CN104639490A - Joint estimation and compensation method for frequency-dependent IQ (In-phase Quadrature) mismatch and channel - Google Patents

Abstract

The invention belongs to wireless communication technology fields, in particular to a kind of method of joint channel estimation and frequency dependence IQ (In-phase Quadrature, IQ) imbalance compensation in wireless communication system. This hair by channel and introduces receiving end frequency dependence IQ imbalance, obtains receiving signal and its frequency-domain expression by sending training sequence; Initiation parameter α R, β R, hI and hQ; Obtain response of the conjugation h* [n] of response Hk and channel h [n] of the channel h [n] on frequency domain on frequency domain Obtain the estimated value of β R, α R, hI=[hI1, hI2, hI3] and hQ=[hQ1, hQ2, hQ3]; These steps of originally transmitted signal are recovered to realize. The unbalanced parameter of IQ and estimation channel that the present invention is separated simultaneously, the IQ imbalance parameter estimated is used to carry out unified compensation as preset parameter, it no longer needs to carry out duplicate parameter Estimation to IQ imbalance parameter, relative to previous IQ imbalance compensation method IQ imbalance is considered mostly as a whole with channel, reduces the expense of system-computed. Meanwhile total algorithm of the invention relates generally to linear operation, complexity is low.

Description

The Combined estimator of a kind of frequency dependence IQ imbalance and channel and compensation method

Technical field

The invention belongs to wireless communication technology field, particularly a kind of method of joint channel estimation and frequency dependence IQ (In-phase Quadrature, IQ) imbalance compensation in wireless communication system.

Background technology

Radio communication needs carrier modulation usually, in view of the imperfection of analogue device in reality, AFE (analog front end) (FE, front-end) homophase and orthogonal (IQ, In-phase Quadrature) two paths of signals can produce the imbalance of signal in the process of modulation /demodulation, namely the amplitude of IQ two-way local oscillation signal is no longer identical, phase difference is also not equal to (IQ is uneven), thus cause systematic function to decline, this is even more serious in the system (as millimeter-wave communication system) that carrier frequency is higher, especially when adopting high order modulation or radio-frequency front-end to adopt the Direct Conversion structure of low cost to reduce costs when high frequency communication system.In general, also the technology of some analog domains is had to can be used to reduce the unbalanced impact of IQ, as wiring technique, use different circuit topological structure etc., but these technology often increase equipment size, power consumption and cost, and can not effectively eliminate IQ imbalance.By contrast, estimate not need to do various balance or compromise as analog domain with compensation to IQ imbalance by Digital Signal Processing at numeric field, have huge advantage.Therefore, in digital baseband, carry out IQ imbalance compensation is necessary and crucial.

The unbalanced reason of IQ is caused to be divided into two kinds: one is and frequency independently IQ imbalance (frequency-independent I/Q mismatch), this imbalance is that the local oscillation signal produced due to frequency mixer inaccurately causes, in frequency mixer, local oscillation signal phase difference is not that 90 ° or amplitude difference cause IQ two paths of signals that orthogonal demodulation can not occur, and in this unbalanced effect and signal, each frequency component is separate.Another kind of IQ imbalance and frequency dependence, namely the unbalanced effect of IQ is different according to the difference of the frequency content in signal, the gap mainly coming from circuit devcie Frequency Response in IQ two-way that this imbalance produces, the difference of such as, filter freguency response in two branch roads, amplifier, digital to analog converter (DAC, Digital-to-Analog Converter) or analog to digital converter (ADC, Analog-to-Digital Converter) export equal, these all can cause I Q two-way that the unmatched situation with frequency dependence occurs.

At present, some common IQ imbalance compensation schemes, are roughly divided into two kinds: blind estimate or non-blind algorithm for estimating.About blind estimate algorithm, such as, with Interference Cancellation (IC, Interference Cancellation) and blind source separating (BSS, Blind Source Separation) based on backoff algorithm, by analyze IQ imbalance IQ imbalance is compensated on the impact of signal statistics.The method is without any need for known array, and also do not need to estimate the uneven parameter of IQ, but usually need a large amount of symbols and longer adaptive iteration process, synchronous signal statistical property is subject to the destruction of multipath.And for non-blind algorithm for estimating, based on signal detection theory, the uneven parameter of IQ also can be estimated by sending known training sequence.Generally can pass through based on system-level algorithm and adopt traditional least square (LS, Least Squares), expectation maximization (EM, Expectation Maximization) etc. criterion, realize estimating accurately and rapidly IQ imbalance and compensating.This compensation scheme is less than blind estimate operand, is easy to realize, and is therefore widely used.But conventional non-blind algorithm for estimating is faced with and depends on perfect channel estimation, have particular requirement to training sequence thus applicability is limited, cannot by uneven for IQ parameter and channel separation be opened or cannot carry out the problems such as effective compensation to frequency dependence IQ imbalance.For appeal problem, a kind of method studying general joint channel estimation and frequency dependence IQ imbalance compensation has important practical significance.

Summary of the invention

The invention provides Combined estimator and the compensation method of a kind of frequency dependence IQ imbalance and channel, specific as follows:

S1, make length be N training sequence x [n] by channel h [n], consider receiving terminal frequency dependence IQ unbalanced introducing, the frequency domain presentation of receiving terminal Received signal strength is: wherein, X _kfor x [n] is through N _sfrequency-region signal after the FFT of point, for the conjugated signal x of x [n] ^*[n] is through N _sfrequency-region signal after the FFT of point, H _kfor channel h [n] is through N _sfrequency domain response after the FFT of point, for the conjugation h of channel h [n] ^*[n] is through N _sfrequency domain response after the FFT of point, 0≤k≤N _s-1, for noise, 0≤N _s≤ N and N _sfor integer, A _k, B _kfor parameter alpha uneven with receiving terminal frequency dependence IQ _r, β _r, h _i=[h _i1, h _i2, h _i3] and h _q=[h _q1, h _q2, h _q3] parameter that is associated, physical relationship expression formula is as follows:

$A_k=\frac{({\mathrm{\α}}_R+{{\mathrm{\β}}_R}^{*})H_{I\·k}+({\mathrm{\α}}_R-{{\mathrm{\β}}_R}^{*})H_{Q\·k}}{2},B_k=\frac{({\mathrm{\β}}_R+{{\mathrm{\α}}_R}^{*})H_{I\·k}+({\mathrm{\β}}_R-{{\mathrm{\α}}_R}^{*})H_{Q\·k}}{2}$

H _ikwith H _qkbe respectively h _iwith h _qn _spoint FFT: $H_{I\·k}=h_{I1}+h_{I2}\exp (-j\frac{2\mathrm{\πk}}{N_s})+h_{I3}\exp (-j\frac{4\mathrm{\πk}}{N_s}),$

$H_{Q\·k}=h_{Q1}+h_{Q2}\exp (-j\frac{2\mathrm{\πk}}{N_s})+h_{Q3}\exp (-j\frac{4\mathrm{\πk}}{N_s})$

Described parameter alpha _r, β _r, h _iand h _qbetween separate, α _rand β _rrepresent the uneven impact to received signal of receiving terminal IQ, h _i=[h _i1, h _i2, h _i3] and h _q=[h _q1, h _q2, h _q3] equivalently represented for considering the one of the uneven frequency dependence of IQ, $j=\sqrt{-1};$

Parameter alpha described in S2, initialization S1 _r, β _r, h _iand h _q, make α _r=1, β _r=0, h _i=[1,0,0], h _q=[1,0,0];

S3, maximal possibility estimation is carried out to channel h [n] described in S1, obtain the response H of channel h [n] on frequency domain _kwith the conjugation h of channel h [n] ^*[n] response on frequency domain

S4, by H described in S3 _kwith substitute into described in S1 in, to β _restimate, obtain β _restimated value;

S5, by β described in S4 _restimated value substitute into obtain α _restimated value;

S6, by H described in S3 _kwith β described in S4 _restimated value and S5 described in α _restimated value substitute into described in S1 in, by the mode of iteration successively to h described in S1 _i=[h _i1, h _i2, h _i3] and h _q=[h _q1, h _q2, h _q3] estimate, obtain h _i=[h _i1, h _i2, h _i3] and h _q=[h _q1, h _q2, h _q3] estimated value;

S7, the unbalanced all parameters of frequency dependence IQ estimation obtained substitute into described in S1 in, estimated channel by maximum-likelihood criterion, the estimated value obtained is as the channel estimating finally determined;

S8, transmission information sequence, through channel and the unbalanced impact of IQ, obtain Received signal strength, utilize the channel estimating finally determined described in S7 to compensate described Received signal strength, recovers not by the signal that receiving terminal IQ imbalance affects the channel estimating that utilization obtains can obtain after removing channel effect namely original transmission signal is recovered.

Further, N described in S1 _s=512.

Further, described in S3, the concrete steps of maximal possibility estimation are carried out to channel h [n] as follows:

S31, channel frequency domain response is carried out Fourier expansion.Make F _{k, n}represent the FFT column vector corresponding to a kth subcarrier, namely (0≤k≤N _s-1,0≤n≤N), h is the column vector representing that channel time domain impacts, and has H _k=F _k ^th, the frequency-region signal then received can be expressed as:

$Y_k=A_k\·X_k\·{F_k}^T\·{h+B}_k\·X_{N_s-k}^{*}\·F_k^T\·h^{*}+{\stackrel{\‾}{W}}_k$

S32, according to maximum-likelihood criterion, the maximum-likelihood estimator obtaining h is $\hat{h}=\underset{h}{\arg \min }\left\{{(Y_k-A_k\·X_k\·{F_k}^T\·{h-B}_k\·X_{N_s-k}^{*}\·F_k^T\·h^{*})}^{*}{{\mathrm{\σ}}_{{\stackrel{\‾}{W}}_k}^2}^{-1}(Y_k-A_k\·X_k\·{F_k}^T\·{h-B}_k\·X_{N_s-k}^{*}\·F_k^T\·h^{*})\right\},$ To described in h and h ^*get differential respectively and result is set to 0, obtaining about h and h ^*equation group be $\left[\begin{array}{cc}{b}_1& c_1\\ {c_1}^{*}& {b_1}^{*}\end{array}\right]\left[\begin{array}{c}h\\ h^{*}\end{array}\right]=\left[\begin{array}{c}{a}_1\\ {a_1}^{*}\end{array}\right],$ Wherein, a ₁, b ₁and c ₁for known quantity;

S33, according to S32, equation group is tried to achieve channel time domain and is impacted the estimated value of h;

S34, calculate the response H of channel h on frequency domain _kwith the conjugation h of channel h ^*response on frequency domain

Further, described in S4 to β _restimate, concrete steps are as follows:

S41, by H described in S3 _kwith substitute into described in S1 in, keep α _r=1, h _i=[1,0,0] and h _qthe value of=[1,0,0] is constant;

S42, according to S1, to have:

$A_k=\frac{({\mathrm{\α}}_R+{{\mathrm{\β}}_R}^{*})H_{I\·k}+({\mathrm{\α}}_R-{{\mathrm{\β}}_R}^{*})H_{Q\·k}}{2}$

$B_k=\frac{({\mathrm{\β}}_R+{{\mathrm{\α}}_R}^{*})H_{I\·k}+({\mathrm{\β}}_R-{{\mathrm{\α}}_R}^{*})H_{Q\·k}}{2}$

$H_{I\·k}=h_{I1}+h_{I2}\exp (-j\frac{2\mathrm{\πk}}{N_s})+h_{I3}\exp (-j\frac{4\mathrm{\πk}}{N_s})$

$H_{Q\·k}=h_{Q1}+h_{Q2}\exp (-j\frac{2\mathrm{\πk}}{N_s})+h_{Q3}\exp (-j\frac{4\mathrm{\πk}}{N_s})$

S43, according to maximum-likelihood criterion, obtain β _rmaximum-likelihood estimator be

${\widehat{\mathrm{\β}}}_R=\underset{{\mathrm{\β}}_R}{\arg \min }\left\{{(Y_k-A_k\·X_k\·H_k-B_k\·X_{N_s-k}^{*}\·H_{N_s-k}^{*})}^{*}{{\mathrm{\σ}}_{{\stackrel{\‾}{W}}_k}^2}^{-1}(Y_k-A_k\·X_k\·H_k-B_k\·X_{N_s-k}^{*}\·H_{N_s-k}^{*})\right\};$

S44, to described in S43 middle β _rand β _r ^*get differential and result is set to 0, obtaining about β _rand β _r ^*equation group;

Solving equations described in S45, simultaneous S44, obtains β _r=(a ₂b ₂ ^*-a ₂ ^*c ₂)/(b ₂b ₂ ^*-c ₂c ₂ ^*), namely obtain β _restimated value.

The invention has the beneficial effects as follows:

The present invention is uneven all applicable with frequency dependence IQ imbalance compensation for frequency independence IQ, h under frequency independence IQ imbalance _i=[1,0,0], h _q=[1,0,0], only needs to consider α _r, β _r.The present invention is applicable to SC-FDE system and ofdm system simultaneously, and algorithm of the present invention based on training sequence, but to training sequence without particular requirement, is applicable to the communication system under many various criterions, has good invention value and practical significance.

The present invention obtains the unbalanced parameter of IQ separated simultaneously and estimates channel, carry out unifying to compensate as preset parameter by the uneven parameter of the IQ estimated, without the need to carrying out the parameter Estimation of repetition again to the uneven parameter of IQ, mostly integrally consider uneven for IQ with channel relative to IQ imbalance compensation method in the past, reduce the expense of system-computed.Meanwhile, total algorithm of the present invention relates generally to linear operation, and complexity is low.

Accompanying drawing explanation

Fig. 1 is present system model structure figure.

Fig. 2 is receiving terminal frequency dependence IQ unbalanced construction figure of the present invention.

Fig. 3 is algorithm flow chart of the present invention.

Fig. 4 is algorithm bit error rate (BER) performance chart of the present invention.

Embodiment

Below in conjunction with embodiment and accompanying drawing, describe technical scheme of the present invention in detail.

Consider that IQ is uneven, the system model of SC-FDE and OFDM as shown in Figure 1.Both flow processs are closely similar, just the difference of FFT module position, but the application purpose of its FFT/IFFT difference causes them to have the difference of essence.For OFDM, IFFT/FFT module, signal is carried out multi-carrier modulation and be converted into parallel transmission, its essence is a kind of modulation.And SC-FDE, FFT and IFFT are carried out successively at receiving terminal, carrying out equalization operation between, is the distortion produced by channel is put in frequency domain make up, and essence is the equilibrium of receiving terminal.

Suppose original data stream be u [n] obtain after serioparallel exchange symbolic blocks u=that length is N [u [nN], u [nN+1] ..., u [(n+1) N-1]] ^Τ, wherein, n=1,2 ...Then, by a square formation Γ, precoding is carried out to described symbolic blocks u: for OFDM, Γ=Ι _n, Ι _nrepresent N rank unit matrix.N rank FFT matrix is represented for SC-FDE, Γ=F, F: wherein 0≤k, n≤N-1.

By frequency-region signal be transformed into time domain: wherein, F ^Ηrepresent N rank IFFT matrix.Before x, intubating length is N _cpcyclic Prefix (Cyclic Prefix, CP) to form length be N _snew sequence: wherein, N _sthe matrix of × N for adding CP matrix, N _s=N+N _cp, 0 _{m × n}represent the null matrix of m × n.After parallel-serial conversion, become sequence of scalars x [n], wherein, n=kN _s+ l-1.

Ignore the unbalanced impact of transmitting terminal IQ, through channel and noise effect, the signal received at receiving terminal is: $r\left[n\right]=x\left[n\right]\⊗h\left[n\right]+w\left[n\right].$

Fig. 2 is receiving terminal frequency dependence IQ unbalanced construction figure of the present invention.

First consider to be subject to the unbalanced interference of receiving terminal frequency independence IQ, then Received signal strength generation distortion becomes: $y\left[n\right]={\mathrm{\α}}_{R}r\left[n\right]+{\mathrm{\β}}_R{r}^{*}\left[n\right]={\mathrm{\α}}_{R}x\left[n\right]\⊗h\left[n\right]+{\mathrm{\β}}_{R}x{\left[n\right]}^{*}\⊗h{\left[n\right]}^{*}+\stackrel{\‾}{w}\left[n\right],$ Wherein, α _r=cos (Δ φ _r)-j ε _rsin (Δ φ _r), β _r=ε _rcos (Δ φ _r)+j sin (Δ φ _r), by α _r≈ 1, β _r≈ 0 is known, variance and σ _w ²approximately equal.

The impact h of the frequency response difference of branch _i, h _qcome equivalently represented, consider the uneven frequency dependence of IQ, Received signal strength is finally: $Y_k=A_k\·X_k\·H_k+B_k\·X_{N_s-k}^{*}\·H_{N_s-k}^{*}+{\stackrel{\‾}{W}}_k,$ Wherein, $A_k=\frac{({\mathrm{\α}}_R+{{\mathrm{\β}}_R}^{*})H_{I\·k}+({\mathrm{\α}}_R-{{\mathrm{\β}}_R}^{*})H_{Q\·k}}{2},$ h _ikand H _qkbe respectively h _iand h _qn _sthe FFT of point. the N of channel h [n] _spoint FFT, i.e. channel frequency domain response (CFR, channel frequency response), k represents a kth subcarrier. h ^*the N of [n] _spoint FFT.In like manner, X _k, with x [n], x respectively ^*the N of [n] _spoint FFT and AWGN vector n _spoint FFT.After IQ imbalance is compensated, frequency domain MMSE equalizing coefficient E _kwith Y _kbe multiplied, just removable channel effect.

As shown in Figure 3, to estimate the uneven parameter of IQ, first consider the situation of unknown parameter in Received signal strength frequency-domain expression, wherein comprise channel frequency domain response H _k, α _r, β _r, h _i=[h _i1, h _i2, h _i3] and h _q=[h _q1, h _q2, h _q3], separate between parameter, consider in actual conditions have α for the uneven parameter of frequency dependence IQ _r≈ 1, β _r≈ 0, h _i≈ [1,0,0], h _q≈ [1,0,0], then might as well using these parameters as known quantity process, now unknown parameter only has channel, and maximum-likelihood criterion thus can be adopted to estimate it, the channel estimating now made is under desirable assumed condition, therefore can not as final channel estimation value.After obtaining channel estimating, be classified as and estimated parameter, selected next one parameter β to be estimated _r, now will parameter β be estimated _ras unknown quantity, all the other parameters substitute in Received signal strength frequency-domain expression as known quantity, then to β _rcarry out maximal possibility estimation, obtain its estimated value, and replace β _rinitial value, then by β _rbe classified as and estimate parameter.By the mode of this iteration, making only to need in Received signal strength frequency-domain expression to estimate parameter is unknown quantity, then estimates it based on maximum-likelihood criterion, until obtain the estimated value of all parameters.Now, the estimated value of all parameters is brought in expression formula, again channel is estimated, make its estimated value more accurate.Really the information sequence part transmitted is needed to carry out IQ imbalance compensation and channel equalization, to recover original transmission signal finally by the uneven parameter of the IQ obtained and channel to received signal.

S1, make length be N training sequence x [n] by channel h [n], consider receiving terminal frequency dependence IQ unbalanced introducing, the frequency domain presentation of receiving terminal Received signal strength is: wherein, X _kfor x [n] is through N _sfrequency-region signal after the FFT of point, for the conjugated signal x of x [n] ^*[n] is through N _sfrequency-region signal after the FFT of point, H _kfor channel h [n] is through N _sfrequency domain response after the FFT of point, for the conjugation h of channel h [n] ^*[n] is through N _sfrequency domain response after the FFT of point, 0≤k≤N _s-1, for noise, 0≤N _s≤ N and N _sfor integer, A _k, B _kfor parameter alpha uneven with receiving terminal frequency dependence IQ _r, β _r, h _i=[h _i1, h _i2, h _i3] and h _q=[h _q1, h _q2, h _q3] parameter that is associated, physical relationship expression formula is as follows:

$A_k=\frac{({\mathrm{\α}}_R+{{\mathrm{\β}}_R}^{*})H_{I\·k}+({\mathrm{\α}}_R-{{\mathrm{\β}}_R}^{*})H_{Q\·k}}{2},B_k=\frac{({\mathrm{\β}}_R+{{\mathrm{\α}}_R}^{*})H_{I\·k}+({\mathrm{\β}}_R-{{\mathrm{\α}}_R}^{*})H_{Q\·k}}{2}$

H _ikwith H _qkbe respectively h _iwith h _qn _spoint FFT: $H_{I\·k}=h_{I1}+h_{I2}\exp (-j\frac{2\mathrm{\πk}}{N_s})+h_{I3}\exp (-j\frac{4\mathrm{\πk}}{N_s}),$

$H_{Q\·k}=h_{Q1}+h_{Q2}\exp (-j\frac{2\mathrm{\πk}}{N_s})+h_{Q3}\exp (-j\frac{4\mathrm{\πk}}{N_s}),$

Described parameter alpha _r, β _r, h _iand h _qbetween separate, α _rand β _rrepresent the uneven impact to received signal of receiving terminal IQ, h _i=[h _i1, h _i2, h _i3] and h _q=[h _q1, h _q2, h _q3] equivalently represented for considering the one of the uneven frequency dependence of IQ, $j=\sqrt{-1};$

Parameter alpha described in S2, initialization S1 _r, β _r, h _iand h _q, make α _r=1, β _r=0, h _i=[1,0,0], h _q=[1,0,0];

S3, maximal possibility estimation is carried out to channel h [n] described in S1, obtain the response H of channel h [n] on frequency domain _kwith the conjugation h of channel h [n] ^*[n] response on frequency domain specific as follows:

S31, for the ease of process, channel frequency domain response is carried out Fourier expansion, then the frequency-region signal received is expressed as wherein, F _{k, n}represent the FFT column vector corresponding to a kth subcarrier, namely (0≤k≤N _s-1,0≤n≤N), h is the column vector representing that channel time domain impacts, and has H _k=F _k ^th, $H_{N_s-k}^{*}=F_k^T\·h^{*};$

S32, according to maximum-likelihood criterion, the maximum-likelihood estimator obtaining h is

$\hat{h}=\underset{h}{\arg \min }\left\{{(Y_k-A_k\·X_k\·{F_k}^T\·{h-B}_k\·X_{N_s-k}^{*}\·F_k^T\·h^{*})}^{*}{{\mathrm{\σ}}_{{\stackrel{\‾}{W}}_k}^2}^{-1}(Y_k-A_k\·X_k\·{F_k}^T\·{h-B}_k\·X_{N_s-k}^{*}\·F_k^T\·h^{*})\right\},$ To described in h and h ^*get differential respectively and result is set to 0, obtaining about h and h ^*equation group be $\left[\begin{array}{cc}{b}_1& c_1\\ {c_1}^{*}& {b_1}^{*}\end{array}\right]\left[\begin{array}{c}h\\ h^{*}\end{array}\right]=\left[\begin{array}{c}{a}_1\\ {a_1}^{*}\end{array}\right],$ Wherein, a ₁, b ₁and c ₁for known quantity;

S33, according to S32, equation group is tried to achieve channel time domain and is impacted the estimated value of h;

S34, calculate the response H of channel h on frequency domain _kwith the conjugation h of channel h ^*response on frequency domain

S4, by H described in S3 _kwith substitute into described in S1 in, to β _restimate, obtain β _restimated value, be specially:

S41, by H described in S3 _kwith substitute into described in S1 in, keep α _r=1, h _i=[1,0,0] and h _qthe value of=[1,0,0] is constant;

S42, according to S1, to have:

$A_k=\frac{({\mathrm{\α}}_R+{{\mathrm{\β}}_R}^{*})H_{I\·k}+({\mathrm{\α}}_R-{{\mathrm{\β}}_R}^{*})H_{Q\·k}}{2}$

$B_k=\frac{({\mathrm{\β}}_R+{{\mathrm{\α}}_R}^{*})H_{I\·k}+({\mathrm{\β}}_R-{{\mathrm{\α}}_R}^{*})H_{Q\·k}}{2}$

$H_{I\·k}=h_{I1}+h_{I2}\exp (-j\frac{2\mathrm{\πk}}{N_s})+h_{I3}\exp (-j\frac{4\mathrm{\πk}}{N_s})$

$H_{Q\·k}=h_{Q1}+h_{Q2}\exp (-j\frac{2\mathrm{\πk}}{N_s})+h_{Q3}\exp (-j\frac{4\mathrm{\πk}}{N_s})$

S43, according to maximum-likelihood criterion, obtain β _rmaximum-likelihood estimator be

${\widehat{\mathrm{\β}}}_R=\underset{{\mathrm{\β}}_R}{\arg \min }\left\{{(Y_k-A_k\·X_k\·H_k-B_k\·X_{N_s-k}^{*}\·H_{N_s-k}^{*})}^{*}{{\mathrm{\σ}}_{{\stackrel{\‾}{W}}_k}^2}^{-1}(Y_k-A_k\·X_k\·H_k-B_k\·X_{N_s-k}^{*}\·H_{N_s-k}^{*})\right\};$

S44, to described in S43 middle β _rand β _r ^*get differential and result is set to 0, obtaining about β _rand β _r ^*equation group;

Solving equations described in S45, simultaneous S44, obtains β _r=(a ₂b ₂ ^*-a ₂ ^*c ₂)/(b ₂b ₂ ^*-c ₂c ₂ ^*), namely obtain β _restimated value.

S5, by β described in S4 _restimated value substitute into obtain α _restimated value;

S6, except current solve for parameter, all estimates of parameters obtained and the initial value also not carrying out the parameter estimated are substituted in the frequency-domain expression of Received signal strength, the estimated value of solve for parameter is obtained by maximum-likelihood criterion, current value is replaced by the estimated value obtained, listed in the estimates of parameters obtained, then from untreated parameter, select next solve for parameter parameter in the same way estimate, until estimate all parameters.Can successively to h by the mode of such iteration _i=[h _i1, h _i2, h _i3] and h _q=[h _q1, h _q2, h _q3] estimate.By H described in S3 _kwith β described in S4 _restimated value and S5 described in α _restimated value substitute into described in S1 in, by the mode of iteration successively to h described in S1 _i=[h _i1, h _i2, h _i3] and h _q=[h _q1, h _q2, h _q3] estimate, obtain h _i=[h _i1, h _i2, h _i3] and h _q=[h _q1, h _q2, h _q3] estimated value;

S7, the unbalanced all parameters of frequency dependence IQ estimation obtained substitute into described in S1 in, estimated channel by maximum-likelihood criterion, the estimated value obtained is as the channel estimating finally determined;

S8, transmission information sequence, through channel and the unbalanced impact of IQ, obtain Received signal strength, utilize the channel estimating finally determined described in S7 to compensate described Received signal strength, recovers not by the signal that receiving terminal IQ imbalance affects described signal ignore noise item.The channel estimating that utilization obtains can obtain after removing channel effect namely original transmission signal is recovered.

Fig. 4 is the algorithm flow using the system model structure of Fig. 1, the IQ imbalance model structure of Fig. 2 and Fig. 3, is applied in concrete communication system, emulates the bit error rate of algorithm of the present invention in SC-FDE system (BER) performance chart obtained.Wherein, Fig. 4 (a) and Fig. 4 (b) are illustrated respectively in different bit signal to noise ratio E in sighting distance (LOS) channel model of IEEE 802.15.ad standard channel definition and non line of sight (NLOS) channel model _b/ N ₀(dB) performance chart.The analogue system of this example belongs to high-frequency high-speed ultra-wideband communication system, its main simulation parameter is: carrier frequency is 60GHz, bit rate is 1.76Gbps, 16QAM modulates, the roll-off factor sending and receive roll-off filter is 0.25, system bandwidth is 2.16GHz, and the uneven parameter of receiving terminal frequency dependence IQ is ε _r=1dB, Δ φ _r=5 ⁰, h _i=[1.02,0.04,0.03], h _q[1.03,0.02,0.012], physical layer frame structure adopts the frame format defined in 802.11ad standard.Lead code is mainly used in interblock interference, automatic growth control, frequency deviation are estimated, synchronous, channel estimating and modulation system represent etc., by short training sequence (STF, Short Training Field) and channel estimation sequence (CEF, Channel Estimation Field) composition.From Fig. 4, we can see, time not to IQ imbalance compensation, and the poor performance of system, and to after IQ imbalance compensation, systematic function is improved clearly.

Claims (4)

1. the uneven Combined estimator with channel of frequency dependence IQ and a compensation method, is characterized in that, comprise the steps:

S1, make length be N training sequence x [n] by channel h [n], consider receiving terminal frequency dependence IQ unbalanced introducing, the frequency domain presentation of receiving terminal Received signal strength is: wherein, X _kfor x [n] is through N _sfrequency-region signal after the FFT of point, for the conjugated signal x of x [n] ^*[n] is through N _sfrequency-region signal after the FFT of point, H _kfor channel h [n] is through N _sfrequency domain response after the FFT of point, for the conjugation h of channel h [n] ^*[n] is through N _sfrequency domain response after the FFT of point, 0≤k≤N _s-1, for noise, 0≤N _s≤ N and N _sfor integer, A _k, B _kfor parameter alpha uneven with receiving terminal frequency dependence IQ _r, β _r, h _i=[h _i1, h _i2, h _i3] and h _q=[h _q1, h _q2, h _q3] parameter that is associated, physical relationship expression formula is as follows:

$A_k=\frac{({\mathrm{\α}}_R+{{\mathrm{\β}}_R}^{*})H_{I\·k}+({\mathrm{\α}}_R-{{\mathrm{\β}}_R}^{*})H_{Q\·k}}{2},$ $B_k=\frac{({\mathrm{\β}}_R+{{\mathrm{\α}}_R}^{*})H_{I\·k}+({\mathrm{\β}}_R-{{\mathrm{\α}}_R}^{*})H_{Q\·k}}{2}$

H _ikwith H _qkbe respectively h _iwith h _qn _spoint FFT: $H_{I\·k}=h_{I1}+h_{I2}\exp (-j\frac{2\mathrm{\πk}}{N_s})+h_{I3}\exp (-j\frac{4\mathrm{\πk}}{N_s}),$ $H_{Q\·k}=h_{Q1}+h_{Q2}\exp (-j\frac{2\mathrm{\πk}}{N_s})+h_{Q3}\exp (-j\frac{4\mathrm{\πk}}{N_s})$

Described parameter alpha _r, β _r, h _iand h _qbetween separate, α _rand β _rrepresent the uneven impact to received signal of receiving terminal IQ, h _i=[h _i1, h _i2, h _i3] and h _q=[h _q1, h _q2, h _q3] equivalently represented for considering the one of the uneven frequency dependence of IQ, $j=\sqrt{-1};$

Parameter alpha described in S2, initialization S1 _r, β _r, h _iand h _q, make α _r=1, β _r=0, h _i=[1,0,0], h _q=[1,0,0];

S3, maximal possibility estimation is carried out to channel h [n] described in S1, obtain the response H of channel h [n] on frequency domain _kwith the conjugation h of channel h [n] ^*[n] response on frequency domain

S4, by H described in S3 _kwith substitute into described in S1 in, to β _restimate, obtain β _restimated value;

S5, by β described in S4 _restimated value substitute into obtain α _restimated value;

S6, by H described in S3 _kwith β described in S4 _restimated value and S5 described in α _restimated value substitute into described in S1 in, by the mode of iteration successively to h described in S1 _i=[h _i1, h _i2, h _i3] and h _q=[h _q1, h _q2, h _q3] estimate, obtain h _i=[h _i1, h _i2, h _i3] and h _q=[h _q1, h _q2, h _q3] estimated value;

S7, the unbalanced all parameters of frequency dependence IQ estimation obtained substitute into described in S1 in, estimated channel by maximum-likelihood criterion, the estimated value obtained is as the channel estimating finally determined;

S8, transmission information sequence, through channel and the unbalanced impact of IQ, obtain Received signal strength, utilize the channel estimating finally determined described in S7 to compensate described Received signal strength, recovers not by the signal that receiving terminal IQ imbalance affects the channel estimating that utilization obtains can obtain after removing channel effect namely original transmission signal is recovered.

2. the uneven Combined estimator with channel of a kind of frequency dependence IQ according to claim 1 and compensation method, is characterized in that: N described in S1 _s=512.

3. the uneven Combined estimator with channel of a kind of frequency dependence IQ according to claim 1 and compensation method, is characterized in that: carry out the concrete steps of maximal possibility estimation to channel h [n] described in S3 as follows:

S31, channel frequency domain response is carried out Fourier expansion, make F _{k, n}represent the FFT column vector corresponding to a kth subcarrier, namely (0≤k≤N _s-1,0≤n≤N), h is the column vector representing that channel time domain impacts, and has H _k=F _k ^th, the frequency-region signal then received can be expressed as: $Y_k=A_k\·X_k\·{F_k}^T\·h+B_k\·X_{N_s-k}^{*}\·F_k^T\·h^{*}+{\stackrel{\‾}{W}}_k$

S32, according to maximum-likelihood criterion, the maximum-likelihood estimator obtaining h is

$\hat{h}=\underset{h}{\arg \min }\left\{{(Y_k-A_k\·X_k\·{F_k}^T\·h-B_k\·X_{N_s-k}^{*}\·F_k^T\·h^{*})}^{*}{\mathrm{\σ}}_{{\stackrel{\‾}{W}}_k}^{2-1}(Y_k-A_k\·X_k\·{F_k}^T\·h-B_k\·X_{N_s-k}^{*}\·F_k^T\·h^{*})\right\},$ To described in h and h ^*get differential respectively and result is set to 0, obtaining about h and h ^*equation group be $\left[\begin{array}{cc}{b}_1& c_1\\ {c_1}^{*}& {b_1}^{*}\end{array}\right]\left[\begin{array}{c}h\\ h^{*}\end{array}\right]=\left[\begin{array}{c}{a}_1\\ {a_1}^{*}\end{array}\right],$ Wherein, a ₁, b ₁and c ₁for known quantity;

S33, according to S32, equation group is tried to achieve channel time domain and is impacted the estimated value of h;

S34, calculate the response H of channel h on frequency domain _kwith the conjugation h of channel h ^*response on frequency domain

4. the uneven Combined estimator with channel of a kind of frequency dependence IQ according to claim 1 and compensation method, is characterized in that: to β described in S4 _restimate, concrete steps are as follows:

S41, by H described in S3 _kwith substitute into described in S1 in, keep α _r=1, h _i=[1,0,0] and h _qthe value of=[1,0,0] is constant;

S42, according to S1, to have:

$A_k=\frac{({\mathrm{\α}}_R+{{\mathrm{\β}}_R}^{*})H_{I\·k}+({\mathrm{\α}}_R-{{\mathrm{\β}}_R}^{*})H_{Q\·k}}{2}$

$B_k=\frac{({\mathrm{\β}}_R+{{\mathrm{\α}}_R}^{*})H_{I\·k}+({\mathrm{\β}}_R-{{\mathrm{\α}}_R}^{*})H_{Q\·k}}{2}$

$H_{I\·k}=h_{I1}+h_{I2}\exp (-j\frac{2\mathrm{\πk}}{N_s})+h_{I3}\exp (-j\frac{4\mathrm{\πk}}{N_s}),$

$H_{Q\·k}=h_{Q1}+h_{Q2}\exp (-j\frac{2\mathrm{\πk}}{N_s})+h_{Q3}\exp (-j\frac{4\mathrm{\πk}}{N_s})$

S43, according to maximum-likelihood criterion, obtain β _rmaximum-likelihood estimator be

$\hat{h}=\underset{h}{\arg \min }\left\{{(Y_k-A_k\·X_k\·{F_k}^T\·h-B_k\·X_{N_s-k}^{*}\·F_k^T\·h^{*})}^{*}{\mathrm{\σ}}_{{\stackrel{\‾}{W}}_k}^{2-1}(Y_k-A_k\·X_k\·{F_k}^T\·h-B_k\·X_{N_s-k}^{*}\·F_k^T\·h^{*})\right\},$

S44, to described in S43 middle β _rand β _r ^*get differential and result is set to 0, obtaining about β _rand β _r ^*equation group;

Solving equations described in S45, simultaneous S44, obtains β _r=(a ₂b ₂ ^*-a ₂ ^*c ₂)/(b ₂b ₂ ^*-c ₂c ₂ ^*), namely obtain β _restimated value.