CN103905355A - Virtual time reversal underwater sound OFDM channel equalization method - Google Patents

Abstract

The invention relates to a virtual time reversal underwater sound OFDM channel equalization method which is characterized by comprising the steps that 1. detecting signals for signal estimation are added into signals sent by a sending end; 2. a receiving end completes a synchronous process, and the received detecting signals are extracted, and according to the received detecting signals, channel impulse response is estimated; 3. according to the estimated channel impulse response, virtual time reversal OFDM channel equalization is completed; and 4. signals after equalization are subjected to serial-parallel conversion, circulating prefixes and circulating postfixes are removed, and a demodulation process is completed.

Description

A kind of virtual time reversal underwater sound OFDM channel equalization method

Technical field

The present invention relates to a kind of virtual time reversal underwater sound OFDM channel equalization method.

Background technology

Underwater acoustic channel have limited bandwidth, multipath serious interference, space-variant, time become, the feature such as change frequently, especially for Shallow Water Acoustic Channels, be subject to the impact of sea-floor relief and velocity of sound distribution, channel is very complicated, the many ways of channel time delay often reaches milliseconds up to a hundred, and underwater sound communication has been brought to serious interference.OFDM(OFDM) to have band efficiency high, anti-frequency selective fading, the advantages such as equilibrium is simple, be widely used in Short Range High Speed underwater sound communication, but intersymbol interference (ISI) and the inter-carrier interference (ICI) brought in order to overcome long many ways time delay, need to add and be greater than the most mostly Cyclic Prefix of elongatedness when way of channel, and insert closeer pilot tone and add error correction coding etc., this will cause the degradation of OFDM band efficiency, therefore, how to reduce many ways channel the interference of OFDM is become to one of key issue of high-speed underwater sound communication.

Channel equalization technique is the effective ways that overcome the interference bringing on the many ways of channel, and conventional OFDM channel equalization method is the frequency domain channel equalization algorithm based on pilot tone, by inserting known sequence in time domain or frequency domain, and estimates channel frequency response at receiving terminal.Existing Channel Equalization Algorithm, as least square (LS), least mean-square error (MMSE) etc., all need to add the Cyclic Prefix that is greater than the most most way of channel time delay to overcome ISI and ICI, and add closeer pilot tone to estimate channel, seriously limited the traffic rate of OFDM.

Summary of the invention

The object of the invention is to provide a kind of virtual time reversal underwater sound OFDM channel equalization method, can effectively shorten channel length, reduces the intersymbol interference that bring on many ways, improves OFDM band efficiency.

Realize the object of the invention technical scheme:

A kind of virtual time reversal underwater sound OFDM channel equalization method adds and is greater than channel Cyclic Prefix and the cyclic suffix of elongatedness when way the most mostly in OFDM symbol, it is characterized in that:

Step 1: add the detectable signal for Signal estimation during transmitting terminal transmits;

Step 2: receiving terminal completes synchronizing process, extracts the detectable signal receiving; According to the detectable signal receiving, estimate channel impulse response;

Step 3: according to the channel impulse response of estimating, complete anti-OFDM channel equalization when virtual;

Step 4: to the train of signal after equilibrium conversion, go Cyclic Prefix and cyclic suffix, complete demodulating process.

In step 1, select pseudo random sequence through after OFDM modulation as detectable signal.

In step 2, according to the detectable signal receiving, utilize matching pursuit algorithm to estimate amplitude, time delay and the phase place of channel, estimate channel impulse response.

In step 3, the channel impulse response time reversal of estimation, and receive signal convolution, complete anti-OFDM channel equalization when virtual.

The beneficial effect that the present invention has:

During transmitting terminal of the present invention transmits, add the detectable signal for Signal estimation, receiving terminal, according to the detectable signal receiving, is estimated channel impulse response, completes anti-OFDM channel equalization when virtual; Utilize virtual time reversal Channel Equalization Algorithm, can effectively shorten channel length, improve received signal to noise ratio, reduce ISI and ICI that many ways channel brings, improve OFDM band efficiency, realize adaptive channel equalizer.The present invention utilizes matching pursuit algorithm to estimate amplitude, time delay and the phase place of channel, and estimated accuracy is high, can estimate accurately channel impulse response, and inverse channel provides channel information accurately when virtual.

Accompanying drawing explanation

Fig. 1 is the basic principle figure of VTRM technology;

Fig. 2 is the frame assumption diagram that transmits;

Fig. 3 is that VTRM realizes block diagram for the receiver of OFDM channel equalization;

Fig. 4 is the schematic diagram that adds Cyclic Prefix and cyclic suffix in OFDM symbol.

Embodiment

Step 1: according to Virtual time reversal mirror (VTRM) principle, add the detectable signal for Signal estimation during transmitting terminal transmits;

As shown in Figure 1, the basic principle of VTRM technology is, before transmitting information signal, first sends a detectable signal, estimates channel impulse response according to detectable signal, then its time-reversal signal done to convolution with receiving signal, obtains the signal after anti-when virtual.

According to VTRM principle, design transmits frame structure as shown in Figure 2, frame head adopts linear frequency modulation (LFM) signal to carry out frame Timing Synchronization, add a pure-tone pulse (CW) signal for the estimating Doppler factor below, be used for eliminating the impact of Doppler frequency deviation at receiving terminal, being detectable signal afterwards, for estimating channel impulse response, is finally OFDM symbol.Between each signal, all leave certain protection interval, and protection interval is greater than the channel length of way time delay the most mostly.

Step 2: receiving terminal completes synchronizing process, extracts the detectable signal receiving; According to the detectable signal receiving, utilize matching pursuit algorithm to estimate amplitude, time delay and the phase place of channel, estimate channel impulse response.

Step 3: according to the channel impulse response of estimating, the signal after convolution is re-started synchronously, complete anti-OFDM channel equalization when virtual;

Step 4: to the train of signal after equilibrium conversion, go Cyclic Prefix and cyclic suffix, complete demodulating process.

When concrete enforcement, as shown in Figure 3, the signal of reception finds behind the initial position of signal by synchronizing signal, extract CW signal, obtain Doppler's compressibility factor by frequency measurement, according to Doppler factor, the signal receiving is become to sampling processing, eliminate the impact of Doppler frequency deviation on signal.Then from become the signal sampling, extract detectable signal, estimate channel impulse response, its time reversal and reception signal are done to convolution, complete VTRM channel equalization, train of signal after equilibrium conversion, go Cyclic Prefix and cyclic suffix, carry out FFT and separate the inverse mapping of mediation constellation, complete OFDM demodulating process.

Wherein, there is a key technology to affect the performance of VTRM for OFDM channel equalization, in OFDM symbol, should add cyclic suffix, and elongatedness when the way the most mostly of cyclic suffix and circulating prefix-length inverse channel all should be greater than time.Cyclic suffix by several data Replicas before OFDM symbolic blocks to after symbolic blocks, as shown in Figure 4.

If actual channel impulse response is h (t), the channel impulse response of estimation is

defining virtual channel impulse response time reversal is can be considered the channel of the final process of signal.Suppose direct sound wave amplitude maximum, length is that the discrete channel of L can be expressed as h=[h (0) h (1) ... h (L-1)], time inverse channel can be expressed as h '=[h ' (1-L) ... h ' (1) h ' (0) h ' (1) ... h ' (L-1)], its length is 2L-1, wherein, the sound ray of h ' (0) after for the stack of each path, amplitude maximum, can see time, inverse channel is a non-minimum phase channel, before the sound ray of amplitude maximum, still has many ways component.If when synchronous take the sound ray of energy maximum as initial time, ISI and ICI that many ways component after the maximum sound ray of energy causes can be overcome by Cyclic Prefix, and ISI and ICI that the many ways component after the maximum sound ray of energy causes need to be overcome by cyclic suffix.

How lower surface analysis once cyclic suffix overcomes ISI and ICI, and p OFDM symbol establishing current demodulation is r ^p, corresponding transmitting is expressed as s ^p, a previous and rear OFDM symbol table is shown s ^p-1, s ^p+1if, not adding Cyclic Prefix and cyclic suffix, after the channel that is h through impulse response, receiving signal can be expressed as

$r^p=h_N^p{s}^P+h_N^{p-1}{s}^{p-1}+h_N^{p+1}{s}^{p+1}+n^p---\left(1\right)$

Wherein, r ^p, s ^p, s ^p-1, s ^p+1and n ^pfor N × 1 dimensional vector,

with

for N × N ties up matrix, can be expressed as respectively

$h_N^{p-1}=\left(\begin{array}{ccccccc}0& \·\·\·& h\′(L-1)& h\′(L-2)& \·\·\·& h\′\left(2\right)& h\′\left(1\right)\\ 0& \·\·\·& 0& h\′(L-1)& \·\·\·& h\′\left(3\right)& h\′\left(2\right)\\ \·& \·& \·& \·& \·& \·& \·\\ \·& \·& \·& \·& \·& \·& \·\\ \·& \·& \·& \·& \·& \·& \·\\ 0& \·\·\·& 0& \·\·\·& \·\·\·& 0& h\′(L-1)\\ \·& \·& \·& \·& \·\·\·& \·& \·\\ \·& \·& \·& \·& \·\·\·& \·& \·\\ \·& \·& \·& \·& \·\·\·& \·& \·\\ 0& \·\·\·& 0& 0& \·\·\·& 0& 0\end{array}\right)---\left(3\right)$

$h_N^{p+1}=\left(\begin{array}{ccccccc}0& 0& \·\·\·& 0& 0& \·\·\·& 0\\ \·& \·& \·& \·& \·& \·& \·\\ \·& \·& \·& \·& \·& \·& \·\\ \·& \·& \·& \·& \·& \·& \·\\ h\′(1-L)& 0& \·\·\·& \·\·\·& 0& \·\·\·& 0\\ \·& \·\·\·& \·\·\·& & & \·\·\·& \·\\ \·& \·& \·& & & \·\·\·& \·\\ \·& \·& \·& & \·\·\·& \·\·\·& \·\\ h\′(-2)& \·& \·& h\′(1-L)& 0& \·\·\·& 0\\ h\′(-1)& h\′(-2)& \·\·\·& h\′(2-L)& h\′(1-L)& \·\·\·& 0\end{array}\right)---\left(4\right)$

From formula (1), can see Section 1 for desired signal, Section 2

and Section 3

for previous symbol and a rear intersymbol interference that symbol brings, n ^pfor noise item.Wherein, distracter

can be by adding Cyclic Prefix to overcome, and distracter need to add cyclic suffix to overcome.Be greater than Cyclic Prefix and the cyclic suffix of channel length L if add, receive signal remove circulation before and after sew after, can be expressed as

$r^p={\stackrel{\‾}{h}}_N{s}^p+n^p---\left(5\right)$

Wherein,

be circular matrix, can be expressed as

Can see, after removing and sewing before and after circulation, make in time domain the linear convolution of original transmitted signal and channel impulse response become circular convolution.Can find out from formula (5), the output of current sign piece is only relevant with the input of current sign piece, irrelevant with a previous and rear symbolic blocks, eliminate by Cyclic Prefix and cyclic suffix the ICI that bring on ISI that previous symbol and a rear symbol bring and many ways.

The key of VTRM channel equalization is channel estimating, the present invention adopts match tracing (Matching Pursuit, MP) algorithm is estimated channel impulse response, compared with the method that copies correlation estimation channel with conventional signal, estimated accuracy is high, and can estimate channel phase information, introduce the specific implementation process of MP algorithm below.

Consider the normal linear model using of Sparse Problems

y=Ax+v (7)

Wherein, x ∈ R ^mfor sparse signal to be estimated, y ∈ R ⁿfor observation vector, v ∈ R ⁿfor Gaussian noise vector, A ∈ R ^{n × M}, and N<M, A can be expressed as

A=[a ₁,a ₂,...,a _M] (8)

Wherein, a _i∈ R ⁿ, i=1,2 ..., M, claims that A is dictionary or former word bank, a conventionally _ifor the atom in dictionary.The basic thought of MP algorithm is in iterative process each time, from dictionary, find the atom mating most with signal to build sparse approaching, then obtain signal residual error, and the atom that continuation is selected and signal residual error is mated most in remaining atom, through after iteration repeatedly, be restructural sparse signal by observation vector and the atom selected.MP algorithm does not require that the atom in dictionary is orthogonal, but requires two norms || a _i|| ₂=1.

If the residual error after the p time iteration is r _p, be initialized as r ₀=y, the atom of the coupling of selecting from dictionary is

each selection in the former word bank of residue and the atom of residual signals inner product minimum,

$s_p=\arg \underset{j\&Element;I_{p-1}}{\max }\frac{<a_{s_j},r_{p-1}>}{{\left|\right|a_{s_j}\left|\right|}^2}---\left(9\right)$

Wherein, I _p-1∈ { s ₁, s ₂..., s _p-1the set of the selected matched atoms index of front p-1 iteration,

The p time iterative estimate goes out the element of signal x

can be expressed as

${\hat{x}}_p=\frac{<a_{s_p},r_{p-1}>}{{\left|\right|a_{s_p}\left|\right|}^2}---\left(10\right)$

Residual signals can be expressed as

$r_p=r_{p-1}{\hat{x}}_p{a}_{s_p}---\left(11\right)$

Work as residual signals || r _p|| ²when < ε, iteration stops, and ε is given residual error thresholding, is an amount relevant with input signal-to-noise ratio.According to above-mentioned analysis, the concrete steps of summing up MP algorithm are as follows:

1. initialization: set residual error thresholding ε, r ₀=y

2. select the atom of coupling: $s_1=\arg \underset{j=1,...,M}{\max }\frac{<a_{s_j},r_0>}{{\left|\right|a_j\left|\right|}^2},j=1,...M,I_1=\left\{s_1\right\}$

3. obtain the signal component of estimating:

4. residual error: $r_1=r_0-{\hat{x}}_1{a}_{s_1}$

5. the p time iteration, p>1

6. from remain former word bank, mate: $s_p=\arg \underset{j=1,...,M,j\&NotElement;I_{p-1}}{\max }\frac{<a_{s_j},r_{p-1}>}{{\left|\right|a_{s_j}\left|\right|}^2},I_p=\{I_{p-1},s_p\}$

7. the signal component of the p time iterative estimate:

8. the residual error of the p time iteration:

In order to use MP algorithm to estimate channel impulse response, first should construct a sparse signal model, consider that detectable signal x (n) is the channel of h (n) through channel impulse response, receiving signal y (n) can be expressed as

$y\left(n\right)=x\left(n\right)\&CircleTimes;h\left(n\right)+v\left(n\right),n=\mathrm{0,1},...,N---\left(12\right)$

Wherein,

represent convolution, Fourier transform is done in formula (12) both sides simultaneously, can be expressed as

Y=XH+V (13)

Wherein, Y and X are respectively the Fourier transforms of y (n) and x (n), and H is channel frequency response matrix, are the Fourier transforms of channel impulse response, can be expressed as

$H\left[i\right]=\underset{l=0}{\overset{L-1}{\mathrm{\&Sigma;}}}h\left[l\right]e^{-j2\mathrm{\&pi;il}/N},i=\mathrm{0,1},...,N-1---\left(14\right)$

Bring formula (14) into formula (13), can be expressed as

$Y=\hat{X}\mathrm{Fh}+V---\left(15\right)$

Wherein, be the diagonal matrix being made up of X, h can be expressed as

h=[h(0),h(1),...,h(L)] ^T (16)

Wherein, [] ^trepresent transposition, F is Fourier transform matrix, can be expressed as

Can see from derivation above, formula (15) meets the representation of sparse signal, and Y can be expressed as observing matrix,

can be expressed as dictionary, because X and Y are transmitting and the frequency domain representation that receives detectable signal, be complex matrix, therefore can estimate complex gain and the time delay of channel by MP algorithm.Now detectable signal should be chosen in frequency domain and have the signal of good autocorrelation, the present invention select a pseudo random sequence after OFDM modulation as detectable signal.

Claims (4)

1. a virtual time reversal underwater sound OFDM channel equalization method adds and is greater than the most mostly Cyclic Prefix and the cyclic suffix of elongatedness when way of channel in OFDM symbol, it is characterized in that:

Step 1: add the detectable signal for Signal estimation during transmitting terminal transmits;

Step 2: receiving terminal completes synchronizing process, extracts the detectable signal receiving; According to the detectable signal receiving, estimate channel impulse response;

Step 3: according to the channel impulse response of estimating, complete anti-OFDM channel equalization when virtual;

Step 4: to the train of signal after equilibrium conversion, go Cyclic Prefix and cyclic suffix, complete demodulating process.

2. virtual time reversal underwater sound OFDM channel equalization method according to claim 1, is characterized in that: in step 1, select pseudo random sequence through after OFDM modulation as detectable signal.

3. virtual time reversal underwater sound OFDM channel equalization method according to claim 2, is characterized in that: in step 2, according to the detectable signal receiving, utilize matching pursuit algorithm to estimate amplitude, time delay and the phase place of channel, estimate channel impulse response.

4. virtual time reversal underwater sound OFDM channel equalization method according to claim 3, is characterized in that: in step 3, the channel impulse response time reversal of estimation, and receive signal convolution, complete anti-OFDM channel equalization when virtual.

Images