CN101388687B - Chip implementing fast high-efficient non-cooperative signal-noise ratio direct spreading signal detection method - Google Patents

Abstract

The invention relates to a method for detecting direct sequence signals, which can be realized by a chip and has rapid speed, high efficiency, none cooperation, low signal-to-noise ratio. The method comprises the following steps: pre-collating received data, pre-calculating FFT, estimating consolidation biased autocorrelation functions, synthesizing consolidation unbiased autocorrelation functions, calculating four-level cross-correlation statistics quantity, and supposing examination. The method uses a covering principle (MCT) with FFT as a core and uses maximum cost to calculate main elements in a detector with optimum principles, and then an optimum detector can be rapidly and approximately realized through detecting symbolic code period parameters. The saved calculation amount and the size of a memory of the method are two magnitude orders more than that of the optimum detector. The mode of modulating signals has wide application range, wide Doppler frequency shift is tolerated, the constant false alarm rate and the detecting precession of the parameter are high, the detecting performance is much better than an existing symbolic code period detector, and the problem that OPSK or higher digital phase modulation DSSS signals can not be detected by a carrier frequency detector under low signal-to-noise ratio is solved.

Description

The attainable fast high-efficient non-cooperative signal-noise ratio direct spreading signal detecting method of a kind of chip

Technical field

This method belongs to communication technical field, specifically is applied to frequency range management, and signal is intercepted and the distress message analysis, also can be used for the sensor network of self-organizing or as the implementation of spread spectrum communication network of new generation.Be under low signal-to-noise ratio very, the directly approximate a kind of Fast implementation that realizes the Direct Sequence Spread Spectrum Signal optimum detector of all unknown parameters.

DSSS (DSSS, be called for short directly expand) signal is as a kind of signal that typically has secured feature, in many non-collaboration application occasions, if can obtain the parameter such as the carrier frequency f of DSSS signal in advance unlike the cooperation recipient _c, chip rate R, pseudo-code cycle T _pAnd pseudo-code sequence is such, and then signal is difficult to be detected under low signal-to-noise ratio.The attainable detection method of non-cooperation chip that this patent proposes has realized a kind of detection principle of novelty, has fast, the suitable characteristics that signal to noise ratio is low, the adaptation parameter scope is wide of detection speed, also can obtain the pseudo-code cycle T simultaneously _pThe accurate estimation of this key parameter, estimated accuracy is far above the unit sampling interval.The application scenario that this patent algorithm relates to comprises: radio monitoring, signal of communication is scouted, satellite-signal intercepting and capturing etc.Its application simultaneously also comprises some new DSSS communication system of design, such as parameter or (with) the DSSS communications platform that becomes during pseudo-code, thereby obtain good anti-interception capability and antijamming capability; Perhaps as a kind of effective detection mode under the cooperative communication sync fail state, like emergency processing in danger.

Background technology

All the time, utilize algorithm that the pseudo-code periodic characteristic studies the DSSS input seldom, what a large amount of on the contrary Weak-signal detectors was directed against is common psk signal.Mainly contain two reasons: at first, support with theoretical about the systematic method of the detection of psk signal, such as the cyclic spectral detector; Simultaneously in engineering, also there are some practical algorithm, like the carrier frequency detector, the rate thread detector; Realize simple; Obtained number of applications, but there is great limitation in these detectors when being applied to the DSSS input: the carrier frequency detector is very responsive to Doppler frequency displacement and narrow band interference, and is not suitable for high-order PSK such as the QPSK modulation of widespread usage; The rate thread detector detects poor performance under low signal-to-noise ratio, and its performance is subject to pulse shaping and channel circumstance influence.On the other hand; Up to the present, the pseudo-code cycle detection device performance that existing method proposes is not good, and operand is bigger than normal; Do not have advantage compared with the psk signal detector, the main purpose of therefore studying pseudo-code cycle detection device is still in order to be used for estimating the pseudo-code cycle after detecting completion.Simultaneously, because the ratio of precision of estimating is lower, also be difficult to as reliable estimated value of pseudo-code cycle.At first briefly introduce existing non-cooperation DSSS signal detecting method below:

1. existing P SK signal detector is used for the detection of non-cooperation DSSS signal

Existing P SK signal detector mainly contains two kinds, and a kind of is the carrier frequency detector, if DSSS signal to be detected is the BPSK modulation system, then behind the signal square, its periodogram has spectral line and occurs at twice carrier frequency place, detects this spectral line and has or not and get final product; Another kind is the rate thread detector, and after signal and the delay conjugation of self matched, its periodogram had spectral line and occurs at the chip rate place, also can be used to detect the DSSS signal and have or not.If DSSS signal modulation system to be detected is the QPSK or the psk signal of high-order more; Quadratic method lost efficacy; The carrier frequency spectral line can more recover on high-order (QPSK the is a quadravalence) meaning, but well-known, exponent number is high more; The ability that suppresses noise more a little less than, so the carrier frequency detector is not suitable for the low signal-to-noise ratio occasion of high order modulation.On the other hand, rate thread detector performance is not good.Such as in practical application, the carrier frequency detector of frequency multiplication detects-20dB the low energy of BPSK-DSSS signal, and the rate thread detector only can detect to-10dB.Therefore, as far as the BPSK-DSSS signal, the carrier frequency detector is a kind of feasible scheme.But for the PSK-DSSS signal of other form, the carrier frequency detector performance under low signal-to-noise ratio (＜-10dB) can't bear to use.The theory diagram of two kinds of detectors is respectively like Fig. 1, shown in Figure 2.

2. have the method for non-cooperation DSSS signal pseudo-code cycle detection now

For describing conveniently, consider following DSSS signal discrete model:

Y, (k) expression is by the DSSS discrete signal of Gaussian noise pollution, and p is the average power of DSSS signal, { c (k) }, k=0,1;, L-1 is that length is the PSK pseudo-code sequence of L, and mod (k, L) expression k gets remainder to L, { s (k) }; K=0,1 ..., the PSK information code sequence that M-1 is

The maximum integer that expression is not more than, f _cWith

_cBe respectively carrier frequency frequency and first phase, n (k) is independent identically distributed multiple Gaussian noise sequence, and N representes that total sample counts.

At present, existing non-cooperation DSSS signal pseudo-code cycle detection algorithm mainly comprises:

(1). the fluctuation correlation method

Become a plurality of non-overlapped data windows (to satisfy the long L of window the DSSS division of signal ₁＞2L)

I=1,2 ..., M ₁(M ₁Be the data window number), obtain the auto-correlation function of each data window signal $r_i\left(\mathrm{\τ}\right)=\frac{1}{L_1}\underset{k=1}{\overset{L_1}{\mathrm{\Σ}}}{\stackrel{\→}{y}}_i\left(k\right){\stackrel{\→}{y}}_i^{*}(k-\mathrm{\τ}),$ A sample as signal auto-correlation function; Y wherein ^*The complex conjugate of expression y; Auto-correlation function to all data windows are obtained is made even all, tries to achieve the estimated value of signal auto-correlation function mould square $\mathrm{\ρ}\left(\mathrm{\τ}\right)=\frac{1}{M_1}\underset{i=1}{\overset{M_1}{\mathrm{\Σ}}}{\left|r_i\left(\mathrm{\τ}\right)\right|}^2;$ ρ (τ) is carried out peak value searching, in the ideal case, only peak value occurs, therefore carry out peak value and detect, and obtain the PN sign indicating number cycle according to the Minimum distance estimation, between the peak value at the integral multiple place in pseudo-code cycle, as shown in Figure 3.

(2). the secondary power spectrometry

Similar relevant fluctuation method becomes a plurality of non-overlapped data windows with the DSSS division of signal, tries to achieve the periodogram of signal in the window and estimates ${\mathrm{\φ}}_i\left(\mathrm{\ω}\right)=\frac{1}{L_1}{\left|\mathrm{FT}\left({\stackrel{\→}{y}}_i\left(k\right)\right)\right|}^2$ (FT representes Fourier Tranform); To fenestrate in the secondary power spectral line property of signal average, obtain $\mathrm{\ρ}\left(\mathrm{\τ}\right)=\frac{1}{M_1{L}_1}\underset{i=1}{\overset{M_1}{\mathrm{\Σ}}}\left[\mathrm{FT}\left({\mathrm{\φ}}_i\left(\mathrm{\ω}\right)\right)\right];$ ρ (τ) is carried out peak value searching, go out the PN sign indicating number cycle according to the distance detecting between the peak value.Utilize the corresponding relation of periodogram estimation and auto-correlation function estimated power spectrum; Therefore can prove; This method is consistent with the essence of fluctuation correlation method; But owing to directly utilize periodogram to have the problem (available correlation diagram is estimated alternative the elimination, but so just in full accord with relevant fluctuation method) of peak aliasing, the correlation method that therefore do not fluctuate is convenient.

(3). the cepstrum method

Same similar relevant fluctuation method becomes a plurality of non-overlapped data windows with the DSSS division of signal, and signal in each window is asked cepstrum $C_i\left(\mathrm{\τ}\right)={\left|\mathrm{FT}\right[\mathrm{Ln}{\left|\mathrm{FT}\right[{\stackrel{\→}{y}}_i\left(k\right)\left]\right|}^2\left]\right|}^2;$ The cepstrum of trying to achieve in each window is asked linear averaging $\mathrm{\ρ}\left(\mathrm{\τ}\right)=\frac{1}{M_1}\underset{i=1}{\overset{M_1}{\mathrm{\Σ}}}{C}_i\left(\mathrm{\τ}\right);$ ρ (τ) is carried out peak value searching, go out the PN sign indicating number cycle according to the distance estimations between the peak value.

Emulation shows, and is best with fluctuation correlation method effect in the above algorithm.Comprehensive said method is not difficult to find out, this several method has all utilized the periodic characteristic of the quite sharp-pointed and pseudo-code sequence in pseudo-code sequence auto-correlation peak.The shortcoming of above algorithm mainly is that performance is not good when low signal-to-noise ratio, can only accomplish-15dB (under code length L≤1023 situation) above (seeing table 1).These methods are difficult to be used as the main detection means of low signal-to-noise ratio DSSS signal.In addition, if consider in the practical application sample rate and chip rate often an odd lot doubly concern, so the pseudo-code cycle that above-mentioned algorithm estimates only is the integer approximation value in true pseudo-code cycle.Because error is bigger, so these algorithms also are difficult to accurate estimating of task of competent pseudo-code cycle.

In a word, there are many weakness in existing DSSS signal detecting method, and particularly performance also has suitable distance apart from practicability when low signal-to-noise ratio and high-order digit modulation.

Summary of the invention:

The objective of the invention is to overcome foregoing problems, on the DSSS signal optimum detector basis of unknown all parameters, provide a kind of based on the attainable method for quick of chip.

The core of this method is to have designed a kind of Rapid Realization mode of DSSS signal optimum detector dexterously---netted covering table fast algorithm (fast Mesh-Covered table algorithm; Be called for short the MCT algorithm); It has utilized FFT computing capability efficiently, the FFT arithmetic element is stitched together according to certain pattern carries out computing then.It is quite superior that this method detects performance: can realize the very detection in low signal-to-noise ratio DSSS signal pseudo-code cycle by enough quite short samples; Also existing other method under the same detection performance of amount of calculation simultaneously, and can obtain the pseudo-code cycle parameter of estimated accuracy simultaneously far above existing other method.

Different with aforementioned pseudo-code cycle detection algorithm based on the frequency domain processing; This detection method is regarded the DSSS signal as the sign indicating number sequence that Memorability is arranged of carrying bulk redundancy information; Through some specific computings on the time domain, make full use of redundant information and remove noise, thereby obtain good inhibition noise ability.

Describe for convenient, discrete DSSS signal y (k) can be write as column vector form y, uses y _iI element among the expression column vector y, i=1,2 ..., N, and point of each pseudo-code chip samples establish L ₀For a near value the pseudo-code period L, might as well establish $L\∈[\frac{2}{3}{L}_0,\frac{4}{3}{L}_0],$ M ₀=N/L ₀Through the theoretical derivation of optimum detector, structure statistic ρ (L ₀) as detection statistic:

Wherein γ is a decision threshold;

hypothesis when being no signal,

is the hypothesis when signal is arranged.The performance of this detector is outstanding, the detector that robustness is very good, widely applicable.Theoretical research shows: it is that the PSK-DSSS signal is tending towards optimum detector (the LMPI test under 0 in signal to noise ratio; Local maximal potential invariant test); Be its optimal performance at any noise variance, all be consistent under carrier frequency and the first phase, pseudo-code sequence and information code sequence.Simultaneously, this detector also is CFAR (CFAR) and influence that do not receive the PSK modulation type.Its computation complexity is approximately O (NM).

Amount of calculation that this detector is required and memory space are all very big, and are difficult to practical application.The MCT implementation of the present invention's design can be similar to and realize this non-cooperative signal-noise ratio direct spreading signal detector.Technical scheme of the present invention is as shown in Figure 4; Signal gets into receiver through antenna; After intermediate frequency quadrature reception and A/D sampling,, utilize the fpga chip implementation algorithm then with the sampled data input store; Generate the decision statistic amount at last and accomplish the detection of signal, estimate the pseudo-code cycle parameter of DSSS signal simultaneously.It is characterized in that this method comprises following steps successively in processor:

Step 1. receive the data preliminary finish, in processor, data are carried out preliminary treatment, can be through filtering and albefaction function, inhibition out-of-band noise, and the noise contribution that the makes the actual reception data suitable theoretical model of coincidence detection method more.Then that data are calculative structural segmented according to the MCT algorithm, and every section zero padding storage respectively, so that make processor can efficiently handle corresponding FFT computing.

Step 2. FFT calculates in advance.Every segment data is carried out the FFT computing respectively, deposit the result in memory.Because this group will be the result will use repeatedly, once all realize in advance, and computational efficiency is improved greatly.

Step 3. becoming during estimation has partial autocorrelation function.Be exactly in fact according to the MCT algorithm with corresponding data conjugation dot product and get the IFFT conversion.

Step 4. become no partial autocorrelation function when synthetic.According to the MCT algorithm, the adjacent partial autocorrelation function that has is combined into no partial autocorrelation function.

Step 5. calculate each item quadravalence cross covariance statistic in the optimal detection amount.Promptly after repeated execution of steps three and step 4, institute become sometimes obtain on the no partial autocorrelation function basis pseudo-code cycle the quadravalence cross covariance function of possible value correspondence.

Step 6. the check judgement.Detection statistic that obtains at last and thresholding are compared.Because this detector is CFAR, promptly thresholding does not rely on DSSS signal parameter and background noise size, so threshold value adopts the mode that presets to be solidificated in the program.In this method be to the pretreated method feature of data:

Signal gets into the intermediate frequency quadrature receiver through antenna, and through the A/D sampling, behind digital filtering and the noise whitening, sampled data is carried out data rearrangement then, and input store is so that the utilization subsequent treatment.Below will receive data and represent (y with vectorial y _nN point among the expression y).Order $L^{\′}=\underset{k\∈N}{\min }\left(p^k\right)>{2L}_0-1,$ Define L ' * (M respectively ₀-1) dimension matrix A ⁽¹⁾, A ⁽²⁾And matrix B ⁽¹⁾, B ⁽²⁾P is the substrate of the FFT conversion supported of platform, is generally 2.

$A^{\left(1\right)}=\left[\begin{array}{cccc}{y}_0& y_{L_0}& \·\·\·& y_{(M_0-2)L_0}\\ y_1& y_{L_0+1}& \·\·\·& y_{(M_0-2)L_0+1}\\ \·& \·& & \·\\ \·& \·& \·\·\·& \·\\ \·& \·& & \·\\ y_{L_0-1}& y_{2L_0-1}& \·\·\·& y_{(M_0-1)L_0-1}\\ 0& 0& 0& 0\\ \·& \·& \·& \·\\ \·& \·& \·& \·\\ \·& \·& \·& \·\\ 0& 0& 0& 0\end{array}\right],$ $B^{\left(1\right)}=\left[\begin{array}{cccc}{y}_{L_0/2}& y_{{3L}_0/2}& \·\·\·& y_{(M_0-3/2)L_0}\\ y_{L_0/2+1}& y_{{3L}_0/2+1}& \·\·\·& y_{(M_0-3/2)L_0+1}\\ \·& \·& & \·\\ \·& \·& \·\·\·& \·\\ \·& \·& & \·\\ y_{{3L}_0/2-1}& y_{5L_0/2-1}& \·\·\·& y_{(M_0-1/2)L_0-1}\\ 0& 0& 0& 0\\ \·& \·& \·& \·\\ \·& \·& \·& \·\\ \·& \·& \·& \·\\ 0& 0& 0& 0\end{array}\right]$

$A^{\left(2\right)}=\left[\begin{array}{cccc}{\left|y_0\right|}^2& {\left|y_{L_0}\right|}^2& \·\·\·& {\left|y_{(M_0-2)L_0}\right|}^2\\ {\left|y_1\right|}^2& {\left|y_{L_0+1}\right|}^2& \·\·\·& {\left|y_{(M_0-2)L_0+1}\right|}^2\\ \·& \·& & \·\\ \·& \·& \·\·\·& \·\\ \·& \·& & \·\\ {\left|y_{L_0-1}\right|}^2& {\left|y_{2L_0-1}\right|}^2& \·\·\·& {\left|y_{(M_0-1)L_0-1}\right|}^2\\ 0& 0& 0& 0\\ \·& \·& \·& \·\\ \·& \·& \·& \·\\ \·& \·& \·& \·\\ 0& 0& 0& 0\end{array}\right],$ $B^{\left(2\right)}=\left[\begin{array}{cccc}{\left|y_{L_0/2}\right|}^2& {\left|y_{3L_0/2}\right|}^2& \·\·\·& {\left|y_{(M_0-3/2)L_0}\right|}^2\\ {\left|y_{L_0/2+1}\right|}^2& {\left|y_{3/L_0/2+1}\right|}^2& \·\·\·& {\left|y_{(M_0-3/2)L_0+1}\right|}^2\\ \·& \·& & \·\\ \·& \·& \·\·\·& \·\\ \·& \·& & \·\\ {\left|y_{{3L}_0/2-1}\right|}^2& {\left|y_{5L_0/2-1}\right|}^2& \·\·\·& {\left|y_{(M_0-1/2)L_0-1}\right|}^2\\ 0& 0& 0& 0\\ \·& \·& \·& \·\\ \·& \·& \·& \·\\ \·& \·& \·& \·\\ 0& 0& 0& 0\end{array}\right]$

And establish $A^{\left(1\right)}=\begin{array}{c}\left[\begin{array}{ccc}{a}_0^{\left(1\right)}& \·\·\·& a_{M_0-2}^{\left(1\right)}\end{array}\right],\end{array}$ $A^{\left(2\right)}=\begin{array}{c}\left[\begin{array}{ccc}{a}_0^{\left(2\right)}& \·\·\·& a_{M_0-2}^{\left(2\right)}\end{array}\right],\end{array}$ $B^{\left(1\right)}=\begin{array}{c}\left[\begin{array}{ccc}{b}_0^{\left(1\right)}& \·\·\·& b_{M_0-2}^{\left(1\right)}\end{array}\right],\end{array}$ $B^{\left(2\right)}=\begin{array}{c}\left[\begin{array}{ccc}{b}_0^{\left(2\right)}& \·\·\·& b_{M_0-2}^{\left(2\right)}\end{array}\right].\end{array}$ A ⁽¹⁾With B ⁽¹⁾, A ⁽²⁾With B ⁽²⁾L just in time staggers ₀/ 2, so that in step 4, can effectively suppress information code initial phase difference to detecting the fluctuation of performance.

In this method preparatory FFT Calculation Method is characterized as:

Realize A respectively by the fpga chip module ⁽¹⁾, A ⁽²⁾, B ⁽¹⁾, B ⁽²⁾Base-p FFT the computing of each row.Obtain: ${\tilde{A}}^{\left(1\right)}=\begin{array}{c}\left[\begin{array}{ccc}{\tilde{a}}_0^{\left(1\right)}& \·\·\·& {\tilde{a}}_{M_0-2}^{\left(1\right)}\end{array}\right],\end{array}$ ${\tilde{A}}^{\left(2\right)}=\begin{array}{c}\left[\begin{array}{ccc}{\tilde{a}}_0^{\left(2\right)}& \·\·\·& {\tilde{a}}_{M_0-2}^{\left(2\right)}\end{array}\right],\end{array}$ ${\tilde{B}}^{\left(1\right)}=\begin{array}{c}\left[\begin{array}{ccc}{\tilde{b}}_0^{\left(1\right)}& \·\·\·& {\tilde{b}}_{M_0-2}^{\left(1\right)}\end{array}\right],\end{array}$ ${\tilde{B}}^{\left(2\right)}=\begin{array}{c}\left[\begin{array}{ccc}{\tilde{b}}_0^{\left(2\right)}& \·\·\·& {\tilde{b}}_{M_0-2}^{\left(2\right)}\end{array}\right],\end{array}$ Then the result is sent back in the memory.

Become when estimating in this method method feature that partial autocorrelation function is arranged into: being this step of intuitivism apprehension, describing (shown in Figure 5 in the description of drawings) with figure here, is (m, the corresponding y that is defined as of some n) with coordinate on the plane _m ^*y _nWith | y _m| ²| y _n| ²( ^*The expression conjugation), m=0 ..., N-1, n=0 ..., N-1.And definition has inclined to one side estimation auto-correlation function:

$r_{m,n}^{\left(1\right)}\left(k\right)=\left\{\begin{array}{cc}\underset{i=0}{\overset{L_0-k}{\mathrm{\&Sigma;}}}{y}^{*}(m+k+i)y(n+i)& 0\&le;k\&le;L_0-1\\ \underset{i=1}{\overset{L_0-k}{\mathrm{\&Sigma;}}}{y}^{*}(m+i)y(n-k+i)& -L_0+1\&le;k<0\end{array}\right.$

With

$r_{m,n}^{\left(2\right)}\left(k\right)=\left\{\begin{array}{cc}\underset{i=0}{\overset{L_0-k}{\mathrm{\&Sigma;}}}|{y(m+k+i)|}^2{\left|y(n+i)\right|}^2& 0\&le;k\&le;L_0-1\\ \underset{i=1}{\overset{L_0-k}{\mathrm{\&Sigma;}}}{\left|y(m+i)\right|}^2{\left|y(n-k+i)\right|}^2& -L_0+1\&le;k<0\end{array}\right.$

Use length to be L ₀Empty frame each point on the plane is carried out piecemeal (among Fig. 5 with L ₀=4 is example), calculate the inclined to one side estimation auto-correlation function of having of each empty frame r _{M, n} ⁽¹⁾(k) and r _{M, n} ⁽²⁾(k) r _{M, n} ⁽¹⁾(k).Calculating has the process of inclined to one side estimation auto-correlation function: read successively from memory by the fpga chip module

With And their corresponding element multiplied each other, and do base-p IFFT conversion, the sequence that obtains is [r _{M, n} ⁽¹⁾(0) ... R _{M, n} ⁽¹⁾(L ₀-1) 0 ... 0r _{M, n} ⁽¹⁾(L ₀+ 1) ... R _{M, n} ⁽¹⁾(1)] ^TIn like manner, respectively with each vector With Corresponding element multiplies each other, and does base-p IFFT conversion, and the sequence that obtains is [r _{M, n} ⁽²⁾(0) ... R _{M, n} ⁽²⁾(L ₀-1) 0 ... 0r _{M, n} ⁽²⁾(L ₀+ 1) ... R _{M, n} ⁽²⁾(1)] ^T

The method feature that becomes no partial autocorrelation function in this method when synthetic into:

Calculate the auto-correlation function that (part) nothing is estimated partially With Every L on the same diagonal line about being or in neighbouring two empty frames ₀The sum of individual next-door neighbour's element (being the sum of per 4 next-door neighbours' element among Fig. 5), (m n) is defined as y when coordinate points _m ^*y _nThe time tried to achieve and do When coordinate points (m n) is defined as | y _m| ²| y _n| ²The time tried to achieve and do

All of trying to achieve like this

With

Can represent that also the coordinate meaning that Fig. 6 and Fig. 5 represent is in full accord with Fig. 6, promptly more corresponding one by one, only in order more succinctly intuitively to represent each element with [].Work as m=0, L ₀..., (M-2) L ₀, n=L ₀/ 2,3L ₀/ 2 ..., (M-3/2) L ₀The time, have $\stackrel{\&OverBar;}{\mathrm{\&alpha;}}(m+k,n)=r_{m,n}^{\left(1\right)}\left(k\right)+r_{m+L_0,n}^{\left(1\right)}(k-L_0),k=1,\&CenterDot;\&CenterDot;\&CenterDot;,L_0-1.$ This pair relationhip shows with empty tiltedly frame table respectively in Fig. 6, and has marked

as example.

$\stackrel{\&OverBar;}{\mathrm{\&alpha;}}(m,n)=r_{m,n}^{\left(1\right)}\left(k\right),k=0$

$\stackrel{\&OverBar;}{\mathrm{\&alpha;}}(m,n-k)=r_{m,n}^{\left(1\right)}\left(k\right)+r_{m,n+L_0}^{\left(1\right)}(k+L_0),k=-L_0+1,\&CenterDot;\&CenterDot;\&CenterDot;,-1.$ This pair relationhip shows with empty tiltedly frame table respectively in Fig. 6, and has marked

as example.

In like manner:

$\stackrel{\&OverBar;}{\mathrm{\&beta;}}(m+k,n)=r_{m,n}^{\left(2\right)}\left(k\right)+r_{m+L_0,n}^{\left(2\right)}(k-L_0),k=1,\&CenterDot;\&CenterDot;\&CenterDot;,L_0-1$

$\stackrel{\&OverBar;}{\mathrm{\&beta;}}(m,n)=r_{m,n}^{\left(2\right)}\left(k\right),k=0$

$\stackrel{\&OverBar;}{\mathrm{\&beta;}}(m,n-k)=r_{m,n}^{\left(2\right)}\left(k\right)+r_{m,n+L_0}^{\left(2\right)}(k+L_0),k=-L_0+1,\&CenterDot;\&CenterDot;\&CenterDot;,-1$

In full accord during their pair relationhip and calculating

in Fig. 6.

The method feature that calculates each item quadravalence cross covariance statistic in the optimal detection amount in this method is:

Calculate judgement sequence γ (k):

$\mathrm{\&gamma;}\left(k\right)=\frac{1}{2}[\underset{j=0}{\overset{M_0-2}{\mathrm{\&Sigma;}}}({\left|\stackrel{\&OverBar;}{\mathrm{\&alpha;}}(j{L}_0,j{L}_0+k)\right|}^2-\stackrel{\&OverBar;}{\mathrm{\&beta;}}(j{L}_0,{\mathrm{jL}}_0+k)),$

$+\underset{j=1}{\overset{M_0-2}{\mathrm{\&Sigma;}}}({\left|\stackrel{\&OverBar;}{\mathrm{\&alpha;}}((j+1/2)L_0-k,(j+1/2)L_0)\right|}^2-\stackrel{\&OverBar;}{\mathrm{\&beta;}}((j+1/2)L_0-k,(j+1/2)L_0))]$

Wherein, k=2L ₀/ 3+1 ..., (M ₀-3/2) L ₀, being about to satisfy n-m is owning of identical value $({\left|\stackrel{\&OverBar;}{\mathrm{\&alpha;}}(m,n)\right|}^2-\stackrel{\&OverBar;}{\mathrm{\&beta;}}(m,n))/2$ The item addition.

The method feature of check judgement is in this method:

Calculate detection statistic ρ ':

wherein round () is the round function,

be the maximum integer that is not more than.At last, calculate detection statistic

L among the maximum ρ (L) is the estimated value in pseudo-code cycle.

The present invention has reached intended purposes, and the DSSS signal has successfully been realized detection, and has obtained pseudo-code cycle estimated value.

Through test, this detector and existing relevant fluctuation method pseudo-code cycle detection device and carrier frequency detector are carried out performance relatively, pseudo-code sequence length L=1023 of establishing the BPSK-DSSS signal, carrier frequency f _c=1/4 (adopting the normalization numerical frequency), pseudo-code bit rate R _c=1/6, sample length N=30K carries out Monte-Carlo Simulation 10000 times, false alarm probability P _Fa=0.01, simulation result is as shown in the table, and visible detector of the present invention more also has good detection performance at sample length, and its performance obviously is superior to existing other detector under the same sample length situation.

Table 1. detector performance relatively

Signal to noise ratio (dB)	-16	-17	-18
				Detection probability P d(relevant fluctuation method)	71.5％	35.5％	18.2％
Detection probability P d(carrier frequency detector)	29.8％	6.5％	1.2％
				Detection probability P d(MCT that the present invention proposes detects)	100％	92.4％	37.5％

Essence of the present invention is: the present invention directly utilizes and has simplified the optimum detector of DSSS signal, has finally obtained a kind of performance in theory near the optimum attainable detector of chip simultaneously.

The DSSS signal optimum detector of foundation of the present invention is a kind of detection principle of novelty, and it is to be suggested first and to utilize at present.It is applied widely that it has the modulation signal form, the parameter transformation strong robustness, and CFAR, the outstanding feature that accuracy of detection is high is superior to existing detector especially greatly on the detection performance.Its appearance has also solved a QPSK or the DSSS signal of the digital phase moudlation of the high-order theoretical difficult problem that under low signal-to-noise ratio, can't use the carrier frequency detector to detect more especially.Simultaneously, owing to do not detect the carrier frequency parameter, it also can tolerate the Doppler frequency shift of wide range.More than these advantages, all be to utilize the detection principle that the present invention proposes under low signal-to-noise ratio, to realize first.

The present invention has invented a kind of chip that is used to simplify DSSS signal optimum detector can realize the MCT Fast implementation.It implements the search and the detection in pseudo-code cycle in the doubling time variable range in given pseudo-code cycle.It uses a kind of covering principle (MCT) based on FFT dexterously, calculates major ingredients in the optimum detector with very little calculation cost, thus the approximate apace optimum detector that realizes.Each FFT operation can be realized by the IP kernel parallel pipelining process among the FPGA just, in our design, contains the fft processing unit of three complete independent parallels.Because approximate, it has the loss of 1dB approximately than optimum detector on input signal-to-noise ratio, but improved realizability greatly.In our design, need the total size of memory to be about ten times of sample lengths based on FPGA.In the pseudo-code cycle is 10 ³During the order of magnitude, can obtain amount of calculation than optimum detector and reduce at least more than about two one magnitude, memory space also reduces the saving of 2 one magnitude.Under cycle, performance improves also bigger in longer pseudo-code.Therefore be particularly suitable for the fast detecting of faint non-cooperation PSK-DSSS signal based on the pseudo-code cycle detection device of MCT fast algorithm; Also be fit to emission source with respect to recipient's high-speed motion; Practical matter such as noise background is unknown or variable have high practical value.

Description of drawings:

Fig. 1: quadratic method detects BPSK-DSSS signal basic principle block diagram.

Fig. 2: the rate thread detector detects BPSK-DSSS signal basic principle block diagram.

Fig. 3: the fluctuation correlation method detects DSSS signal basic principle block diagram.

Fig. 4: the attainable fast high-efficient non-cooperative signal-noise ratio direct spreading signal detection algorithm of chip system principle diagram.

Fig. 5: the realization schematic diagram of optimum pseudo-code cycle detection device.

Fig. 6: the realization principle schematic of MCT fast method.

Embodiment:

The attainable fast high-efficient non-cooperative signal-noise ratio direct spreading signal detecting method of a kind of chip; Its core is to have designed a kind of Rapid Realization mode of DSSS signal optimum detector dexterously---netted covering table fast algorithm (fast Mesh-Covered tablealgorithm; Be called for short the MCT algorithm); It has utilized FFT computing capability efficiently, the FFT arithmetic element is stitched together according to certain pattern carries out computing then.It is quite superior that this method detects performance: can realize the very detection in low signal-to-noise ratio DSSS signal pseudo-code cycle by enough quite short samples; Also existing other method under the same detection performance of amount of calculation simultaneously, and can obtain the pseudo-code cycle parameter of estimated accuracy simultaneously far above existing other method.

Different with aforementioned pseudo-code cycle detection algorithm based on the frequency domain processing; This detection method is regarded the DSSS signal as the sign indicating number sequence that Memorability is arranged of carrying bulk redundancy information; Through some specific computings on the time domain, make full use of redundant information and remove noise, thereby obtain good inhibition noise ability.

Describe for convenient, discrete DSSS signal y (k) can be write as column vector form y, uses y _iI element among the expression column vector y, i=1,2 ..., N, and point of each pseudo-code chip samples establish L ₀For a near value the pseudo-code period L, might as well establish $L\&Element;[\frac{2}{3}{L}_0,\frac{4}{3}{L}_0],$ M ₀=N/L ₀Through the theoretical derivation of optimum detector, structure statistic ρ (L ₀) as detection statistic:

Wherein γ is a decision threshold;

hypothesis when being no signal,

is the hypothesis when signal is arranged.The performance of this detector is outstanding, the detector that robustness is very good, widely applicable.Theoretical research shows: it is that the PSK-DSSS signal is tending towards optimum detector (the LMPI test under 0 in signal to noise ratio; Local maximal potential invariant test); Be its optimal performance at any noise variance, all be consistent under carrier frequency and the first phase, pseudo-code sequence and information code sequence.Simultaneously, this detector also is CFAR (CFAR) and influence that do not receive the PSK modulation type.Its computation complexity is approximately O (NM).

Amount of calculation that this detector is required and memory space are all very big, and are difficult to practical application.The MCT implementation of the present invention's design can be similar to and realize this non-cooperative signal-noise ratio direct spreading signal detector.Technical scheme of the present invention is as shown in Figure 4; Signal gets into receiver through antenna; After intermediate frequency quadrature reception and A/D sampling,, utilize the fpga chip implementation algorithm then with the sampled data input store; Generate the decision statistic amount at last and accomplish the detection of signal, estimate the pseudo-code cycle parameter of DSSS signal simultaneously.It is characterized in that this method comprises following steps successively in processor:

Step 1. receive the data preliminary finish, in processor, data are carried out preliminary treatment, can be through filtering and albefaction function, inhibition out-of-band noise, and the noise contribution that the makes the actual reception data suitable theoretical model of coincidence detection method more.Then that data are calculative structural segmented according to the MCT algorithm, and every section zero padding storage respectively, so that make processor can efficiently handle corresponding FFT computing.

Step 2. FFT calculates in advance.Every segment data is carried out the FFT computing respectively, deposit the result in memory.Because this group will be the result will use repeatedly, once all realize in advance, and computational efficiency is improved greatly.

Step 3. becoming during estimation has partial autocorrelation function.Be exactly in fact according to the MCT algorithm with corresponding data conjugation dot product and get the IFFT conversion.

Step 4. become no partial autocorrelation function when synthetic.According to the MCT algorithm, the adjacent partial autocorrelation function that has is combined into no partial autocorrelation function.

Step 5. calculate each item quadravalence cross covariance statistic in the optimal detection amount.Promptly after repeated execution of steps three and step 4, institute become sometimes obtain on the no partial autocorrelation function basis pseudo-code cycle the quadravalence cross covariance function of possible value correspondence.

Step 6. the check judgement.Detection statistic that obtains at last and thresholding are compared.Because this detector is CFAR, promptly thresholding does not rely on DSSS signal parameter and background noise size, so threshold value adopts the mode that presets to be solidificated in the program.

In this method be to the pretreated method feature of data:

Signal gets into the intermediate frequency quadrature receiver through antenna, and through the A/D sampling, behind digital filtering and the noise whitening, sampled data is carried out data rearrangement then, and input store is so that the utilization subsequent treatment.Below will receive data and represent (y with vectorial y _nN point among the expression y).Order $L^{\&prime;}=\underset{k\&Element;N}{\min }\left(p^k\right)>{2L}_0-1,$ Define L ' * (M respectively ₀-1) dimension matrix A ⁽¹⁾, A ⁽²⁾And matrix B ⁽¹⁾, B ⁽²⁾P is the substrate of the FFT conversion supported of platform, is generally 2.

$A^{\left(1\right)}=\left[\begin{array}{cccc}{y}_0& y_{L_0}& \&CenterDot;\&CenterDot;\&CenterDot;& y_{(M_0-2)L_0}\\ y_1& y_{L_0+1}& \&CenterDot;\&CenterDot;\&CenterDot;& y_{(M_0-2)L_0+1}\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ y_{L_0-1}& y_{2L_0-1}& \&CenterDot;\&CenterDot;\&CenterDot;& y_{(M_0-1)L_0-1}\\ 0& 0& 0& 0\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ 0& 0& 0& 0\end{array}\right],$ $B^{\left(1\right)}=\left[\begin{array}{cccc}{y}_{L_0/2}& y_{{3L}_0/2}& \&CenterDot;\&CenterDot;\&CenterDot;& y_{(M_0-3/2)L_0}\\ y_{L_0/2+1}& y_{{3L}_0/2+1}& \&CenterDot;\&CenterDot;\&CenterDot;& y_{(M_0-3/2)L_0+1}\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ y_{{3L}_0/2-1}& y_{5L_0/2-1}& \&CenterDot;\&CenterDot;\&CenterDot;& y_{(M_0-1/2)L_0-1}\\ 0& 0& 0& 0\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ 0& 0& 0& 0\end{array}\right]$

$A^{\left(2\right)}=\left[\begin{array}{cccc}{\left|y_0\right|}^2& {\left|y_{L_0}\right|}^2& \&CenterDot;\&CenterDot;\&CenterDot;& {\left|y_{(M_0-2)L_0}\right|}^2\\ {\left|y_1\right|}^2& {\left|y_{L_0+1}\right|}^2& \&CenterDot;\&CenterDot;\&CenterDot;& {\left|y_{(M_0-2)L_0+1}\right|}^2\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ {\left|y_{L_0-1}\right|}^2& {\left|y_{2L_0-1}\right|}^2& \&CenterDot;\&CenterDot;\&CenterDot;& {\left|y_{(M_0-1)L_0-1}\right|}^2\\ 0& 0& 0& 0\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ 0& 0& 0& 0\end{array}\right],$ $B^{\left(2\right)}=\left[\begin{array}{cccc}{\left|y_{L_0/2}\right|}^2& {\left|y_{3L_0/2}\right|}^2& \&CenterDot;\&CenterDot;\&CenterDot;& {\left|y_{(M_0-3/2)L_0}\right|}^2\\ {\left|y_{L_0/2+1}\right|}^2& {\left|y_{3/L_0/2+1}\right|}^2& \&CenterDot;\&CenterDot;\&CenterDot;& {\left|y_{(M_0-3/2)L_0+1}\right|}^2\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ {\left|y_{{3L}_0/2-1}\right|}^2& {\left|y_{5L_0/2-1}\right|}^2& \&CenterDot;\&CenterDot;\&CenterDot;& {\left|y_{(M_0-1/2)L_0-1}\right|}^2\\ 0& 0& 0& 0\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ 0& 0& 0& 0\end{array}\right]$

And establish $A^{\left(1\right)}=\begin{array}{c}\left[\begin{array}{ccc}{a}_0^{\left(1\right)}& \&CenterDot;\&CenterDot;\&CenterDot;& a_{M_0-2}^{\left(1\right)}\end{array}\right],\end{array}$ $A^{\left(2\right)}=\begin{array}{c}\left[\begin{array}{ccc}{a}_0^{\left(2\right)}& \&CenterDot;\&CenterDot;\&CenterDot;& a_{M_0-2}^{\left(2\right)}\end{array}\right],\end{array}$ $B^{\left(1\right)}=\begin{array}{c}\left[\begin{array}{ccc}{b}_0^{\left(1\right)}& \&CenterDot;\&CenterDot;\&CenterDot;& b_{M_0-2}^{\left(1\right)}\end{array}\right],\end{array}$ $B^{\left(2\right)}=\begin{array}{c}\left[\begin{array}{ccc}{b}_0^{\left(2\right)}& \&CenterDot;\&CenterDot;\&CenterDot;& b_{M_0-2}^{\left(2\right)}\end{array}\right].\end{array}$ A ⁽¹⁾With B ⁽¹⁾, A ⁽²⁾With B ⁽²⁾L just in time staggers ₀/ 2, so that in step 4, can effectively suppress information code initial phase difference to detecting the fluctuation of performance.

In this method preparatory FFT Calculation Method is characterized as:

Realize A respectively by the fpga chip module ⁽¹⁾, A ⁽²⁾, B ⁽¹⁾, B ⁽²⁾Base-p FFT the computing of each row.Obtain: ${\tilde{A}}^{\left(1\right)}=\begin{array}{c}\left[\begin{array}{ccc}{\tilde{a}}_0^{\left(1\right)}& \&CenterDot;\&CenterDot;\&CenterDot;& {\tilde{a}}_{M_0-2}^{\left(1\right)}\end{array}\right],\end{array}$ ${\tilde{A}}^{\left(2\right)}=\begin{array}{c}\left[\begin{array}{ccc}{\tilde{a}}_0^{\left(2\right)}& \&CenterDot;\&CenterDot;\&CenterDot;& {\tilde{a}}_{M_0-2}^{\left(2\right)}\end{array}\right],\end{array}$ ${\tilde{B}}^{\left(1\right)}=\begin{array}{c}\left[\begin{array}{ccc}{\tilde{b}}_0^{\left(1\right)}& \&CenterDot;\&CenterDot;\&CenterDot;& {\tilde{b}}_{M_0-2}^{\left(1\right)}\end{array}\right],\end{array}$ ${\tilde{B}}^{\left(2\right)}=\begin{array}{c}\left[\begin{array}{ccc}{\tilde{b}}_0^{\left(2\right)}& \&CenterDot;\&CenterDot;\&CenterDot;& {\tilde{b}}_{M_0-2}^{\left(2\right)}\end{array}\right],\end{array}$ Then the result is sent back in the memory.

Become when estimating in this method method feature that partial autocorrelation function is arranged into:

As shown in Figure 5, with coordinate on the plane (m, the corresponding y that is defined as of some n) _m ^*y _nWith | y _m| ²| y _n| ²( ^*The expression conjugation), m=0 ..., N-1, n=0 ..., N-1.And definition has inclined to one side estimation auto-correlation function:

$r_{m,n}^{\left(1\right)}\left(k\right)=\left\{\begin{array}{cc}\underset{i=0}{\overset{L_0-k}{\mathrm{\&Sigma;}}}{y}^{*}(m+k+i)y(n+i)& 0\&le;k\&le;L_0-1\\ \underset{i=1}{\overset{L_0-k}{\mathrm{\&Sigma;}}}{y}^{*}(m+i)y(n-k+i)& -L_0+1\&le;k<0\end{array}\right.$

With

$r_{m,n}^{\left(2\right)}\left(k\right)=\left\{\begin{array}{cc}\underset{i=0}{\overset{L_0-k}{\mathrm{\&Sigma;}}}|{y(m+k+i)|}^2{\left|y(n+i)\right|}^2& 0\&le;k\&le;L_0-1\\ \underset{i=1}{\overset{L_0-k}{\mathrm{\&Sigma;}}}{\left|y(m+i)\right|}^2{\left|y(n-k+i)\right|}^2& -L_0+1\&le;k<0\end{array}\right.$

Use length to be L ₀Empty frame each point on the plane is carried out piecemeal (among Fig. 5 with L ₀=4 is example), calculate the inclined to one side estimation auto-correlation function of having of each empty frame r _{M, n} ⁽¹⁾(k) and r _{M, n} ⁽²⁾(k) r _{M, n} ⁽¹⁾(k).Calculating has the process of inclined to one side estimation auto-correlation function: read successively from memory by the fpga chip module

With

And their corresponding element multiplied each other, and do base-p IFFT conversion, the sequence that obtains is [r _{M, n} ⁽¹⁾(0) ... R _{M, n} ⁽¹⁾(L ₀-1) 0 ... 0r _{M, n} ⁽¹⁾(L ₀+ 1) ... R _{M, n} ⁽¹⁾(1)] ^TIn like manner, respectively with each vector

With

Corresponding element multiplies each other, and does base-p IFFT conversion, and the sequence that obtains is [r _{M, n} ⁽²⁾(0) ... R _{M, n} ⁽²⁾(L ₀-1) 0 ... 0r _{M, n} ⁽²⁾(L ₀+ 1) ... R _{M, n} ⁽²⁾(1)] ^T

The method feature that becomes no partial autocorrelation function in this method when synthetic into:

Calculate the auto-correlation function that (part) nothing is estimated partially

With Every L on the same diagonal line about being or in neighbouring two empty frames ₀The sum of individual next-door neighbour's element (being the sum of per 4 next-door neighbours' element among Fig. 5), (m n) is defined as y when coordinate points _m ^*y _nThe time tried to achieve and do

When coordinate points (m n) is defined as | y _m| ²| y _n| ²The time tried to achieve and do

All of trying to achieve like this

With

Can represent that also the coordinate meaning that Fig. 6 and Fig. 5 represent is in full accord with Fig. 6, promptly more corresponding one by one, only in order more succinctly intuitively to represent each element with [].Work as m=0, L ₀..., (M-2) L ₀, n=L ₀/ 2,3L ₀/ 2 ..., (M-3/2) L ₀The time, have $\stackrel{\&OverBar;}{\mathrm{\&alpha;}}(m+k,n)=r_{m,n}^{\left(1\right)}\left(k\right)+r_{m+L_0,n}^{\left(1\right)}(k-L_0),k=1,\&CenterDot;\&CenterDot;\&CenterDot;,L_0-1.$ This pair relationhip shows with empty tiltedly frame table respectively in Fig. 6, and has marked

as example.

$\stackrel{\&OverBar;}{\mathrm{\&alpha;}}(m,n)=r_{m,n}^{\left(1\right)}\left(k\right),k=0$

$\stackrel{\&OverBar;}{\mathrm{\&alpha;}}(m,n-k)=r_{m,n}^{\left(1\right)}\left(k\right)+r_{m,n+L_0}^{\left(1\right)}(k+L_0),k=-L_0+1,\&CenterDot;\&CenterDot;\&CenterDot;,-1.$ This pair relationhip shows with empty tiltedly frame table respectively in Fig. 6, and has marked

as example.

In like manner:

$\stackrel{\&OverBar;}{\mathrm{\&beta;}}(m+k,n)=r_{m,n}^{\left(2\right)}\left(k\right)+r_{m+L_0,n}^{\left(2\right)}(k-L_0),k=1,\&CenterDot;\&CenterDot;\&CenterDot;,L_0-1$

$\stackrel{\&OverBar;}{\mathrm{\&beta;}}(m,n)=r_{m,n}^{\left(2\right)}\left(k\right),k=0$

$\stackrel{\&OverBar;}{\mathrm{\&beta;}}(m,n-k)=r_{m,n}^{\left(2\right)}\left(k\right)+r_{m,n+L_0}^{\left(2\right)}(k+L_0),k=-L_0+1,\&CenterDot;\&CenterDot;\&CenterDot;,-1$

In full accord during their pair relationhip and calculating

in Fig. 6.

The method feature that calculates each item quadravalence cross covariance statistic in the optimal detection amount in this method is:

Calculate judgement sequence γ (k):

$\mathrm{\&gamma;}\left(k\right)=\frac{1}{2}[\underset{j=0}{\overset{M_0-2}{\mathrm{\&Sigma;}}}({\left|\stackrel{\&OverBar;}{\mathrm{\&alpha;}}(j{L}_0,j{L}_0+k)\right|}^2-\stackrel{\&OverBar;}{\mathrm{\&beta;}}(j{L}_0,{\mathrm{jL}}_0+k)),$

$+\underset{j=1}{\overset{M_0-2}{\mathrm{\&Sigma;}}}({\left|\stackrel{\&OverBar;}{\mathrm{\&alpha;}}((j+1/2)L_0-k,(j+1/2)L_0)\right|}^2-\stackrel{\&OverBar;}{\mathrm{\&beta;}}((j+1/2)L_0-k,(j+1/2)L_0))]$

Wherein, k=2L ₀/ 3+1 ..., (M ₀-3/2) L ₀, being about to satisfy n-m is owning of identical value $({\left|\stackrel{\&OverBar;}{\mathrm{\&alpha;}}(m,n)\right|}^2-\stackrel{\&OverBar;}{\mathrm{\&beta;}}(m,n))/2$ The item addition.

The method feature of check judgement is in this method:

Calculate detection statistic ρ ':

wherein round () is the round function,

be the maximum integer that is not more than.At last, calculate detection statistic

L among the maximum ρ (L) is the estimated value in pseudo-code cycle.

The present invention has reached intended purposes, and the DSSS signal has successfully been realized detection, and has obtained pseudo-code cycle estimated value.

Through test, this detector and existing relevant fluctuation method pseudo-code cycle detection device and carrier frequency detector are carried out performance relatively, pseudo-code sequence length L=1023 of establishing the BPSK-DSSS signal, carrier frequency F _c=1/4 (adopting the normalization numerical frequency), pseudo-code bit rate R _c=1/6, sample length N=30K carries out Monte-Carlo Simulation 10000 times, false alarm probability P _Fa=0.01, simulation result is as shown in the table, and visible detector of the present invention more also has good detection performance at sample length, and its performance obviously is superior to existing other detector under the same sample length situation.

Table 1. detector performance relatively

Signal to noise ratio (dB)	-16	-17	-18
				Detection probability P d(relevant fluctuation method)	71.5％	35.5％	18.2％
Detection probability P d(carrier frequency detector)	29.8％	6.5％	1.2％

Detection probability P d(MCT that the present invention proposes detects)

100％

92.4％

37.5％

Claims (7)

1. attainable non-cooperative signal-noise ratio direct spreading signal detection method of chip; Through using a kind of covering principle MCT based on FFT; Design a kind of chip and can realize the MCT Fast implementation; In given DSSS (DSSS) the doubling time variable range in signal pseudo-code cycle, implement the search and the detection in pseudo-code cycle, calculate major ingredients in the optimum detector with very little calculation cost; Thereby the approximate optimum detector that realizes the DSSS signal, it is characterized in that: the MCT Fast implementation also comprises following steps successively in processor:

1. in processor through filtering and albefaction function; Suppress out-of-band noise; And the noise contribution that makes the actual reception data theoretical model that is suitable for of coincidence detection method more, then that data are calculative structural segmented according to the MCT algorithm, and every section zero padding storage respectively; So that make processor can efficiently handle corresponding FFT computing, achieve a butt joint and receive the preliminary treatment of data;

2. FFT calculates in advance: every segment data is carried out the FFT computing respectively, deposit the result in memory, because this group result will use repeatedly, once all realize in advance, computational efficiency is improved greatly;

3. according to the MCT algorithm with corresponding data conjugation dot product and get the IFFT conversion, become the estimation that partial autocorrelation function is arranged during realization;

4. according to the MCT algorithm, become no partial autocorrelation function when having partial autocorrelation function to be combined into adjacent;

5. calculate each item quadravalence cross covariance statistic in the optimal detection amount, promptly after repeated execution of steps three and step 4, institute become sometimes obtain on the no partial autocorrelation function basis pseudo-code cycle the quadravalence cross covariance function of possible value correspondence;

The detection statistic that 6. will obtain at last and thresholding relatively because this detector is CFAR, promptly thresholding does not rely on DSSS signal parameter and background noise size, so threshold value adopts the mode that presets to be solidificated in the program, realizes that the final inspection of this method is adjudicated.

2. the attainable non-cooperative signal-noise ratio direct spreading signal detection method of a kind of chip according to claim 1 is characterized in that: data are carried out pretreated method is:

Signal gets into the intermediate frequency quadrature receiver through antenna, and through the A/D sampling, behind digital filtering and the noise whitening, sampled data is carried out data rearrangement then, and input store is so that the utilization subsequent treatment; Below will receive data and represent y with vectorial y _nN point among the expression y; Order

Define L ' * (M respectively ₀-1) dimension matrix A ⁽¹⁾, A ⁽²⁾And matrix B ⁽¹⁾, B ⁽²⁾, L wherein ₀Be a near value the pseudo-code period L, M ₀=N/L ₀Be to receive the pseudo-code number of cycles that sample of signal contains, N representes that total sample counts; P is the substrate of the FFT conversion supported of platform;

$A^{\left(1\right)}=\left[\begin{array}{cccc}{y}_0& y_{L_0}& \&CenterDot;\&CenterDot;\&CenterDot;& y_{(M_0-2)L_0}\\ y_1& y_{L_0+1}& \&CenterDot;\&CenterDot;\&CenterDot;& y_{(M_0-2)L_0+1}\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ y_{L_0-1}& y_{2L_0-1}& \&CenterDot;\&CenterDot;\&CenterDot;& y_{(M_0-1)L_0-1}\\ 0& 0& 0& 0\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ 0& 0& 0& 0\end{array}\right],$ $B^{\left(1\right)}=\left[\begin{array}{cccc}{y}_{L_0/2}& y_{3L_0/2}& \&CenterDot;\&CenterDot;\&CenterDot;& y_{(M_0-3/2)L_0}\\ y_{L_0/2+1}& y_{3L_0/2+1}& \&CenterDot;\&CenterDot;\&CenterDot;& y_{(M_0-3/2)L_0+1}\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ y_{{3L}_0/2-1}& y_{5L_0/2-1}& \&CenterDot;\&CenterDot;\&CenterDot;& y_{(M_0-1/2)L_0-1}\\ 0& 0& 0& 0\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ 0& 0& 0& 0\end{array}\right]$

$A^{\left(2\right)}=\left[\begin{array}{cccc}{\left|y_0\right|}^2& {\left|y_{L_0}\right|}^2& \&CenterDot;\&CenterDot;\&CenterDot;& {\left|y_{(M_0-2)L_0}\right|}^2\\ {\left|y_1\right|}^2& {\left|y_{L_0+1}\right|}^2& \&CenterDot;\&CenterDot;\&CenterDot;& {\left|y_{(M_0-2)L_0+1}\right|}^2\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ {\left|y_{L_0-1}\right|}^2& {\left|y_{2L_0-1}\right|}^2& \&CenterDot;\&CenterDot;\&CenterDot;& {\left|y_{(M_0-1)L_0-1}\right|}^2\\ 0& 0& 0& 0\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ 0& 0& 0& 0\end{array}\right],$ $B^{\left(2\right)}=\left[\begin{array}{cccc}{\left|y_{L_0/2}\right|}^2& {\left|y_{3L_0/2}\right|}^2& \&CenterDot;\&CenterDot;\&CenterDot;& {\left|y_{(M_0-3/2)L_0}\right|}^2\\ {\left|y_{L_0/2+1}\right|}^2& {\left|y_{{3L}_0/2+1}\right|}^2& \&CenterDot;\&CenterDot;\&CenterDot;& {\left|y_{(M_0-3/2)L_0+1}\right|}^2\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ {\left|y_{{3L}_0/2-1}\right|}^2& {\left|y_{5L_0/2-1}\right|}^2& \&CenterDot;\&CenterDot;\&CenterDot;& {\left|y_{(M_0-1/2)L_0-1}\right|}^2\\ 0& 0& 0& 0\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ 0& 0& 0& 0\end{array}\right]$

And establish $A^{\left(1\right)}=\left[\begin{array}{ccc}{a}_0^{\left(1\right)}& \&CenterDot;\&CenterDot;\&CenterDot;& a_{M_0-2}^{\left(1\right)}\end{array}\right],$ $A^{\left(2\right)}=\left[\begin{array}{ccc}{a}_0^{\left(2\right)}& \&CenterDot;\&CenterDot;\&CenterDot;& a_{M_0-2}^{\left(2\right)}\end{array}\right],$ $B^{\left(1\right)}=\left[\begin{array}{ccc}{b}_0^{\left(1\right)}& \&CenterDot;\&CenterDot;\&CenterDot;& b_{M_0-2}^{\left(1\right)}\end{array}\right],$ $B^{\left(2\right)}=\left[\begin{array}{ccc}{b}_0^{\left(2\right)}& \&CenterDot;\&CenterDot;\&CenterDot;& b_{M_0-2}^{\left(2\right)}\end{array}\right];$ A ⁽¹⁾With B ⁽¹⁾, A ⁽²⁾With B ⁽²⁾L just in time staggers ₀/ 2, so that in step 4, can effectively suppress information code initial phase difference to detecting the fluctuation of performance.

3. the attainable non-cooperative signal-noise ratio direct spreading signal detection method of a kind of chip according to claim 2 is characterized in that: to preparatory FFT Calculation Method be:

Realize A respectively by the fpga chip module ⁽¹⁾, A ⁽²⁾, B ⁽¹⁾, B ⁽²⁾Base-pFFT the computing of each row;

Obtain: ${\tilde{A}}^{\left(1\right)}=\left[\begin{array}{ccc}{\tilde{a}}_0^{\left(1\right)}& \&CenterDot;\&CenterDot;\&CenterDot;& {\tilde{a}}_{M_0-2}^{\left(1\right)}\end{array}\right],$ ${\tilde{A}}^{\left(2\right)}=\left[\begin{array}{ccc}{\tilde{a}}_0^{\left(2\right)}& \&CenterDot;\&CenterDot;\&CenterDot;& {\tilde{a}}_{M_0-2}^{\left(2\right)}\end{array}\right],$ ${\tilde{B}}^{\left(1\right)}=\left[\begin{array}{ccc}{\tilde{b}}_0^{\left(1\right)}& \&CenterDot;\&CenterDot;\&CenterDot;& {\tilde{b}}_{M_0-2}^{\left(1\right)}\end{array}\right],$ ${\tilde{B}}^{\left(2\right)}=\left[\begin{array}{ccc}{\tilde{b}}_0^{\left(2\right)}& \&CenterDot;\&CenterDot;\&CenterDot;& {\tilde{b}}_{M_0-2}^{\left(2\right)}\end{array}\right],$ Then the result is sent back in the memory.

4. the attainable non-cooperative signal-noise ratio direct spreading signal detection method of a kind of chip according to claim 3 is characterized in that: become when estimating method that partial autocorrelation function is arranged into:

With coordinate on the plane (m, corresponding being defined as of some n)

With | y _m| ²| y _n| ², wherein ^*The expression conjugation, m=0 ..., N-1, n=0 ..., N-1; And definition has inclined to one side estimation auto-correlation function:

$r_{m,n}^{\left(1\right)}\left(k\right)=\left\{\begin{array}{cc}\underset{i=0}{\overset{L_0-k}{\mathrm{\&Sigma;}}}{y}_{(m+k+i)}^{*}{y}_{(n+i)}& 0\&le;k\&le;L_0-1\\ \underset{i=1}{\overset{L_0-k}{\mathrm{\&Sigma;}}}{y}_{(m+i)}^{*}{y}_{(n-k+i)}& -L_0+1\&le;k<0\end{array}\right.$

With

$r_{m,n}^{\left(2\right)}\left(k\right)=\left\{\begin{array}{cc}\underset{i=0}{\overset{L_0-k}{\mathrm{\&Sigma;}}}{\left|y_{(m+k+i)}\right|}^2{\left|y_{(n+i)}\right|}^2& 0\&le;k\&le;L_0-1\\ \underset{i=1}{\overset{L_0-k}{\mathrm{\&Sigma;}}}{\left|y_{(m+i)}\right|}^2{\left|y_{(n-k+i)}\right|}^2& -L_0+1\&le;k<0\end{array}\right.$

Use length to be L ₀Empty frame each point on the plane is carried out piecemeal, calculate the inclined to one side estimation auto-correlation function of having of each empty frame

With

Calculating has the process of inclined to one side estimation auto-correlation function: read successively from memory by the fpga chip module With

And their corresponding element multiplied each other, and do base-pIFFT conversion, the sequence that obtains is ${\left[\begin{array}{ccccccccc}{r}_{m,n}^{\left(1\right)}\left(0\right)& \&CenterDot;\&CenterDot;\&CenterDot;& r_{m,n}^{\left(1\right)}(L_0-1)& 0& \&CenterDot;\&CenterDot;\&CenterDot;& 0& r_{m,n}^{\left(1\right)}(-L_0+1)& \&CenterDot;\&CenterDot;\&CenterDot;& r_{m,n}^{\left(1\right)}(-1)\end{array}\right]}^T;$ In like manner, respectively with each vector

With Corresponding element multiplies each other, and does base-pIFFT conversion, and the sequence that obtains does ${\left[\begin{array}{ccccccccc}{r}_{m,n}^{\left(2\right)}\left(0\right)& \&CenterDot;\&CenterDot;\&CenterDot;& r_{m,n}^{\left(2\right)}(L_0-1)& 0& \&CenterDot;\&CenterDot;\&CenterDot;& 0& r_{m,n}^{\left(2\right)}(-L_0+1)& \&CenterDot;\&CenterDot;\&CenterDot;& r_{m,n}^{\left(2\right)}(-1)\end{array}\right]}^T.$

5. the attainable non-cooperative signal-noise ratio direct spreading signal detection method of a kind of chip according to claim 4 is characterized in that: the method that becomes no partial autocorrelation function when synthetic into:

The auto-correlation function that the nothing of calculating section is estimated partially With (m n) is defined as when coordinate points

The time tried to achieve and do

When coordinate points (m n) is defined as | y _m| ²| y _n| ²The time tried to achieve and do

Work as m=0, L ₀..., (M-2) L ₀, n=L ₀/ 2,3L ₀/ 2 ..., (M-3/2) L ₀The time, have

$\stackrel{\&OverBar;}{\mathrm{\&alpha;}}(m+k,n)=r_{m,n}^{\left(1\right)}\left(k\right)+r_{m+L_0,n}^{\left(1\right)}(k-L_0),$ k＝1，…，L ₀-1，

$\stackrel{\&OverBar;}{\mathrm{\&alpha;}}(m,n)=r_{m,n}^{\left(1\right)}\left(k\right),$ k＝0

$\stackrel{\&OverBar;}{\mathrm{\&alpha;}}(m,n-k)=r_{m,n}^{\left(1\right)}\left(k\right)+r_{m,n+L_0}^{\left(1\right)}(k+L_0),$ k＝-L ₀+1，…，-1，

In like manner:

$\stackrel{\&OverBar;}{\mathrm{\&beta;}}(m+k,n)=r_{m,n}^{\left(2\right)}\left(k\right)+r_{m+L_0,n}^{\left(2\right)}(k-L_0),$ k＝1，…，L ₀-1

$\stackrel{\&OverBar;}{\mathrm{\&beta;}}(m,n)=r_{m,n}^{\left(2\right)}\left(k\right),$ k＝0

$\stackrel{\&OverBar;}{\mathrm{\&beta;}}(m,n-k)=r_{m,n}^{\left(2\right)}\left(k\right)+r_{m,n+L_0}^{\left(2\right)}(k+L_0),$ k＝-L ₀+1，…，-1。

6. the attainable non-cooperative signal-noise ratio direct spreading signal detection method of a kind of chip according to claim 5 is characterized in that: the method to each item quadravalence cross covariance statistic in the compute optimal detection limit is:

Calculate judgement sequence γ (k):

$\mathrm{\&gamma;}\left(k\right)=\frac{1}{2}[\underset{j=0}{\overset{M_0-2}{\mathrm{\&Sigma;}}}({\left|\stackrel{\&OverBar;}{\mathrm{\&alpha;}}(j{L}_0,j{L}_0+k)\right|}^2-\stackrel{\&OverBar;}{\mathrm{\&beta;}}(j{L}_0,j{L}_0+k)),$

$+\underset{j=1}{\overset{M_0-2}{\mathrm{\&Sigma;}}}({\left|\stackrel{\&OverBar;}{\mathrm{\&alpha;}}((j+1/2)L_0-k,(j+1/2)L_0)\right|}^2-\stackrel{\&OverBar;}{\mathrm{\&beta;}}((j+1/2)L_0-k,(j+1/2)L_0))]$

Wherein, k=2L ₀/ 3+1 ..., (M ₀-3/2) L ₀, being about to satisfy n-m is owning of identical value $({\left|\stackrel{\&OverBar;}{\mathrm{\&alpha;}}(m,n)\right|}^2-\stackrel{\&OverBar;}{\mathrm{\&beta;}}(m,n))/2$ The item addition.

7. the attainable non-cooperative signal-noise ratio direct spreading signal detection method of a kind of chip according to claim 6 is characterized in that the method for check judgement being:

Calculate detection statistic ρ ':

wherein round () is the round function;

maximum integer for being not more than calculates detection statistic at last

L among the maximum ρ (L) is the estimated value in pseudo-code cycle; Wherein η is a decision threshold;

hypothesis when being no signal,

is the hypothesis when signal is arranged.

Images