CN1622539A - Decision feedback and segment iteration based channel estimation method and implementing device thereof - Google Patents

Abstract

The present invention is channel estimation method and device based on deciding feedback and section iteration. The method includes the following steps: processing waveband signal to obtain training sequence field and data field; performing initial channel estimation and threshold post-treatment of the training sequence field to obtain the initial static value of the time interval impact response; detecting the data subfields with the initial static value to obtain demodulated bit sequences for different users; performing the iterating processing of the data subfields with the bit sequences and the initial static value to obtain the optimized estimated value of the channel impact response; performing data detection with the optimized estimated value to obtain the optimized bit sequence. The device includes data receiving and separating unit, initial channel estimating and post-treating unit, sectional channel estimating unit and data detecting unit. The present invention has raised channel estimating accuracy and mobile communication system performance.

Description

Channel estimation methods and implement device thereof based on decision-feedback and segment iteration

Technical field

The present invention relates to channel estimation methods and implement device thereof in the digital mobile communication system, especially relate in a kind of UTRA TDD (Universal Terrestrial Radio Access-Time DivisionDuplex is based on the general land wireless receiving of time division duplex) mobile communication system channel estimation methods and implement device thereof based on training sequence.

Background technology

UTRA TDD system (for example TD-SCDMA system) is a kind of third generation CDMA digital mobile communication system based on time division duplex, wherein relates to many key technologies, as associated detection technique, and intelligent antenna beam shaping technology, synchronous control technique etc.The realization of these technology all is to be based upon on the basis of channel estimating, accuracy of channel estimation has very significant effects to the performance of joint-detection, wave beam forming and Synchronization Control, and therefore good channel estimation technique becomes the good necessary condition of decision UTRA TDD systematic function.

Existing UTRA TDD system channel estimation method is generally Steiner B., the channel estimation methods that article " the Low cost channelestimation in the uplink receiver of CDMA mobile radio systems " lining that people such as BAIER P deliver in periodical " Frequenz " the 47th volume in December, 1993 proposes, promptly adopt the B.Steiner estimator, by training sequence rationally is set, construct the Toepliz matrix of right circulation, thereby replace complicated matrix inversion (pseudoinverse) computing with the FFT and IFFT (fast fourier transform and the inverse fast fourier transform) computing of low-cost.

The B.Steiner estimator that uses in the said method is a kind of channel estimator of low-cost, its estimated accuracy is subjected to the influence of additive noise in the channel, ratio error is bigger mutually with actual value to cause estimated value, influence systematic function, therefore interim at " communication journal " the 23rd volume the 10th in 2002, the Kang Shaoli of Beijing Jiaotong University, people such as Qiu Zhengding publish an article " improvement of low-cost channel estimation methods in the TD-SCDMA system " noise-reduction method based on Threshold detection is proposed, utilize very little operand can curb some noise taps of channel impulse response sequence, thereby partly eliminate the influence that additive noise brings effectively.

Yet above-mentioned two kinds of channel estimation methods all are based on time slot, promptly utilize the middle trained sequence of a time slot burst structure partly to do channel estimating one time, with this secondary channel results estimated this time slot data division are carried out joint-detection then.Because the random process that becomes when weak channel is in the mobile environment, especially when user's high-speed motion (for example speed is greater than per hour 120 kilometers), real channel circumstance changes very big in a time slot (being 675 μ s in as TD-SCDMA), the truth that can not represent channel with the amount of a static state, so handle in the conference (GSPx) at the global signal in April, 2003, proposed a kind of result of adjacent time-slots channel estimating that utilizes in the meeting paper of delivering by Dae Soon Cho and Deuk Su Lyu " ImprovedJoint Channel Estimation Method with Interpolation " and done the UTRA TDD system channel estimation method that approach based on linear interpolation obtains the channel impulse response of relatively accurate time slot edge.But, this method of utilizing adjacent time-slots channel estimating interpolation is the user during for single time slot services and infeasible, the subframe because two time slots that this user is adjacent are separated by (subframe of TD-SCDMA is 5ms), high-speed motion user's channel relevancy is very weak, can not be used as interpolation; Even and if the channel variation of adjacent two time slots of the user of multislot services simple linear change often, thereby linear interpolation still can be introduced many errors.

Summary of the invention

The technical issues that need to address of the present invention provide a kind of channel estimation methods and implement device thereof based on decision-feedback and segment iteration, adopt this method and implement device thereof can improve accuracy of channel estimation greatly, and then improve the performance of whole mobile communication system.

In order to solve the problems of the technologies described above, the invention provides a kind of channel estimation methods based on decision-feedback and segment iteration, it comprises the steps:

(a) band signal that comes from the wireless interface receiving end is carried out digitized processing, the training sequence field and the data field that obtain being separated from each other, wherein data field comprises the plurality of data son field that is separated from each other;

(b) the training sequence field that obtains in the step (a) is carried out initial channel estimation and thresholding reprocessing, obtain the initial static value of this time slot channel impulse response;

(c) utilize the initial static value of the initial channel impulse response that obtains in the step (b) respectively all data subfields that step (a) obtains to be done Data Detection, and carry out hard decision, obtain the bit sequence after each user's demodulation;

(d) utilize the initial static value of the initial channel impulse response that obtains in the bit sequence that obtains in the step (c) and the step (b) respectively each data subfield to be carried out iterative processing, obtain the pairing preferred channels impulse response of each data subfield estimated value;

(e) utilize the pairing preferred channels impulse response of each data subfield estimated value that obtains in the step (d) respectively data subfield separately to be done Data Detection, comprehensively all Data Detection results obtain each user's optimal bit sequence.

Further, this method also has following characteristics, and described step (d) can be further divided into following steps:

(d1) at each data subfield, bit sequence is carried out segmentation, obtain and the corresponding bit subsequence of each data subfield, and each bit subsequence modulated spread spectrum reconstruct chip sequence separately respectively, with the training sequence of this chip sequence, construct Toepliz matrix separately respectively as data subfield;

(d2) determine the iteration gain coefficient of each Toepliz matrix;

(d3) with the initial static value of the initial channel impulse response that obtains in the step (b) initial value as iteration, respectively each data subfield and with the iteration gain coefficient substitution iterative equation of the corresponding Toepliz separately of data subfield matrix, Toepliz matrix, iterate until satisfying stopping criterion for iteration, obtain the preferred channels impulse response of each data subfield correspondence.

Further, this method also has following characteristics: described iterative equation is

h(n+1)＝h(n)+μ·G ^H·[e-G·h(n)]

Wherein, the estimated value of channel impulse response after h (n+1) the expression iteration; H (n) represents current estimated value of channel impulse response respectively; N represents iterations, and primary iteration value h (0) is the initial static value of the initial channel impulse response that obtains in the step (b); E represents data subfield; G represents the Toepliz matrix; μ represents the iteration gain coefficient of G; G ^HThe conjugate transpose of expression G.

Further, this method also has following characteristics, and the span of described iteration gain coefficient satisfies following condition:

$0<\mathrm{\&mu;}<\frac{2}{{\mathrm{\&lambda;}}_{\max }}$

Wherein, λ _MaxBe spatial correlation matrix R _G=G ^HThe eigenvalue of maximum of G.

Further, this method also has following characteristics: the value of described iteration gain coefficient is μ=1/trace (R _G) or μ=2/trace (R _G).

Further, this method also has following characteristics, and described stopping criterion for iteration is:

‖h(n+1)-h(n)‖＜ε

Wherein, the span of ε is $0.001<\frac{\mathrm{\&epsiv;}}{\left|\right|h\left(n\right)\left|\right|}<0.01.$

Further, this method also has following characteristics: the digitized processing in the described step (a) comprises sampling after band signal amplification, filtering, the down-conversion, be converted to the base-band digital received signal through A/D, and according to the time slot burst structure training sequence field and the data field of base-band digital received signal are separated, wherein data field comprises the plurality of data son field that is separated from each other.

Further, this method also has following characteristics: Data Detection adopts the method for joint-detection or adopts the Rake receiver to add the serial or parallel interference cancellation process in the described step (e).

In order to solve the problems of the technologies described above, the present invention also provides a kind of implement device of the channel estimation methods based on decision-feedback and segment iteration, this device includes Data Receiving separative element, initial channel estimation and post-processing unit, segmentation channel estimating unit and Data Detection unit, the Data Receiving separative element is used for the band signal of digitized processing wireless interface receiving end, the training sequence field and the data field that obtain being separated from each other, this data field comprise the plurality of data son field that is separated from each other; Initial channel estimation and post-processing unit are used for the training sequence field is carried out initial channel estimation and thresholding reprocessing, obtain the initial static value of this time slot channel impulse response; The Data Detection unit is used for each data subfield and corresponding preferred channels impulse response estimated value thereof are done Data Detection, and carries out hard decision, obtains bit subsequence and optimal bit sequence after each user's demodulation; The segmentation channel estimating unit is used in conjunction with bit sequence and initial static value each data subfield being carried out iterative processing, obtains the pairing preferred channels impulse response of each data subfield estimated value.

Further, this implement device also has following characteristics: described segmentation channel estimating unit includes the Toepliz matrix and generates subelement, iteration coefficient generation subelement, channel impulse response renewal subelement and iteration termination judgement subelement, the Toepliz matrix generates subelement and is used to receive each bit subsequence, spreading code and scrambling code information, the bit subsequence is carried out spread spectrum reconstruct generate and the corresponding Toepliz matrix of each data subfield; The iteration coefficient produces subelement and receives the information that generates subelement from the Toepliz matrix, generates Toepliz matrix iteration gain coefficient; Channel impulse response upgrades subelement and carries out iteration respectively according to each data subfield and Toepliz matrix, iteration gain coefficient and the current estimated value of channel impulse response corresponding with it, draws the estimated value of channel impulse response after the iteration; The iteration result that subelement is upgraded in the impulse response of iteration termination judgement subelement receive channel judges whether this iteration result satisfies iterated conditional, if satisfy the preferred channels impulse response of then exporting each data subfield correspondence.

Compared with prior art, channel estimation methods and the implement device thereof that the present invention is based on decision-feedback and segment iteration has the following advantages:

The present invention utilizes decision-feedback and interior each son field of data of segment iteration time slot to obtain the preferred channels impulse response estimated value of each data subfield, overcome and adopted the existing channel method of estimation can't follow the tracks of the irregular defective of channel in the time slot that causes owing to user's high-speed motion, thereby greatly reduce channel estimation errors, improve the order of accuarcy of Data Detection and the accuracy of channel estimating, and then improved the performance of whole mobile communication system.

Description of drawings

Fig. 1 is the schematic diagram of conventional structure of time slot in the UTRA TDD mobile communication system;

Fig. 2 is based on the structured flowchart of the channel estimation methods implement device of decision-feedback and segment iteration among the present invention;

Fig. 3 is the refined structure block diagram of segmentation channel estimating unit among Fig. 2;

Fig. 4 is based on the schematic flow sheet of the channel estimation methods of decision-feedback and segment iteration among the present invention;

Fig. 5 is the realization flow figure of segmentation channel estimation module among Fig. 4;

Fig. 6 is the error rate of system simulation performance comparison diagram that adds the threshold judgement post-processing approach among the present invention based on the channel estimation methods of decision-feedback and segment iteration and traditional B.Steiner estimator.

Embodiment

For understanding the present invention in depth, the present invention is described in detail below in conjunction with drawings and the specific embodiments.

Present embodiment is described specific embodiments of the present invention at the base station receiver of TD-SCDMA system (time division multiplexing-synchonism CDMA mobile communication system), similarly, the present invention is applicable to the base station and the terminal of other UTRA TDD systems, does the mobile communication system of channel estimating to such an extent as to can extend to any training sequence that utilizes.

As shown in Figure 2, the implement device based on the channel estimation methods of decision-feedback and segment iteration includes Data Receiving separative element, initial channel estimation and post-processing unit, segmentation channel estimating unit and Data Detection unit among the present invention; Wherein, the Data Receiving separative element is used for the band signal e of digitized processing wireless interface receiving end, the training sequence field e_mid that obtains being separated from each other and comprise the data field e_data of the plurality of data son field that is separated from each other, and training sequence field e_mid and data field e_data are sent to initial channel estimation and post-processing unit, segmentation channel estimating unit and Data Detection unit respectively; Initial channel estimation and post-processing unit are used for training sequence field e_mid is carried out initial channel estimation and thresholding reprocessing, obtain an initial static value h0 of this time slot channel impulse response, and this initial static value h0 is sent to segmentation channel estimating unit and Data Detection unit; The segmentation channel estimating unit is used for the initial static value h0 that comes from initial channel estimation and post-processing unit data field e_data being carried out segment iteration in conjunction with the bit sequence d0 from the Data Detection unit and handles, obtain the pairing preferred channels impulse response of each data subfield estimated value h, and this preferred channels impulse response estimated value h is sent to the Data Detection unit; The Data Detection unit is used for data field e_data and corresponding preferred channels impulse response estimated value h thereof are done Data Detection, and carry out hard decision, obtain after each user's demodulation bit subsequence d0 and at the optimal bit sequence d of preferred channels impulse response estimated value h, and this bit subsequence d0 fed back to the segmentation channel estimating unit, optimal bit sequence d is exported.

As shown in Figure 3, the segmentation channel estimating unit includes the Toepliz matrix and generates subelement, iteration coefficient generation subelement, channel impulse response renewal subelement and iteration termination judgement subelement, and the Toepliz matrix generates subelement and is used to receive each bit subsequence d0 _i, spreading code c and scrambling code information PN, to bit subsequence d0 _iCarrying out spread spectrum reconstruct generates and the corresponding Toepliz matrix of each data subfield G _i, and with this Toepliz matrix G _iBe sent to the iteration coefficient and produce subelement and channel impulse response renewal subelement; The iteration coefficient produces subelement and receives the Toepliz matrix G that generates subelement from the Toepliz matrix _i, generate Toepliz matrix iteration gain coefficient, and send it to channel impulse response renewal subelement; Channel impulse response upgrades subelement according to each data subfield e _iAnd the Toepliz matrix G corresponding with it _i, G _iIteration gain coefficient, initial static value h0, current estimated value of channel impulse response h _i(n) carry out iteration respectively, draw the estimated value of channel impulse response h after the iteration _i(n+1) and send it to iteration termination judgement subelement and judge; The iteration of iteration termination judgement subelement receive channel impulse response renewal subelement is h as a result _i(n+1), judge whether this iteration result satisfies iterated conditional, if satisfy then with the preferred channels impulse response h of this iteration result as each data subfield correspondence _{I, opt}, and, then feed back this iteration h as a result if not with its output _i(n+1) continue iteration as current estimated value of channel impulse response, until satisfying iterated conditional.

As shown in Figure 4, the channel estimation methods based on decision-feedback and segment iteration comprises the steps: among the present invention

Step 1, Data Receiving separative element A amplifies the band signal e that the antenna place receives, filtering, sample after the down-conversion, carry out A/D then and be converted to the base-band digital received signal, and according to the time slot burst structure, training sequence field e_mid and data field e_data are separated, this data field e_data includes the plurality of data son field that is separated from each other, shape is as [e1 e2 ... ei_max], the chip sequence length of the maximum segment number of this data subfield after according to segmentation is not less than 128, and minimum segments is that 2 (being the Data1 and the Data2 of the definition of TD-SCDMA time slot burst structure) are set;

The burst structure of TD-SCDMA as shown in Figure 1,144 chips in the middle of the burst are training sequence Midamble, the data division of each 352 chip lengths of both sides, 16 last chip GP are guard interval;

Step 2, the training sequence field e_mid after the separation delivers to initial channel estimation and post-processing unit carries out initial channel estimation and thresholding reprocessing, obtains an initial static value h0 of this time slot channel impulse response;

Step 3, the initial static value h0 of the initial channel impulse response that Data Detection unit by using step 2 obtains does Data Detection to isolated all data division e_data of step 1 (TD-SCDMA is defined as the Data1 and the Data2 of two 352 chip long data pieces), and carry out hard decision, obtain the bit sequence d0 after each user's the demodulation;

Step 4 is at each data subfield e _i, bit sequence d0 is carried out segmentation, obtain and the corresponding bit subsequence of each data subfield d _i, and all data of buffer memory, initialization i=1; Then, utilize bit subsequence d _i, initial static value h0 and the spreading code c, the scrambler PN that are used for spread spectrum reconstruct be respectively to each data subfield e _iCarry out iterative processing, obtain the pairing preferred channels impulse response of each data subfield ei estimated value h _{I, opt}As shown in Figure 5, this step can further be subdivided into following steps:

Step 41 is at each data subfield e _i, bit sequence d0 is carried out segmentation, obtain and the corresponding bit subsequence of each data subfield d _i, and all data of buffer memory, initialization i=1; Utilize bit subsequence d then _i, spreading code c and scrambler PN modulate spread spectrum and reconstruct corresponding chip sequence, and with the chip sequence of this reconstruct training sequence as this data block correspondence, constructs new Toepliz matrix G _i

Step 42 is determined each Toepliz matrix G _iIteration gain coefficient μ _i, μ _iSpan satisfy following condition:

$0<{\mathrm{\&mu;}}_i<\frac{2}{{\mathrm{\&lambda;}}_{\max }}$

Wherein, λ _MaxBe spatial correlation matrix R _G=G _i ^HG _iEigenvalue of maximum, in order to avoid the process that complicated matrix exgenvalue is found the solution, we can make μ in the practical application _i=1/trace (R _Gi), or adopt μ _i=2/trace (R _Gi) speed of iteration convergence is faster;

Step 43 is with the initial static value h (0) of the initial channel impulse response initial value as iteration, i.e. initialization h _i(0)=h (0), and make n=0, data subfield e _iAnd with the corresponding Toepliz matrix of data subfield G _i, the Toepliz matrix iteration gain coefficient μ _iSubstitution iterative equation, this iterative equation are h _i(n+1)=h _i(n)+μ _iG _i ^H[e _i-G _iH _i(n)], iterate, obtain data subfield e until satisfying stopping criterion for iteration _iCorresponding preferred channels impulse response h _{I, opt}, wherein, stopping criterion for iteration is ‖ h (n+1)-h (n) ‖＜ε, and ε is a very little positive number, and its span can be

$0.001<\frac{\mathrm{\&epsiv;}}{\left|\right|h\left(n\right)\left|\right|}<0.01;$

Step 5, the preferred channels impulse response h of each data subfield of Data Detection unit by using _{I, opt}Carry out Data Detection respectively, obtain corresponding optimal bit subsequence di, opt, and then relatively judge whether the equal iterative detection of all data subfields is finished, judge that promptly whether i is smaller or equal to i_max, if i is less than i_max, buffer memory di then, opt, and i added 1, take out corresponding segment data and carry out iterative detection; If i, then stops circulation, the testing result [d of all data subfields of cascade greater than i_max _{1, opt}d _{2, opt}D _{I_max, opt}], obtain the optimal bit sequence d after each the user's demodulation of this time slot _Opt

Wherein, Data Detection can adopt MUD method (as joint-detection), also can adopt the Rake receiver to add the serial or parallel interference cancellation process.

Because TD-SCDMA time slot burst structure includes two data block Data1 and Data2, therefore data field can be divided into two data son fields, be example with data subfield Data1, further describe and how to construct Toepliz matrix G _i

If the bit sequence of k user's correspondence is d among the data subfield Data1 _k, the symbol sebolic addressing after the corresponding modulating is s _k, spreading code is c _k, scrambler is PN, then can obtain the chip sequence after this user's spectrum-spreading and scrambling in data subfield Data1:

m _k＝s _k(c _k·*PN)

Wherein symbol  represents the Kronecker product in order to spread spectrum, and .* represents dot product in order to scrambling, and the current data block chip sequence (establishing length is L) after k user's reconstruct can constitute the Toepliz matrix:

$G_k=\left[\begin{array}{cccc}{m}_{k,1}& 0& & 0\\ m_{k,2}& m_{k,1}& & 0\\ ...& m_{k,2}& ...& 0\\ m_{k,L}& ...& ...& m_{k,1}\\ 0& m_{k,L}& ...& m_{k,2}\\ 0& 0& ...& ...\\ 0& 0& & m_{k,L}\end{array}\right]$

Thus, the system transmissions equation of this data block can be expressed as:

$e=(\underset{i=1}{\overset{K}{\mathrm{\&Sigma;}}}{G}_i\&CenterDot;h_i)+n=\left[\begin{array}{cccc}{G}_1& G_2& ...& G_K\end{array}\right]\&CenterDot;\left[\begin{array}{c}{h}_1\\ h_2\\ ...\\ h_K\end{array}\right]+n=\mathrm{Gh}+n$

Wherein e is data subfield Data1, and n is an additive noise, and h is the channel impulse response of all users in the notebook data son field, and K represents number of users.

Adopt LMS algorithm (least mean square algorithm) to carry out the estimation of preferred channels impulse response in the present embodiment, similarly, the present invention also can adopt as RLS algorithm (recursive least-squares algorithm), Kalman filter (Kalman filter) and wait other iterative manner to estimate the preferred channels impulse response of current data block in the time slot.

Include in the Data Receiving separative element of implement device beyond radio circuit, intermediate-frequency circuit and FPGA (field programmable gate array) device among the present invention, all processing units are all realized by DSP.

Figure 6 shows that among the present invention and add that based on the channel estimation methods of decision-feedback and segment iteration and traditional B.Steiner estimator the threshold judgement post-processing approach contrasts at the simulation performance aspect the error rate of system, simulated environment is that 5 users, spreading factor are 8 TD-SCDMA system, business is the 12.2k speech business, coded system adopts 1/2 convolution code, channel model is 3GPP (3G (Third Generation) Moblie partnership projects group) test channel case3, user movement speed is 120km/h (promptly per hour 120 kilometers), and receiving terminal adopts joint detection algorithm.

From Fig. 6, can be clearly seen that, to obtain system performance gain about 2dB based on the channel estimation methods of decision-feedback and segment iteration among the present invention, and segments only is 2 in simulation process, be the Data1 and the Data2 part of TD-SCDMA system time gap burst self, if increase the segmentation number, channel estimation methods then of the present invention will obtain more systematic and can gain, promptly have higher channel estimation accuracy, thereby improve the performance of whole mobile communication system.

Claims (10)

1, a kind of channel estimation methods based on decision-feedback and segment iteration is characterized in that, this method comprises the steps:

(a) band signal that comes from the wireless interface receiving end is carried out digitized processing, the training sequence field and the data field that obtain being separated from each other, wherein data field comprises the plurality of data son field that is separated from each other;

(b) the training sequence field that obtains in the step (a) is carried out initial channel estimation and thresholding reprocessing, obtain the initial static value of this time slot channel impulse response;

(c) utilize the initial static value of the initial channel impulse response that obtains in the step (b) respectively all data subfields that step (a) obtains to be done Data Detection, and carry out hard decision, obtain the bit sequence after each user's demodulation;

(d) utilize the initial static value of the initial channel impulse response that obtains in the bit sequence that obtains in the step (c) and the step (b) respectively each data subfield to be carried out iterative processing, obtain the pairing preferred channels impulse response of each data subfield estimated value;

(e) utilize the pairing preferred channels impulse response of each data subfield estimated value that obtains in the step (d) respectively data subfield separately to be done Data Detection, comprehensively all Data Detection results obtain each user's optimal bit sequence.

2, the channel estimation methods based on decision-feedback and segment iteration according to claim 1 is characterized in that, described step (d) can be further divided into following steps:

(d1) at each data subfield, bit sequence is carried out segmentation, obtain and the corresponding bit subsequence of each data subfield, and each bit subsequence modulated spread spectrum reconstruct chip sequence separately respectively, with the training sequence of this chip sequence, construct Toepliz matrix separately respectively as data subfield;

(d2) determine the iteration gain coefficient of each Toepliz matrix;

(d3) with the initial static value of the initial channel impulse response that obtains in the step (b) initial value as iteration, respectively each data subfield and with the iteration gain coefficient substitution iterative equation of the corresponding Toepliz separately of data subfield matrix, Toepliz matrix, iterate until satisfying stopping criterion for iteration, obtain the preferred channels impulse response of each data subfield correspondence.

3, the channel estimation methods based on decision-feedback and segment iteration according to claim 2, it is characterized in that: described iterative equation is

h(n+1)＝h(n)+μ·G ^H·[e-G·h(n)]

Wherein, the estimated value of channel impulse response after h (n+1) the expression iteration; H (n) represents current estimated value of channel impulse response respectively; N represents iterations, and primary iteration value h (0) is the initial static value of the initial channel impulse response that obtains in the step (b); E represents data subfield; G represents the corresponding Toepliz matrix with e; μ represents the iteration gain coefficient of G; G ^HThe conjugate transpose of expression G.

4, the channel estimation methods based on decision-feedback and segment iteration according to claim 3 is characterized in that, the span of described iteration gain coefficient satisfies following condition:

$0<\mathrm{\&mu;}<\frac{2}{{\mathrm{\&lambda;}}_{\max }}$

Wherein, λ _MaxBe spatial correlation matrix R _G=G ^HThe eigenvalue of maximum of G.

5, the channel estimation methods based on decision-feedback and segment iteration according to claim 4 is characterized in that: the value of described iteration gain coefficient is μ=1/trace (R _G) or μ=2/trace (R _G).

6, the channel estimation methods based on decision-feedback and segment iteration according to claim 2 is characterized in that, described stopping criterion for iteration is:

‖h(n+1)-h(n)‖＜ε

Wherein, the span of ε is $0.001<\frac{\mathrm{\&epsiv;}}{\left|\right|h\left(n\right)\left|\right|}<0.01.$

7, the channel estimation methods based on decision-feedback and segment iteration according to claim 1, it is characterized in that: the digitized processing in the described step (a) comprises sampling after band signal amplification, filtering, the down-conversion, be converted to the base-band digital received signal through A/D, and according to the time slot burst structure training sequence field and the data field of base-band digital received signal are separated, wherein data field comprises the plurality of data son field that is separated from each other.

8, the channel estimation methods based on decision-feedback and segment iteration according to claim 1 is characterized in that: Data Detection adopts the method for joint-detection or adopts the Rake receiver to add the serial or parallel interference cancellation process in the described step (e).

9, a kind of implement device of the channel estimation methods based on decision-feedback and segment iteration, it is characterized in that: this device includes Data Receiving separative element, initial channel estimation and post-processing unit, segmentation channel estimating unit and Data Detection unit, the Data Receiving separative element is used for the band signal of digitized processing wireless interface receiving end, the training sequence field and the data field that obtain being separated from each other, this data field comprise the plurality of data son field that is separated from each other; Initial channel estimation and post-processing unit are used for the training sequence field is carried out initial channel estimation and thresholding reprocessing, obtain the initial static value of this time slot channel impulse response; The Data Detection unit is used for each data subfield and corresponding preferred channels impulse response estimated value thereof are done Data Detection, and carries out hard decision, obtains bit subsequence and optimal bit sequence after each user's demodulation; The segmentation channel estimating unit is used in conjunction with bit sequence and initial static value each data subfield being carried out iterative processing, obtains the pairing preferred channels impulse response of each data subfield estimated value.

10, implement device according to claim 9, it is characterized in that: described segmentation channel estimating unit includes the Toepliz matrix and generates subelement, iteration coefficient generation subelement, channel impulse response renewal subelement and iteration termination judgement subelement, the Toepliz matrix generates subelement and is used to receive each bit subsequence, spreading code and scrambling code information, the bit subsequence is carried out spread spectrum reconstruct generate and the corresponding Toepliz matrix of each data subfield; The iteration coefficient produces subelement and receives the information that generates subelement from the Toepliz matrix, generates Toepliz matrix iteration gain coefficient; Channel impulse response upgrades subelement and carries out iteration respectively according to each data subfield and Toepliz matrix, iteration gain coefficient and the current estimated value of channel impulse response corresponding with it, draws the estimated value of channel impulse response after the iteration; The iteration result that subelement is upgraded in the impulse response of iteration termination judgement subelement receive channel judges whether this iteration result satisfies iterated conditional, if satisfy the preferred channels impulse response of then exporting each data subfield correspondence.

Images