CN100403657C - TD-SCDMA system frequency offset compensating method and apparatus based on exercising sequence - Google Patents

Abstract

The present invention discloses a TD-SCDMA system frequency offset compensating method based on an exercising sequence, which comprises a data receiving sequence, a training sequence, a spreading factor, symbols and a cost function, wherein the data receiving sequence is selected; the training sequence is intercepted from the data receiving sequence; the spreading factor which carries out phase compensation to the training sequence at different frequency deviation points and carries out channel estimation and determines the training sequence; the symbols carry out dispreading to the training sequence; the dispread symbols of the training sequence of each frequency deviation point are obtained at each frequency deviation point, and the cost function processes the abnormal dispreading symbols and calculates each frequency point and each user. The function value of the minimum cost is calculated; the optimum frequency point is determined, and the current frequency offset estimating values are obtained. The present invention carries out smooth correction for the frequency offset estimating values of the current time slot; the impulse response of mixed channels is obtained by the channel estimation, spreading codes, scrambling and weighted values, and the frequency deviation compensation is carried out for the impulse response of the mixing channel of each user. The present invention can carry out the frequency offset estimation and the weighted correction, and the base station can rapidly and accurately carry out the dispreading and the demodulation.

Description

TD-SCDMA system frequency deviation compensation method and device based on training sequence

Technical field

The present invention relates to a kind of frequency bias compensation method in radio mobile communication field, particularly a kind of TD-SCDMA system frequency deviation compensation method and device based on training sequence.

Background technology

In mobile communication system, the frequency shift (FS) that Doppler causes can reach the hundreds of hertz, and the maximum frequency difference of being introduced by carrier wave can reach KHz, so receiving terminal must carry out frequency correction.Frequency calibration function is finished by the frequency correction loop, and device commonly used has automatic frequency control (AFC, AutoFrequency Control) and phase-locked loop (PLL, Phase Lock Loop).

The frequency correction loop is to come contrary frequency skew, the frequency shift (FS) that can effectively proofread and correct follow-up bursty data by adjusting local frequency.But this correction can't have any improvement to current reception data.Therefore, need carry out the frequency difference compensating operation to current reception data.

In the prior art scheme, as application number is that 97115151.2 patents of invention disclose a kind of " carrier wave recovers and compensates in the spread spectrum communication system method and device thereof ", it is the data hard decision demodulation of being come out by despreading, again carry out quadriphase PSK (QPSK then, Quadrature Phase Shift Keying) modulation, ask for the bit of readjustment and the phase difference between the unsentenced bit, ask frequency deviation with following formula again:

${\mathrm{cita}}^{\left(k_{\mathrm{VRU}}\right)}=\frac{1}{2M}\underset{i=0}{\overset{M-1}{\mathrm{\Σ}}}(\frac{{\mathrm{cita}}_{2,i}^{\left(k_{\mathrm{VRU}}\right)}}{4.5+i}-\frac{{\mathrm{cita}}_{\mathrm{1,21}-i}^{\left(k_{\mathrm{VRU}}\right)}}{5.5+i})---\left(1\right)$

Following formula can be represented by Fig. 1, wherein:

$\left\{\begin{array}{c}{\mathrm{\θ}}_1={\mathrm{\θ}}_0+\mathrm{\Δf}\·\mathrm{\Δ}{t}_1\\ {\mathrm{\θ}}_2={\mathrm{\θ}}_0+\mathrm{\Δf}\·\mathrm{\Δ}{t}_2\end{array}\right.---\left(2\right)$

$\mathrm{\Δf}=\frac{{\mathrm{\θ}}_0-{\mathrm{\θ}}_1}{\mathrm{\Δ}{t}_1}=\frac{{\mathrm{\θ}}_2-{\mathrm{\θ}}_0}{\mathrm{\Δ}{t}_2}---\left(3\right)$

Here think 4.5+i and 5.5+i approximately equal in (1) formula.

After (1) formula of utilization is tried to achieve frequency deviation, be converted into phase place again and refill each symbol, give the decoding unit of back as soft output information.

In the prior art scheme, when frequency shift (FS) is bigger, insert (TD-SCDMA in TD SDMA, Time Division-Synchronous Code Division Multiple Access) in burst time, the subscriber channel environment is constant in the time of no longer can being assumed to be, and the channel impulse response h (t) that obtains by training sequence no longer is applicable to whole burst.Therefore, after carrying out joint-detection, utilize unknown user data to carry out the hard decision demodulation, obtain compensate of frequency deviation, there are the following problems:

1. when frequency shift (FS) is bigger, at TD-SCDMA in burst time, the subscriber channel environment is constant in the time of no longer can being assumed to be, the channel impulse response h (t) that obtains by training sequence no longer is applicable to whole burst, therefore, the decreased performance of joint-detection, thus error rates of data increases after causing despreading.

Since after the despreading error rates of data increase, thereby cause demodulation performance to descend.

3. because demodulation is declared in employing firmly, inevitably introduced and declared loss firmly.

4. when the integral multiple variation appearred in phase place, this scheme can't effectively compensate.

5. under some error situation, declare back demodulation planisphere firmly and appear at round dot, can bring bigger error.

Therefore, before joint-detection, frequency deviation is revised, can effectively be reduced or avoid the problems referred to above.

Summary of the invention

The objective of the invention is to propose a kind of TD-SCDMA system frequency deviation compensation method and device based on training sequence, this method intercepts out training sequence symbols from the digital spread spectrum signal that receives, carry out despreading, demodulation after adding default frequency deviation, because training sequence symbols is known, therefore, can carry out frequency offset estimating according to the cost function of the training sequence symbols of each frequency.And the method for employing weighting correction, the frequency offset estimating of current time slots is revised, make the base station carry out despreading, demodulation fast and accurately.On digital signal processing platform, use method and the less memory cell simplified and obtain accurate frequency offset estimating.

The object of the present invention is achieved like this:

The invention discloses a kind of TD-SCDMA system frequency deviation compensation method, comprise the steps: based on training sequence

(1) selects a receiving data sequence;

(2) from receiving data sequence, intercept training sequence;

(3) at different frequency difference points training sequence is carried out phase compensation, and carry out channel estimating, determine the spreading factor Q of training sequence;

(4) training sequence is carried out despreading, obtain the symbol after the training sequence despreading of each frequency difference point;

(5) handle unusual despread symbols, calculate each frequency difference point, each user's cost function;

(6) calculate the minimum cost functional value, determine best frequency difference point, obtain the frequency offset estimating value of current time slots;

(7) the frequency offset estimating value of current time slots is smoothly revised;

(8) by receiving data sequence is carried out channel estimating, spread spectrum, scrambling, weighting, obtain the mixed channel impulse response;

(9) utilize the frequency offset estimating value of the level and smooth current time slots of revising of process that compensate of frequency deviation is carried out in the mixed channel impulse response.

The unusual despread symbols of described processing comprises the mismark that identifies after the despreading, and abandons mismark.

Described TD-SCDMA system frequency deviation compensation method based on training sequence, default compensate of frequency deviation value is e ^{-j2 π f} _i, f wherein _i∈ f, the scope of f is [f _Max, f _Max].

Described frequency offset estimating value to current time slots is smoothly revised and is specially: behind the current time slots frequency offset estimating value weighting P, the addition after the level and smooth last time slot frequency offset estimating value weighting of revising (1-P) with preserving obtains level and smooth revised frequency offset estimating value.

The scope of the frequency difference point of described receiving data sequence is [f _Max, f _Max], the frequency difference point is spaced apart $f_d=\frac{{2f}_{\max }}{\mathrm{\λ}},$ λ is for estimating number of times.

The spreading factor Q=128/N of described training sequence, N are the training sequence symbols number.

The cost function of described training sequence is:

$C^{i,k}\left(\mathrm{Re}\right)=\frac{1}{N}\underset{n=1}{\overset{N}{\mathrm{\Σ}}}|\mathrm{Re}\left(d_n^{i,k}\right)-\mathrm{Re}\left({\hat{d}}_n^{i,k}\right)|$

$C^{i,k}\left(\mathrm{Im}\right)=\frac{1}{N}\underset{n=1}{\overset{N}{\mathrm{\Σ}}}|\mathrm{Im}\left(d_n^{i,k}\right)-\mathrm{Im}\left({\hat{d}}_n^{i,k}\right)|$

Wherein, training sequence

${\hat{d}}_n^{i,k}=\{{\hat{d}}_1^{i,k},{\hat{d}}_2^{i,k},\·\·\·,{\hat{d}}_N^{i,k}\},$ i＝1，......，λ，

d _n ^{I, k}The actual value of expression training sequence symbols,

The estimated value of expression training sequence symbols, i is an i frequency traversal point, and k is k user, and n=1......N, N are the training sequence symbols number.

Described level and smooth correction is to adopt forgetting factor smoothly to revise to current time slots frequency offset estimating value with through the frequency offset estimating value of the level and smooth last time slot of revising:

${\stackrel{\‾}{\mathrm{\θ}}}_{\mathrm{correct}}^{k,n}=(1-p){\stackrel{\‾}{\mathrm{\θ}}}_{\mathrm{correct}}^{k,n-1}+p{\mathrm{\θ}}_{\mathrm{correct}}^{k,n}$

Wherein, ${\mathrm{\θ}}_{\mathrm{correct}}^{k,n}=2\mathrm{\π}{f}_{\mathrm{correct}}^k\·t$ Be the frequency offset estimating value of n time slot of k user, t is a slot length, f _Correct ^kBe the best estimate frequency difference of k user's correspondence, θ _Correct ^{K, n-1}Be the level and smooth frequency offset estimating value of revising of warp of n-1 time slot of k user, θ _Correct ^{K, n}Be the level and smooth frequency offset estimating value of revising of warp of n time slot of k user, p is the average forgetting factor, is made as between the 0.75-0.9.

The invention also discloses a kind of TD-SCDMA system frequency deviation compensation device based on training sequence, this device comprises that frequency deviation is preset device, multiplier, despreading module, channel estimator, mismark deletes and calculation cost function module, cost function value compare judging module, frequency deviation estimating modules, the level and smooth correcting module of frequency deviation, demodulation module, mixing impulse response module, joint-detection module, spreading code, scrambler, weights generator, wherein, default device of described frequency deviation and multiplier are used for the training sequence that intercepts from receiving data sequence is carried out the frequency deviation pre-estimation; Described channel estimator carries out channel estimating to the training sequence through the overdeviation pre-estimation; Described despreading module is used to obtain training sequence symbols; Deletion of described mismark and calculation cost function module are used to delete the error training sequence symbol after the despreading, and the calculation cost function; Described cost function value comparison judging module is used for the result after a plurality of frequency deviation pre-estimation despreadings of comparison same subscriber, chooses each user's best frequency deviation pre-estimation; Described frequency deviation estimating modules is used to calculate each user's of current time slots frequency deviation; The level and smooth correcting module of described frequency deviation is used for frequency deviation is carried out smoothing processing, obtains level and smooth back frequency offset estimating value; Described channel estimator, spreading code, scrambler, weights generator and multiplier are used to obtain the mixed channel impulse response; Described multiplier also is used for compensate of frequency deviation is carried out in the mixed channel impulse response, the mixed channel impulse response after being compensated; Described joint-detection module, demodulation module are used for data are carried out despreading, demodulation.

The present invention utilizes the method for frequency deviation traversal, before joint-detection, training sequence is carried out despreading, obtains the cost function of the training sequence symbols of different frequency deviation points, because training sequence symbols is known, therefore, can goes out mismark to despreading and delete processing.Compared with prior art, advantage of the present invention is as follows:

1. made full use of the Given information of training sequence, therefore, can delete processing going out mismark after the despreading;

2. because training sequence symbols is known, only need the cost function of calculation training sequence after the default frequency deviation point despreading of difference, demodulation, do not need to carry out demodulation and declare the loss of having avoided demodulation to declare firmly firmly;

3. owing to before joint-detection, carried out compensate of frequency deviation, therefore, improved the performance of joint-detection;

4. owing to adopt the method for frequency deviation traversal, therefore need only in the frequency deviation of setting, just can effectively compensate frequency deviation;

5. bigger frequency deviation is had compensating action preferably, particularly when phase overturn occurring, the compensate of frequency deviation effect is obvious;

6. can effectively improve the SNR of system (Signal Noise Ratio)-BER (Bit ErrorRatio) performance;

7. because the symbol numbers of training sequence can dynamically be set, the symbolic number scope is: 1～9, so amount of calculation is little, practical;

8. known owing to training sequence symbols, so this scheme has been estimated reference role to SNR.

Description of drawings

Fig. 1 is phase place frequency offset calculation schematic diagram in the existing compensate of frequency deviation scheme;

Fig. 2 is the flow chart of the frequency bias compensation method among the present invention;

Fig. 3 is frequency deviation smoothing processing module declaration figure among the present invention.

Embodiment

In the TD-SCDMA system, no matter in the uplink synchronous process, still in the traffic channel transmission process, the known array effective length is 128chips in a time slot.Can utilize this known array to carry out Frequency Estimation.

If known transmission sequence is:

S＝{s ₁，s ₂，...，s _n} (4)

Through air traffic channel, corresponding receiving terminal data sequence is:

R＝{r ₁，r ₂，...，r _n，...，r _n+W} (5)

According to the agreement of TD-SCDMA, the training sequence that receives at receiving terminal can be equal to data division, is expressed as:

{w _Q ^kd ₁ ^k，ic ₁ ^k，iv ₁，w _Q ^kd ₁ ^k，ic ₂ ^k，iv ₂，...，w _Q ^kd ₁ ^k，ic _Q ^k，iv _Q，w _Q ^kd ₂ ^k，ic ₁ ^k，iv _Q+1，......，w _Q ^kd _16/Q ^k，ic _Q ^k，iv ₁₆，...} (6)

Wherein the symbol value is as follows:

$w_Q^k\∈\{1,j,-1,-j\}---\left(7\right)$

v _i∈{1，j，-1，-j} (8)

c _i∈{1，-1} (9)

Utilize scrambler, spread spectrum principles, sequence can be expressed as

{w _Q ^kd ₁ ^k，i(c ₁ ^k，iv ₁，c ₂ ^k，iv ₂，...，c _Q ^k，iv _Q)，w _Q ^kd ₂ ^k，i(c ₁ ^k，iv _Q+1，...，c _Q ^k，iv _2Q)，...，w _Q ^kd _16/Q ^k，i(c ₁ ^k，iv _16-Q+1，...，c _Q ^k，iv ₁₆)，...} (10)

If the possible range of this receiving sequence frequency difference is [f _Max, f _Max], suppose that carrying out λ time in this scope estimates that then frequency difference is spaced apart $f_d=\frac{{2f}_{\max }}{\mathrm{\λ}},$ The frequency difference point of traversal is

f＝{0，±f _d，±2f _d，..，±f _max} (11)

Default compensate of frequency deviation value is

F wherein _i∈ f, the scope of f is [f _Max, f _Max].

At different frequency difference points training sequence is carried out phase compensation, because training sequence can think that spreading factor is the data sequence of Q (Q=128/N), therefore training sequence is carried out data separates spread spectrum.Can obtain following training data symbol sebolic addressing at each frequency difference point

${\hat{d}}^{i,k}=\{{\hat{d}}_1^{i,k},{\hat{d}}_2^{i,k},{\hat{d}}_N^{i,k}\},i=1,\·\·\·,\mathrm{\λ}---\left(12\right)$

Wherein i is an i frequency traversal point, and k is k user, and N is the training sequence symbols number.Because training symbol sequence is known, the cost function of this training symbol sequence is:

$C^{i,k}\left(\mathrm{Re}\right)=\frac{1}{N}\underset{n=1}{\overset{N}{\mathrm{\Σ}}}|\mathrm{Re}\left(d_n^{i,k}\right)-\mathrm{Re}\left({\hat{d}}_n^{i,k}\right)|---\left(13\right)$

$C^{i,k}\left(\mathrm{Im}\right)=\frac{1}{N}\underset{n=1}{\overset{N}{\mathrm{\Σ}}}|\mathrm{Im}\left(d_n^{i,k}\right)-\mathrm{Im}\left({\hat{d}}_n^{i,k}\right)|---\left(14\right)$

Each compensate of frequency deviation point is carried out traversal search, and its cost function is set C.

$C^k=\{C^{-f_{\mathrm{mxa}},k},\·\·\·,C^{0,k},\·\·\·C^{f_{\max },k}\}---\left(15\right)$

Wherein best compensate of frequency deviation point cost function is:

$\left|\right|C_f^k\left|\right|=\min \left(\right|\left|C^k\right|\left|\right)---\left(16\right)$

The frequency difference of minimum value correspondence point is exactly the best estimate frequency difference of this receiving sequence in the above-mentioned cost function.The frequency offset estimating value of its n time slot is:

${\mathrm{\θ}}_{\mathrm{correct}}^{k,n}=2\mathrm{\π}{f}_{\mathrm{correct}}^k\·t---\left(17\right)$

Wherein t is a slot length, f _Correct ^kIt is the best estimate frequency difference of k user's correspondence.

After estimating the best frequency deviation of current time slots,, need smoothly revise the frequency offset estimating of current time slots for effectively utilizing frequency deviation information in the past and reducing amount of calculation.Adopt the iteration forgetting factor that the frequency offset estimating value of current time slots and former time slot is smoothly revised:

${\stackrel{\‾}{\mathrm{\θ}}}_{\mathrm{correct}}^{k,n}=(1-p){\stackrel{\‾}{\mathrm{\θ}}}_{\mathrm{correct}}^{k,n-1}+p{\mathrm{\θ}}_{\mathrm{correct}}^{k,n}---\left(18\right)$

Wherein P is the average forgetting factor, generally is made as between 1.75～0.9.

After obtaining revised frequency offset estimating, before joint-detection, mixed channel is impacted the corresponding compensate of frequency deviation that carries out.

Concrete implementation step is as follows:

1. the scope of the frequency difference point of selective reception data sequence is [f _Max, f _Max], the frequency difference point is spaced apart $f_d=\frac{{2f}_{\max }}{\mathrm{\λ}},$ λ is for estimating number of times;

2. from receive data, intercept training sequence, at different frequency difference points training sequence is carried out the preset phase compensation, and carry out channel estimating;

3. determine the spreading factor Q of training sequence, Q=128/N;

4. training sequence is carried out despreading, obtain the symbol after the training sequence despreading of each frequency difference point;

Therefore 5. because training sequence symbols is known, can identifies the mismark after the despreading, and abandon mismark;

6. calculate the cost function of symbol after each frequency difference point, each user's the training sequence despreading;

7. calculate the minimum cost functional value, judge best frequency difference point, obtain the frequency offset estimating value of current time slots;

8. the frequency offset estimating value of current time slots is smoothly revised;

9. by receiving data sequence is carried out channel estimating, spread spectrum, scrambling, weighting, obtain the mixed channel impulse response;

10. compensate of frequency deviation is carried out in the impulse response of K user's mixed channel respectively, carry out joint-detection, demodulation after the compensation.

Referring to Fig. 2 is frequency bias compensation method of the present invention and device, and this device comprises that mainly frequency deviation presets device 22, multiplier 23, despreading module 24, channel estimator 25, mismark deletion and calculation cost function module 26, cost function value is judging module 28 relatively, frequency deviation estimating modules 29, the level and smooth correcting module 210 of frequency deviation, demodulation module 214 mixes impulse response module 212, joint-detection module 213, spreading code, scrambler, weights generator 211.

According to design objective and requirement, can obtain the possible range of receiving sequence frequency deviation in advance, set the frequency difference point and be spaced apart $f_d=\frac{{2f}_{\max }}{\mathrm{\λ}},$ λ is for estimating number of times.During reception, can obtain receiving data 21 from antenna end,

From receive data, intercept training sequence, use the training preface of the default 22 pairs of interceptings of device of multiplier 23 and frequency deviation

Row carry out the frequency deviation pre-estimation, adopt 25 pairs of training sequences through the overdeviation pre-estimation of channel estimator to carry out channel estimating, obtain training sequence symbols by despreading module 24; Because training sequence symbols 27 is known, so can delete mismark after the despreadings and calculation cost function by mismark deletion and calculation cost function module 26; By the result after the default despreading of a plurality of frequency deviations of judging module 28 comparison same subscriber, choose each user's best frequency deviation preset 216; Calculate each user's of current time slots frequency deviation 217 by frequency deviation estimating modules 29; Carry out smoothing processing by 210 pairs of frequency deviations of the level and smooth correcting module of frequency deviation, obtain level and smooth back frequency offset estimating value 218.

By channel estimator 25, spreading code, scrambler, weights generator 211 and multiplier 23 obtain mixed channel impulse response 212; With the frequency offset estimating value 218 after level and smooth, carry out compensate of frequency deviation by 23 pairs of mixed channel impulse responses of multiplier 212, the mixed channel impulse response 215 after being compensated; Through joint-detection module 213,214 pairs of data of demodulation module are carried out despreading, demodulation.

Referring to Fig. 3, the frequency deviation smoothing processing is described, the frequency deviation smoothing processing comprises current time slots frequency offset estimation result 31, weighting P34, weighting (1-P) 32, the last time slot frequency offset estimation result 33 of preservation, level and smooth revised frequency offset estimating 36.

After obtaining current time slots frequency offset estimation result 31, it is carried out level and smooth weighted, weights are respectively P34 and (1-P) 32, utilize formula 18, calculate the level and smooth revised frequency offset estimating of current time slots.

In sum, outstanding feature of the present invention is to adopt the method for frequency deviation traversal, takes full advantage of training Sequence will train sequence to be seen as N symbol, introduce cost function, carry out frequency deviation with known symbolic information Estimate, the frequency deviation of current time slots is estimated smoothly to revise, and before joint-detection, frequency deviation is compensated. When fundamentally having avoided after the despreading unknown demodulates information, firmly declare the error of bringing, improved joint-detection Performance and entire system performance, have higher breakthrough.

Claims (9)

1. the TD-SCDMA system frequency deviation compensation method based on training sequence is characterized in that this method comprises the steps:

(1) selects a receiving data sequence;

(2) from receiving data sequence, intercept training sequence;

(3) at different frequency difference points training sequence is carried out phase compensation, and carry out channel estimating, determine the spreading factor Q of training sequence;

(4) training sequence is carried out despreading, obtain the symbol after the training sequence despreading of each frequency difference point;

(5) handle unusual despread symbols, calculate each frequency difference point, each user's cost function;

(6) calculate the minimum cost functional value, determine best frequency difference point, obtain the frequency offset estimating value of current time slots;

(7) the frequency offset estimating value of current time slots is smoothly revised;

(8) by receiving data sequence is carried out channel estimating, spread spectrum, scrambling, weighting, obtain the mixed channel impulse response;

(9) utilize the frequency offset estimating value of the level and smooth current time slots of revising of process that compensate of frequency deviation is carried out in the mixed channel impulse response.

2. the TD-SCDMA system frequency deviation compensation method based on training sequence as claimed in claim 1 is characterized in that, the unusual despread symbols of described processing comprises the mismark that identifies after the despreading, and abandons mismark.

3. the TD-SCDMA system frequency deviation compensation method based on training sequence as claimed in claim 1, it is characterized in that, described frequency offset estimating value to current time slots is smoothly revised and is specially: behind current time slots frequency offset estimating value weighting P, with the addition after the level and smooth last time slot frequency offset estimating value weighting of revising (1-P) of preserving, obtain level and smooth revised frequency offset estimating value, wherein P is the average forgetting factor.

4. the TD-SCDMA system frequency deviation compensation method based on training sequence as claimed in claim 1 is characterized in that the scope of the frequency difference point of described receiving data sequence is [f _Max, f _Max], the frequency difference point is spaced apart $f_d=\frac{2f_{\max }}{\mathrm{\λ}},$ λ is for estimating number of times.

5. the TD-SCDMA system frequency deviation compensation method based on training sequence as claimed in claim 4 is characterized in that, utilizes default compensate of frequency deviation value at different frequency difference points training sequence to be carried out phase compensation, and default compensate of frequency deviation value complex representation is e ^{-j2 π fi}, f wherein _i∈ f, the scope of f is [f _Max, f _Max].

6. the TD-SCDMA system frequency deviation compensation method based on training sequence as claimed in claim 4 is characterized in that the spreading factor Q=128/N of described training sequence, N are the training sequence symbols number.

7. the TD-SCDMA system frequency deviation compensation method based on training sequence as claimed in claim 4 is characterized in that the cost function of described training sequence is:

$C^{i,k}\left(\mathrm{Re}\right)=\frac{1}{N}\underset{n=1}{\overset{N}{\mathrm{\Σ}}}|\mathrm{Re}\left(d_n^{i,k}\right)-\mathrm{Re}\left({\hat{d}}_n^{i,k}\right)|$

$C^{i,k}\left(\mathrm{Im}\right)=\frac{1}{N}\underset{n=1}{\overset{N}{\mathrm{\Σ}}}|\mathrm{Im}\left(d_n^{i,k}\right)-\mathrm{Im}\left({\hat{d}}_n^{i,k}\right)|$

Wherein, training sequence

${\hat{d}}_n^{i,k}=\{{\hat{d}}_1^{i,k},{\hat{d}}_2^{i,k},\·\·\·,{\hat{d}}_N^{i,k}\},i=1,\·\·\·,\mathrm{\λ},$

d _n ^{I, k}The actual value of expression training sequence symbols,

The estimated value of expression training sequence symbols, i is an i frequency traversal point, k is k user, n=1 ..., N, N are the training sequence symbols number.

8. the TD-SCDMA system frequency deviation compensation method based on training sequence as claimed in claim 7, it is characterized in that described level and smooth correction is to adopt forgetting factor smoothly to revise to current time slots frequency offset estimating value with through the frequency offset estimating value of the level and smooth last time slot of revising:

${\stackrel{\‾}{\mathrm{\θ}}}_{\mathrm{correct}}^{k,n}=(1-p){\stackrel{\‾}{\mathrm{\θ}}}_{\mathrm{correct}}^{k,n-1}+p{\mathrm{\θ}}_{\mathrm{correct}}^{k,n}$

Wherein, ${\mathrm{\θ}}_{\mathrm{correct}}^{k,n}=2\mathrm{\π}{f}_{\mathrm{correct}}^k\·t$ Be the frequency offset estimating value of n time slot of k user, t is a slot length, f _Correct ^kBe the best frequency difference point of k user's correspondence, θ _Correct ^{K, n-1}Be the level and smooth frequency offset estimating value of revising of warp of n-1 time slot of k user, θ _Correct ^{K, n}Be the level and smooth frequency offset estimating value of revising of warp of n time slot of k user, p is the average forgetting factor, is made as between the 0.75-0.9.

9. TD-SCDMA system frequency deviation compensation device based on training sequence, it is characterized in that, this device comprises that frequency deviation is preset device, multiplier, despreading module, channel estimator, mismark deletes and calculation cost function module, cost function value compare judging module, frequency deviation estimating modules, the level and smooth correcting module of frequency deviation, demodulation module, joint-detection module, spreading code, scrambler, weights generator, wherein, default device of described frequency deviation and multiplier are used for the training sequence that intercepts from receiving data sequence is carried out the frequency deviation pre-estimation; Described channel estimator carries out channel estimating to the training sequence through the overdeviation pre-estimation; Described despreading module is used to obtain training sequence symbols; Deletion of described mismark and calculation cost function module are used to delete the error training sequence symbol after the despreading, and the calculation cost function; Described cost function value comparison judging module is used for the result after a plurality of frequency deviation pre-estimation despreadings of comparison same subscriber, chooses each user's best frequency deviation pre-estimation; Described frequency deviation estimating modules is used to calculate each user's of current time slots frequency deviation; The level and smooth correcting module of described frequency deviation is used for frequency deviation is carried out smoothing processing, obtains level and smooth back frequency offset estimating value; Described channel estimator, spreading code, scrambler, weights generator and multiplier are used to obtain the mixed channel impulse response; Described multiplier also is used for compensate of frequency deviation is carried out in the mixed channel impulse response, the mixed channel impulse response after being compensated; Described joint-detection module, demodulation module are used for data are carried out despreading, demodulation.

Images