and the B (k) simultaneously divided by σ _s ², the expression formula of forward direction system and bucking-out system becomes:

C(k)＝snrH ^*(k)[1+snr|H(k)| ²] ^-1

B(k)＝[1+snr|H(k)| ²] ^-1

Wherein $\mathrm{snr}={\mathrm{\&sigma;}}_s^2/{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="n"\; filenames="A20071004274500141.png,A20071004274500151.png"\; state="\{\{state\}\}">n</figure-callout>}}^2$ Be system's received signal to noise ratio, can be estimated at the receiving terminal of system that H (k) can be estimated at the receiving terminal of system equally, Shang Mian C (k) just can obtain in reality system with B (k) like this.

The construction design method of above-mentioned improved model is: after FFT handled, through a finite impulse response filter group with several the single taps of FFT point, hard decision device A was sent in the output one tunnel of bank of filters, and one the tunnel sends into one group of adder.The output of hard decision device is again through a finite impulse response filter group with several the single taps of FFT point, the output of this bank of filters is organized the input of adder as above-mentioned this equally, the output of this group adder is sent into hard decision device B, the output of the hard decision device B output of frequency-domain equalizer improved model just; The finite impulse response filter group also can replace with one group of multiplier, and one tunnel input of each multiplier is the circuit-switched data after handling through FFT, and another road input is a tap weights coefficient of former limited impact response filter group; Suppose that counting of FFT is N, the initial model structure of so whole frequency-domain equalizer is made up of finite impulse response filter group and a hard decision device of a N tap, and the improved model structure of whole frequency-domain equalizer is made of jointly finite impulse response filter group, a N adder and two hard decision devices of two N taps.Wherein can replace the inverse filterbank of N tap with N multiplier.

The above-mentioned implementation method of improved model structure in application is: the structure of frequency-domain equalizer is in case determine, so each realization of handling is exactly a tap coefficient of determining the finite impulse response filter group, in fact just estimates the value of channel impulse response value and signal to noise ratio; The algorithm of channel estimating is to adopt least-squares algorithm, obtains channel estimation values of pilot frequency positions by the method for inserting pilot tone in each symbol, and the estimated value of all the other positions utilizes the method for interpolation to obtain; Estimation for signal to noise ratio is to adopt online signal-to-noise ratio (SNR) estimation scheme, this signal-to-noise ratio (SNR) estimation scheme is based on statistics ratio to one group of specific observable reception data and obtains the signal-to-noise ratio (SNR) estimation value, this scheme has been used the SNR estimator of Sa Musi and the inferior design of Weir, one group of specific observable reception data is to have continued to use the pilot data in the channel estimating, and the signal to noise ratio of other position also utilizes the method for interpolation to obtain.

Performance simulation result such as Fig. 4, Fig. 5 and shown in Figure 6 relatively of the ofdm system frequency-domain equalizer primary mold of present embodiment, improved model and traditional frequency-domain equalizer.

Claims (6)

1. the frequency domain balancer design method in the ofdm system is characterized in that design procedure is:

1) method for solving of primary mold parameter and result;

2) method for solving and the result of the improved model parameter of primary mold and bucking-out system formation;

3) the final realization expression formula of each model parameter;

4) structural design of improved model;

5) realization of improved model structure in application.

2. the frequency domain balancer design method in the ofdm system according to claim 1, the method for solving that it is characterized in that described primary mold is: the frequency-region signal that the ofdm system receiver receives be frequency-domain equalizer be input as X (k)=H (k) S (k)+N (k), frequency-domain equalizer is output as Y (k)=C (k) X (k), the mean square error of frequency-domain equalizer $\mathrm{\&epsiv;}\left(k\right)=E\left[{\left|E\left(k\right)\right|}^2\right]={\mathrm{\&sigma;}}_s^2{|1-C\left(k\right)H\left(k\right)|}^2+{\mathrm{\&sigma;}}_n^2{\left|C\left(k\right)\right|}^2,$ Defining residual interference power is ${\mathrm{\&epsiv;}}_I\left(k\right)={\mathrm{\&sigma;}}_s^2{|1-C\left(k\right)H\left(k\right)|}^2,$ Residual noise power is ${\mathrm{\&epsiv;}}_N\left(k\right)={\mathrm{\&sigma;}}_n^2{\left|C\left(k\right)\right|}^2,$ Under the MMSE criterion, the frequency-domain equalizer primary mold tap weights coefficient results of trying to achieve when making mean square error get minimum value is:

$C\left(k\right)={\mathrm{\&sigma;}}_s^2{H}^{*}\left(k\right){[{\mathrm{\&sigma;}}_n^2+{\mathrm{\&sigma;}}_s^2{\left|H\left(k\right)\right|}^2]}^{-1}$

3. the frequency domain balancer design method in the ofdm system according to claim 1, the method for solving that it is characterized in that the improved model parameter that described primary mold and bucking-out system constitute is: identical in the input of frequency-domain equalizer and the primary mold, be output as Y (k)=C (k) X (k)+B (k) S (k), the mean square error of frequency-domain equalizer $\mathrm{\&epsiv;}\left(k\right)=E\left[{\left|E\left(k\right)\right|}^2\right]={\mathrm{\&sigma;}}_s^2{|1-C\left(k\right)H\left(k\right)-B\left(k\right)|}^2+{\mathrm{\&sigma;}}_n^2{\left|C\left(k\right)\right|}^2,$ Under the MMSE criterion, the frequency-domain equalizer bucking-out system tap weights coefficient results of trying to achieve when making mean square error get minimum value is:

$B\left(k\right)=1-C\left(k\right)H\left(k\right)={\mathrm{\&sigma;}}_n^2{[{\mathrm{\&sigma;}}_n^2+{\mathrm{\&sigma;}}_s^2{\left|H\left(k\right)\right|}^2]}^{-1}$

The forward direction system tap weights coefficient results of improved model is identical with primary mold tap weights coefficient C (k).

4. the frequency domain balancer design method in the ofdm system according to claim 1, the final realization expression formula that it is characterized in that described each model parameter is: with the expression formula further abbreviation of the C of forward direction system (k) with bucking-out system B (k), become the function expression of certain value that can be obtained in the practical application, with the molecule denominator among C (k) and the B (k) simultaneously divided by σ _s ², the expression formula of forward direction system and bucking-out system becomes:

C(k)＝snrH ^*(k)[1+snr|H(k)| ²] ^-1

B(k)＝[1+snr|H(k)| ²] ^-1

Wherein $\mathrm{snr}={\mathrm{\&sigma;}}_s^2/{\mathrm{\&sigma;}}_n^2$ Be system's received signal to noise ratio, can be estimated at the receiving terminal of system that H (k) can be estimated at the receiving terminal of system equally, Shang Mian C (k) just can obtain in reality system with B (k) like this.

5. the frequency domain balancer design method in the ofdm system according to claim 1, the construction design method that it is characterized in that described improved model is: after FFT handles, through a finite impulse response filter group with several the single taps of FFT point, hard decision device A is sent in the output one tunnel of bank of filters, and one the tunnel sends into one group of adder.The output of hard decision device is again through a finite impulse response filter group with several the single taps of FFT point, the output of this bank of filters is organized the input of adder as above-mentioned this equally, the output of this group adder is sent into hard decision device B, the output of the hard decision device B output of frequency-domain equalizer improved model just; The finite impulse response filter group also can replace with one group of multiplier, and one tunnel input of each multiplier is the circuit-switched data after handling through FFT, and another road input is a tap weights coefficient of former limited impact response filter group; Suppose that counting of FFT is N, the initial model structure of so whole frequency-domain equalizer is made up of finite impulse response filter group and a hard decision device of a N tap, and the improved model structure of whole frequency-domain equalizer is made of jointly finite impulse response filter group, a N adder and two hard decision devices of two N taps.Wherein can replace the inverse filterbank of N tap with N multiplier.

6. the frequency domain balancer design method in the ofdm system according to claim 1, it is characterized in that the implementation method of described improved model structure in application is: the structure of frequency-domain equalizer is in case determine, so each realization of handling is exactly a tap coefficient of determining the finite impulse response filter group, in fact just estimates the value of channel impulse response value and signal to noise ratio; The algorithm of channel estimating is to adopt least-squares algorithm, obtains channel estimation values of pilot frequency positions by the method for inserting pilot tone in each symbol, and the estimated value of all the other positions utilizes the method for interpolation to obtain; Estimation for signal to noise ratio is to adopt online signal-to-noise ratio (SNR) estimation scheme, this signal-to-noise ratio (SNR) estimation scheme is based on statistics ratio to one group of specific observable reception data and obtains the signal-to-noise ratio (SNR) estimation value, this scheme has been used the SNR estimator of Sa Musi and the inferior design of Weir, one group of specific observable reception data is to have continued to use the pilot data in the channel estimating, and the signal to noise ratio of other position also utilizes the method for interpolation to obtain.

Images