CN101447969A

CN101447969A - Channel estimation method of multi-band orthogonal frequency division multiplexing ultra wide band system

Info

Publication number: CN101447969A
Application number: CNA2008101642240A
Authority: CN
Inventors: 李有明; 李新苗; 徐铁锋
Original assignee: Ningbo University
Current assignee: Ningbo University
Priority date: 2008-12-31
Filing date: 2008-12-31
Publication date: 2009-06-03
Anticipated expiration: 2028-12-31
Also published as: CN101447969B

Abstract

The invention discloses a channel estimation method of multi-band orthogonal frequency division multiplexing ultra wide band system. The invention has the advantages that m sequence having better auto-correlation characteristics is taken as a time-domain training sequence and a cyclic prefix is appended; impulse response estimation value is obtained by doing mutual-correlation operation to receiving signals the cyclic prefix of which is eliminated and the training sequence and doing auto-correlation operation to each training sequence on a receiving terminal; that an auto-correlation matrix of the m sequence having diagonally dominant characteristics is utilized; one diagonal decomposition or three diagonal decomposition is respectively carried out to the auto-correlation matrix of the m sequence firstly; then an approximation method of first-order inverse matrix is adopted to effectively avoid complex inverse operation; then the operation amount is reduced by an order of magnitude; the performance approaches normal time-domain channel estimation method; and the invention is the fast and effective channel estimation method of the ultra wide band system, and is easy to be realized.

Description

Channel estimation method of multi-band orthogonal frequency division multiplexing ultra-wideband system

Technical Field

The present invention relates to a channel estimation method, and more particularly, to a channel estimation method for a multi-band orthogonal frequency division multiplexing ultra-wideband system.

Background

Ultra Wide Band (UWB) technology is a high-speed and short-distance wireless personal communication technology with great potential, and has attracted great attention in recent years in both academic and industrial circles, and has become a hotspot for research and development in the field of wireless communication at present. The ultra-wideband technology combines the Multi-Band orthogonal frequency Division Multiplexing (MB-OFDM) technology to form the Multi-Band orthogonal frequency Division Multiplexing ultra-wideband (MB-OFDM UWB) technology, which can effectively resist the characteristics of Multi-path fading, various narrowband interferences, flexible utilization of frequency spectrum resources and the like, and becomes one of the mainstream realization schemes of the ultra-wideband technology. The application prospect of the multi-band orthogonal frequency division multiplexing ultra-wideband technology is very attractive, for example, the technology is widely applied to a plurality of fields such as a high-speed wireless personal area network, a wireless Ethernet interface link, an intelligent wireless local area network, an outdoor peer-to-peer network, a sensing, positioning and identifying network and the like, and is particularly applied to the field of digital home electronic products. At present, many companies select the application of wireless home electronic products as a breakthrough of the multiband Orthogonal Frequency Division Multiplexing (OFDM) ultra-wideband technology.

In order to obtain ideal performance, technologies such as coherent detection, demodulation, equalization and the like need to be adopted in the multi-band orthogonal frequency division multiplexing ultra-wideband system, and the technologies all need to utilize information of a channel, so that accurate channel estimation information plays a crucial role in ensuring reliable data transmission in a multi-band orthogonal frequency division multiplexing ultra-wideband communication environment. Because the ultra-wideband signal occupies a large bandwidth, has short signal duration and high transmission rate, the requirements of high estimation precision and low calculation complexity are provided for the channel estimation technology. Therefore, how to perform fast and effective channel estimation in a multiband orthogonal frequency division multiplexing ultra-wideband system is a big challenge faced by the existing multiband orthogonal frequency division multiplexing ultra-wideband technology.

The multi-band orthogonal frequency division multiplexing ultra-wideband system mostly adopts a method of frequency domain pilot frequency domain channel estimation, namely, pilot frequency is inserted in a frequency domain, and channel estimation is carried out in the frequency domain. Such a channel estimation method includes the steps of: firstly, inserting pilot frequency in proper position of frequency domain of transmitting end, utilizing pilot frequency data to obtain channel information of pilot frequency position by utilizing correspondent channel estimation rule

Then passes through an interpolator, and uses the interpolation mode to carry out the comparisonInterpolating in the whole frequency domain to obtain the whole channel estimation value

Finally, the channel estimation value and the received data are sent to an equalizer, and the received data can be equalized to obtain the estimation value of the original transmitted data.

Currently, the channel information for the pilot positions is usually obtained based on Least Square (LS) criterion or Minimum Mean Square Error (MMSE) criterion. The frequency domain pilot frequency domain channel estimation method based on the least square criterion is simple in calculation process and easy to implement, but the method does not consider the influence of noise, so that the accuracy of channel estimation is not high. The frequency domain pilot frequency domain channel estimation method based on the minimum mean square error criterion can obtain good performance because of utilizing the frequency domain autocorrelation characteristic of the channel, but matrix inversion is involved in the estimation process of the method, the calculation complexity of the method is increased, and the feasibility of the method is poor. In summary, some existing frequency domain pilot frequency domain channel estimation methods have the problems that the calculation complexity is high, the methods are difficult to be used in practice, and the calculation accuracy is low due to the fact that the channel characteristics of the non-pilot position need to use an interpolation mode.

The existing time domain channel estimation methods mainly include estimation methods based on Discrete Fourier Transform (DFT) filtering methods and Maximum Likelihood criteria (ML), which can reduce the mean square error value of channel estimation to some extent, but have the disadvantage that channel length (or finite delay spread of a channel) information needs to be accurately obtained before channel estimation, thereby increasing the duration and computational complexity of the channel estimation process, and limiting the two methods in practical application.

Bowei Song et al propose a time domain channel estimation method based on m-sequence, and the working flow of a multi-band orthogonal frequency division multiplexing ultra-wideband system applying the method is shown in FIG. 1. At a transmitting end, an input data signal is subjected to Quadrature Phase Shift modulation (QPSK) to obtain a modulation signal, the modulation signal is subjected to serial-to-parallel conversion, Inverse Fourier Transform (IFFT) and parallel-to-serial conversion to form a plurality of OFDM symbols, and a length L is inserted into every fixed number of OFDM symbols_PThe m sequence s is used as a training sequence for time domain channel estimation, and the length is L according to the quality of the channel characteristics_CAssuming that the multi-band orthogonal frequency division multiplexing ultra-wideband system is synchronous, the training sequence obtained after the Cyclic Prefix is added and the input data signal are transmitted through an ultra-wideband channel after being modulated by a carrier; at the receiving end, firstly, the cyclic prefix in the training sequence after being influenced by the channel fading and the Gaussian white noise is removed, and then the training sequence after being influenced by the channel fading and the Gaussian white noise after the cyclic prefix is removed

M sequence s circularly right shifted by i bit from m sequence sⁱThe correlation operation is carried out, and the correlation operation is carried out,

wherein k is 0, 1, …, L_p+L_c1, h denotes a matrix vector composed of coefficients of respective multipaths of the channel,

h_jfor the jth multipath coefficient of the channel, h should satisfy the condition: { h_j＝0|L≤j≤L_C-1}, L is the order of the channel, C_P(i, j) is the m-sequence s circularly right-shifted by j bits^jAnd m sequence s circularly right-shifted by i bitⁱThe second term is white gaussianAnd the noise sequence n and the m sequence s are normalized cross correlation coefficients, n is white Gaussian noise, and n (k) is white Gaussian noise at the kth moment. The amplitude of the noise is compressed to 1/L_PDouble, i.e.

Thus, can be used for

Approximately as C ≈ C_Ph, wherein C_PIs an autocorrelation matrix of m-sequences s, inserted into the m-sequences s of length L_PThen the autocorrelation matrix C_PThe normalized autocorrelation function over one period satisfies:

<math> <mrow> <mfenced open='' close=''> <mtable> <mtr> <mtd> <msub> <mi>C</mi> <mi>P</mi> </msub> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> <mo>=</mo> <mrow> <mo>(</mo> <mn>1</mn> <mo>/</mo> <msub> <mi>L</mi> <mi>P</mi> </msub> <mo>)</mo> </mrow> <munderover> <mi>Σ</mi> <mrow> <mi>k</mi> <mo>=</mo> <mn>0</mn> </mrow> <mrow> <msub> <mi>L</mi> <mi>P</mi> </msub> <mo>-</mo> <mn>1</mn> </mrow> </munderover> <msup> <mi>s</mi> <mi>j</mi> </msup> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> </mrow> <msup> <mi>s</mi> <mi>i</mi> </msup> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> </mrow> </mtd> </mtr> <mtr> <mtd> <mo>=</mo> <mrow> <mfenced open='{' close=''> <mtable> <mtr> <mtd> <mn>1</mn> <mo>,</mo> </mtd> <mtd> <mi>i</mi> <mo>=</mo> <mi>j</mi> </mtd> </mtr> <mtr> <mtd> <mo>-</mo> <mn>1</mn> <mo>/</mo> <msub> <mi>L</mi> <mi>P</mi> </msub> <mo>,</mo> </mtd> <mtd> <mi>i</mi> <mo>&NotEqual;</mo> <mi>j</mi> </mtd> </mtr> </mtable> </mfenced> </mrow> </mtd> </mtr> </mtable> </mfenced> </mrow></math>

. Thereby obtaining the impulse response estimated value of the channel by utilizing the autocorrelation characteristic of the m sequence

\tilde{h} = C_{p}^{- 1} C .

The method skillfully utilizes the autocorrelation characteristic of the m sequence to obtain the estimated value of the channel impulse response, has high estimation precision, and can flexibly adjust the overhead of the training sequence according to the transmission rate requirement of the multi-band orthogonal frequency division multiplexing ultra-wideband communication system so as to obtain the compromise between the estimation precision and the overhead. But from

\tilde{h} = C_{p}^{- 1} C

It can be seen that to obtain the channel impulse response, the matrix C_PIs necessary, whereas the autocorrelation matrix C_P＝[C_P(i，j)]，i＝0，1，…L_p-1，j＝0，1，…L_p-1 is an L_PThe order square matrix, if it needs to be inverted, has high computational complexity (the computational complexity is) The high computational complexity poses a significant obstacle to the application of this method.

Disclosure of Invention

The technical problem to be solved by the present invention is to provide a channel estimation method with low computation complexity, which is suitable for a multi-band orthogonal frequency division multiplexing ultra-wideband system, aiming at the defects existing in the prior art.

The technical scheme adopted by the invention for solving the technical problems is as follows: multi-band orthogonal frequency divisionThe channel estimation method of the multiplexing ultra-wideband system comprises the following steps: firstly, carrying out quadrature phase shift modulation processing on an input data signal at a transmitting end to obtain a modulation signal; secondly, performing serial-parallel conversion, inverse Fourier transform and parallel-serial conversion on the modulation signals in sequence to form a plurality of OFDM symbols; thirdly, inserting a length L into a plurality of formed OFDM symbols every set number of OFDM symbols_PThe m sequence s is taken as a training sequence, and a length L is added in front of the training sequence according to the channel characteristics_CThe cyclic prefix of (a) is obtained as a training sequence to which the cyclic prefix is added, and x is represented by x ═ x (0), x (1), … x (L)_P+L_C-1)](ii) a Finally, the training sequence x with the attached cyclic prefix and the formed OFDM symbol are transmitted to a receiving end through an ultra-wideband channel after being subjected to carrier modulation processing, and the training sequence x with the attached cyclic prefix and the OFDM symbol are influenced by channel fading and Gaussian white noise in the transmission process; at the receiving end, defining the training sequence x with attached cyclic prefix received by the receiving end and influenced by channel fading and white Gaussian noise as the first received signal, defining the OFDM symbol received by the receiving end and influenced by channel fading and white Gaussian noise as the second received signal, and representing the first received signal as the second received signal by using a tapped delay line model

Wherein k is 0, 1, …, L_p+L_c-1, r (k) is the first received signal at time k, h denotes the received signalThe coefficients of the various multipaths of a track constitute a matrix vector,

h_tfor the t-th multipath coefficient of the channel, h should satisfy the condition: { h_t＝0|L≤t≤L_C-1}, L is the order of the channel, x is the training sequence after cyclic prefix is added, x (k-t) is the training sequence after cyclic prefix is added at the k-t time, n is white gaussian noise, and n (k) is white gaussian noise at the k time; firstly, the first received signal r (k) is processed with carrier wave removing modulation, and the first received signal processed with carrier wave removing modulation is processed with cyclic prefix removing processing to obtain

Wherein k is 0, 1, …, L_p+L_c-1，

H represents a matrix vector composed of respective multipath coefficients of the channel for the first received signal at the k-th time after the cyclic prefix is removed,

h_jfor the jth multipath coefficient of the channel, h should satisfy the condition: { h_j＝0|L≤j≤L_C-1}, L is the order of the channel, n is white Gaussian noise, n (k) is white Gaussian noise at the k-th time, s^jCyclically shift the m sequence s by j bits^j(k) The sequence at the k-th moment after the m sequence s is circularly shifted by j bits to the right; seventhly, calculating the first received signal after the cyclic prefix is removed

M sequence s circularly right shifted by i bit from m sequence sⁱCross correlation matrix C and autocorrelation matrix C of each training sequence s_P，C＝[C(i，j)]C (i, j) is the first received signal after the cyclic prefix is removed

M sequence s circularly right-shifted by i from m sequence sⁱThe normalized cross-correlation coefficient of (a),

C_P＝[C_P(i，j)]，C_P(i, j) is the m sequence s circularly right shifted by j bits^jAnd m sequence s circularly right-shifted by i bitⁱThe normalized auto-correlation coefficient of (a),

wherein i is 0, 1, …, L_p，j＝0，1，…，L_p，k＝0，1，…，L_p+L_c-1，

Is the first received signal at the k-th time after the cyclic prefix removal, s^j(k) For m-sequences s cyclically shifted by j bits at time k, sⁱ(k) And circularly right shifting the m sequence s by i bit to the k time sequence. And according to the first received signal after removing cyclic prefix

M sequence s circularly right shifted by i bit from m sequence sⁱCross correlation matrix C and autocorrelation matrix C of each training sequence s_PCalculating an impulse response estimate of the channel

\tilde{h} = C_{p}^{- 1} C,

Wherein,

is an autocorrelation matrix C_PThe inverse matrix of (d); according to the autocorrelation matrix C_PThe said autocorrelation matrix C_PDecomposing the matrix into the sum of a first matrix and a second matrix, recording the first matrix as D, and recording the second matrix as E, C_PD + E, said first matrix D and said second matrix E satisfying | D^-1E‖<1, calculating the autocorrelation matrix C_PInverse matrix of

Wherein the symbol "II" is a norm symbol, I is an identity matrix, D^-1Is the inverse of the first matrix D, m is 1, 2, …, ∞; then according to

Computing

To a first order approximation of the first order,

the first matrix D is formed by the autocorrelation matrix C_PThe second matrix E is the diagonal matrix formed by the autocorrelation matrix C_PThe diagonal matrix is marked as D₁Said off-diagonal matrix is denoted as E₁To obtain

For the autocorrelation matrix C_PThe coefficient of (a) is normalized, and the diagonal matrix D is obtained after the normalization₁Is an identity matrix I; according to

To obtain

C_{p}^{- 1} = I - E_{1} .

The first matrix D is formed by the autocorrelation matrix C_PThe second matrix E is a three-diagonal matrix formed by the autocorrelation matrix C_PThe non-tri-diagonal matrix composed of elements except the tri-diagonal elements is marked as D₃And recording the non-tri-diagonal matrix as E₃To obtain

Combining the three diagonal matrix D₃Decomposed into autocorrelation matrix C_PAnd a diagonal matrix composed of diagonal elements of and said autocorrelation matrix C_PThe diagonal element of (2) is the sum of diagonal matrices composed of diagonal elements of 0, and the diagonal matrix is denoted as D₁And said diagonal matrix is denoted as D₂Calculating said tri-diagonal matrix D₃Inverse matrix of

Wherein, I is an identity matrix,

is a diagonal matrix D₁M 1, 2, …, ∞; then according to

Computing

To a first order approximation of the first order,

for the autocorrelation matrix C_PThe coefficient of (A) is normalized, and the diagonal matrix D is obtained after the normalization₁Is an identity matrix I, based on

To obtain

D_{3}^{- 1} = I - D_{2};

Then according to

And

D_{3}^{- 1} = I - D_{2},

to obtain

Compared with the prior art, the invention has the advantages that the m sequence with better autocorrelation property is adopted as the time domain training sequence, the first receiving signal without cyclic prefix and the training sequence are subjected to cross correlation operation at the receiving end, the impulse response estimation value of the channel is obtained by the autocorrelation operation of each training sequence, the autocorrelation matrix of the m sequence is utilized to have diagonal dominance property, firstly, the autocorrelation matrix of the m sequence is subjected to one-diagonal decomposition or three-diagonal decomposition respectively, and then, the approximation method of the first-order inverse matrix is adopted, so that the complex inversion operation is effectively avoided, the operation amount is reduced by one order of magnitude, and the performance approaches to the conventional time domain channel estimation method, thereby being a quick and effective channel estimation method of the ultra-wideband system and being easy to realize.

Drawings

Fig. 1 is a schematic diagram of a working flow of a multi-band orthogonal frequency division multiplexing ultra-wideband system;

FIG. 2 is a graph showing bit error rate of LS algorithm and conventional time domain channel estimation method corresponding to m sequences of different lengths varying with signal-to-noise ratio;

FIG. 3 is L_PA performance comparison graph of a 31-hour LS algorithm, a conventional time domain estimation method, a one-pair angular decomposition method and a three-pair angular decomposition method is obtained;

FIG. 4 is L_PA performance comparison graph of a 15-hour conventional time domain estimation method, a pair of angle decomposition method and a three-pair angle decomposition method of the invention.

Detailed Description

The invention is described in further detail below with reference to the accompanying examples.

A channel estimation method of a multi-band orthogonal frequency division multiplexing ultra-wideband system comprises the following steps:

firstly, at a transmitting end, the prior quadrature phase shift modulation (QPSK) technology is adopted to perform quadrature phase shift modulation processing on an input data signal to obtain a modulation signal.

Secondly, serial-parallel conversion and Fourier conversion are carried out on the modulation signals in sequenceInverse fourier transform (IFFT) and parallel-to-serial conversion processes form a plurality of OFDM symbols. In this embodiment, each OFDM symbol uses 128 subcarriers, the frequency interval between adjacent subcarriers is 4.1254MHz, and the duration of each OFDM symbol is T₀＝242.4ns。

Thirdly, inserting a length L into a plurality of formed OFDM symbols every set number of OFDM symbols_PThe m sequence s is used as a training sequence, and a length L is added in front of the training sequence according to the channel characteristics_CThe Cyclic Prefix (CP) of (1) is obtained as a cyclic prefix-added training sequence, which is represented by x ═ x (0), x (1), … x (L)_P+L_C-1)]Wherein L is_PIs the length of the m-sequence s, i.e. the training sequence, L_CIs the length of the cyclic prefix. In this embodiment, the selected set number is 4, that is, a length L is inserted every 4 OFDM symbols_PM-sequence s of (1).

And finally, the training sequence x with the cyclic prefix and the formed OFDM symbol are transmitted to a receiving end through an ultra-wideband channel after being subjected to carrier modulation processing, and the training sequence x and the OFDM symbol with the cyclic prefix are influenced by channel fading and white Gaussian noise in the transmission process.

At the receiving end, defining the training sequence x with attached cyclic prefix received by the receiving end and influenced by channel fading and white Gaussian noise as the first received signal, defining the OFDM symbol received by the receiving end and influenced by channel fading and white Gaussian noise as the second received signal, and representing the first received signal as the second received signal by using a tapped delay line model

Wherein k is 0, 1, …, L_p+L_c1, r (k) is the first received signal at time k, h represents a matrix vector formed by the multipath coefficients of the channel,

h_tfor the t-th multipath coefficient of the channel, h should satisfy the condition: { h_t＝0|L≤t≤L_C-1}, L is the order of the channel, x is the training sequence with the cyclic prefix appended thereto, x (k-t) is the training sequence with the cyclic prefix appended thereto at the k-t time, n is white gaussian noise, and n (k) is white gaussian noise at the k time.

Firstly, the first received signal r (k) is processed with carrier wave removing modulation, and the first received signal processed with carrier wave removing modulation is processed with cyclic prefix removing processing to obtain

Wherein k is 0, 1, …, L_p+L_c-1，

h_jfor the jth multipath coefficient of the channel, h should satisfy the condition: { h_j＝0|L≤j≤L_C-1}, L is the order of the channel, n is white Gaussian noise, n (k) is white Gaussian noise at the k-th time, s^jCyclically shift the m sequence s by j bits^j(k) The m-sequence s is circularly shifted by j bits to the right to obtain the sequence at the k-th moment.

Seventhly, calculating the first received signal after the cyclic prefix is removed

C_P＝[C_P(i，j)]，C_P(i, j) is the m sequence s circularly right shifted by j bits^jAnd the normalized autocorrelation coefficient of the m-sequence si after the m-sequence s is circularly right-shifted by i bits,

Before being recycledFirst received signal, s, at time k after affixation^j(k) For m-sequences s cyclically shifted by j bits at time k, sⁱ(k) And circularly right shifting the m sequence s by i bit to the k time sequence.

And according to the first received signal after removing cyclic prefix

\tilde{h} = C_{p}^{- 1} C,

An autocorrelation matrix C representing each training sequence s_PThe inverse matrix of (c). In this step, an impulse response estimate of the channel is calculated

Previously based on an autocorrelation matrix C_PThe diagonal dominance of (C) is given by the autocorrelation matrix C_PDecomposing into the sum of the first matrix and the second matrix, and recording the first matrix as D and the second matrix as E, if there is C_PD + E, satisfying | D in the first matrix D and the second matrix E^-1E‖<1 hour, calculate the autocorrelation matrix C_PInverse matrix of

Wherein the symbol "II" is a norm symbol, I is an identity matrix, D^-1Is the inverse of the first matrix D, m is 1, 2, …, ∞; if only considerTo a first order approximation, is based on

Computing

To a first order approximation of the first order,

finally utilize

Computing an impulse response estimate for a channel

\begin{matrix} \tilde{h} = C_{p}^{- 1} C \\ = (D^{- 1} - D^{- 1} E D^{- 1}) C \end{matrix} .

In order to reduce the complexity of calculation, the invention providesTwo solutions C are obtained_PFast approximation method of inverse matrix: a one-diagonal decomposition method and a three-diagonal decomposition method.

A pair of angle decomposition methods: to autocorrelation matrix C_PDecomposed into a matrix of autocorrelation C_PAnd a diagonal matrix composed of diagonal elements of and an autocorrelation matrix C_PThe sum of off-diagonal matrices composed of off-diagonal elements, and the diagonal matrix is denoted as D₁Let the off-diagonal matrix be denoted as E₁Then C is_P＝D₁+E₁Wherein

L_Pis the length of the training sequence inserted at the transmitting end; due to the autocorrelation matrix C_PThe opposite-angle dominance of (a) is,

| | D_{1}^{- 1} E_{1} | | < 1,

thus, it is possible to provide

There are the following expansions:

where | is the norm symbol,

is composed of

To the m-th power of (a), I is the identity matrix,

is a diagonal matrix D₁M 1, 2, …, ∞; according to

Is obtained by expansion ofTo a first order approximation of

From

It can be seen that the autocorrelation matrix C is solved using a one-to-one angular decomposition method, involving only the inversion of the diagonal matrix_PThe coefficient of (A) is normalized, and the diagonal matrix D is obtained after the normalization₁Is an identity matrix I, thus

To a first order approximation of

C_{p}^{- 1} = I - E_{1},

From

C_{p}^{- 1} = I - E_{1}

Can derive the calculation

The inversion process is not needed, the computational complexity is greatly reduced, and the performance of the method completely depends on the autocorrelation matrix C_PThe opposite angle dominance of.

A three diagonal decomposition method: to autocorrelation matrix C_PDecomposed into a matrix of autocorrelation C_PAnd a tri-diagonal matrix composed of tri-diagonal elements of and an autocorrelation matrix C_PThe sum of the non-tri-diagonal matrices composed of elements other than the tri-diagonal elements, and the tri-diagonal matrix is denoted as D₃Let the non-tri-diagonal matrix be denoted as E₃Then C is_P＝D₃+E₃Wherein

L_Pis the length of the training sequence inserted at the transmitting end; like a diagonal decomposition method, due to the autocorrelation matrix C_PThe opposite-angle dominance of (a) is,

| | D_{3}^{- 1} E_{3} | | < 1,

thus, it is possible to provideThere are the following expansions:

where | is the norm symbol,

is composed of

To the m-th power of (a), I is the identity matrix,

is a three diagonal matrix D₃M 1, 2, …, ∞; according to

Is obtained by expansion ofCan be expressed as a first approximation

Combining the three diagonal matrices D₃Decomposed into a matrix of autocorrelation C_PAnd a diagonal matrix composed of diagonal elements of and an autocorrelation matrix C_PThe diagonal matrix is represented as D₁Let a diagonal matrix be D₂Then there is D₃＝D₁+D₂Wherein

computing a tri-diagonal matrix D₃Inverse matrix of

Wherein, I is an identity matrix,

is a diagonal matrix D₁M 1, 2, …, ∞; then according to

ComputingTo a first order approximation of the first order,

for autocorrelation matrix C_PThe coefficient of (A) is normalized, and the diagonal matrix D is obtained after the normalization₁Is an identity matrix I, based on

To obtain

D_{3}^{- 1} = I - D_{2};

Then according to

And

D_{3}^{- 1} = I - D_{2},

to obtain

Known three diagonal decomposition method calculation

And an inversion process is also not needed, so that the computational complexity is greatly reduced.

Table 1 lists the magnitudes of the computational complexity of the existing time-domain channel estimation method based on m-sequences (referred to as the direct inversion method in table 1), the one-pair angular decomposition method and the three-pair angular decomposition method of the present invention.

Table 1 calculation complexity comparison table

As can be seen from table 1, the processing method using the one-diagonal decomposition and the three-diagonal decomposition proposed in the present invention can greatly reduce the computational complexity of the existing direct inversion method.

The invention utilizes the diagonal dominance characteristic of a special training sequence, namely the autocorrelation matrix of the m sequence, firstly carries out one-diagonal decomposition and three-diagonal decomposition on the autocorrelation matrix of the m sequence, and then adopts an approximation method of a first-order inverse matrix. The computer simulation result verifies the effectiveness of the invention.

Inserting a training sequence every 4 OFDM symbols, respectively inserting L _P15, 31, 63 and 127, and m sequences with different lengths. FIG. 2 compares the corresponding different lengthsThe bit error rate of the conventional time domain channel estimation method (i.e. direct inversion method) and the frequency domain LS channel estimation method of the m sequence is a curve which changes along with the signal-to-noise ratio. As can be easily seen from fig. 2, under the same snr condition, the shorter the m-sequence length is, the higher the bit error rate of the conventional time domain channel estimation method is, the worse the performance is, while the longer the m-sequence length is, the lower the bit error rate thereof is, the better the performance is. The purpose of the simulation of fig. 2 is to choose the appropriate length of the training sequence. However, considering the amount of calculation and the system performance, the pilot length is not suitable to be too long in practice. Since m sequences are respectively taken as L_P＝31、L_PThe performance is similar when the length is 64, so that the length L should be considered in practice_PThe m-sequence of 31 is suitable.

FIG. 3 compares the m-sequence length L _P31, the bit error rate versus signal-to-noise ratio curves of the conventional time domain estimation method, the one-to-angle decomposition method, the three-to-angle decomposition method, and the frequency domain LS channel estimation method. As can be seen from fig. 3, the performance of the one-diagonal decomposition method and the three-diagonal decomposition method is very similar to that of the conventional time domain estimation method, but the computational complexity of the first two methods is reduced by an order of magnitude.

In order to further compare the performance of the one-diagonal decomposition method and the three-diagonal decomposition method with the performance difference from the conventional time domain estimation method, fig. 4 shows that the three methods have poor m-sequence autocorrelation characteristics (m-sequence length L)_P15) was simulated. As shown in FIG. 4, the performance of the proposed diagonal decomposition method is slightly reduced, but the performance of the three-diagonal decomposition method is obviously better than that of the one-pair diagonal decomposition method.

Claims

1. A channel estimation method of a multi-band orthogonal frequency division multiplexing ultra-wideband system comprises the following steps: firstly, carrying out quadrature phase shift modulation processing on an input data signal at a transmitting end to obtain a modulation signal; secondly, performing serial-parallel conversion, inverse Fourier transform and parallel-serial conversion on the modulation signals in sequence to form a plurality of OFDM symbols; thirdly, inserting a length L into a plurality of formed OFDM symbols every set number of OFDM symbols_PThe m sequence s is taken as a training sequence, and a length L is added in front of the training sequence according to the channel characteristics_CThe cyclic prefix of (a) is obtained as a training sequence to which the cyclic prefix is added, and x is represented by x ═ x (0), x (1), … x (L)_P+L_C-1)](ii) a Finally, the training sequence x with the attached cyclic prefix and the formed OFDM symbol are transmitted to a receiving end through an ultra-wideband channel after being subjected to carrier modulation processing, and the training sequence x with the attached cyclic prefix and the OFDM symbol are influenced by channel fading and Gaussian white noise in the transmission process; at the receiving end, defining the training sequence x with attached cyclic prefix received by the receiving end and influenced by channel fading and white Gaussian noise as the first received signal, defining the OFDM symbol received by the receiving end and influenced by channel fading and white Gaussian noise as the second received signal, and representing the first received signal as the second received signal by using a tapped delay line model

Wherein k is 0, 1, …, L_p+L_c1, r (k) is the first received signal at time k, h represents a matrix vector composed of the coefficients of the various multipaths of the channel,

Wherein k is 0, 1, …, L_p+L_c-1，

<math> <mrow> <mrow> <mfenced open='' close='' separators=' '> <mtable> <mtr> <mtd> <msub> <mi>C</mi> <mi>P</mi> </msub> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> <mo>=</mo> <mrow> <mo>(</mo> <mn>1</mn> <mo>/</mo> <msub> <mi>L</mi> <mi>P</mi> </msub> <mo>)</mo> </mrow> <munderover> <mi>Σ</mi> <mrow> <mi>k</mi> <mo>=</mo> <mn>0</mn> </mrow> <mrow> <msub> <mi>L</mi> <mi>P</mi> </msub> <mo>-</mo> <mn>1</mn> </mrow> </munderover> <msup> <mi>s</mi> <mi>j</mi> </msup> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> </mrow> <msup> <mi>s</mi> <mi>i</mi> </msup> <mrow> <mo>(</mo> <mi>k</mi> <mo>)</mo> </mrow> </mtd> </mtr> <mtr> <mtd> <mo>=</mo> <mrow> <mfenced open='{' close=''> <mtable> <mtr> <mtd> <mn>1</mn> <mo>,</mo> </mtd> <mtd> <mi>i</mi> <mo>=</mo> <mi>j</mi> </mtd> </mtr> <mtr> <mtd> <mo>-</mo> <mn>1</mn> <mo>/</mo> <msub> <mi>L</mi> <mi>P</mi> </msub> <mo>,</mo> </mtd> <mtd> <mi>i</mi> <mo>&NotEqual;</mo> <mi>j</mi> </mtd> </mtr> </mtable> </mfenced> </mrow> </mtd> </mtr> </mtable> </mfenced> </mrow> <mo>,</mo> </mrow></math>

\tilde{h} = C_{p}^{- 1} C,

Wherein,

is an autocorrelation matrix C_PThe inverse matrix of (d); characterized in that it is based on said autocorrelation matrix C_PThe said autocorrelation matrix C_PDecomposing the matrix into the sum of a first matrix and a second matrix, recording the first matrix as D, and recording the second matrix as E, C_PD + E, said first matrix D and said second matrix E satisfying | D^-１E‖<1, calculating the autocorrelation matrix C_PInverse matrix of

Computing

To a first order approximation of the first order,

2. the method as claimed in claim 1, wherein the first matrix D is the autocorrelation matrix C_PThe second matrix E is the diagonal matrix formed by the autocorrelation matrix C_PThe diagonal matrix is marked as D₁Said off-diagonal matrix is denoted as E₁To obtain

To obtain

C_{p}^{- 1} = I - E_{1} .

3. The method as claimed in claim 1, wherein the first matrix D is the autocorrelation matrix C_PThe second matrix E is a three-diagonal matrix formed by the autocorrelation matrix C_PThe non-tri-diagonal matrix composed of elements except the tri-diagonal elements is marked as D₃And recording the non-tri-diagonal matrix as E₃To obtain

Combining the three diagonal matrix D₃Decomposed into autocorrelation matrix C_PAnd a diagonal matrix composed of diagonal elements of and said autocorrelation matrix C_PThe diagonal element of (2) is the sum of diagonal matrices composed of two pairs of diagonal elements of 0, and the diagonal matrix is marked as D₁And said diagonal matrix is denoted as D₂Calculating said tri-diagonal matrix D₃Inverse matrix of

Wherein, I is an identity matrix,

is a diagonal matrix D₁M 1, 2, …, ∞; then according to

Computing

To a first order approximation of the first order,

To obtain

D_{3}^{- 1} = I - D_{2};

Then according to

And

D_{3}^{- 1} = I - D_{2},

to obtain