CN101001219A - Fast convergence rate adaptive blind estimation method for characteristic parameter - Google Patents

Abstract

This invention provides a character parameter adaptive blind estimation method with quick convergency speed, which applies a time-varying step to improve the convergency speed of the algorithm and gets the best filter weight coefficient, applies soft judgement weighing to maintain the Rodust of hard judgement errors and channel noises, which is a channel blind estimation method with quicker convergency speed and higher accuracy than the current recursion least square technology and is used in all kinds of communication systems applying OFDM, and in CDMA and TDMS systems, at the same time, such method can be used in all RLS algorithms and its derived algorithms and devices to estimate other character parameters.

Description

The adaptive blind estimation method for characteristic parameter of fast convergence rate

Technical field:

The present invention relates to a kind of adaptive blind estimation method for characteristic parameter of fast convergence rate.Especially relate to wireless mobile channel self adaptation from the estimation method, belong to the digital mobile communication field of using the electromagnetic wave technology, particularly Digital Television, single carrier ofdm communication system, multi-carrier OFDM communication system, wireless lan (wlan) etc. adopt the channel estimation technique in the OFDM modulated digital communication system.Simultaneously, the present invention not only can be used for carrying out channel estimating in code division multiple access (CDMA) and time division multiple access (TDMA) system, and thought of the present invention also can be used for all RLS methods with and the method for deriving in estimate other characteristic parameters in fields such as communication, radar, space flight, remote-control romote-sensing, sonar, image processing, computer vision, biomedical engineering.

Background technology:

In communication system, transmission rate is limited.In the Channel Transmission process, because the influence of multipath, pulse signal is widened, thereby has caused intersymbol interference (ISI).Do not overlap in order to make between the transmission data, the speed of system transmissions just must be restricted.And in multi-carrier mobile communication system, high-speed data-flow be divided into a plurality of branch roads with different subcarriers with low speed transmissions.On each subcarrier, the symbol rate of modulation is lower than channel delay spread, so just overcome ISI.

The most popular multi-transceiver technology is exactly OFDM at present.Channel estimation technique then is one of key technology in the ofdm system, and its success or failure are directly connected to systematic function.Channel estimating is according to adopting Given information whether can be divided into: use the channel estimation methods of protecting channel estimation methods at interval, use pilot tone and the blind Channel Estimation Based of not having prior information.Can adopt and on the subcarrier of OFDM symbol, insert training sequence with certain cycle and realize, also can adopt the mode of on each OFDM symbol, inserting training sequence.The former is called block-type pilot channel estimation, is mainly used in the estimation to slow fading channel.The latter is called the Comb Pilot channel estimating, is mainly used in the estimation of fast fading channel.These two kinds of methods all can be further divided into least square (Least Square, LS) method or least mean-square error (Minimum Mean-Square-Error, MMSE) method etc.Yet these methods have all taken useful bandwidth and have reduced data transmission efficiency.Blind Channel Estimation Based is not owing to using the information that takies system's valuable bandwidth, so certain advantage is arranged in the communication system of high data rate.

In the blind algorithm for estimating of channel, best blind algorithm for estimating must possess following 3 conditions.The first, the convergence of the blind algorithm for estimating of channel must be fast, and promptly channel must accurately estimate in the least possible symbol; The second, the adaptive adjustment capability of algorithm must be very strong so that can follow the trail of channel variation adaptively to satisfy the needs of communication; Three, the computation complexity of blind algorithm and hardware implementation complexity should be lower, to strengthen its practicality.

In adaptive filter system, the RLS algorithm has follow-up control fast owing to adopted the LS criterion to time varying channel.Yet the RLS algorithm also has some intrinsic defectives: in order to reduce the noise in the prediction, when Prediction Parameters trend actual value, the Kalman gain vector in the RLS algorithm approaches 0, and so just may following the tracks of not, upper signal channel changes.Some scholars have successively proposed various improvement algorithms.Such as: the windowing RLS algorithm that index is forgotten, its shortcoming are very responsive to the disturbance of channel and noise; Reinitialize iteration correlation matrix method, the key of this algorithm is how to detect the variation of channel parameter.On this basis, document D.J.Pauk and B.K.Jun, Self-perturbing Recursive Least Squares Algorithm with Fast TrackingCapability, Electrics Letters 12 ^ThMarch 1992,28 (6). the SPRIS algorithm has been proposed, document J.Jiang and R.Cook, Fast parameter tracking RLS algorithm with hign noise immunity, Electronices Letters 22 ^NdOct.1996:32 (22). the MRIS algorithm has been proposed, document Kwong.Scop Eoon, B.E.Jun and D.J.Park, Fast tracking andnoise immunited RLS algorithm based on Kalman fliter, Electronics Letters, 5 ^ThDec.1996:32 (25) has proposed the ISPRLS algorithm based on Kalman filtering.These improved RLS algorithms have follow-up control preferably for the sudden change of channel, and noise is also had certain inhibitory action.When these algorithms are used for channel equalization, when signal to noise ratio is higher, can follow the tracks of channel apace, when signal to noise ratio was relatively poor, the channel sudden change will be submerged in the noise, can't follow the tracks of channel.Document ParkD.J.etal, Fast trucking RLS algorithm using novel variable forgetting factor with unity zone, Electronics letters[J], 1991,9,27 (23): the RLS algorithm of the variable forgetting factor that 2150-2151. provides is based on regular RLS, in the process of following the trail of channel variation, adjust algorithm that forgetting factor derives out and have fast convergence rate, advantage that follow-up control is strong, but require very high the computational accuracy of equipment.

Summary of the invention:

The objective of the invention is: propose a kind of adaptive blind estimation method for characteristic parameter of fast convergence rate, this method is particularly useful for carrying out in the mobile radio system channel estimating.With respect to the channel estimation methods of prior art, this method will improve convergence rate, strengthen adaptive ability and precision of channel estimation, realize simple.

To achieve these goals, the present invention proposes a kind of adaptive blind estimation method for characteristic parameter of fast convergence rate.Its technical scheme is: the constant step-length that will have now in the RLS technology changes by the timely tracking characteristics parameter of Nonlinear Processing, at first allow the filter weight coefficient be a bigger value, by the time after the filter weight coefficient converged to the best weights coefficient, step-length reduced to obtain better estimated performance.Simultaneously, in order to keep the robustness of hard decision errors and interchannel noise, also judgement is weighted, this mainly is to utilize soft weight decisions (Soft Decision weighted the be called for short SDW) method of receiver soft decision information function as weight coefficient.Thought in this patent, has proposed a kind of adaptive blind estimation method for characteristic parameter of fast convergence rate thus, that is: time change step length soft-decision weighted least-squares method (TVCPSDWRLS).

The adaptive blind estimation method for characteristic parameter of a kind of fast convergence rate that the present invention proposes can estimate channel state parameter effectively and be used for demodulation, thereby effectively improve systematic function.

Be that example describes the art of this patent with OFDM mobile radio system model below.In this system, channel estimating model of the present invention and principle are described in detail as follows:

For a sub-carrier number is N _cOfdm system, S _iThe frequency-region signal of k symbol of [k] i piece of expression (block).Suppose in system, not have inter-block-interference (inter-block-interference, IBI).In order to simplify, can omit piece mark i.Time domain OFDM signal s[n] can be expressed as:

$s\left[n\right]=\frac{1}{\sqrt{N}}\underset{k=0}{\overset{N_c-1}{\mathrm{\&Sigma;}}}S\left[k\right]{\mathrm{<figure-callout\; id="2"\; label="e\; j"\; filenames="A20061003273200111.png"\; state="\{\{state\}\}">e</figure-callout>}}^j---\left(1\right)$

Then, length is N _gProtection be filled in the information symbol front at interval, have:

$\tilde{s}\left[n\right]=\mathrm{Gs}{[n+N_c-N_g]}_{N_c},0\&le;n\&le;N_c+N_g-1---\left(2\right)$

In the formula, G is the gain of power amplifier, [n] _NcN is divided by N in expression _cAfter remainder.General hypothesis G=1.

For frequency selective fading channels, we can constant finite impulse response filter h[n of time spent] represent.So received signal can be expressed as:

$\tilde{y}\left[n\right]=\tilde{x}{\left[n\right]}^{*}h\left[n\right]+v\left[n\right],0\&le;n\&le;N_c+N_g-1---\left(3\right)$

In the formula, * represents linear convolution, h[n] be the impulse response of transmitter filter, frequency-selective channel, filter for receiver, v[n] be the zero-mean additive noise.After eliminating Cyclic Prefix, linear convolution just becomes circular convolution (representing with ),

At this moment, have:

y[n]＝x[n]h[n]+v[n]，0≤n≤N _c-1 (4)

With matrix (4) formula of rewriting be:

Y＝Xh+v (5)

When information that receiver does not transmit, at this moment, be exactly blind estimation.

As shown in Figure 4, have:

Y(n)＝b ₁u ₁(n)+b ₂u _s(n)+...+b _Mu _M(n)+w(n) (6)

$\hat{Y}\left(n\right)={\mathrm{\&beta;}}_1{x}_1\left(n\right)+{\mathrm{\&beta;}}_2{x}_2\left(n\right)+...+{\mathrm{\&beta;}}_M{x}_M\left(n\right)---\left(7\right)$

$e\left(n\right)=Y\left(n\right)-\hat{Y}\left(n\right)=Y\left(n\right)-({\mathrm{\&beta;}}_1{x}_1\left(n\right)+{\mathrm{\&beta;}}_2{x}_2\left(n\right)+...+{\mathrm{\&beta;}}_M{x}_M\left(n\right))---\left(8\right)$

Wherein, u _i(n), i=1,2 ..., M is the discrete time random process of real number, x _i(n), i=1,2 ..., M is detected u _i(n), b ₁..., b _{M, β 1 ..., β M ∈ R}, M is a positive integer, and w (n) is that average is zero real number discrete time random process.After the sampling period, we can be expressed in matrix as the data that obtain at observing system N:

Y＝[y(1)，y(2)，...，y(N)] ^T (9)

w＝[w(1)，w(2)，...，w(N)] ^T (10)

e＝[e(1)，e(2)，...，e(N)] ^T (11)

b＝[b ₁，b ₂，...，b _M] ^T (12)

β＝[β ₁，β ₂，...，β _M] ^T (13)

$U=\left[\begin{array}{ccc}{u}_1\left(1\right)& ...& u_M\left(1\right)\\ ...& ...& ...\\ u_1\left(N\right)& ...& u_M\left(N\right)\end{array}\right]---\left(14\right)$

$X=\left[\begin{array}{ccc}{x}_1\left(1\right)& ...& x_M\left(1\right)\\ ...& ...& ...\\ x_1\left(N\right)& ...& x_M\left(N\right)\end{array}\right]---\left(15\right)$

That is:

Y＝Ub+w (16)

e＝Y-Xβ (17)

The definition cost function is:

J(β)＝e ^TRe＝(Y-Xβ) ^TR(Y-Xβ) (18)

Wherein, R is the weight coefficient matrix of the N * N size of a positive definite symmetry, and is main relevant with X and U.Main application and the estimation attribute according to reality of determining to the R value decided.

In order to keep the robustness of hard decision errors and interchannel noise, we are weighted direct judgement.The judgement Linear Estimation device of weighting is that the decision error information or the difference of receiver soft decision information can further be divided into two subclasses of desirable weight decisions and soft weight decisions again according to weighting function.

Theorem 1: suppose to have accurate decision error information, can utilize the weight coefficient of this information like this as the judgement of RLS method.The influence that this coefficient will make noise and mistake in judgment produce reduces.

Suppose θ _iAnd φ _iArgument in the demodulation process process when being soft-decision and hard decision respectively, definition p _iThe normalized value between [0,1] for difference between reflection soft-decision and the hard decision has:

$p_i=1-\frac{|{\mathrm{\&phi;}}_i-{\mathrm{\&theta;}}_i|}{\mathrm{\&pi;}/S}---\left(19\right)$

Wherein, S is alternative number of symbols.

Because channel can be expressed as the tap time-delay of a plurality of time delays, so have:

u(n)＝[u(n)，u(n-1)，...，u(n-M+1)] ^T (20)

X(n)＝[x(n)，x(n-1)，...，x(n-M+1)] ^T (21)

Correspondingly, the weight at moment n need reflect the accuracy of M judgement in the past.So the set of the weight that this is possible is:

a _n＝p _np _n-1…p _nM+1 (22)

The soft weight decisions RLS estimator of theorem 2:(time change step length) the soft weight decisions RLS estimator of time change step length algorithm is determined to formula (27) by formula (23), wherein a _nDetermine by formula (22).

${\widehat{\mathrm{\&beta;}}}_n={\widehat{\mathrm{\&beta;}}}_{n-1}+{\mathrm{\&mu;}}_n{a}_n{H}_n^{-1}X\left(n\right)e\left(n\right)---\left(23\right)$

H _n＝λH _n-1+a _nX(n)X ^T(n) (24)

$e\left(n\right)=Y\left(n\right)-X{\left(n\right)}^T{\widehat{\mathrm{\&beta;}}}_{n-1}---\left(25\right)$

Wherein: μ _n=α _n* μ ₀(26)

${\mathrm{\&alpha;}}_n=C^{\frac{1}{1+a{n}^b}}---\left(27\right)$

In the present invention, this method adopts time change step length can make algorithm the convergence speed faster, and hypothesis has accurate decision error information in this method, can utilize the weight coefficient of this information as the judgement of RLS method like this.The influence that this coefficient will make noise and mistake in judgment produce reduces.

Prove also through the ofdm communication system link simulation, compare that the present invention has fast convergence rate, estimated accuracy height, characteristics that computation complexity is low with other conventional methods.

The present invention is applicable to the system that all adopt ofdm system to modulate, and is particularly useful for the channel estimation technique in the ofdm system.Though technical scheme of the present invention is primarily aimed at ofdm system, but, any engineer with knowledge background such as signal processing, communications can design corresponding channel estimating apparatus at code division multiple access, time division multiple access according to the present invention, and these all should be included among inventive concept and the scope.Simultaneously, the thought of this patent also can be used to adopt the RLS method with and deriving method estimate other characteristic parameters in multiple fields such as communication, radar, space flight, these methods also should be included in thought of the present invention and the scope.

Description of drawings:

Fig. 1 is the art of this patent block diagram.As shown in the figure, the art of this patent comprises 12 modules, and wherein initial value is provided with 7, time change step length structure 8, upgrades matrix construction 9, soft decision information structure 10, control information 11, characteristic parameter and estimate that 12 are this patented technology and routine techniques difference.

Emulation major parameter from Fig. 2 to Fig. 3 is: channel model is the Rummler channel and the wireless mobile channel of aforesaid standard, QPSK modulation, forgetting factor λ=0.99 of RLS algorithm, RLS constant step size mu ₀=0.002, RLS time change step length factor a=0.01, b=1.970, c=5.0.

In emulation, adopt two kinds of different channel circumstances.That is: Rummler channel and wireless mobile Rayleigh channel.The Rummler channel is made up of three multipaths, and wherein preceding two time delay is more close, and therefore, it is that two multipaths are formed that this channel model can be regarded as, that is: a direct-view (LOS) footpath and a reflection footpath.And certain attenuation law is satisfied in each tap of wireless mobile Rayleigh channel response, and this decay can respond with a single pole low-pass filter and describe, and can be expressed as:

G(v)＝A(1-(v/f _m) ²) ^-1/2 (28)

Wherein, A is the decay of tap, and v is a translational speed.f _mFor the 3dB frequency, represent with Doppler frequency sometimes.

Fig. 2 is recurrence least square blind Channel Estimation Based and the art of this patent convergence and channel estimating performance comparison diagram.

For the performance of the art of this patent and prior art relatively, we adopt the absolute difference between real impulse response and its estimated value squared and on average describe.That is:

$\mathrm{Error}=\frac{1}{N_c-N_0-1}\underset{n=N_0}{\overset{N_c}{\mathrm{\&Sigma;}}}{|\mathrm{\&beta;}-{\widehat{\mathrm{\&beta;}}}_n|}^T|\mathrm{\&beta;}-\widehat{\mathrm{\&beta;}}|---\left(29\right)$

Wherein, N ₀Be than a constantly bigger integer of the initial transient response of estimator.

Relatively should scheme as can be known, the art of this patent has improved the convergence rate and the performance for estimating channel of recurrence least square channel estimation methods.Though Fig. 1 only is a simulated example, its conclusion is of universal significance.

Fig. 3 is blind recurrence least square (BRLS) method of estimation, time change step length soft-decision weighting recurrence least square blind (TVCPSDWRLS) method of estimation channel estimation case when difference delay expansion.More as can be seen, when difference postponed expansion, the art of this patent TVCPSDWRLS algorithm performance was excellent especially by this.

Fig. 4 directly adjudicates the channel estimation system block diagram.

Embodiment:

Below by concrete enforcement technical scheme of the present invention is further described.

Concrete steps are:

1, transmitting terminal is sent into the OFDM base band signal modulated, produces protection at interval, and by D/A and formed filter, generation transmits.

2, at receiving terminal, received signal by A/D and low pass filter after, protection is at interval deleted, obtains received signal matrix Y.

Y＝Xh+v (30)

3, setup parameter μ ₀, a, b value, calculate step-length matrix μ _nWherein:

μ _n＝α _n×μ ₀ (31)

${\mathrm{\&alpha;}}_n=C^{\frac{1}{1+{\mathrm{an}}^b}}---\left(32\right)$

4, suppose to have accurate decision error information, can utilize the weight coefficient of this information like this as the judgement of RLS method.The influence that this coefficient will make noise and mistake in judgment produce reduces.

Suppose θ _iAnd φ _iArgument in the demodulation process process when being soft-decision and hard decision respectively, definition p _iThe normalized value between [0,1] for difference between reflection soft-decision and the hard decision has:

$p_i=1-\frac{|{\mathrm{\&phi;}}_i-{\mathrm{\&theta;}}_i|}{\mathrm{\&pi;}/S}---\left(32\right)$

Wherein, S is alternative number of symbols.

Because channel can be expressed as the tap time-delay of a plurality of time delays, so have:

u(n)＝[u(n)，u(n-1)，...，u(n-M+1)] ^T (33)

X(n)＝[x(n)，x(n-1)，...，x(n-M+1)] ^T (34)

Correspondingly, the weight at moment n need reflect the accuracy of M judgement in the past.So the set of the soft-decision weight that this is possible is:

a _n＝p _np _n-1…p _n-M+1 (35)

5, the value of setup parameter λ calculates the renewal matrix H _n

Wherein, for the soft weight decisions RLS method of time change step length, a _nDetermine have by formula (35):

H _n＝λH _n-1+a _nX(n)X ^T(n) (36)

6, setup parameter β ₀Value, calculate error matrix e (n).Wherein:

$e\left(n\right)=Y\left(n\right)-X{\left(n\right)}^T{\widehat{\mathrm{\&beta;}}}_{n-1}---\left(37\right)$

7, by loop iteration, estimate the channel characteristics parameter

${\widehat{\mathrm{\&beta;}}}_n={\widehat{\mathrm{\&beta;}}}_{n-1}+{\mathrm{\&mu;}}_n{a}_n{H}_n^{-1}X\left(n\right)e\left(n\right)---\left(38\right)$

Claims (6)

1, the present invention relates to a kind of adaptive blind estimation method for characteristic parameter of fast convergence rate, it is characterized in that comprising the steps:

Step 1 is for input signal moment vector u (n), and behind unknown feature parameter model D to be estimated, the output signal vector is Y (n), and wherein: W is a noise.

Y(n)＝u(n) ^T·D+W (1)

Step 2 setup parameter μ ₀, a, b value, wherein, α _nBe the nonlinear time-varying parameter, the step-length matrix μ that becomes when calculating _nHave:

μ _n＝α _n×μ ₀ (2)

${\mathrm{\&alpha;}}_n=C^{\frac{1}{1+{\mathrm{an}}^b}}---\left(3\right)$

Step 3 hypothesis has accurate decision error information, can utilize the weight coefficient of this information as the judgement of RLS method like this.The influence that this coefficient will make noise and mistake in judgment produce reduces.

Suppose θ _iAnd φ _iArgument in the demodulation process process when being soft-decision and hard decision respectively, definition p _iThe normalized value between [0,1] for difference between reflection soft-decision and the hard decision has:

$p_i=1-\frac{|{\mathrm{\&phi;}}_i-{\mathrm{\&theta;}}_i|}{\mathrm{\&pi;}/S}---\left(4\right)$

Wherein, S is alternative number of symbols.

Because the unknown characteristics parameter system can be expressed as the tap time-delay of a plurality of time delays, suppose that the transmission data are expressed as u (n), x (n) is the output result of detected transmission data, that is:

u(n)＝[u(n)，u(n-1)，...，u(n-M+1)] ^T (5)

X(n)＝[x(n)，x(n-1)，...，x(n-M+1)] ^T (6)

Correspondingly, the weight at moment n need reflect the accuracy of M judgement in the past.So the set of the weight that this is possible is:

a _n＝p _np _n-1…p _n-M+1 (7)

The value of step 4 setup parameter λ calculates the renewal matrix H _n

H _n＝λH _n-1+a _nX(n)X ^T(n) (8)

Step 5 setup parameter β ₀Value, calculate error matrix e (n).Wherein:

e(n)＝Y(n)-X(n) ^Tβ _n-1 (9)

Step 6 estimates characteristic parameter β by loop iteration _n

${\mathrm{\&beta;}}_n={\mathrm{\&beta;}}_{n-1}+{\mathrm{\&mu;}}_n{a}_n{H}_n^{-1}X\left(n\right)e\left(n\right)---\left(10\right)$

2, said as claim 1, a kind of adaptive blind estimation method for characteristic parameter of fast convergence rate is characterized in that: become when step-length is.Its core concept is: allow the filter weight coefficient be a bigger value earlier, after the filter weight coefficient rapidly converged to the best weights coefficient by the time, step change reduced to obtain better estimated performance.

3, the adaptive blind estimation method for characteristic parameter of a kind of fast convergence rate as claimed in claim 1, it is characterized in that:, adopt the soft weight decisions method of receiver soft decision information function as weight coefficient in order to keep the robustness of hard decision errors and interchannel noise.

4, said as claim 1, a kind of adaptive blind estimation method for characteristic parameter of fast convergence rate is characterized in that: can be used for employing OFDM such as single carrier, multicarrier and protocol of wireless local area network and digital television protocol and carry out channel estimating in the communication system for modulation.

5, said as claim 1, a kind of adaptive blind estimation method for characteristic parameter of fast convergence rate, it is characterized in that: though technical scheme of the present invention is primarily aimed at ofdm system, but, any engineer with knowledge background such as signal processing, communications can design corresponding channel estimating apparatus at code division multiple access, time division multiple access according to the present invention, and these all should be included among inventive concept and the scope.

6, said as claim 1, a kind of adaptive blind estimation method for characteristic parameter of fast convergence rate is characterized in that: the serial of methods that the thought of the time change step length of this patent and ideal decision weighted thought can be used for all existing RLS methods and be derived by the RLS method; Simultaneously, this thought not only can be used for channel estimating, and can be used to relate to the characteristic parameter estimation in multiple fields such as communication, seismic prospecting, sonar, image processing, computer vision, biomedical engineering, vibration engineering, radar, remote-control romote-sensing, space flight.

Images