CN101521521A - High-speed predictive-code auxiliary method for suppressing narrow-band interference of spread-spectrum system - Google Patents

Abstract

The invention provides a high-speed predictive-code auxiliary method for suppressing the narrow-band interference of a spread-spectrum system. The method comprises the following steps: (1) receiving a wireless communication signal; (2) down converting the frequency of the wireless communication signal into a mid-frequency signal, digitalizing the mid-frequency signal to provide a digital signal, demodulating the digital signal to provide a baseband signal, enabling a baseband signal envelope r(t) to pass through CHIRP matched filtering so as to provide a sampling signal r(m) and filtering the sampling signal r(m) at high speed. The invention provides a blind P-LMS-LMS method by parallelizing an LMS prediction module and an LMS code auxiliary module which are serially processed in a traditional S-LMS-LMS method at the cost of slight performance declination, and the blind P-LMS-LMS method suppresses and processes the narrow-band interference at high speed.

Description

The high-speed predictive-code auxiliary method that suppresses the spread spectrum system narrow band interference

Technical field

The present invention relates to suppress in straight expansion-code division multiple access (DS-CDMA) wireless communication system the method for all kinds of narrow band interference, wherein all kinds of narrow band interference comprise the class in audio disturbances, digital narrow band interference or the autoregression random process.Present invention is specifically related to the auxiliary slow problem of (S-LMS-LMS) method processing speed of existing blind serial-lowest mean square prediction-lowest mean square sign indicating number that solves.

Background technology

The reason that spread spectrum system is used widely in wireless channel is frequency selective fading and its superior function in being total to the road channel that it can effectively cause anti-multipath, and wherein representative core technology is straight expansion-code division multiple access (DS-CDMA) technology.Narrow band interference is often to get involved the interference of spread spectrum system, and wherein narrow band interference is modeled as audio disturbances (single-tone disturbs or multitone disturbs), digital narrow band interference and autoregression random process usually.Though self possesses certain anti-interference capability spread spectrum system, effectively interference mitigation technology can significantly improve systematic function.

Initial spread spectrum Anti-Jamming Technique originates from the seventies in 20th century, until the end of the eighties, the main focus of Anti-Jamming Technique is on the directly-enlarging system Suppression of narrow band interference based on prediction/estimation filtering and frequency domain filtering, scientific research personnel's achievement based on Milstein is the highest, summary as document " L B Milstein; Interference rejection techniques in spread spectrumcommunication, IEEE Proceedings, 1988 ".Enter the mid-90, arrival along with the CDMA research boom, the cdma system of focus having been transferred to based on technology such as linear prediction, nonlinear prediction and Multiuser Detection with the scientific research personnel headed by Poor and the Rusch disturbs in the associating inhibition more, summary as document " H V Poor; L A Rusch; Narrowband interference suppression in spread spectrum CDMA, IEEE Personal Communication, 1994 ".20th century, scientific research personnel such as Wang further develop into the cdma system interference mitigation technology Predicting Technique, transform domain technology and sign indicating number ancillary technique, summary as document " Xiaodong Wang; H V Poor; Wireless CommunicationSystems-Advanced Techniques for Signal Reception; Beijing:Publishing House of ElectronicsIndustry, 2005 ".In above each technology, Predicting Technique comprises linear prediction and nonlinear prediction two big branches, its research concentrates in the improvement that receives structure, but owing to taked to pursue the processing mode of bit, the typical problem of existence is an error rate height, as document " J Wang; L B Milstein; Adaptive LMS filters for cellular CDMA overlay situations, IEEE Select Areas Commun, 1996 ".Under this viewpoint, utilize the signal code feature to carry out the sign indicating number ancillary technique that piece handles and seem especially effective, it is that multi-user system is disturbed one of the most promising technology of inhibition, the achievement of Poor and Wang is the highest in this field.In the last few years, the shortcoming that needs the known disturbances priori at the sign indicating number ancillary technique, developed the blind coding ancillary technique, as document " S Buzzi; M Lops, A M Tulino, Blind adaptive multiuser detection forasynchronous dual-rate DS/CDMA systems; IEEE Select Areas Commun, 2001 ".The blind coding ancillary technique has realized need not the blind Detecting of priori, but can't suppress strong narrow band interference.Ho is for the proposition of new thought that Predicting Technique and sign indicating number ancillary technique are combined, it is a wonderful work that has disturbed the inhibition field since 20th century, as document " KC Ho; Xiaoning Lu; Vandana Mehta; Adaptive blind narrowband interference cancellation formulti-user detection, IEEE Trans Commun, 2007 ".Auxiliary more traditional Predicting Technique of (S-LMS-LMS) technical performance of blind serial-lowest mean square prediction-lowest mean square sign indicating number that Ho proposes and conventional code ancillary technique improve greatly, and the problem of existence is that the processing speed of serial structure is slower.

The method of the slow problem of processing speed significantly needs when accordingly, overcoming existing blind S-LMS-LMS method DS-CDMA system narrow band interference is suppressed.

Summary of the invention

The object of the present invention is to provide a kind of high-speed predictive-code auxiliary method that can overcome the inhibition spread spectrum system narrow band interference of the slow problem of auxiliary (S-LMS-LMS) method processing speed of existing blind serial-lowest mean square prediction-lowest mean square sign indicating number.

The object of the present invention is achieved like this:

(1) receive wireless communication signals, comprise DS-CDMA signal, white noise and narrow band interference, wherein DS-CDMA signal spread-spectrum sign indicating number is chosen short code, short code code length N≤63; Narrow band interference comprises the class in audio disturbances, digital narrow band interference or the autoregression random process.

(2) described wireless communication signals is down-converted to intermediate-freuqncy signal.

(3) the described intermediate-freuqncy signal of digitlization is to provide digital signal.

(4) with described digital demodulation signal, so that baseband signal to be provided.

(5) with described baseband signal envelope r (t) by cutting general matched filtering so that sampled signal r to be provided (m).

(6) described sampled signal r (m) is carried out the blind P-LMS-LMS filtering of high speed, comprising:

The LMS predictive filtering

At first described sampled signal r (m) is carried out the LMS predictive filtering, establishing the predictive filtering exponent number is M, then M right-safeguarding vector more new formula be

$w(m+1)=w\left(m\right)+2\mathrm{\μ}{\tilde{r}}^{*}\left(m\right)r_M\left(m\right)$

Predicated error wherein, promptly LMS predictive filtering signal is

$\tilde{r}\left(m\right)=r\left(m\right)-\hat{r}\left(m\right)=r\left(m\right)-w^H\left(m\right)r_M\left(m\right)$

In the formula: r _M(m)=[r (m-1) ... r (m-M)] ^TBe the filter input vector; μ is the prediction step factor, satisfies 0＜μ＜1/ λ _Max, λ wherein _MaxBe correlation matrix $R_{r_M{r}_M}=\frac{1}{N}{\mathrm{\&Sigma;}}_{k=1}^{\mathrm{<figure-callout\; id="2r"\; label="N\; R"\; filenames="A200910071731E00171.png"\; state="\{\{state\}\}">N</figure-callout>}}{R}_{}{}_{2\mathrm{<figure-callout\; id="2r"\; label="r"\; filenames="A200910071731E00171.png"\; state="\{\{state\}\}">r</figure-callout>}2r}(k:k+M-1,k:k+M-1)$ Eigenvalue of maximum, the definition $R_{}$ ${}_{2\mathrm{<figure-callout\; id="2r"\; label="r"\; filenames="A200910071731E00171.png"\; state="\{\{state\}\}">r</figure-callout>}2r}=\left[\begin{array}{cc}{R}_{\mathrm{rr}}& R_{\mathrm{rr}}\\ R_{\mathrm{rr}}& R_{\mathrm{rr}}\end{array}\right],$ R _Rr=E{r (n) r ^H(n) }, r (n)=[r (nN+N-1) ... r (nN)] ^TPrediction weight vector initial condition is w (0)=0.

The LMS sign indicating number is assisted filtering

With described LMS predictive filtering signal

At processing time ［ nT _b, (n+1) T _b] interior windowing to be to extract LMS predictive filtering vector $\tilde{r}\left(n\right)={[\tilde{r}(\mathrm{nN}+N-1)\&CenterDot;\&CenterDot;\&CenterDot;\tilde{r}\left(\mathrm{nN}\right)]}^T,$ Described LMS predictive filtering vector is carried out the auxiliary filtering of LMS sign indicating number, and (the right-safeguarding vector of N * K) more new formula is

$Q(n+1)=Q\left(n\right)+2\widetilde{\mathrm{\&mu;}}\tilde{r}\left(n\right)e^H\left(n\right)$

In the formula: $e\left(n\right)=b\left(n\right)-Q^H\left(n\right)\tilde{r}\left(n\right)$ Be error vector, b (n)=[b ₀(n) ... b _K-1(n)] ^TBe user's bit vectors,

Be signal flow (1 or-1);

Be the auxiliary step factor of sign indicating number;

Bring described error vector into described weight vector more new formula, obtain

$Q(n+1)=Q\left(n\right)+2\widetilde{\mathrm{\&mu;}}\tilde{r}\left(n\right)[b^T\left(n\right)-{\tilde{r}}^H\left(n\right)Q\left(n\right)]$

In formula

Decompose

$\tilde{r}\left(n\right)=W^H\left(n\right)\{\left[\begin{array}{c}S\\ 0\end{array}\right]\mathrm{Pb}\left(n\right)+\left[\begin{array}{c}0\\ S_M\end{array}\right]\mathrm{Pb}(n-1)\}+\tilde{i}\left(n\right)+\widetilde{\mathrm{\&epsiv;}}\left(n\right)$

In the formula: $P=\mathrm{diag}(\sqrt{{\mathrm{<figure-callout\; id="0"\; label="P"\; filenames="A200910071731E00201.png"\; state="\{\{state\}\}">P</figure-callout>}}_0}\&CenterDot;\&CenterDot;\&CenterDot;\sqrt{P_{K-1}})$ Be signal power diagonal matrix, P _kBe signal power, wherein K is the CDMA number of users; S=[s ₀S _K-1] be CDMA spreading code matrix, wherein s _k=[c _{K, N-1}C _{K, 0}] T/N is the spreading code vector, { c _{K, i}: i=0 ..., N-1} is direct sequence spread spectrum codes (1 or-1), N is a spreading gain; Definition S _MFor 1～M of S is capable, 1～N row, definition simultaneously

Utilize E{b (n) b ^T(n) }=and I, E{b (n-1) b ^T(n) }=0, obtain

$\tilde{r}\left(n\right)b^T\left(n\right)=E\left\{\tilde{r}\left(n\right)b^T\left(n\right)\right\}=W_1^H\left(n\right)\mathrm{SP}$

And then obtain need not the weight vector recurrence formula of the blind S-LMS-LMS method of training sequence

$Q(n+1)=Q\left(n\right)+2\widetilde{\mathrm{\&mu;}}[W_1^H\left(n\right)\mathrm{SP}-\tilde{r}\left(n\right){\tilde{r}}^H\left(n\right)Q\left(n\right)]$

The key that realizes parallel processing is to make the sign indicating number supplementary module of blind S-LMS-LMS method need not the result of calculation of prediction module

Because at W ₂(n) (in the dimension of the M * N) element, only there is nonzero value in the littlest triangle battle array of below, so can be with described

Middle W ₂(n) part of participation computing is approximate casts out, and be need not

Participate in the parallel weight vector recurrence formula of computing directly

In the following formula, Q at every turn more new capital need wait for that w upgrades formation W N time ₁, significantly not shortening the processing time, the key that shortens the processing time is that the renewal of w and the renewal of Q are carried out synchronously, analyzes the auxiliary recurrence formula of described blind predictive code, predict is finished the renewal of weight vector w for the first time, can constitute W ₁First row, the premultiplication matrix is equivalent to do line translation, can get W ₁ ^H(n) SP and W ₁ ^H(n) first of r (n) row, right multiply matrix is equivalent to do rank transformation, can get r ^H(n) W ₁(n) first row have promptly obtained W ₁ ^H(n) r (n) r ^H(n) W ₁(n) first row, first row are made important hypothesis here, make W ₁ ^H(n) r (n) r ^H(n) W ₁(n) remainder data of first row is 0, just can finish the recursion of Q first row; In like manner, when predict is finished the renewal of the weight vector w second time, utilize the renewal result of weight vector for the first time, can obtain first and second data of Q second row, suppose that the remainder data that Q second goes is 0, just finished the recursion of Q second row; The rest may be inferred, and the each more new capital of w and the renewal of Q delegation are carried out synchronously, obtain the weight vector recurrence formula of the blind P-LMS-LMS method of high speed

Q(n+1)＝Q(n)+ ₂μ[W ₁ ^H(n)SP-tril(W ₁ ^H(n)r(n)r ^H(n)W ₁(n)Q(n))]

In the formula: the triangle battle array is taken off in tril () expression; The auxiliary step factor of sign indicating number satisfies $0<\widetilde{\mathrm{\&mu;}}<1/{\widetilde{\mathrm{\&lambda;}}}_{\max },$ Wherein

Be correlation matrix $R_{\tilde{r}\tilde{r}}=E\left\{\tilde{r}\left(n\right){\tilde{r}}^H\left(n\right)\right\}$ Eigenvalue of maximum; The auxiliary weight vector initial condition of sign indicating number of the blind P-LMS-LMS method of high speed is Q (0)=0.

The present invention's beneficial effect compared with prior art is:

The present invention has realized that white noise exists under the situation the high speed filtering of all kinds of narrow band interference of DS-CDMA system, and wherein all kinds of narrow band interference comprise the class in audio disturbances, digital narrow band interference or the autoregression random process.Auxiliary (P-LMS-LMS) method of blind parallel-lowest mean square prediction-lowest mean square sign indicating number that provides has overcome the auxiliary slow problem of (S-LMS-LMS) method processing speed of existing blind serial-lowest mean square prediction-lowest mean square sign indicating number, is easy to Project Realization.

Description of drawings

Fig. 1 describes the DS-CAMA system reception block diagram that existing blind S-LMS-LMS method suppresses narrow band interference;

Fig. 2 describes the DS-CAMA system reception block diagram that the blind P-LMS-LMS method of high speed suppresses narrow band interference;

Fig. 3 a-1 to Fig. 3 c describes the performance simulation correlation curve that the blind P-LMS-LMS method of high speed and existing blind S-LMS-LMS method suppress all kinds of narrow band interference; Wherein Fig. 3 a-1 and Fig. 3 a-2 describe to receive the performance simulation correlation curve of audio disturbances, and Fig. 3 a-1 describes to receive performance simulation correlation curve, Fig. 3 a-2 that single-tone disturbs and describes to receive the performance simulation correlation curve that multitone disturbs; Fig. 3 b describes to receive the performance simulation correlation curve of digital narrow band interference; Fig. 3 c describes to be received from the performance simulation correlation curve that returns random process;

Fig. 4 describes algorithm implementing procedure figure of the present invention.

Embodiment

In the detailed description of the present invention, with reference to appended drawing, the specific exemplary embodiment of these accompanying drawing explainations invention can be implemented in these exemplary embodiments below.These embodiment describe with sufficient details, implement the present invention to allow those skilled in the art, but can utilize other embodiment, and can make logic, machinery, electrical equipment with other change, and do not depart from standard of the present invention.Therefore, following detailed should not be considered restrictive, and scope of the present invention is limited by appended claims only.

The embodiment of conversion identification detection method comprises the following steps:

(1) receive wireless communication signals, comprise DS-CDMA signal, white noise and narrow band interference, wherein DS-CDMA signal spread-spectrum sign indicating number is chosen short code, short code code length N≤63; Narrow band interference comprises the class in audio disturbances, digital narrow band interference or the autoregression random process.

(2) described wireless communication signals is down-converted to intermediate-freuqncy signal.

(3) the described intermediate-freuqncy signal of digitlization is to provide digital signal.

(4) with described digital demodulation signal, so that baseband signal to be provided.

(5) with described baseband signal envelope r (t) by cutting general matched filtering so that sampled signal r to be provided (m), cut general matched filtering sampling formulate and be

$r\left(m\right)=\frac{1}{T_c}{\&Integral;}_{m{T}_c}^{(m+1)T_c}r\left(t\right)\mathrm{dt}$

T wherein _cFor direct sequence spread spectrum is cut general speed.

(6) described sampled signal r (m) is carried out the blind P-LMS-LMS filtering of high speed, comprising:

The LMS predictive filtering

At first described sampled signal r (m) is carried out the LMS predictive filtering, establishing the predictive filtering exponent number is M, then M right-safeguarding vector more new formula be

$w(m+1)=w\left(m\right)+2\mathrm{\&mu;}{\tilde{r}}^{*}\left(m\right)r_M\left(m\right)$

Predicated error wherein, promptly LMS predictive filtering signal is

$\tilde{r}\left(m\right)=r\left(m\right)-\hat{r}\left(m\right)=r\left(m\right)-w^H\left(m\right)r_M\left(m\right)$

In the formula: r _M(m)=[r (m-1) ... r (m-M)] ^TBe the filter input vector; μ is the prediction step factor, satisfies 0＜μ＜1/ λ _Max, λ wherein _MaxBe correlation matrix $R_{r_M{r}_M}=\frac{1}{N}{\mathrm{\&Sigma;}}_{k=1}^{\mathrm{<figure-callout\; id="2r"\; label="N\; R"\; filenames="A200910071731E00171.png"\; state="\{\{state\}\}">N</figure-callout>}}{R}_{}{}_{2\mathrm{<figure-callout\; id="2r"\; label="r"\; filenames="A200910071731E00171.png"\; state="\{\{state\}\}">r</figure-callout>}2r}(k:k+M-1,k:k+M-1)$ Eigenvalue of maximum, the definition $R_{}$ ${}_{2\mathrm{<figure-callout\; id="2r"\; label="r"\; filenames="A200910071731E00171.png"\; state="\{\{state\}\}">r</figure-callout>}2r}=\left[\begin{array}{cc}{R}_{\mathrm{rr}}& R_{\mathrm{rr}}\\ R_{\mathrm{rr}}& R_{\mathrm{rr}}\end{array}\right],$ R _Rr=E{r (n) r ^H(n) }, r (n)=[r (nN+N-1) ... r (nN)] ^TPrediction weight vector initial condition is w (0)=0.

The LMS sign indicating number is assisted filtering

With described LMS predictive filtering signal

At processing time [nT _b, (n+1) T _b] interior windowing to be to extract LMS predictive filtering vector $\tilde{r}\left(n\right)={[\tilde{r}(\mathrm{nN}+N-1)\&CenterDot;\&CenterDot;\&CenterDot;\tilde{r}\left(\mathrm{nN}\right)]}^T,$ Described LMS predictive filtering vector is carried out the auxiliary filtering of LMS sign indicating number, and (the right-safeguarding vector of N * K) more new formula is

$Q(n+1)=Q\left(n\right)+2\widetilde{\mathrm{\&mu;}}\tilde{r}\left(n\right)e^H\left(n\right)$

In the formula: $e\left(n\right)=b\left(n\right)-Q^H\left(n\right)\tilde{r}\left(n\right)$ Be error vector, b (n)=[b ₀(n) ... b _K-1(n)] ^TBe user's bit vectors,

Be signal flow (1 or-1);

Be the auxiliary step factor of sign indicating number;

Bring described error vector into described weight vector more new formula, obtain

$Q(n+1)=Q\left(n\right)+2\widetilde{\mathrm{\&mu;}}\tilde{r}\left(n\right)[b^T\left(n\right)-{\tilde{r}}^H\left(n\right)Q\left(n\right)]$

In formula

Decompose

$\tilde{r}\left(n\right)=W^H\left(n\right)\{\left[\begin{array}{c}\mathrm{<figure-callout\; id="0"\; label="S"\; filenames="A200910071731E00201.png"\; state="\{\{state\}\}">S</figure-callout>}\\ 0\end{array}\right]\mathrm{Pb}\left(n\right)+\left[\begin{array}{c}0\\ S_M\end{array}\right]\mathrm{Pb}(n-1)\}+\tilde{i}\left(n\right)+\widetilde{\mathrm{\&epsiv;}}\left(n\right)$

In the formula: $P=\mathrm{diag}(\sqrt{{\mathrm{<figure-callout\; id="0"\; label="P"\; filenames="A200910071731E00201.png"\; state="\{\{state\}\}">P</figure-callout>}}_0}\&CenterDot;\&CenterDot;\&CenterDot;\sqrt{P_{K-1}})$ Be signal power diagonal matrix, P _kBe signal power, wherein K is the CDMA number of users; S=[s ₀S _K-1] be CDMA spreading code matrix, wherein s _k=[c _{K, N-1}C _{K, 0}] ^T/ N is the spreading code vector, { c _{K, i}: i=0 ..., N-1} is direct sequence spread spectrum codes (1 or-1), N is a spreading gain; Definition S _MFor 1～M of S is capable, 1～N row, definition simultaneously

Utilize E{b (n) b ^T(n) }=and I, E{b (n-1) b ^T(n) }=0, obtain

$\tilde{r}\left(n\right)b^T\left(n\right)=E\left\{\tilde{r}\left(n\right)b^T\left(n\right)\right\}=W_1^H\left(n\right)\mathrm{SP}$

And then obtain need not the weight vector recurrence formula of the blind S-LMS-LMS method of training sequence

$Q(n+1)=Q\left(n\right)+2\widetilde{\mathrm{\&mu;}}[W_1^H\left(n\right)\mathrm{SP}-\tilde{r}\left(n\right){\tilde{r}}^H\left(n\right)Q\left(n\right)]$

The key that realizes parallel processing is to make the sign indicating number supplementary module of blind S-LMS-LMS method need not the result of calculation of prediction module

Because at W ₂(n) (in the dimension of the M * N) element, only there is nonzero value in the littlest triangle battle array of below, so can be with described

Middle W ₂(n) part of participation computing is approximate casts out, and obtains need not the parallel weight vector recurrence formula that r (n) participates in computing directly

$Q(n+1)=Q\left(n\right)+2\widetilde{\mathrm{\&mu;}}[W_1^H\left(n\right)\mathrm{SP}-W_1^H\left(n\right)r\left(n\right)r^H\left(n\right)W_1\left(n\right)Q\left(n\right)]$

In the following formula, Q at every turn more new capital need wait for that w upgrades formation W N time ₁, significantly not shortening the processing time, the key that shortens the processing time is that the renewal of w and the renewal of Q are carried out synchronously, analyze the auxiliary recurrence formula of described blind predictive code, predict is finished the renewal of weight vector w for the first time, can constitute first row of W1, the premultiplication matrix is equivalent to do line translation, can get W ₁ ^H(n) SP and W ₁ ^H(n) first of r (n) row, right multiply matrix is equivalent to do rank transformation, can get r ^H(n) W ₁(n) first row have promptly obtained W ₁ ^H(n) r (n) r ^H(n) W ₁(n) first row, first row are made important hypothesis here, make W ₁ ^H(n) r (n) r ^H(n) W ₁(n) remainder data of first row is 0, just can finish the recursion of Q first row; In like manner, when predict is finished the renewal of the weight vector w second time, utilize the renewal result of weight vector for the first time, can obtain first and second data of Q second row, suppose that the remainder data that Q second goes is 0, just finished the recursion of Q second row; The rest may be inferred, and the each more new capital of w and the renewal of Q delegation are carried out synchronously, obtain the weight vector recurrence formula of the blind P-LMS-LMS method of high speed

Q(n+1)＝Q(n)+2μ[W ₁ ^H(n)SP-tril{W ₁ ^H(n)r(n)r ^H(n)W ₁(n)Q(n))]

In the formula: the triangle battle array is taken off in tril () expression; The auxiliary step factor of sign indicating number satisfies $0<\widetilde{\mathrm{\&mu;}}<1/{\widetilde{\mathrm{\&lambda;}}}_{\max },$ Wherein

Be correlation matrix $R_{\tilde{r}\tilde{r}}=E\left\{\tilde{r}\left(n\right){\tilde{r}}^H\left(n\right)\right\}$ Eigenvalue of maximum; The auxiliary weight vector initial condition of sign indicating number of the blind P-LMS-LMS method of high speed is Q (0)=0.

Fig. 1 describes the DS-CAMA system reception block diagram that existing blind S-LMS-LMS method suppresses narrow band interference, is made up of the auxiliary filtration module 103 of sign indicating number of down conversion module 101, LMS predictive filtering module 102 and blind S-LMS-LMS method.

As shown in the figure, the wireless communication signals 105 that antenna 104 receives comprises useful signal DS-CDMA signal and narrow band interference, and wherein narrow band interference comprises the class in audio disturbances-single-tone interference or multitone interference, digital narrow band interference or the autoregression random process.Antenna 104 is coupled to down conversion module 101.In down conversion module 101, at first handle wireless communication signals 105 by band pass filter 106, this filter is optimally selected the frequency wanted, for example with the frequency of DS-CDMA signal correction connection.Band pass filter 106 is coupled to amplifier 107, and this amplifier amplifies the signal from band pass filter 106.Blender 108 mixes the output of amplifier 107 with oscillator signal from local oscillator 109.Like this, the output signal of blender 108 down-conversion amplifiers 107 is to provide intermediate-freuqncy signal 110.After initial down-conversion, intermediate-freuqncy signal 110 is transformed into digital signal 112 by modulus a/d transducer 111.Because the QPSK signal can be regarded as the stack of two quadrature 2PSK signals, so the coherent carrier that intersects with two-way goes demodulation.Wherein one road signal enters blender 115 and mixes with signal from digital controlled oscillator 113, another road signal enter blender 116 and with mix from the signal of digital controlled oscillator 113 after pi/2 phase shift 114.The signal of blender 115 and blender 116 outputs is connected low pass filter 117 and low pass filter 118 respectively.Thereafter, the output signal of low pass filter 117 and the output signal of low pass filter 118 are connected sampling decision device 119 and sampling decision device 120 respectively, and the output signal of will sample decision device 119 and sampling decision device 120 becomes baseband signal 122 outputs after parallel/serial device 121 conversion.

Like this, band pass filter 106, amplifier 107, blender 108, local oscillator 109 have been finished the process that is down-converted to intermediate-freuqncy signal, modulus a/d transducer 111, digital controlled oscillator 113, pi/2 phase shift 114, blender 115,116, low pass filter 117,118, sampling decision device 119,120, parallel/serial device 121 have been finished the process that is down-converted to baseband signal by intermediate-freuqncy signal.Above process has been finished down conversion module 101.

Existing blind S-LMS-LMS method is at first carried out LMS predictive filtering 102.Baseband signal 122 envelope r (t) are cut general matched filtering sampling 123, sampled signal r (m) 124 is provided.Sampled signal r (m) 124 respectively by delay line 1131, delay line 2132... delay line M13M, is obtained inhibit signal r (m-1) 141, inhibit signal r (m-2) 142... inhibit signal r (m-M) 14M.Inhibit signal r (m-1) 141, inhibit signal r (m-2) 142... inhibit signal r (m-M) 14M are sent into the estimation of linear combiner $\begin{array}{ccc}1& {\mathrm{<figure-callout\; id="1"\; label="w"\; filenames="A200910071731E00171.png,A200910071731E00181.png"\; state="\{\{state\}\}">w</figure-callout>}}_1^{*}r(m-1)& 151\end{array},$ Estimate $\begin{array}{ccc}2& {\mathrm{<figure-callout\; id="2"\; label="w"\; filenames="A200910071731E00192.png"\; state="\{\{state\}\}">w</figure-callout>}}_2^{*}r(m-2)& 152...\end{array}$ Estimate 15M obtains the linear combination signal through adder 161 $\hat{r}\left(m\right)=w^H\left(m\right)r_M\left(m\right).$ Subsequently, sampled signal r (m) 124 is added computing, obtain the linear combination signal through adder 161 $\hat{r}\left(m\right)=w^H\left(m\right)r_M\left(m\right)$ Subtract computing, send into adder 162 simultaneously, obtain predicated error, be i.e. LMS predictive filtering signal $\begin{array}{cc}\tilde{r}\left(m\right)=r\left(m\right)-w^H\left(m\right)r_M\left(m\right)& 163\end{array}.$ At last, by inhibit signal r (m-1) 141, inhibit signal r (m-2) 142... inhibit signal r (m-M) 14M and LMS predictive filtering signal $\begin{array}{cc}\tilde{r}\left(m\right)=r\left(m\right)-w^H\left(m\right)r_M\left(m\right)& 163\end{array}$ Jointly as factor of influence control prediction weight vector recurrence formula $\begin{array}{cc}w(m+1)=w\left(m\right)+2\mathrm{\&mu;}{\tilde{r}}^{*}\left(m\right)r_M\left(m\right)& 164\end{array},$ The weight coefficient that obtains upgrading

171, weight coefficient

172... weight coefficient

17M.Wherein: μ is the prediction step factor, satisfies 0＜μ＜1/ λ _Max, λ wherein _MaxBe correlation matrix $R_{r_M{r}_M}=\frac{1}{N}{\mathrm{\&Sigma;}}_{k=1}^{\mathrm{<figure-callout\; id="2r"\; label="N\; R"\; filenames="A200910071731E00171.png"\; state="\{\{state\}\}">N</figure-callout>}}{R}_{}{}_{2\mathrm{<figure-callout\; id="2r"\; label="r"\; filenames="A200910071731E00171.png"\; state="\{\{state\}\}">r</figure-callout>}2r}(k:k+M-1,k:k+M-1)$ Eigenvalue of maximum, the definition $R_{}$ ${}_{2\mathrm{<figure-callout\; id="2r"\; label="r"\; filenames="A200910071731E00171.png"\; state="\{\{state\}\}">r</figure-callout>}2r}=\left[\begin{array}{cc}{R}_{\mathrm{rr}}& R_{\mathrm{rr}}\\ R_{\mathrm{rr}}& R_{\mathrm{rr}}\end{array}\right],$ R _Rr=E{r (n) r ^H(n) }, r (n)=[r (nN+N-1) ... r (nN)] ^TPrediction weight vector initial condition is w (0)=0.

Like this, cut general matched filtering sampling 123; Delay line 1131, delay line 2132... delay line M 13M; Estimate $\begin{array}{ccc}1& {\mathrm{<figure-callout\; id="1"\; label="w"\; filenames="A200910071731E00171.png,A200910071731E00181.png"\; state="\{\{state\}\}">w</figure-callout>}}_1^{*}r(m-1)& 151\end{array},$ Estimate $\begin{array}{ccc}2& {\mathrm{<figure-callout\; id="2"\; label="w"\; filenames="A200910071731E00192.png"\; state="\{\{state\}\}">w</figure-callout>}}_2^{*}r(m-2)& 152...\end{array}$ Estimate $\begin{array}{ccc}M& w_M^{*}r(m-M)& 15M\end{array};$ Adder 161; Adder 162 and the control of prediction weight vector recursion $\begin{array}{cc}w(m+1)=w\left(m\right)+2\mathrm{\&mu;}{\tilde{r}}^{*}\left(m\right)r_M\left(m\right)& 164\end{array}$ Constituted LMS predictive filtering 102 jointly.

So far, blind S-LMS-LMS method is carried out the auxiliary filtering 103 of LMS sign indicating number.With LMS predictive filtering signal $\begin{array}{cc}\tilde{r}\left(m\right)=r\left(m\right)-w^H\left(m\right)r_M\left(m\right)& 163\end{array}$ Carry out windowing storage 181, obtain windowing LMS predictive filtering vector $\begin{array}{cc}\tilde{r}\left(n\right)={[\tilde{r}(\mathrm{nN}+N-1)\&CenterDot;\&CenterDot;\&CenterDot;\tilde{r}\left(\mathrm{nN}\right)]}^{\mathrm{<figure-callout\; id="182"\; label="T"\; filenames="A200910071731E00171.png"\; state="\{\{state\}\}">T</figure-callout>}}& 182\end{array}.$ To windowing LMS predictive filtering vector $\begin{array}{cc}\tilde{r}\left(n\right)={[\tilde{r}(\mathrm{nN}+N-1)\&CenterDot;\&CenterDot;\&CenterDot;\tilde{r}\left(\mathrm{nN}\right)]}^T& 182\end{array}$ Carrying out bit estimates $\begin{array}{cc}{Q}^H\left(n\right)\tilde{r}\left(n\right)& 183\end{array},$ Obtain bit estimated signal 184.Subsequently, by bit estimated signal 184 with by the control of prediction weight vector recursion $\begin{array}{cc}w(m+1)=w\left(m\right)+2\mathrm{\&mu;}{\tilde{r}}^{*}\left(m\right)r_M\left(m\right)& 164\end{array}$ The prediction weight vector w that obtains is through W ₁The W1 that maker 185 obtains is jointly as the auxiliary weight vector recurrence formula of factor of influence control code $\begin{array}{cc}Q(n+1)=Q\left(n\right)+2\widetilde{\mathrm{\&mu;}}[W_1^H\left(n\right)\mathrm{SP}-\tilde{r}\left(n\right){\tilde{r}}^H\left(n\right)Q\left(n\right)]& 186\end{array},$ Obtain the auxiliary recursion weight vector Q 187 of sign indicating number.Wherein, the auxiliary step factor of sign indicating number satisfies $0<\widetilde{\mathrm{\&mu;}}<1/{\widetilde{\mathrm{\&lambda;}}}_{\max },$ Wherein Be correlation matrix $R_{\tilde{r}\tilde{r}}=E\left\{\tilde{r}\left(n\right){\tilde{r}}^H\left(n\right)\right\}$ Eigenvalue of maximum; The auxiliary weight vector initial condition of sign indicating number of blind S-LMS-LMS method is Q (0)=0.At last, to bit estimated signal 184 number of winning the confidence real parts 188, obtain the subscriber signal estimated value $\begin{array}{cc}\hat{b}\left(n\right)=\mathrm{Re}\left(Q^H\left(n\right)\tilde{r}\left(n\right)\right)& 189\end{array},$ Re (●) the expression number of winning the confidence real part in the formula.

Like this, the windowing storage 181; Bit is estimated $\begin{array}{cc}{Q}^H\left(n\right)\tilde{r}\left(n\right)& 183\end{array};$ W ₁Maker 185; The auxiliary weight vector recursion control of sign indicating number $\begin{array}{cc}Q(n+1)=Q\left(n\right)+2\widetilde{\mathrm{\&mu;}}[W_1^H\left(n\right)\mathrm{SP}-\tilde{r}\left(n\right){\tilde{r}}^H\left(n\right)Q\left(n\right)]& 186\end{array}$ With the number of winning the confidence real part 188 constituted jointly blind S-LMS-LMS method the sign indicating number auxiliary filtration module 103.

Fig. 2 describes the DS-CAMA system reception block diagram that the blind P-LMS-LMS method of high speed suppresses narrow band interference, is made up of the auxiliary filtration module 203 of sign indicating number of down conversion module 201, LMS predictive filtering module 202 and the blind P-LMS-LMS method of high speed.And the auxiliary filtration module 203 of sign indicating number of LMS predictive filtering module 202 and the blind P-LMS-LMS method of high speed carries out synchronously.

As shown in the figure, the wireless communication signals 205 that antenna 204 receives comprises useful signal DS-CDMA signal and narrow band interference, and wherein narrow band interference comprises the class in audio disturbances-single-tone interference or multitone interference, digital narrow band interference or the autoregression random process.Antenna 204 is coupled to down conversion module 201.In down conversion module 201, at first handle wireless communication signals 205 by band pass filter 206, this filter is optimally selected the frequency wanted, for example with the frequency of DS-CDMA signal correction connection.Band pass filter 206 is coupled to amplifier 207, and this amplifier amplifies the signal from band pass filter 206.Blender 208 mixes the output of amplifier 207 with oscillator signal from local oscillator 209.Like this, the output signal of blender 208 down-conversion amplifiers 207 is to provide intermediate-freuqncy signal 210.After initial down-conversion, intermediate-freuqncy signal 210 is transformed into digital signal 212 by modulus a/d transducer 211.Because the QPSK signal can be regarded as the stack of two quadrature 2PSK signals, so the coherent carrier that intersects with two-way goes demodulation.Wherein one road signal enters blender 215 and mixes with signal from digital controlled oscillator 213, another road signal enter blender 216 and with mix from the signal of digital controlled oscillator 213 after pi/2 phase shift 214.The signal of blender 215 and blender 216 outputs is connected low pass filter 217 and low pass filter 218 respectively.Thereafter, the output signal of low pass filter 217 and the output signal of low pass filter 218 are connected sampling decision device 219 and sampling decision device 220 respectively, and the output signal of will sample decision device 219 and sampling decision device 220 becomes baseband signal 222 outputs after parallel/serial device 221 conversion.

Like this, band pass filter 206, amplifier 207, blender 208, local oscillator 209 have been finished the process that is down-converted to intermediate-freuqncy signal, modulus a/d transducer 211, digital controlled oscillator 213, pi/2 phase shift 214, blender 215,216, low pass filter 217,218, sampling decision device 219,220, parallel/serial device 221 have been finished the process that is down-converted to baseband signal by intermediate-freuqncy signal.Above process has been finished down conversion module 201.

The blind P-LMS-LMS method of high speed of the present invention is carried out LMS predictive filtering 202 and the auxiliary filtering 203 of LMS sign indicating number synchronously.For LMS prediction module 202, baseband signal 222 envelope r (t) are cut general matched filtering sampling 223, sampled signal r (m) 224 is provided.Sampled signal r (m) 224 respectively by delay line 1231, delay line 2232... delay line M23M, is obtained inhibit signal r (m-1) 241, inhibit signal r (m-2) 242... inhibit signal r (m-M) 24M.Inhibit signal r (m-1) 241, inhibit signal r (m-2) 242... inhibit signal r (m-M) 24M are sent into the estimation of linear combiner $\begin{array}{ccc}1& {\mathrm{<figure-callout\; id="1"\; label="w"\; filenames="A200910071731E00171.png,A200910071731E00181.png"\; state="\{\{state\}\}">w</figure-callout>}}_1^{*}r(m-1)& 251\end{array},$ Estimate $\begin{array}{ccc}2& {\mathrm{<figure-callout\; id="2"\; label="w"\; filenames="A200910071731E00192.png"\; state="\{\{state\}\}">w</figure-callout>}}_2^{*}r(m-2)& 252...\end{array}$ Estimate $\begin{array}{ccc}M& w_M^{*}r(m-M)& 25M\end{array},$ Obtain the linear combination signal through adder 261 $\hat{r}\left(m\right)=w^H\left(m\right)r_M\left(m\right).$ Subsequently, sampled signal r (m) 224 is added computing, obtain the linear combination signal through adder 261 $\hat{r}\left(m\right)=w^H\left(m\right)r_M\left(m\right)$ Subtract computing, send into adder 262 simultaneously, obtain predicated error, be i.e. LMS predictive filtering signal $\begin{array}{cc}\tilde{r}\left(m\right)=r\left(m\right)-w^H\left(m\right)r_M\left(m\right)& 263\end{array}.$ At last, by inhibit signal r (m-1) 241, inhibit signal r (m-2) 242... inhibit signal r (m-M) 24M and LMS predictive filtering signal $\tilde{r}\left(m\right)=r\left(m\right)-w^H\left(m\right)r_M\left(m\right)$ 263 jointly as factor of influence control prediction weight vector recurrence formula $\begin{array}{cc}w(m+1)=w\left(m\right)+2\mathrm{\&mu;}{\tilde{r}}^{*}\left(m\right)r_M\left(m\right)& 264\end{array},$ The weight coefficient that obtains upgrading 271, weight coefficient

272... weight coefficient

27M.Wherein: μ is the prediction step factor, satisfies 0＜μ＜1/ λ _Max, λ wherein _MaxBe correlation matrix $R_{r_M{r}_M}=\frac{1}{N}{\mathrm{\&Sigma;}}_{k=1}^{\mathrm{<figure-callout\; id="2r"\; label="N\; R"\; filenames="A200910071731E00171.png"\; state="\{\{state\}\}">N</figure-callout>}}{R}_{}{}_{2\mathrm{<figure-callout\; id="2r"\; label="r"\; filenames="A200910071731E00171.png"\; state="\{\{state\}\}">r</figure-callout>}2r}(k:k+M-1,k:k+M-1)$ Eigenvalue of maximum, the definition $R_{}$ ${}_{2\mathrm{<figure-callout\; id="2r"\; label="r"\; filenames="A200910071731E00171.png"\; state="\{\{state\}\}">r</figure-callout>}2r}=\left[\begin{array}{cc}{R}_{\mathrm{rr}}& R_{\mathrm{rr}}\\ R_{\mathrm{rr}}& R_{\mathrm{rr}}\end{array}\right],$ R _Rr=E{r (n) r ^H(n) }, r (n)=[r (nN+N-1) ... r (nN)] ^TPrediction weight vector initial condition is w (0)=0.

Like this, cut general matched filtering sampling 223; Delay line 1231, delay line 2232... delay line M23M; Estimate $\begin{array}{ccc}1& {\mathrm{<figure-callout\; id="1"\; label="w"\; filenames="A200910071731E00171.png,A200910071731E00181.png"\; state="\{\{state\}\}">w</figure-callout>}}_1^{*}r(m-1)& 251\end{array},$ Estimate $\begin{array}{ccc}2& {\mathrm{<figure-callout\; id="2"\; label="w"\; filenames="A200910071731E00192.png"\; state="\{\{state\}\}">w</figure-callout>}}_2^{*}r(m-2)& 252...\end{array}$ Estimate $\begin{array}{ccc}M& w_M^{*}r(m-M)& 25M\end{array};$ Adder 261; Adder 262 and prediction weight vector recurrence formula $\begin{array}{cc}w(m+1)=w\left(m\right)+2\mathrm{\&mu;}{\tilde{r}}^{*}\left(m\right)r_M\left(m\right)& 264\end{array}$ Constituted LMS prediction module 202 jointly.

The blind P-LMS-LMS method of high speed of the present invention is finished the auxiliary filtering 203 of LMS sign indicating number when carrying out LMS predictive filtering 202.For LMS sign indicating number supplementary module 203, sampled signal r (m) 224 carries out windowing storage 281, obtains windowing vector of samples r (n)=[r (nN+N-1) ... r (nN)] ^T282.Subsequently, by windowing vector of samples r (n)=[r (nN+N-1) ... r (nN)] ^T282 and the control of prediction weight vector recursion $w(m+1)=w\left(m\right)+2\mathrm{\&mu;}{\tilde{r}}^{*}\left(m\right)r_M\left(m\right)$ 264 weight vectors that obtain are jointly as the auxiliary weight vector recurrence formula Q (n+1) of factor of influence control code=Q (n)+2 μ [W ₁ ^H(n) SP-tril (W ₁ ^H(n) r (n) r ^H(n) W ₁(n) Q (n))] 283, and each more new capital of w and the renewal of Q delegation carry out synchronously, obtains recursion weight vector Q 284.Wherein, the auxiliary step factor of sign indicating number satisfies $0<\widetilde{\mathrm{\&mu;}}<1/{\widetilde{\mathrm{\&lambda;}}}_{\max },$ Wherein

Be correlation matrix $R_{\tilde{r}\tilde{r}}=E\left\{\tilde{r}\left(n\right){\tilde{r}}^H\left(n\right)\right\}$ Eigenvalue of maximum; The auxiliary weight vector initial condition of sign indicating number of the blind P-LMS-LMS method of high speed is Q (0)=0.Then, with recursion weight vector Q284 with by the control of prediction weight vector recursion $\begin{array}{cc}w(m+1)=w\left(m\right)+2\mathrm{\&mu;}{\tilde{r}}^{*}\left(m\right)r_M\left(m\right)& 264\end{array}$ The prediction weight vector w that obtains is through W ₁The W that maker 285 obtains ₁286 carry out bit estimates Q ^H(n) W ₁ ^H(n) r (n) 287 obtains bit estimated signal 288.At last, to bit estimated signal 288 number of winning the confidence real parts 289, obtain the subscriber signal estimated value $\begin{array}{cc}\hat{b}\left(n\right)=\mathrm{Re}\left(Q^H\left(n\right)W_1^H\left(n\right)r\left(n\right)\right)& 290\end{array}$ , Re (●) the expression number of winning the confidence real part in the formula.

Like this, the windowing storage 281; The auxiliary weight vector recursion control of sign indicating number Q (n+1)=Q (n)+2 μ [W ₁ ^H(n) SP-tril (W ₁ ^H(n) r (n) r ^H(n) W ₁(n) Q (n))] 283; W ₁Maker 785; Bit is estimated $\begin{array}{cc}{Q}^H\left(n\right)\tilde{r}\left(n\right)& 287\end{array}$ With the number of winning the confidence real part 289 constituted jointly the blind P-LMS-LMS method of high speed the sign indicating number auxiliary filtration module 103.

Fig. 3 describes the performance simulation correlation curve that the blind P-LMS-LMS method of high speed and existing blind S-LMS-LMS method suppress all kinds of narrow band interference, wherein Fig. 3 a describes to receive the performance simulation correlation curve of audio disturbances, Fig. 3 a-1 describes to receive the performance simulation correlation curve that single-tone disturbs, and Fig. 3 a-2 describes to receive the performance simulation correlation curve that multitone disturbs; Fig. 3 b describes to receive the performance simulation correlation curve of digital narrow band interference; Fig. 3 c describes to be received from the performance simulation correlation curve that returns random process.Selecting system mean square error average reflection system is defined as the rejection of narrow band interference

$J\left(n\right)=E\left\{{[b\left(n\right)-\hat{b}\left(n\right)]}^H[b\left(n\right)-\hat{b}\left(n\right)]\right\}/K$

User's mean square error average can reflect disturb to suppress filtering after, the multi-user's decoded signal estimated value before the conclusive judgement and the average mean square error of actual value can also embody the average error rate of system.When user's mean square error average much smaller than 1 the time, can be considered to disturb and suppress the filtering success, and when its near 1 or greater than 1 the time, can be considered and disturb that to suppress filter effect relatively poor or fail.The mean square error average is more little, and filter effect is good more.Simulated conditions is set at: 3 CDMA users, spreading code N=63; The audio disturbances frequency accidental, the cycle of digital narrow band interference is 4 with sending signal data cycle ratio; Second order autoregression random process autoregressive coefficient φ ₁=-1.98, φ ₂=0.9801; Step factor μ=0.00005, $\widetilde{\mathrm{\&mu;}}=0.005;$ DS-CDMA signal and white noise power ratio, promptly signal to noise ratio snr=-15dB, DS=CDMA signal and narrow band interference power ratio, i.e. signal interference ratio SJR=-20dB.

As shown in the figure, Fig. 3 a-1 describes to receive the performance simulation correlation curve that single-tone disturbs, J 301 expression user mean square error averages, and n 302 expressions send signal message quantity.When interference was disturbed for single-tone, user's mean square error average performance of the blind P-LMS-LMS method 303 more existing blind S-LMS-LMS methods 304 of high speed descended to some extent, but steady-state value can normally be used still much smaller than 1.Fig. 3 a-2 describes to receive the performance simulation correlation curve that multitone disturbs-3 sounds to disturb, and J305 represents user's mean square error average, and n306 represents to send signal message quantity.When disturbing when disturbing for multitone, equally, user's mean square error average 301 performances of the blind P-LMS-LMS method 307 more existing blind S-LMS-LMS methods 308 of high speed descend to some extent, but steady-state value can normally use still much smaller than 1, and are identical with single-tone interference conclusion.Simultaneously, because multitone disturbs the reduction of disturbing predictability than single-tone, make stable state mean square error average 305 performances of blind P-LMS-LMS method 307 of high speed and existing blind S-LMS-LMS method 308 descend to some extent.Figure a has proved that the blind P-LMS-LMS method of high speed is to drop to the conclusion that cost has exchanged the high speed processing effect for by a small margin with systematic function.

Fig. 3 b describes to receive the performance simulation correlation curve of digital narrow band interference, and J309 represents user's mean square error average, and n310 represents to send signal message quantity.Different with audio disturbances, when disturbing to digital narrow band interference, steady-state system mean square error average 309 performances of blind P-LMS-LMS method 311 of high speed and existing blind S-LMS-LMS method 312 are suitable, and existing blind S-LMS-LMS method 312 convergence rates are fast slightly, but the two is very nearly the same.

Fig. 3 c describes to be received from the performance simulation correlation curve that returns random process, and J313 represents user's mean square error average, and n314 represents to send signal message quantity.When disturbing to the autoregression random process, user's mean square error average performance of the blind P-LMS-LMS method 315 more existing blind S-LMS-LMS methods 316 of high speed descends to some extent, but steady-state value can normally be used still much smaller than 1.Can draw the blind P-LMS-LMS method of high speed equally is to drop to the conclusion that cost has exchanged the high speed processing effect for by a small margin with systematic function.

Fig. 4 is the flow chart of method.In step 401, wireless communication signals 205 is down-converted to intermediate-freuqncy signal 210.In step 402, at first use modulus a/d transducer 211 that intermediate-freuqncy signal 210 is digitized as digital signal, carry out QPSK demodulation output baseband signal 222 thereafter.In step 403, baseband signal 222 is obtained sampled signal r (m) 224 by cutting general matched filtering sampler 223.In step 404, prediction weight vector and the auxiliary weight vector of sign indicating number are carried out initialization process, w (0)=0, Q (0)=0.In step 405, to prediction weight vector and the auxiliary weight vector iterative processing of sign indicating number, $w(m+1)=w\left(m\right)+2\mathrm{\&mu;}{\tilde{r}}^{*}\left(m\right)r_M\left(m\right),$ $\tilde{r}\left(m\right)=r\left(m\right)-\hat{r}\left(m\right)=r\left(m\right)-w^H\left(m\right)r_M\left(m\right),$ Q(n+1)＝Q(n)+2μ[W ₁ ^H(n)SP-tril(W ₁ ^H(n)r(n)r ^H(n)W ₁(n)Q(n))]。In step 406, do the processing of n=n+1.In step 407, judge n whether greater than sending the signal message sum, if, process ends, if not, as from 407 send return arrow indicated.

Claims (6)

1, a kind of high-speed predictive-code auxiliary method that suppresses the spread spectrum system narrow band interference is characterized in that comprising:

(1) receives wireless communication signals;

(2) described wireless communication signals is down-converted to intermediate-freuqncy signal;

(3) the described intermediate-freuqncy signal of digitlization provides digital signal;

(4) with described digital demodulation signal, provide baseband signal;

(5) with described baseband signal envelope r (t) by cutting general matched filtering so that sampled signal r to be provided (m);

(6) described sampled signal r (m) is carried out high speed filtering.

2, the high-speed predictive-code auxiliary method of inhibition spread spectrum system narrow band interference as claimed in claim 1, it is characterized in that described wireless communication signals comprises DS-CDMA signal, white noise and narrow band interference, wherein DS-CDMA signal spread-spectrum sign indicating number is chosen short code, short code code length N≤63; Narrow band interference comprises the class in audio disturbances, digital narrow band interference or the autoregression random process.

3, the high-speed predictive-code auxiliary method of inhibition spread spectrum system narrow band interference as claimed in claim 1 or 2, it is characterized in that described high speed filtering take blind parallel-lowest mean square prediction-lowest mean square sign indicating number assists (P-LMS-LMS) method, comprises the steps:

Step 1:LMS predictive filtering

At first described sampled signal r (m) is carried out the LMS predictive filtering, establishing the predictive filtering exponent number is M, then M dimension prediction weight vector more new formula be

$w(m+1)=w\left(m\right)+2\mathrm{\&mu;}{\tilde{r}}^{*}\left(m\right)r_M\left(m\right)$

Predicated error wherein, promptly LMS predictive filtering signal is

$\tilde{r}\left(m\right)=r\left(m\right)-\hat{r}\left(m\right)=r\left(m\right)-w^H\left(m\right)r_M\left(m\right)$

In the formula: r _M(m)=[r (m-1) ... r (m-M)] ^TBe the filter input vector; μ is the prediction step factor;

Step 2:LMS sign indicating number is assisted filtering

With described LMS predictive filtering signal

At processing time [nT _b, (n+1) T _b] interior windowing to be to extract LMS predictive filtering vector $\tilde{r}\left(n\right)={[\tilde{r}(\mathrm{nN}+N-1)\&CenterDot;\&CenterDot;\&CenterDot;\tilde{r}\left(\mathrm{nN}\right)]}^T,$ Described LMS predictive filtering vector is carried out the auxiliary filtering of LMS sign indicating number, and (the auxiliary weight vector of dimension sign indicating number of N * K) more new formula is

$Q(n+1)=Q\left(n\right)+2\widetilde{\mathrm{\&mu;}}\tilde{r}\left(n\right)e^H\left(n\right)$

In the formula: $e\left(n\right)=b\left(n\right)-Q^H\left(n\right)\tilde{r}\left(n\right)$ Be error vector, (n)=[b ₀(n) ... b _K-1(n)] ^TBe user's bit vectors,

Be signal flow (1 or-1);

Be the auxiliary step factor of sign indicating number;

Bring described error vector into described weight vector more new formula, obtain

$Q(n+1)=Q\left(n\right)+2\widetilde{\mathrm{\&mu;}}\tilde{r}\left(n\right)[b^T\left(n\right)-{\tilde{r}}^H\left(n\right)Q\left(n\right)]$

In formula

Decompose

$\tilde{r}\left(n\right)=W^H\left(n\right)\{\left[\begin{array}{c}S\\ 0\end{array}\right]\mathrm{Pb}\left(n\right)+\left[\begin{array}{c}0\\ S_M\end{array}\right]\mathrm{Pb}(n-1)\}+\tilde{i}\left(n\right)+\widetilde{\mathrm{\&epsiv;}}\left(n\right)$

In the formula: $P=\mathrm{diag}(\sqrt{P_0}\&CenterDot;\&CenterDot;\&CenterDot;\sqrt{P_{K-1}}),$ Be signal power diagonal matrix, P _kBe signal power, wherein K is the CDMA number of users; S=[s ₀S _K-1] be CDMA spreading code matrix, wherein s _k=[c _{K, N-1}C _{K, 0}] T/N is the spreading code vector, { c _{K, i}: i=0 ..., N-1} is direct sequence spread spectrum codes (1 or-1), N is a spreading gain; Definition S _MFor 1～M of S is capable, 1～N row, definition simultaneously

Utilize E{b (n) b ^T(n) }=and I, E{b (n-1) b ^T(n) }=0, obtain

$\tilde{r}\left(n\right)b^T\left(n\right)=E\left\{\tilde{r}\left(n\right)b^T\left(n\right)\right\}={W_1}^H\left(n\right)\mathrm{SP}$

And then obtain need not the auxiliary weight vector recurrence formula of sign indicating number of the blind S-LMS-LMS method of training sequence

$Q(n+1)=Q\left(n\right)+2\widetilde{\mathrm{\&mu;}}[{W_1}^H\left(n\right)\mathrm{SP}-\tilde{r}\left(n\right){\tilde{r}}^H\left(n\right)Q\left(n\right)]$

The key that realizes parallel processing is to make the sign indicating number supplementary module of blind S-LMS-LMS method need not the result of calculation of prediction module

Because at W ₂(n) (in the dimension of the M * N) element, only there is nonzero value in the littlest triangle battle array of below, so can be with described

The part of middle W2 (n) participation computing is approximate to be cast out, and be need not

Participate in the parallel weight vector recurrence formula of computing directly

$Q(n+1)=Q\left(n\right)+2\widetilde{\mathrm{\&mu;}}[{W_1}^H\left(n\right)\mathrm{SP}-{W_1}^H\left(n\right)r\left(n\right)r^H\left(n\right)W_1\left(n\right)Q\left(n\right)]$

In the following formula, Q at every turn more new capital need wait for that w upgrades formation W N time ₁, significantly not shortening the processing time, the key that shortens the processing time is that the renewal of w and the renewal of Q are carried out synchronously, analyze the auxiliary recurrence formula of described blind predictive code, predict is finished the renewal of weight vector w for the first time, can constitute first row of W1, the premultiplication matrix is equivalent to do line translation, can get W ₁ ^H(n) SP and W ₁ ^H(n) first of r (n) row, right multiply matrix is equivalent to do rank transformation, can get r ^H(n) W ₁(n) first row have promptly obtained W ₁ ^H(n) r (n) r ^H(n) W ₁(n) first row, first row are made important hypothesis here, make W ₁ ^H(n) r (n) r ^H(n) W ₁(n) remainder data of first row is 0, just can finish the recursion of Q first row; In like manner, when predict is finished the renewal of the weight vector w second time, utilize the renewal result of weight vector for the first time, can obtain first and second data of Q second row, suppose that the remainder data that Q second goes is 0, just finished the recursion of Q second row; The rest may be inferred, and the each more new capital of w and the renewal of Q delegation are carried out synchronously, obtains the auxiliary weight vector recurrence formula Q (n+ of sign indicating number of the blind P-LMS-LMS method of high speed ₁)=Q (n)+ ₂μ [W ₁ ^H(n) SP-tril (W ₁ ^H(n) r (n) r ^H(n) W ₁(n) Q (n))]

Wherein the triangle battle array is taken off in tril () expression.

4, the high-speed predictive-code auxiliary method of inhibition spread spectrum system narrow band interference as claimed in claim 3 is characterized in that the prediction step factor satisfies 0＜μ＜1/ λ _Max, λ wherein _MaxBe correlation matrix $R_{r_M{r}_M}=\frac{1}{N}{\mathrm{\&Sigma;}}_{k=1}^N{R}_{2r2r}(k:k+M-1,k:k+M-1)$ Eigenvalue of maximum, the definition $R_{2r2r}=\left[\begin{array}{cc}{R}_{\mathrm{rr}}& R_{\mathrm{rr}}\\ R_{\mathrm{rr}}& R_{\mathrm{rr}}\end{array}\right],$ R _Rr=E{r (n) r ^H(n) }, r (n)=[r (nN+N-1) ... r (nN)] ^TThe auxiliary step factor of sign indicating number satisfies $0<\widetilde{\mathrm{\&mu;}}<1/{\widetilde{\mathrm{\&lambda;}}}_{\max },$ Wherein

Be correlation matrix $R_{\tilde{r}\tilde{r}}=E\left\{\tilde{r}\left(n\right){\tilde{r}}^H\left(n\right)\right\}$ Eigenvalue of maximum.

5, the high-speed predictive-code auxiliary method of inhibition spread spectrum system narrow band interference as claimed in claim 3 is characterized in that the auxiliary weight vector initial condition of sign indicating number of predicting weight vector and the blind P-LMS-LMS method of high speed is respectively w (0)=0 and Q (0)=0.

6, the high-speed predictive-code auxiliary method of inhibition spread spectrum system narrow band interference as claimed in claim 4 is characterized in that the auxiliary weight vector initial condition of sign indicating number of predicting weight vector and the blind P-LMS-LMS method of high speed is respectively w (0)=0 and Q (0)=0.

Images