CN102833016B - Determination method in allusion to homogenous signal time difference under multipath environment - Google Patents

Abstract

The invention relates to a determination method in allusion to homogenous signal time difference under a multipath environment. The determination method comprises the following steps of: establishing a received signal sample, determining residual errors including self-correlation processing and determination of auto-regression parameters and finally determining time difference of two paths of signals through acquiring a cross-correlation spectrum of residual errors of two paths of signals. According to the invention, compared with the prior art that two paths of received homogenous sampled signals are directly cross-correlated, the determination method disclosed by the invention has the characteristics that residual errors of two paths of sampled signals are cross-correlated by an auto-regression method to ensure that related peak shapes are sharp and are still easy to distinguish even though arrival time difference of each signal is shorter, thereby effectively improving the time difference measurement resolution rate and accuracy; and the time difference value measured by the determination method accords with the theoretical value. Therefore, the determination method has the characteristics of being high in resolution ratio of the cross-correlation spectrum of sampled signals under the multipath environment, effectively improving the signal time difference determination accuracy, providing accurate parameters for subsequent positioning of a target signal source and the like.

Description

A kind of under multi-path environment for the assay method of the same source signal time difference

Technical field

The invention belongs to the time difference assay method under multi-path environment in electronic information technical field, particularly a kind of under multi-path environment each signal receiver for the assay method of the time difference of same source signal.

Background technology

Along with the develop rapidly of signal arrival time difference location technology, radio position finding radio directional bearing technology is used widely in radio monitoring field.AOA(weighting vector positioning mode), TOA(ripple reaches timi requirement method), TDOA(ripple reaches digital method) be the cellular system Mobile Location Techniques that has development potentiality at present most.Efficiency or title ripple reach time delay, refer to the time difference causing due to transmission range difference between the same source signal that in receiver array, different receivers receive.Determine that the time difference is an important component part of input and parameter extraction in modern signal processing system, be widely used in the fields such as detection, communication, biomedicine and geophysics.

At present, time difference measurement method is more, and wherein broad sense cross-correlation method is comparatively classical.The method is to receiving Yi road signal and passing through following formula with another road signal of source signal transmitting:

$r_{x_1{x}_2}\left(k\right)=E\left[x_1\left(t\right){x_2}^{*}(t-k)\right]$

Directly carry out cross correlation process; In formula: x ₁and x ₂represent respectively the two paths of signals of same source signal through different paths, t represents current time, and t-k represents k chronomere before current time, represent the cross-correlation function of two paths of signals, and E[] represent to ask desired value computing, ^*represent plural conjugation.By the relatively similarity degree of two paths of signals of cross-correlation function waveform, the time offset of similarity maximum is corresponding time delay value.The method is because it has advantages of that algorithm is simple, amount of calculation is little, is easy to realize, processing gain is high and can be applied to low signal-to-noise ratio environment, and broad sense cross-correlation method has become at present both at home and abroad determining class methods that are most widely used in time difference information.Since Knapp in 1976 and the paper publishing of Cater about broad sense cross-correlation latency measurement, to determine that time delay and the exploration about the broad sense cross-correlation method of parameter are research directions active in signal process field always.When complete when orthogonal between reference signal and noise, noise and noise, the method is the best practice of time delay measurement.But in actual conditions, above-mentioned two complete orthogonality conditions are implacable, now in order to solve the mensuration problem of time delay, broad sense cross-correlation method need to be carried out albefaction processing, i.e. frequency domain weighting processing to the received signal, make reference signal and the peak shape more sharp-pointed (obviously) that receives signal correction, to reach the object that improves the resolution capability of time delay and the stability of time delay measurement.But in broad sense cross-correlation method, the structure of weighting function needs the concrete statistical information of signal and noise; But in practical engineering application, the statistical information of on-site signal and noise is often not yet learnt, can only replace the statistical information relating in weighting function with signal and one group of measured value of noise, thereby its actual performance and the method wish that the theoretical effect obtaining has larger gap conventionally.

In actual application environment, especially city or interior of building have compared with in the situation of multi-obstacle avoidance, in the air through processes such as the reflection of barrier, refraction, scatterings, often arrive the phenomenon of receiver with source signal with different delay through mulitpath, be called multipath effect.Multipath effect is that signal time delay is measured in actual applications inevitably problem.Signal that signal source is launched is through the propagation of mulitpath, after different reflector reflections, be the successively reception of same transducer, thereby formed mixed signal.In this mixed signal, comprise multiple experienced same source signal and noises that time delay, range coefficient and phase place change; Therefore the time difference that Dang Ge road signal arrives, hour by causing, the waveform of each road signal was overlapping, causes the defects such as relevant peaks deformation is wide, summit is level and smooth, and measure error is large, poor accuracy.

Summary of the invention

The object of the invention is the defect existing for background technology, improve design a kind of under multi-path environment for the assay method of the same source signal time difference, do not improve the time difference measure under the prerequisite of signal computational processing, improve time delay resolution, and effectively improve signal receiver and under multi-path environment, receive the accuracy that the signal time difference measures; And then provide the objects such as accurate parameter for the location in echo signal source.

Solution of the present invention is on the basis of background technology broad sense cross-correlation method, for the defect such as accuracy and measurement result precision deficiency of its specifying information of desired signal in practical engineering application, the present invention introduces autoregression method, utilize the stochastic variable linear combination in some moment in early stage to describe after the linear regression method of certain moment stochastic variable, the method is typically expressed as:

$s\left(t\right)=\underset{m=1}{\overset{M}{\mathrm{\Σ}}}{\mathrm{\α}}_{m}s(t-m)+v\left(t\right)$

Wherein s (t) is the signal of time point t, α _mrepresent m rank auto-regressive parameter, s (t-m) is the measuring-signal of front m the chronomere of current demand signal, m=1, and 2 ..., M, M is Autoregressive, v (t) is residual error;

Its auto-correlation function r _vv(k):

$r_{\mathrm{vv}}\left(k\right)=E\left[v\left(t\right)v^{*}(t-k)\right]=\left\{\begin{array}{cc}{\mathrm{\σ}}_v^2,& k=0\\ 0,& k\≠0\end{array}\right.$

There is the sharp-pointed characteristic in white noise coherence spectra peak, in formula: E[] represent to ask desired value computing, ^*represent plural conjugation, t represents current time, and t-k represents k chronomere before current time, represent the variance of white noise;

The signal that two receivers receive is in the method as follows:

x ₁(t)=s(t-t ₀)+n ₁(t)

$x_2\left(t\right)=\underset{j=1}{\overset{J}{\mathrm{\Σ}}}{g}_{2j}s(t-t_j)+n_2\left(t\right)$

Wherein: x ₁and x (t) ₂(t) be the two-way current demand signal that two receivers are received respectively, t is current time, n ₁(t), n ₂(t) be that two receivers receive and uncorrelatedly mutually in signals (to any time t and τ, have ) noise, s (t) represents the signal of signal source transmitting, t ₀represent the time difference that the reception signal of first receiver transmits with respect to signal source, t _jrepresent the time difference transmitting with respect to signal source along the signal of j paths propagation in the reception signal of second receiver, g _2jrepresent the signal amplitude coefficient that j paths is propagated, j=1,2 ..., J, J is number of path, ∑ represents summation;

The inventive method is by two-way sampled signal x ₁and x (t) ₂(t) replace with respectively:

Δx ₁(t)=x ₁(t)-α ₁x ₁(t-1)-α ₂x ₁(t-2)

Δx ₂(t)=x ₂(t)-α ₁x ₂(t-1)-α ₂x ₂(t-2)

Wherein: x ₁and x (t) ₂(t) be the current two paths of signals of receiving respectively of two receivers, x ₁and x (t-1) ₂(t-1) be the signal of the previous chronomere of the current demand signal received of two receivers, x ₁and x (t-2) ₂(t-2) signal of the first two chronomere of the current demand signal of receiving for each receiver, α ₁x ₁(t-1)+α ₂x ₁and α (t-2) ₁x ₂(t-1)+α ₂x ₂(t-2) represent respectively the current demand signal linear combination of variant chronomere received signal before; Δ x ₁and Δ x (t) ₂(t) be respectively x in autoregression method ₁and x (t) ₂(t) residual error;

Again the residual error of two paths of signals is directly carried out to cross correlation process, that is:

$r_{{\mathrm{\Δx}}_2\left(t\right){\mathrm{\Δx}}_1\left(t\right)}\left(k\right)=E\left[{\mathrm{\Δx}}_2\left(t\right){{\mathrm{\Δx}}_1}^{*}(t-k)\right]$

Wherein: represent Δ x ₁(t) with Δ x ₂(t) cross-correlation function between, t represents current time, t-k represents the moment of k unit interval before current time, E[] expression asks desired value computing, ^*represent plural conjugation.

The present invention utilizes residual error coherence spectra in the method to meet the sharp-pointed characteristic of white Gaussian noise coherence spectra peak value, improves thus the resolution of time difference measurement, thereby realizes its goal of the invention; Thereby the inventive method comprises:

Step 1. is set up and is received sample of signal: first determine two receivers 1,2 that carry out time difference measurement, when two receivers are received after the signal of same signal source transmitting, utilized analog to digital converter respectively the analog signal of receiver 1,2 outputs to be converted to digital signal, then to the two ways of digital signals after conversion according to expecting between two paths of signals that the peaked multiple setting-up time interval of the time difference carries out synchronized sampling, obtain two-way and receive the sample of signal 1,2;

Determining of step 2. residual error

2.1. auto-correlation processing: according to background technology, step 1 gained 1,2 two-way sampled signals are carried out to auto-correlation processing, obtain autocorrelation matrix, then matrix is carried out to singular value decomposition, obtain its singular value and corresponding eigenvectors matrix;

2.2. determine auto-regressive parameter: the corresponding characteristic vector of step 2.1 gained zero singular value is normalized, thereby obtains the each rank auto-regressive parameter in autoregression method;

2.3. obtain residual error: by normalization gained auto-regressive parameter substitution following formula in step 2.2:

${\mathrm{\Δx}}_1\left(t\right)=x_1\left(t\right)-\underset{m=1}{\overset{M}{\mathrm{\Σ}}}{\mathrm{\α}}_m{x}_1(t-m)$

${\mathrm{\Δx}}_2\left(t\right)=x_2\left(t\right)-\underset{m=1}{\overset{M}{\mathrm{\Σ}}}{\mathrm{\α}}_m{x}_2(t-m)$

Obtain respectively the residual delta x of two paths of signals 1,2 ₁and Δ x (t) ₂(t), wherein: x ₁(t) signal source of receiving at current time t for first receiver is at time point t-t ₀the same source signal sending, x ₂(t) be second receiver same signal source of receiving at current time t respectively at time point t-t ₁and t-t ₂the mixed signal of the same source signal sending, α _mrepresent m rank auto-regressive parameter, ∑ represents summation;

Determining of step 3. time difference

3.1. obtain the coherence spectra of two paths of signals 1,2 residual errors: pass through following formula according to the residual error of gained two paths of signals in step 2.3:

$r_{{\mathrm{\Δx}}_2\left(t\right){\mathrm{\Δx}}_1\left(t\right)}\left(k\right)=E\left[{\mathrm{\Δx}}_2\left(t\right){{\mathrm{\Δx}}_1}^{*}(t-k)\right]$

Carry out cross correlation process, obtain the coherence spectra of the residual error of two paths of signals; Wherein Δ x ₁(t) with Δ x ₂(t-k) represent respectively the residual error of two paths of signals 1,2, t represents current time, and t-k represents the time of k unit interval before current time, represent Δ x ₁(t) with Δ x ₂(t) cross-correlation function between; The value of the coherence spectra of residual error is determined by following formula:

${\hat{r}}_{{\mathrm{\Δx}}_2\left(t\right){\mathrm{\Δx}}_1\left(t\right)}\left(k\right)=\left\{\begin{array}{cc}\frac{1}{L+k}\underset{t=1}{\overset{L+k}{\mathrm{\Σ}}}{\mathrm{\Δx}}_2\left(t\right){{\mathrm{\Δx}}_1}^{*}(t-k),& -L+1\≤k\≤0\\ \frac{1}{L-k}\underset{t=k+1}{\overset{L}{\mathrm{\Σ}}}{\mathrm{\Δx}}_2\left(t\right){{\mathrm{\Δx}}_1}^{*}(t-k),& L-1\≥k>0\end{array}\right.$

Wherein: it is the cross-correlation function of residual error value, L represents sampling time interval, ∑ represents summation;

3.2. determine the time difference of two paths of signals: in the coherence spectra of step 3.1 gained two-way sampled signal residual error, the corresponding time value of peak-peak is the time difference of two paths of signals; The time difference that this time difference is two receivers 1,2 receives the signal of same signal source transmitting.

Above-mentioned to the two ways of digital signals after conversion according to expecting between two paths of signals that the peaked multiple setting-up time interval of the time difference carries out synchronized sampling, its sampling time interval is not less than 10 times of greatest expected time difference between two paths of signals.

The present invention is because the two-way homology sampled signal that background technology is received is directly carried out cross correlation process; Change into and utilize autoregression method, by the residual error of two-way sampled signal is carried out to cross correlation process, even if make time difference that the sharp-pointed Ge of relevant peak shape road signal arrives hour still be easy to differentiate, thereby effectively improved resolution and the accuracy of time difference measurement; The time difference value that sampled signal residual error coherence spectra of the present invention is surveyed (demonstration) conforms to theoretical value.Thereby the present invention to have the resolution of gained sampled signal coherence spectra under multi-path environment high, effectively improved the accuracy that the signal time difference measures, can be continue after provide the features such as accurate parameter to the location in echo signal source.

Brief description of the drawings

Fig. 1 is for adopting the background technology broad sense cross-correlation method emulation time difference to measure effect schematic diagram (broad sense coherence spectra curve chart);

Fig. 2 be adopt instantiation mode of the present invention with background technology the same terms under the time difference measure effect schematic diagram (the coherence spectra curve chart of residual error);

Negative sign (-) shown in Tu Zhong abscissa chronomere represents the delay of time.

Embodiment:

Present embodiment is to adopt second order autoregression method as example, that is:

s(t)＝α ₁s(t-1)+α ₂s(t-2)+v(t)

α ₁, α ₂represent respectively single order and second order auto-regressive parameter.

Autoregression model stability condition is: α ₂+ α ₁<1 and | α ₂| <1; In present embodiment, adopt autoregression method to produce the signal s (t) that signal source sends, single order and second order auto-regressive parameter are respectively α ₁=1.9114 and α ₂=-0.95, residual error v (t) is the white Gauss noise that variance equals 1.

Signal source is sent after signal, and the signal that wherein first receiver receives is x ₁(t); Second signal x that receiver receives ₂(t), for the signal of signal source transmitting is propagated the mixed signal forming through two paths, wherein in coherence spectra, time difference value corresponding to peak-peak is second receiver and receives that with respect to first receiver signal source sends the time-delay value of signal; As described in the background art, due to the existence of multipath effect, x ₂(t) two signals that experienced the same signal source transmitting of different paths and time delay, different range coefficients in, have been comprised; The time delay of receiving the signal of the same signal source transmitting arriving through two paths with respect to second receiver of first receiver is respectively t ₁-t ₀, t ₂-t ₀, range coefficient is respectively g ₂₁, g ₂₂;

The signal that two receivers receive is as follows:

x ₁(t)＝s(t-t ₀)+n ₁(t)

x ₂(t)=g ₂₁s(t-t ₁)+g ₂₂s(t-t ₂)+n ₂(t)

Wherein: x ₁(t) signal source of receiving at current time t for first receiver is at time point t-t ₀the same source signal sending, x ₂(t) be that second receiver same signal source of receiving at current time t is at time point t-t ₁and t-t ₂the mixed signal of the same source signal sending, n ₁(t), n ₂(t) be respectively two receivers and receive the noise in signal, s (t) represents the signal of signal source transmitting, t ₀the time difference when reception signal that represents first receiver transmits with respect to signal source, t ₁the time difference when reception signal of second receiver of expression transmits with respect to signal source, t ₂the multipath time difference when reception signal of second receiver of expression transmits with respect to signal source; T is set in this example ₀=0, t ₁=-20, t ₂=-30, g ₂₁=1, g ₂₂=0.8;

The flow process of the specific embodiment of the invention is as follows:

Step 1. is set up and is received sample of signal:

Choose two receivers 1,2 that carry out time difference measurement are received between the signal, two paths of signals of same signal source transmitting and expect that the maximum of the time difference is 30 chronomeres, the theoretical time difference of present embodiment is 20 chronomeres, receiver 1,2 carries out respectively after analog-to-digital conversion to the received signal, present embodiment adopts the expection time difference peaked 100 times (i.e. 3000 chronomeres) to set sampling time interval, carry out synchronized sampling, obtains the sample of two-way reception signal 1,2;

Determining of step 2. residual error

2.1. auto-correlation processing:

Two groups of sampled signals that receiver is received are carried out respectively auto-correlation computation processing, obtain corresponding autocorrelation matrix:

$R=\underset{t=4}{\overset{L}{\mathrm{\&Sigma;}}}{\left[\begin{array}{c}{x}_1(t-1)\\ x_1(t-2)\\ x_1(t-3)\end{array}\right]}^{*}\left[\begin{array}{ccc}{x}_1\left(t\right)& x_1(t-1)& x_1(t-2)\end{array}\right]$

Wherein: x ₁(t), t=1,2 ..., L represents the sampled signal of first receiver, and L represents sampling time interval, and for reducing amount of calculation, sampling time interval is elected peaked 100 times of the expection time difference as, L=3000 in present embodiment, ∑ represents summation;

2.2. determine auto-regressive parameter:

By the autocorrelation matrix obtaining in step 2.1, utilize residual sum noise signal self Gaussian Profile and separate characteristic between the two, obtain following formula:

$\left[\begin{array}{ccc}1& -{{\mathrm{\&alpha;}}_1}^{*}& -{{\mathrm{\&alpha;}}_2}^{*}\end{array}\right]R\left[\begin{array}{c}1\\ -{\mathrm{\&alpha;}}_1\\ -{\mathrm{\&alpha;}}_2\end{array}\right]=0$

Based on following Singular Value Decomposition Using theorem: the m × n rank matrix A that is r for Arbitrary Rank, all there are the upper (lower) triangular matrix R in unitary matrice U, V and r rank (L), make $A=U\left(\begin{array}{cc}R& 0\\ 0& 0\end{array}\right)V$ Or $A=U\left(\begin{array}{cc}0& 0\\ 0& L\end{array}\right)V;$ Matrix R is carried out to singular value decomposition:

R=USV

In present embodiment:

$U=\left[\begin{array}{ccc}-0.0666& 0.9168& 0.3938\\ -0.4978& 0.3115& -0.8094\\ -0.8647& -0.2499& 0.4356\end{array}\right],$ $S=\left[\begin{array}{ccc}18750.7& 0& 0\\ 0& 16285.5& 0\\ 0& 0& 29.8\end{array}\right],$ $V=\left[\begin{array}{ccc}0.8988& -0.0642& 0.4336\\ 0.4198& 0.4106& -0.8094\\ -0.1261& 0.9096& 0.3960\end{array}\right];$

Auto-regressive parameter vector is included in singular vector [1-α] corresponding to zero singular value of matrix R ^tin, wherein α=[α ₁α ₂], α ₁, α ₂represent respectively single order and second order auto-regressive parameter value; In present embodiment, choose the corresponding characteristic vector of singular value 29.8 [0.4336-0.8094 0.3960] that approaches zero most ^t, obtain auto-regressive parameter vector [1-α] in order to meet parameter setting in autoregression method, after being normalized ^t=[1-1.8665 0.9131] ^t, wherein α=[1.8665-0.9131] is auto-regressive parameter;

2.3. obtain residual error:

By the single order of step 2.2. gained and second order auto-regressive parameter α ₁=1.8665, α ₂in=-0.9131 substitution residual error expression formula as follows:

Δx ₁(t)=x ₁(t)-α ₁x ₁(t-1)-α ₂x ₁(t-2)

Δx ₂(t)=x ₂(t)-α ₁x ₂(t-1)-α ₂x ₂(t-2)

Wherein, x ₁and x (t) ₂(t) be the sampled signal of two receivers, x ₁and x (t-1) ₂(t-1) be the measuring-signal of the previous chronomere of two receiver current time t, x ₁and x (t-2) ₂(t-2) be the measuring-signal of the first two chronomere of two receiver current demand signals, Δ x ₁and Δ x (t) ₂(t) be the residual error of two paths of signals 1,2;

Determining of step 3. time difference

3.1. obtain the coherence spectra of the residual error of two paths of signals 1,2:

To Δ x ₂(t), Δ x ₁(t) carry out computing cross-correlation, utilize the separate character of receiver noise can obtain following formula:

$r_{{\mathrm{\&Delta;x}}_2\left(t\right){\mathrm{\&Delta;x}}_1\left(t\right)}\left(k\right)=E\left[{\mathrm{\&Delta;x}}_2\left(t\right){{\mathrm{\&Delta;x}}_1}^{*}(t-k)\right]$

$=E[g_{21}v(t-t_1)+g_{22}v(t-t_2)+n_2(t-k)-{\mathrm{\&alpha;}}_1{n}_2(t-1)-{\mathrm{\&alpha;}}_2{n}_2(t-2)){(t-k-t_0)}^{*}]$

$=g_{21}E\left[v(t-t_1)v^{*}(t-k-t_0)\right]+g_{22}E\left[v(t-t_2)v{(t-k-t_0)}^{*}\right]$

$=g_{21}{r}_{\mathrm{vv}}(k-(t_1-t_0))+g_{22}{r}_{\mathrm{vv}}(k-(t_2-t_0))$

Wherein: r _vv(k)=Ev (t) v ^*(t-k)] the auto-correlation function of expression v (t), represent Δ x ₂(t), Δ x ₁(t) the residual error coherence spectra between; The value of the coherence spectra of residual error is:

${\hat{r}}_{{\mathrm{\&Delta;x}}_2\left(t\right){\mathrm{\&Delta;x}}_1\left(k\right)}\left(k\right)=\left\{\begin{array}{cc}\frac{1}{L+k}\underset{t=1}{\overset{L+k}{\mathrm{\&Sigma;}}}{\mathrm{\&Delta;x}}_2\left(t\right){{\mathrm{\&Delta;x}}_1}^{*}(t-k),& -L+1\&le;k\&le;0\\ \frac{1}{L-k}\underset{t=k+1}{\overset{L}{\mathrm{\&Sigma;}}}{\mathrm{\&Delta;x}}_2\left(t\right){{\mathrm{\&Delta;x}}_1}^{*}(t-k),& L-1\&GreaterEqual;k>0\end{array}\right.$

Wherein it is the cross-correlation function of residual error estimated value, L represents sampling time interval, ∑ represent summation.

3.2. determine the time difference of two paths of signals: receiver 1,2 receives the signal of same signal source transmitting, two paths of signals maximum corresponding time of peak value in residual error coherence spectra is the time difference of two-way with source signal; The time difference that this time difference is two receivers 1,2 receives the signal of same signal source transmitting;

Fig. 1 adopts the basic cross-correlation method of background technology to carry out emulation to the time difference between two paths of signals to measure design sketch under the same conditions; Fig. 2 is present embodiment the time difference between two paths of signals is carried out to emulation mensuration design sketch

As can be seen from the figure: the broad sense coherence spectra curve chart that adopts background technology broad sense cross-correlation method to obtain in Fig. 1, the spectrum peak wider (summit is more level and smooth) of broad sense coherence spectra, poor to the definite accuracy of maximal peak point, and-20 chronomeres that time difference value corresponding to peak-peak is about-24 chronomeres, deviation theory value reach 4 chronomeres; And time difference value corresponding to peak-peak is-20 chronomeres in the coherence spectra of Fig. 2 present embodiment residual error, its time difference value of surveying conforms to theoretical value.

Claims (2)

1. Under multi-path environment for the assay method of the same source signal time difference, comprising:

   Step 1. is set up and is received sample of signal: first determine two receivers 1,2 that carry out time difference measurement, when two receivers are received after the signal of same signal source transmitting, utilized analog to digital converter respectively the analog signal of receiver 1,2 outputs to be converted to digital signal, then to the two ways of digital signals after conversion according to expecting between two paths of signals that the peaked multiple setting-up time interval of the time difference carries out synchronized sampling, obtain two-way and receive the sample of signal 1,2;

   Determining of step 2. residual error

   2.1. auto-correlation processing: step 1 gained 1,2 two-way sampled signals are carried out to auto-correlation processing, obtain autocorrelation matrix, then matrix is carried out to singular value decomposition, obtain its singular value and corresponding eigenvectors matrix;

   2.2. determine auto-regressive parameter: the corresponding characteristic vector of step 2.1 gained zero singular value is normalized, thereby obtains the each rank auto-regressive parameter in autoregression method; Described each rank auto-regressive parameter is by the autocorrelation matrix obtaining in step 2.1, utilizes residual sum noise signal self Gaussian Profile and separate characteristic between the two, obtains following formula:

   $\left[\begin{array}{ccc}1& -{{\mathrm{\&alpha;}}_1}^{*}& -{{\mathrm{\&alpha;}}_2}^{*}\end{array}\right]R\left[\begin{array}{c}1\\ -{\mathrm{\&alpha;}}_1\\ -{\mathrm{\&alpha;}}_2\end{array}\right]=0$

   Based on following Singular Value Decomposition Using theorem: the m × n rank matrix A that is r for Arbitrary Rank, all there are the upper (lower) triangular matrix R in unitary matrice U, V and r rank (L), make $A=U\left(\begin{array}{cc}R& 0\\ 0& 0\end{array}\right)V$ Or $A=U\left(\begin{array}{cc}0& 0\\ 0& L\end{array}\right)V;$ Matrix R is carried out to singular value decomposition:

   R＝USV

   Auto-regressive parameter vector is included in singular vector [1-α] corresponding to zero singular value of matrix R ^tin, wherein α=[α ₁α ₂], α ₁, α ₂represent respectively single order and second order auto-regressive parameter value;

   2.3. obtain residual error: by normalization gained auto-regressive parameter substitution following formula in step 2.2:

   $\mathrm{\&Delta;}{x}_1\left(t\right)=x_1\left(t\right)-\underset{m=1}{\overset{M}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_m{x}_1(t-m)$

   $\mathrm{\&Delta;}{x}_2\left(t\right)=x_2\left(t\right)-\underset{m=1}{\overset{M}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_m{x}_2(t-m)$

   Obtain respectively the residual delta x of two paths of signals 1,2 ₁and Δ x (t) ₂(t), wherein: x ₁(t) signal source of receiving at current time t for first receiver is at time point t-t ₀the same source signal sending, x ₂(t) be second receiver same signal source of receiving at current time t respectively at time point t-t ₁and t-t ₂the mixed signal of the same source signal sending, α _mrepresent m rank auto-regressive parameter, ∑ represents summation;

   Determining of step 3. time difference

   3.1. obtain the coherence spectra of two paths of signals 1,2 residual errors: pass through following formula according to the residual error of gained two paths of signals in step 2.3:

   $r_{{\mathrm{\&Delta;x}}_2\left(t\right)\mathrm{\&Delta;}{x}_1\left(t\right)}\left(k\right)=E\left[{\mathrm{\&Delta;x}}_2\left(t\right)\mathrm{\&Delta;}{x_1}^{*}(t-k)\right]$

   Carry out cross correlation process, obtain the coherence spectra of the residual error of two paths of signals; Wherein Δ x ₁(t) with Δ x ₂(t-k) represent respectively the residual error of two paths of signals 1,2, t represents current time, and t-k represents the time of k unit interval before current time, represent Δ x ₁(t) with Δ x ₂(t) cross-correlation function between; The value of the coherence spectra of residual error is determined by following formula:

   ${\hat{r}}_{{\mathrm{\&Delta;x}}_2\left(t\right)\mathrm{\&Delta;}{x}_1\left(t\right)}\left(k\right)=\left\{\begin{array}{c}\frac{1}{L+k}\underset{t=1}{\overset{L+k}{\mathrm{\&Sigma;}}}\mathrm{\&Delta;}{x}_2\left(t\right)\mathrm{\&Delta;}{x_1}^{*}(t-k),-L+1\&le;k\&le;0\\ \frac{1}{L-k}\underset{t=k+1}{\overset{L}{\mathrm{\&Sigma;}}}\mathrm{\&Delta;}{x}_2\left(t\right)\mathrm{\&Delta;}{x_1}^{*}(t-k),L-1\&GreaterEqual;k>0\end{array}\right.$

   Wherein: it is the cross-correlation function of residual error value, L represents sampling time interval, ∑ represents summation;

   3.2. determine the time difference of two paths of signals: in the coherence spectra of step 3.1 gained two-way sampled signal residual error, the corresponding time value of peak-peak is the time difference of two paths of signals; The time difference that this time difference is two receivers 1,2 receives the signal of same signal source transmitting.
2. By described in claim 1 under multi-path environment the assay method for the same source signal time difference, it is characterized in that described in step 1 according to expecting between two paths of signals that the peaked multiple setting-up time interval of the time difference carries out synchronized sampling, its sampling time interval is not less than 10 times of greatest expected time difference between two paths of signals.