CN106656873A - ACO-OFDM channel estimation method based on superimposed training sequence - Google Patents

Abstract

The invention discloses an ACO-OFDM channel estimation method based on a superimposed training sequence. The method comprises the following steps: S1, processing a pseudo noise sequence to get a uni-polarity nonnegative periodic real sequence; S2, generating a local matrix; and S3, calculating required parameter conditions, and completing channel estimation according to the channel estimation method. The method of the invention is of low implementation complexity and high accuracy, and has very high application value.

Description

ACO-OFDM channel estimation methods based on overlying training sequence

Technical field

The invention belongs to Information and Communication Engineering technical field, the training sequence being related in a kind of light wireless communication signal transacting Row superposition and the method for channel estimation process, specifically a kind of channel based on overlying training sequence that can be used for light wireless communication Method of estimation.

Background technology

At present, various radio communication service sustainable growths, wireless frequency spectrum demand is increasing, will to the utilization rate of frequency range Seek also more and more higher.However, the problem is not solved well, and it is on the rise.To reduce frequency spectrum resource pressure, transmission Mode aspect employs orthogonal frequency division multiplexi (OFDM).This is one of multicarrier parallel transmission modulation system.And, because in fact Existing complexity is relatively low, therefore most widely used.From the medium angle being wirelessly transferred, light wireless communication meets the tendency of as a new technology And give birth to.The key technology research such as modulation system corresponding, to light wireless communication, the method for synchronization and channel estimation there has also been quickly Development.

OFDM modulation techniques have a many advantages, such as channel utilization is high, anti-interference, Decay Rate is strong and the scope of application is larger, It is wider etc..But need to intensity modulation in the present invention.Because light intensity can not be negative value, and OFDM technology can only be adjusted to bipolar signal System.For the problem, OFDM technology need to be improved, occur in that the asymmetric amplitude limit light-OFDM (ACO- of multicarrier OFDM) technology.The technology is suitable in light wireless communication the modulated process for only having unipolar signal.And it also has transceiver architecture Simply, modulation depth it is high and by non-linear effects it is little the advantages of.So, how to solve more effectively will be double in modulating system Polar signal is converted to non-negative unipolar signal, is one of key technology of ACO-OFDM modulation.

Channel estimation technique is one of key technology in light wireless communication, and the technology it will be assumed from receiving data The process that certain channel model parameters is estimated, can be divided into blind Channel Estimation, pilot-symbol aided channel estimation and based on superposition instruction Practice channel estimation of sequence etc..And because ACO-OFDM systems have special to the vector for being input into inverse fast Fourier transform (IFFT) Require, make to realize more difficulty within the system based on the channel estimation methods of pilot aided.Now, using based on superposition training The channel estimation methods of sequence are more effectively.But because light intensity signal is all on the occasion of information sequence average is not zero.So, it is right After reception signal first-order statistics are average, it may appear that the problem that can not be extracted the training sequence being superimposed.

The content of the invention

The present invention provides a kind of efficient, stable modulation system for the wireless light communication of existing high speed development and channel is estimated Meter method --- the ACO-OFDM channel estimation methods based on overlying training sequence, to promote optical information more accurate, fast Transmit fastly, can be widely used in living so as to wireless light communication technology, produce and the digital communicating field of aspect such as scientific research in.

The present invention proposes a kind of asymmetric amplitude limit light-OFDM ACO- of multicarrier based on overlying training sequence OFDM channel estimation method, it mainly includes the non-negative cycle reality sequence generating method of unipolarity, the local square for training sequence The generation method and channel estimation methods of battle array, specifically according to the following steps：

Step one, need to process pseudo noise sequence to obtain the non-negative cycle reality sequence of unipolarity.

Wherein, the step will be related to the symmetrical conjugation to pseudo noise sequence, IFFT, parallel serial conversion, amplitude limit and split combination Etc. process, mainly pseudo noise sequence is changed into and can be used for base band transmission and arithmetic number sequence easily detached with data-signal.

Step 2, generates local matrix.

Preferably, by the time-domain signal analysis to ACO-OFDM systems, with Q × Q dimension real number squares that element is all real number b B and circular matrix T' is constituting local matrix T for battle array.The value of Q is the periodic quantity of generation cycle arithmetic number sequence, and T' is by training The circular matrix of Q × Q dimensions of Sequence composition, and T is the real number matrix of Q × Q dimensions, b is time-domain signal s_cThe average of (n).

Step 3, as necessary desired parameters condition is calculated, by above-mentioned channel estimation methods, completes channel estimation.

In addition, in channel estimation emulation experiment, the present invention adds number of subcarriers and power allocation factor two controllable Factor, and channel estimation effect is evaluated with mean square error (MSE), make whole system that there is adjustability and integrality.

Preferably, step one, by pseudo noise sequence (PN sequences) complex vector, then Jing inverse fast Fourier transforms are generated (IFFT) it is changed into sequence of real numbers, sequence ultimately generates cycle arithmetic number sequence after being repeated several times.

Preferably, step one passes sequentially through following steps realization：

PN sequence constellation mapping and parallel serial conversion of step 1.1. to a length of Q, and the value of Q can be the number of 4 integral multiple； Now, gained sequence length be a quarter of former PN sequences, i.e., complex vector p of a length of Q/4, and p be 1 × Q/4 dimension Vector；

Jing step 1.1 gained complex vectors p are made conjugation symmetry transformation by step 1.2., obtain vectorial p'；Vectorial p' only has odd number Item carrying information, and even item is all zero, i.e., when variable m is even number, there is p'(m)=0；Because system signal needs base band to pass It is defeated, therefore need to be symmetrical by complex signal conjugation, then it is changed into sequence of real numbers Jing after IFFT process, then transmitted；And if only if p' to When measurer has conjugate symmetry, IFFT process could be performed, otherwise, repeat step 1.2 makes generation vector have conjugation symmetrical Property；

Step 1.3. makees the vectorial p' in step 1.2 after IFFT process, to be changed into sequence of real numbers；It is performed successively again Parallel serial conversion, amplitude limiting processing, and amplitude limiting processing mainly makes the element value of sequence of real numbers be not less than zero, you can length is generated for Q's Unipolarity nonnegative real number sequence t'(n)；In order to possess periodically, in above-mentioned generation sequence of real numbers t'(n) on the basis of, repeated N_QIt is secondary, N_QFor natural number, non-negative cycle sequence of real numbers t (n) of unipolarity is obtained；And t (n) is a 1 × QN_QDimension, cycle are Q, Element value be all on the occasion of sequence of real numbers；N is natural number, and value is (1, QN_Q) scope integer.

Preferably, step 2 is realized using following steps：

The generating process of step 2.1. real number matrix B；Real number matrix refers to that all elements in matrix are all real number, matrix B is the full real number symmetrical matrix of Q × Q dimension, and Q is the periodic quantity of training sequence in step 1.3；By the orthogonal frequency of input Multiplex technique (OFDM) baseband signal is expressed as：S (n), and s (n) meets the normal distribution that average is zero, i.e. s (n)～N (0, σ²), its average is 0, and variance is σ²；Make amplitude limiting processing to s (n), that is, limit each signal amplitude and be not less than zero, obtain time domain letter Number：s_c(n)；

Step 2.2. is for step 2.1 gained sequence s_c(n) and s (n), because s (n) meets the normal distribution that average is zero, Obtain s_cN () average is：OrderIt is all real number matrix B of b that Q × Q ranks element can be constructed, that is, have：

In real number matrix B that step 2.3. is generated in step 2.2, a circular matrix is superimposed, you can complete local square Battle array builds, and training sequence t (n) that the circular matrix is generated with step 1.3 has as matrix element：

Step 2.4. is added step 2.2 with step 2.3 gained real number matrix B and circular matrix T', obtains local matrix T, Have：

Preferably, the channel estimation methods of step 3 obtain sending signal x (n) and receive signal r by time domain discrete model N (), will receive signal and is divided into Q roads, successively to its Q times of down-sampling, every road is averaged respectively afterwards, obtain quantized result y (n)； Again to its unbiased esti-mator, with reference to local inverse of a matrix matrix, channel estimation process is completed.

Preferably, step 3 is completed using following steps：

Sending signal x (n) of the step 3.1. system is the time-domain signal s of gained in step 2.1_cIn (n) and step 1.3 The sum of training sequence t (n) for obtaining, its expression formula is as follows：

X (n)=s_c(n)+t(n) (4)

After sending signal x (n) channel in step 3.2. step 3.1, the signal that receiving terminal is received is its Jing light multipath After multipath tolerant, each summation for receiving component of signal, its expression formula is as follows：

Wherein, l represents the number of path of light path；H (l) represents the channel impulse response coefficient of l paths；W (n) is represented Additive white Gaussian noise, its feature meets the Gaussian Profile that average is zero for amplitude；And power spectral density satisfaction is uniformly distributed, its Average is only relevant with its frequency span with variance；

Reception signal r (n) of step 3.2 gained is divided into Q roads by step 3.3., i.e., successively to its Q times of down-sampling；It is right afterwards Average respectively per road, must receive signal quantization result is：Y (n)=E [r (kQ+n)].Wherein, k is that value is (0, N_Q) scope Integer, Q be step 1.3 in training sequence periodic quantity；

Step 3.4., can with reference to sending signal x (n) of gained in step 3.1,3.2 and the expression formula of reception signal r (n) Quantized result y (n) of calculation procedure 3.3 is：

Wherein, N_QFor the sequence repetition number described in step 1.3；

Step 3.5. is additive white Gaussian noise because of noise w (n), meets the normal distribution that average is zero according to its amplitude special Point, and by the s of step 2.2 gained_cN () average is the result of b, then the calculating of y (n) can be reduced in step 3.4：

Gained y (n) of step 3.6. solution procedure 3.5；During and if only if Q=L, L represents channel exponent number, and its coefficient matrix During full rank, the formula has unique solution；Otherwise, parameter need to be readjusted so as to unique solution；Because channel exponent number is only estimation The value upper limit, then can make cycle Q be equal to the higher limit that channel exponent number can be obtained；Finally, step 3.5 gained y (n) expression formula can be by Vector is calculated and is expressed as：

Y=Th (8)

Wherein, T is the local matrix constructed in step 2.4；Y and h are the column vectors of Q × 1 dimension.Th is local square Battle array and the multiplication calculating process of channel coefficient vector；

Step 3.7. has steady ergodic because of the quantized sequences in step 3.6；Sequences y is averaged computing, knot Fruit is unrelated with time parameter, therefore the unbiased esti-mator that can calculate the sequences y is：

Wherein, N_QFor the sequence repetition number described in step 1.3, r (n) is the reception signal of step 3.2 gained.

Step 3.8. calculates its inverse matrix T by the local matrix T of step 2.4 gained^-1；With reference to gained sequence in step 3.7 The unbiased esti-mator result of yFinal channel estimation results h can be completed by following formula；Each element in vectorial h is this and estimates Each channel model coefficient obtained by meter method

The present invention is started with by the modulated process principle to ACO-OFDM systems, analyzes the biography of signal in the system channel Defeated feature, using the channel estimation process that the system is completed based on the method for overlying training sequence, having constructed one kind can apply In the channel estimation methods of wireless light communication.The inventive method implementation complexity is relatively low and accuracy is higher, with higher application Value.

For the aforementioned technical problem that prior art is present, the present invention is based on the channel estimation methods of overlying training sequence, Reasonable generation has been carried out to training sequence and local matrix, its not only can solve receiving end signal first-order statistics it is average after can not be by The superposition detached situation of sequence, and its total algorithm is simple, estimated accuracy is high.In addition, added training sequence will not individually occupy Time slot or channel, with higher system transfers efficiency, and the also flexibility with bandwidth, time and power distribution.Therefore it is in nothing There is significant application value the linear light communications field.

Description of the drawings

Fig. 1 is the principle assumption diagram of ACO-OFDM systems in the present invention.

Fig. 2 is the ACO-OFDM system time domain discrete models in the present invention based on overlying training sequence.

Fig. 3 is the process schematic that the non-negative cycle reality sequence of unipolarity is generated in the present invention.

Fig. 4 is local matrix generating process schematic diagram in the present invention.

Fig. 5 is channel estimation method process schematic in the present invention.

Fig. 6 is y (n) sequence unbiased esti-mator process schematics during channel estimation method.

Fig. 7 is circular matrix and real number matrix additive process schematic diagram in the present invention.

Fig. 8 is local matrix in the present invention and channel coefficient vector multiplication calculating process schematic diagram.

Fig. 9 is the graph of a relation of power allocation factor of the present invention and channel estimation mean square error (MSE) performance.

Specific embodiment

Below by way of specific embodiment and combine accompanying drawing the present invention is described in further detail.

ACO-OFDM channel estimation methods based on overlying training sequence provided by the present invention can be applicable to be pretended with light For in the radio optical communication system for modulating object, however it is not limited to the field that following examples are explained in detail.

The present invention sequentially passes through following key steps and is achieved：

Using existing orthogonal frequency division multiplexi (OFDM) modulator approach, the characteristics of according to using light intensity as modulation object, It is developed and is transformed.Construct the ACO-OFDM modulating systems that can be used for light wireless communication.It is determined that after modulating system, being The channel estimation to the system is more preferably completed, the time domain discrete model of the system is established.And then signal Analysis are defeated from being input to The concrete path for going out and processing procedure, so as to obtain the preliminary pass between input signal x (n), channel h (l) and output signal r (n) System.Wherein, in order to ensure the Stability and veracity of channel estimation.In the present invention, to the training sequence for superposition and solution The local matrix of channel parameter has been made to improve and innovation.Training is used as by the non-negative cycle reality sequence of pseudo noise sequence construction unipolarity Sequence, the mode being added with real number matrix B by circular matrix T' constructs local matrix.Finally, dock collection of letters quantification treatment and obtain y N (), obtains to its abbreviation deformation and estimation computingAnd with reference to local inverse of a matrix matrix, complete the system channel estimation.By reality Test simulation result to know：The inventive method has preferable application prospect in ACO-OFDM systems, is that the signal transmission of wireless light communication is carried A kind of feasible efficient channel estimation methods are supplied.

The specific embodiment of the present invention, can pass sequentially through following examples accompanying drawing to describe in detail.

Fig. 1 is the principle assumption diagram of ACO-OFDM systems.As shown in figure 1, ACO-OFDM modulating systems realize principle with OFDM is similar to.First, to the binary sequence constellation mapping being input into, it is translated into complex signal.Secondly, to the complex signal addition Conjugate number constitutes according to this Hermitian symmetrically, and Hermitian symmetrically refers to that the conjugation of sequence, transposed sequence are all the spy of itself Property.Now, the signal value of front N/2 composition conjugate relation corresponding with the signal value of rear N/2 head and the tail, N is the length of input signal.It is right For using light intensity as the ACO-OFDM systems of modulation object, signal input inverse fast Fourier transform (IFFT) before processing is necessary Possesses conjugate symmetry.Again to its serial to parallel conversion after, can be to its IFFT process.Now, signal is then bipolarity sequence of real numbers.So Afterwards, to its asymmetric amplitude limit, threshold judgement and frequency domain filtering, you can obtain baseband signal.Finally, then to gained signal addition CP simultaneously Complete parallel serial conversion, here it is in the present invention ACO-OFDM systems robust modulation process.

Fig. 2 is based on the ACO-OFDM system time domain discrete models of overlying training sequence.The concrete detailed description of the model Process and calculating process of the time-domain signal from baseband signal s (n) to output sequence r (n).Wherein, add at first adder Unipolarity non-negative cycle reality sequence, i.e. training sequence t (n) for entering, will specifically introduce its generating process at Fig. 3.By the figure Can obtain sending sequence, the relational expression between channel and receiving sequence, the ginseng needed for estimating channel information calculating process can be made Said conditions.

Fig. 3 is training sequence, i.e. the non-negative cycle reality sequence generation process schematic diagram of unipolarity.From the figure 3, it may be seen that training sequence It is based on pseudo noise sequence, to generate through a series of signal processing mode.First, pseudo noise sequence (PN sequences) needs Jing Constellation mapping and parallel serial conversion are changed into complex sequences p.Now, to the complex sequences conjugation symmetrical treatment, vectorial p' is obtained.The vector only has Odd term carrying information, and even item is all zero.Secondly, IFFT process is made to vectorial p' so as to be changed into sequence of real numbers.It is right again It performs successively parallel serial conversion, amplitude limiting processing, you can obtain a unipolarity Nonnegative real sequence t'(n).Finally, in order that training sequence Row have periodically, then it repeatedly can be obtained non-negative cycle reality sequence t (n) of required unipolarity.

Fig. 4 is the local matrix generating process schematic diagram for being applied to the system.From the local square in figure, in the present invention Battle array T is constituted after being added by circular matrix T' and real number matrix B.There is T=T'+B.Wherein, circular matrix T' is by Fig. 3 Training sequence t (n) of formation is rearranged, and the exponent number of the circular matrix is Q × Q ranks, and expression is as follows：

In order to coordinate circular matrix to generate local matrix, the exponent number of real number matrix B also should be Q × Q ranks.Now, only need Obtain real number b, you can obtain local matrix.

The sampling process of real number b is illustrated in Fig. 4.First, baseband signal s (n) amplitude limit is obtained into ACO-OFDM time domains letter Number sequence s_c(n).Secondly, according to the characteristics of baseband signal amplitude Normal Distribution, that is, have：S (n)～N (0, σ²).Through limit After width process, the average of time-domain signal sequence is asked for.I.e. value is b,So, the expression formula of real number matrix B is such as Under：

Matrix T' is added with matrix B by adder, local matrix T can be obtained.Its expression formula is as follows：

Fig. 5 is channel estimation method process schematic in the present invention.Channel estimation results in the present invention, are by local Gained vector multiplication calculates gained after matrix and reception signal quantization are processed.First, sending signal x (n) of system is time domain letter Number s_cThe sum of (n) and training sequence t (n).It is represented by：

X (n)=s_c(n)+t(n) (5.1)

After sending signal channel h, must be with channel impulse response coefficient phase convolution.Tire out after signal elder generation multiplication in the figure Plus process.Now, white Gaussian noise w (n) can be mixed into the signal, and output signal is r (n).Expression formula is：

Wherein, l represents the number of path of light path；H (l) represents the channel impulse response coefficient of l paths；W (n) is represented Additive white Gaussian noise.

Secondly, output signal is divided into Q roads, successively to its Q times of down-sampling, every road is averaged respectively afterwards.With reference to send out The number of delivering letters, can obtain quantized result：y(n).Its expression formula is：

And w (n) amplitudes meet the normal distribution and E (s that average is zero_c(n))=b.Thus, quantized result can be reduced to：

During and if only if Q=L, above formula just has unique solution.Now, formula (5.3) is represented by vector form：

Y=Th (5.5)

Finally, sequence is obtained to quantized result y (n) unbiased esti-matorExpression formula is：

Wherein, the value of variable i is (0, N_Q- 1) integer in the range of.And with gained estimated resultWith reference to local square Inverse matrix T of battle array T^-1, you can estimate channel parameter.

Fig. 6 is y (n) sequence unbiased esti-mator process schematics during channel estimation method.Y (n) sequences are to receive signal Quantized result Jing after processing.Within a baseband signal cycle, sample sequence limited length, and sequences y (n) is with steady time The property gone through, therefore can be to its unbiased esti-mator.First, by receiving sequence r (n), n values are the integer of (0, Q-1) scope, and Q is training sequence The row cycle.Secondly, by receiving data point Q roads, successively to its Q times of down-sampling.By taking r (0) as an example, the element sampled successively is respectively r(0)、r(Q)、...、r((N_Q-1)Q).By that analogy, gained sequence number is then Q, respectively r (iQ), r (iQ+ 1) ..., r (iQ+Q-1), the value of variable i is (0, N_Q- 1) integer of scope.Finally, each sequential element adds up respectively, and removes With sequential element numerical value of N_Q.Then Q sequence is changed into Q numerical value, respectivelyWith this Composition unbiased esti-mator sequenceThe value of n is the integer of (0, Q-1) scope.

Fig. 7 is respectively circular matrix T' and real number matrix B in the present invention and is added and matrix T and the calculating of vector h multiplication with Fig. 8 Process schematic.The generation of local matrix is obtained by circular matrix T' and real number matrix B Jing after adder, concrete manifestation The process for obtaining a new matrix is added for two Q × Q rank matrix correspondence positions elements.And local matrix and channel vector are taken advantage of Method is calculated, and vector is multiplied respectively per row and column to show as Q × Q ranks matrix, obtains the process of a numerical value.Finally, obtain Q to Secondary element, that is, obtain column vector：y.

Fig. 9 is the graph of a relation of power allocation factor and channel estimation mean square error (MSE) performance.In the present embodiment, such as It is definite value N=512 to take sub-carrier number, obtains the graph of a relation between power allocation factor and channel estimation.From analysis, as SNR and N Constant, SNR is Signal-to-Noise.In P=0.7 or so, when P is ACO-OFDM systems transmitting terminal generation sending signal, point The power factor matched somebody with somebody, MSE values are minimum.Now, channel estimation accuracy rate is optimum.From Fig. 8 it can also be seen that becoming larger with P values, Square error is presented and first become the big rear trend for reducing.In addition, when signal to noise ratio snr is less than 15dB, the mean square error of channel estimation falls Difference is larger.And work as signal to noise ratio higher than after 15dB, channel estimation mean square error tends to be steady.And when signal to noise ratio is more than 15dB, training Sequence is held essentially constant with data-signal relative power.Because the energy of training sequence is higher, channel estimating performance is better.Institute So that now data-signal is just into the principal element for affecting channel estimating performance.

Multicarrier asymmetric amplitude limit light-OFDM (ACO-OFDM) channel of the present invention based on overlying training sequence Method of estimation.In broadband wireless optic communication, in order to adapt to intensity modulation, signal value is only on the occasion of can adopt ACO- OFDM is modulated.The invention is related to a kind of light wireless communication modulation and channel estimation methods, wherein, modulator approach mainly includes signal The steps such as mapping, symmetrical composition Hamilton, fast discrete inverse Fourier transform and asymmetric amplitude limit；And channel estimation methods bag Include the steps such as the non-negative cycle reality sequence generation of unipolarity, the generation of local matrix and first-order statistics channel estimation.By to being carried The experimental verification of ACO-OFDM channel estimation methods, can obtain the channel estimation effect of the close ideal performance of the method.And the method Also its channel estimating performance can be improved by adjusting power allocation factor and number of sub carrier wave etc..Because of These characteristics, the method Suitable for wireless optical transmission high accuracy channel estimation, and the Detection results that cause because receive signal and be changeable can be avoided unstable etc. Defect, there is larger using value.

Although having described embodiments of the invention, to those skilled in the art, can be without departing from present invention side Various changes, modification, replacement and modification are carried out to these embodiments in the case of method principle and spirit, the scope of the present invention is by institute Attached claim and its equivalent are limited.General principle figure, the dimension of local matrix, work(in i.e. by changing the method for the invention The basic legend such as rate distribution factor numerical value and algorithm parameter, still belong to the category of the method for the invention, are still protected by this patent.

Claims (7)

1. a kind of ACO-OFDM channel estimation methods based on overlying training sequence, is characterized in that carrying out as follows：

Step one, to pseudo noise sequence process, obtains the non-negative cycle reality sequence of unipolarity；

Step 2, generates local matrix；

Step 3, calculates desired parameters condition, by channel estimation methods, completes channel estimation.

2. ACO-OFDM channel estimation methods based on overlying training sequence according to claim 1, it is characterised in that：Step Rapid one, complex vector is generated by pseudo noise sequence, then Jing inverse fast Fourier transforms are changed into sequence of real numbers, sequence be repeated several times after most Throughout one's life into cycle arithmetic number sequence.

3. ACO-OFDM channel estimation methods based on overlying training sequence according to claim 2, it is characterised in that：Institute State step one and pass sequentially through following steps realization：

PN sequence constellation mapping and parallel serial conversion of step 1.1. to a length of Q, and the value of Q is the number of 4 integral multiple；Now, It is the vector that a 1 × Q/4 is tieed up that gained sequence length is a quarter of former PN sequences, i.e., complex vector p of a length of Q/4, and p；

Jing step 1.1 gained complex vectors p are made conjugation symmetry transformation by step 1.2., obtain vectorial p '；Vectorial p ' only has odd term to hold Information carrying ceases, and even item is all zero, i.e., when variable m is even number, there is p ' (m)=0；Because system signal needs base band transmission, therefore Need to be symmetrical by complex signal conjugation, it is changed into sequence of real numbers Jing after IFFT process, then transmitted；And if only if, and p ' vectors have altogether During yoke symmetry, IFFT process could be performed, otherwise, repeat step 1.2 makes generation vector have conjugate symmetry；

Step 1.3. makees the vectorial p ' in step 1.2 after IFFT process, to be changed into sequence of real numbers；Again it is performed successively and is gone here and there Conversion, amplitude limiting processing, and amplitude limiting processing mainly make the element value of sequence of real numbers be not less than zero, you can generate one pole of the length for Q Property nonnegative real number sequence t ' (n)；In order to possess periodically, on the basis of above-mentioned generation sequence of real numbers t ' (n), repeated N_Q It is secondary, N_QFor natural number, non-negative cycle sequence of real numbers t (n) of unipolarity is obtained；And t (n) is a 1 × QN_QDimension, cycle are Q, first Plain value be all on the occasion of sequence of real numbers；N is natural number, and value is (1, QN_Q) scope integer.

4. ACO-OFDM channel estimation methods based on overlying training sequence according to claim 3, it is characterised in that：Step Rapid two, by the time-domain signal analysis to ACO-OFDM systems, with Q × Q dimension real number matrix B and circulation that element is all real number b Matrix T ' is constituting local matrix T；The value of Q is the periodic quantity of generation cycle arithmetic number sequence；T ' is made up of training sequence Q × Q dimension circular matrix, and T be Q × Q dimension real number matrix, b be time-domain signal s_cThe average of (n).

5. ACO-the OFDM channel estimation method based on overlying training sequence according to claim 4, it is characterised in that： Step 2 is realized using following steps：

The generating process of step 2.1. real number matrix B；Real number matrix refers to that all elements in matrix are all real number, and matrix B is The full real number symmetrical matrix of one Q × Q dimension, and Q is the periodic quantity of training sequence in step 1.3；By the orthogonal frequency division multiplexing of input It is expressed as with baseband signal：S (n), and s (n) meets the normal distribution that average is zero, i.e. s (n)～N (0, σ²), its average is 0, Variance is σ²；Make amplitude limiting processing to s (n), that is, limit each signal amplitude and be not less than zero, obtain time-domain signal：s_c(n)；

Step 2.2. is for step 2.1 gained sequence s_cN () and s (n), because s (n) meets the normal distribution that average is zero, obtains s_c N () average is：OrderIt is all real number matrix B of b that Q × Q ranks element can be constructed, that is, have：

$B={\left[\begin{array}{ccccc}b& b& b& \mathrm{...}& b\\ b& b& b& \mathrm{...}& b\\ b& b& b& \mathrm{...}& b\\ \mathrm{...}& \mathrm{...}& \mathrm{...}& \mathrm{...}& \mathrm{...}\\ b& b& b& \mathrm{...}& b\end{array}\right]}_{Q\×Q}---\left(1\right)$

In real number matrix B that step 2.3. is generated in step 2.2, a circular matrix is superimposed, you can complete local matrix structure Build, training sequence t (n) that the circular matrix is generated with step 1.3 has as matrix element：

$T^{\′}=\left[\begin{array}{ccccc}t\left(0\right)& t\left(1\right)& t\left(2\right)& \mathrm{...}& t(Q-1)\\ t(Q-1)& t\left(0\right)& t\left(1\right)& \mathrm{...}& t(Q-2)\\ t(Q-2)& t(Q-1)& t\left(0\right)& \mathrm{...}& t(Q-3)\\ \mathrm{...}& \mathrm{...}& \mathrm{...}& \mathrm{...}& \mathrm{...}\\ t\left(1\right)& t\left(2\right)& t\left(3\right)& \mathrm{...}& t\left(0\right)\end{array}\right]---\left(2\right)$

Step 2.4. is added step 2.2 with step 2.3 gained real number matrix B and circular matrix T ', obtains local matrix T, i.e., Have：

$T=\left[\begin{array}{ccccc}t\left(0\right)+b& t\left(1\right)+b& t\left(2\right)+b& \mathrm{...}& t(Q-1)+b\\ t(Q-1)+b& t\left(0\right)+b& t\left(1\right)+b& \mathrm{...}& t(Q-2)+b\\ t(Q-2)+b& t(Q-1)+b& t\left(0\right)+b& \mathrm{...}& t(Q-3)+b\\ \mathrm{...}& \mathrm{...}& \mathrm{...}& \mathrm{...}& \mathrm{...}\\ t\left(1\right)+b& t\left(2\right)+b& t\left(3\right)+b& \mathrm{...}& t\left(0\right)+b\end{array}\right]---\left(3\right).$

6. ACO-the OFDM channel estimation method based on overlying training sequence according to claim 5, it is characterised in that： The channel estimation methods of step 3, obtain sending signal x (n) and receive signal r (n) by time domain discrete model, will receive signal point For Q roads, i.e., its Q times of down-sampling is averaged respectively afterwards to every road successively, obtain quantized result y (n)；Again to its unbiased esti-mator, With reference to the inverse matrix of local matrix T, channel estimation process is completed.

7. ACO-the OFDM channel estimation method based on overlying training sequence according to claim 6, it is characterised in that： Step 3 is completed using following steps：

Sending signal x (n) of the step 3.1. system is the time-domain signal s of gained in step 2.1_cObtain in (n) and step 1.3 Training sequence t (n) sum, its expression formula is as follows：

X (n)=s_c(n)+t(n) (4)

After sending signal x (n) channel in step 3.2. step 3.1, the signal that receiving terminal is received is its Jing light multiple paths After reflection, each summation for receiving component of signal, its expression formula is as follows：

$r\left(n\right)={\mathrm{\Σ}}_{l=0}^{L-1}h\left(l\right)x(n-1)+w\left(n\right)---\left(5\right)$

Wherein, l represents the number of path of light path；H (l) represents the channel impulse response coefficient of l paths；W (n) represents additivity White Gaussian noise, its feature meets the Gaussian Profile that average is zero for amplitude；And power spectral density satisfaction is uniformly distributed, its average It is only relevant with its frequency span with variance；

Reception signal r (n) of step 3.2 gained is divided into Q roads by step 3.3., i.e., successively to its Q times of down-sampling；Afterwards to every road Average respectively, must receive signal quantization result is：Y (n)=E [r (kQ+n)]. wherein, it is (0, N that k is value_Q) scope it is whole Number, Q is the periodic quantity of training sequence in step 1.3；

Step 3.4. calculates step with reference to sending signal x (n) of gained in step 3.1,3.2 and the expression formula of reception signal r (n) Rapid 3.3 quantized result y (n) is：

$\begin{array}{c}y\left(n\right)=E\[{\Σ}_{l=0}^{L-1}h\left(l\right)\[s_c(kQ+n-l)+t(kQ+n-l)\]+w(kQ+n)\]\\ =E\[{\Σ}_{l=0}^{L-1}h\left(l\right)s_c(kQ+n-l)\]+E\[{\Σ}_{l=0}^{L-1}h\left(l\right)t(kQ+n-l)\]+\\ w(kQ+n)\]+E\[w(kQ+n)\],n=0,1,\mathrm{...},Q-1,k=0,1,\mathrm{...},N_Q-1\end{array}---\left(6\right)$

Wherein, N_QFor the sequence repetition number described in step 1.3；

Step 3.5. is additive white Gaussian noise because of noise w (n), and according to its amplitude the normal distribution feature that average is zero is met, And by the s of step 2.2 gained_cN () average is the result of b, then the computational short cut of y (n) is in step 3.4：

$\begin{array}{c}y\left(n\right)={\Σ}_{l=0}^{L-1}h\left(l\right)B+{\Σ}_{l=0}^{L-1}h\left(l\right)t(kQ+n-l)\\ ={\Σ}_{l=0}^{L-1}h\left(l\right)\[B+t(kQ+n-l)\],n=0,1,\mathrm{...},Q-1,k=0,1,\mathrm{...},N_Q-1\end{array}---\left(7\right)$

Gained y (n) of step 3.6. solution procedure 3.5；During and if only if Q=L, L represents channel exponent number, and its coefficient matrix full rank When, the formula has unique solution；Otherwise, parameter need to be readjusted so as to unique solution；Because channel exponent number is only in estimate Limit, then make cycle Q be equal to the higher limit that channel exponent number can be obtained；Finally, step 3.5 gained y (n) expression formula is by vector calculating It is expressed as：

Y=Th (8)

Wherein, T is the local matrix constructed in step 2.4；Y and h are the column vectors of Q × 1 dimension；Th is matrix with vector Multiplication calculating process；

Step 3.7. has steady ergodic because of the quantized sequences in step 3.6；Sequences y is averaged computing, as a result with Time parameter is unrelated, therefore the unbiased esti-mator for calculating the sequences y is：

$\hat{y}\left(n\right)=1/N_Q{\mathrm{\Σ}}_{i=0}^{N_Q-1}r(iQ+n),n=0,1,...,Q-1---\left(9\right)$

Wherein, N_QFor the sequence repetition number described in step 1.3, r (n) is the reception signal of step 3.2 gained；

Step 3.8. calculates its inverse matrix T by the local matrix T of step 2.4 gained^-1；With reference to gained sequences y in step 3.7 Unbiased esti-mator resultFinal channel estimation results h is completed by following formula；Each element in vectorial h is the estimation side Each channel model coefficient obtained by method

$h=T^{-1}\hat{y}---\left(10\right).$