CN102291160B - TDM (Time Division Multiplexing)-based high-order adaptive wave trap of parallel structure and adaptive wave trap method - Google Patents

Abstract

The invention discloses a TDM (Time Division Multiplexing)-based high-order adaptive wave trap of a parallel structure and an adaptive wave trap method, and belongs to communication technology. The method comprises the following steps of: dividing a high-speed sampling point sequence into N (N is greater than or equal to 2 and is less than or equal to 9) low-speed sampling point sequence branchesbased on a narrow band analytic signal sampling theorem by using a sampling point interleaving TDM technology; performing cascaded high-order adaptive wave trap respectively; and combining by using the TDM technology to reduce the low-speed sampling point sequence branches into a high-speed sequence, wherein the entire wave trap falls to be zero close to a specific frequency omega i; and omega i is amended in a unified way by using a linkage method to realize adaptive iteration. Through the wave trap serving as preprocessing before de-spreading, a plurality of narrow band strong interferencescan be suppressed, and the interference tolerance of the system can be greatly improved. High-order adaptive wave trap can be realized at the sampling rate of over hundreds of megabits, the realization complexity is low, the wave trap effect and the stability are high, a plurality of zero fall combination phenomena can be reduced or avoided, and the number of zero fall frequency points is approximately the same as the number of the actual narrow band strong interferences. The wave trap and the wave trap method are mainly used in wideband or ultra wideband digital communication systems.

Description

Parallel organization high-order adaptive notch filter and adaptive resistance-trap wave method based on TDM

Technical field

The invention belongs to communication technical field, relate to Suppression of narrow band interference in the receiver, a kind of based on time division multiplexing parallel organization high-order adaptive notch filter and the adaptive resistance-trap wave method of (being called for short TDM) specifically, be applied to antagonism arrowband interference in the digital communication system, especially for the preliminary treatment of receiver in the direct sequence spread spectrum communication system before despreading.

Background technology

A system of broadband wireless communication, when the frequency band that is interfered than signal of communication frequency band narrow (for example 1/10) for a long time, this interference can be considered the arrowband and disturbs.Disturb for such arrowband, can utilize the correlation between its sampling point to estimate by various Forecasting Methodologies, and then from receive signal, deduct the interference that estimates, reach the purpose that suppresses interference, referring to: 1. Stefano B, Marco L.Code-aided Interference Suppression for DS/CDMA overlay systems[J] .Proceedings of the IEEE.2002,90 (3): 394-435; 2. Rahul S, Milstein L B.Adaptive Interference Suppression for DS-CDMA[J] .IEEE Transactions on Communications.2002,50 (12): 1902-1905.These Forecasting Methodologies can be summarized as monolateral line of time delay fallout predictor, bilateral line of time delay fallout predictor, interpolative prediction device and lattice filter etc.UWB signal and white noise all are the broadbands, when sample rate during near the nyquist sampling rate, do not have correlation between adjacent chip or the sampling point basically; Be difficult to extract signal correlation by various Forecasting Methodologies; Therefore, receive signal although they and arrowband disturb to be added together to constitute, they do not influence basically by receiving signal estimation estimates that the arrowband disturbs.

The Suppression of narrow band interference technology can be divided into two classes usually: the first kind is to suppress at transform domain; Another kind of is that the adaptive notch filter that adopts time domain to realize disturbs inhibition.Referring to:

[1]Carvalho?P，Dinis?R，Lourenco?D，UWB?Digital?Carrier?User?Codes?for?Narrow?Band?Interference?Cancelation.The?3rd?Int.Conf.on?Signal?Processing?and?Communication?Systems，2009：1-5.

[2]Wei?L，Weixia?Z，Fangmin?X，Zheng?Z，Kyung?S?W.A?Novel?Frequency-band?Coded?Orthogonal?UWB?Chirp?Pulse?Design?for?Cognitive?NBI?Suppression.Int.Symp.on?Microwave，Antenna，Propagation?and?EMC?Technologies?for?Wireless?Communications，2007：1048-1051.

[3]Jeffrey?A?Y，James?S?L.Analysis?of?DFT-based?Frequency?Excision?Algorithms?for?Direct-Sequence?Spread-Spectrum?Communications[J].IEEE?Trans.on?Commun.1998，46(8)：1076-1087.

[4]Xuemei?O，Moeness?G?A.Short-time?Fourier?Transform?Receiver?for?Nonstationary?Interference?Excision?in?Direct?Sequence?Spread?Spectrum?Communications[J].IEEE?Trans.on?Signal?Processing.2001，49(4)：851-863.

[5]Marco?Lops，Antonia?M?T.Automatic?Suppression?of?Narrow-band?Interference?in?Direct-Sequence?Spread-Spectrum?Systems[J].IEEE?Trans.on?Commun.1999，47(8)：1133-1136.

Above-mentioned document [1] adopts special pulse shaping mode, disturbs to suppress the arrowband; Document [2] adopts frequencyband coding quadrature UWB pulse design method, utilizes cognitive radio technology, designs special pulse and disturbs to suppress the arrowband; Document [3] adopts the frequency shearing algorithm based on real-time DFT to suppress the arrowband interference; Document [4] adopts the arrowband interference that suppresses to have non-stationary property based on the method for short time discrete Fourier transform; Become processing method when document [5] adopts, suppress the arrowband interference by carry out nonlinear filtering at transform domain.

Document Sami A, Pertti H.Transform-Selective Interference Suppression Algorithm for Spread-Spectrum Communications[J] .IEEE Signal Processing Letters.2005,12 (1): 49-51. compares and analyzes the interference inhibition strategy of multiple transform domain.

Generally speaking, when being dynamic, the characteristic that the arrowband disturbs becomes, so the trapper that often requires to suppress to disturb has adaptation function---and namely can realize that the oneself adjusts by iterative algorithm, and converge to optimum solution in certain statistical significance.Adaptive algorithm commonly used adopts based on minimum mean square error criterion or the recursive least-squares criterion of gradient at random.

At the hundreds of spread spectrum communications signals more than million, particularly ultra-broad band communication signal, above-mentioned algorithm can not be used directly real-time implementation of hardware (for example FPGA) for spreading rate, and needs to adopt parallel organization to reduce its work clock.Parallel organization has various ways, adopts different parallel organizations, and there are very big difference in its complexity and performance.The problem that the high-order trapper also exists stability problem and a plurality of zero trapping spot to be merged together, though under not high sample rate inaccessible good performances all.Suppressing the arrowband strong jamming in the wideband communication signal that bandwidth reaches hundreds of million, is the difficult problem of fine solution not as yet at present; And the jamming margin of spread spectrum communication system is subjected to the restriction of spread spectrum multiple, only carries out the preliminary treatment of Suppression of narrow band interference before the receiving terminal despreading, and the jamming margin of system is increased substantially.

Summary of the invention

The objective of the invention is to overcome the problem that bad, a plurality of zero trapping spots of existing parallel organization trapper implementation complexity height, trap effect and stability are merged together, provide a kind of implementation complexity low, trap effect and good stability, can reduce or avoid the phenomenon of zero sunken unreasonable merging, make zero to fall into approximate parallel organization high-order adaptive notch filter and the adaptive resistance-trap wave method based on TDM identical with actual arrowband strong jamming number of number.

The technical solution adopted for the present invention to solve the technical problems is: make up the parallel organization high-order adaptive notch filter based on TDM, comprise string and conversion time-division splitter, and string conversion time division multiplexing device and be connected on splitter and multiplexer between N rank adaptive notch filter; The reception signal x (k) of a communication control processor is input to N rank adaptive notch filter concurrently simultaneously, and after the arrowband strong jamming in the adaptive notch filter filtered signal, remerging is one road signal y (k) output; Adaptive notch filter adopts N rank (2≤N≤9) parallel organization, and this parallel organization comprises N bar parallel branch, every route N unit cascaded formation of single order adaptive notch filter; Putting in order of trapper is the cyclic shift that trapper puts in order in its adjacent legs in every branch road; (1≤i≤N) transfer function of single order adaptive notch filter unit is i in every parallel branch

Angular frequency parameter ω wherein _i(abbreviation angular frequency) is by with output signal energy minimum being the adaptive iteration acquisition of criterion ,-π＜ω _i＜π; Parameter r selects 0.8＜r＜1 according to the bandwidth that hope suppresses; It is characterized in that: described string and conversion time-division splitter carry out splitter along separate routes for adopting sample interweaving TDM technology, the described and multiplexer of string conversion time division multiplexing device for adopting the TDM technology to carry out multiple connection; Described reception signal x (k) is an intermediate frequency analytic signal, comprises the zero intermediate frequency analytic signal, and this signal is the digital sampling point sequence of plural form.

At the hundreds of spread-spectrum signal more than million, especially ultra-broad band communication signal, the present invention adopts parallel organization can reduce its work clock for spreading rate, is easy to the realization of hardware (for example FPGA).

The parallel organization of the routine that above-mentioned N rank adaptive notch filter adopts is with reference to Wang Yongchao thesis for the doctorate " some Base-Band Processing key issues research in the wireless communication system " (Xian Electronics Science and Technology University's thesis for the doctorate, 2006.12.) the middle a kind of structure that proposes.The present invention further improves on this architecture basics, has proposed the parallel organization based on sample interweaving TDM technology; String among the present invention and conversion time-division splitter carry out splitter along separate routes for adopting sample interweaving TDM technology, the reception signal x (k) of a communication control processor is input to N bar branch road when adopting sample interweaving TDM technology to carry out shunt, each road signal is equivalent to carry out N: 1 down-sampling, sampling rate have reduced N-1 doubly; Compare with the N road parallel organization of routine, its work clock can reduce N-1 doubly, thereby computation complexity reduces significantly, is conducive to adopt directly real-time implementation of hardware (as FPGA).

The present invention still is a kind of parallel organization high-order adaptive resistance-trap wave method based on TDM, and the adaptive resistance-trap wave process comprises the steps:

Step 1: adopt sample interweaving time-division method along separate routes, the high-speed sampling sampling point sequence x (k) that communication control processor is received is that the sampling point sequence branch road of N bar low speed, x (k) are an intermediate frequency analytic signal along separate routes;

Step 2: various kinds point sequence branch road carries out cascade connection type high-order adaptive notch respectively, obtains N road error signal, and this N road error signal in fact all is to have suppressed the useful signal that the arrowband disturbs; Each adaptive notch filter unit is only in the specific angle frequencies omega _iNeighbouring that the arrowband interference trap is extremely zero, angular frequency _iAdaptive iteration adopt the unification of interlock mode to make amendment;

Step 3: adopting sample interweaving time division multiplexing technology, is one tunnel output with the error signal multiple connection of N road, and the useful signal y (k) that the arrowband disturbs has been eliminated.

The parallel organization high-order adaptive resistance-trap wave method that the present invention is based on TDM adopts sample interweaving TDM technology to carry out the N road time-division along separate routes, and each bar branch road sample rate has reduced N-1 doubly, thereby work clock speed reduces the implementation complexity reduction; The relative bandwidth that the arrowband disturbed when the present invention adopted above-mentioned trap method to carry out trap broadens, simultaneously the transfer function H of used trapper during trap _i(z) limit can be too near unit circle, thereby good stability; Total notch bandwidth always is less than or equal to the summation of each arrowband jamming bandwidth, so the trap frequency band can be narrower, and is less for the useful signal damage, thereby trap is effective.

Realization of the present invention also is: adopt the interlock mode unification realization adaptive iteration of making amendment in the above-mentioned steps 2, specifically include:

3.1 with transfer function H in the n bar branch road _i(z) angular frequency parameter is made as ω _i(k);

3.2 the adaptive iteration index word that estimates is Δ ω _i(k), Δ ω _i(k)=μ Im[T _i(k)], T wherein _i(k) be ω _i(k) Grad in the adaptive iteration computing, μ are step factor;

3.3 to { ω _i(k), i=1,2 ..., when N} carries out actual modification, adopt its average index word Simultaneously to the angular frequency parameter ω in the N bar branch road _i(k) make amendment, realize adaptive iteration.

The present invention adopts the interlock mode to ω _iThe unified modification can be reduced the implementation complexity of parallel organization; Because to N identical transfer function H _i(z) angular frequency _iThe modification unanimity, therefore can reduce angular frequency greatly _iAmount of calculation when carrying out adaptive iteration.

Realization of the present invention also is: in the above-mentioned parallel organization high-order adaptive resistance-trap wave method based on TDM, the input signal of single order adaptive notch filter unit is x _n(k), output signal is error signal e _n(k), angular frequency parameter ω in its transfer function _n(n=1,2,3 ..., adaptive iteration step N) is:

4.1: initialization: make k=0, e _n(k)=x _n(k)=0; ω _n(k)=π/(n-N/2);

4.2: compute gradient value: make k+1 → k, to input vector X _n=[x _n(k), x _n(k-1) ..., x _n(k-N+1)] ^T

$e_n\left(k\right)=x_n\left(k\right)-e^{j{\mathrm{\ω}}_n\left(k\right)}{e}_n(k-1)$

$T_n\left(k\right)=e^{j{\mathrm{\ω}}_n\left(k\right)}{x}_n^{*}\left(k\right)[r{e}_n(k-1)-x_n(k-1)]$

Average: $\stackrel{\‾}{T}\left(k\right)=\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{T}_n\left(k\right)/N;$

4.3: ask angular frequency parameter ω _n(k) mean value: ask

(getting step size mu=0.0001～0.1);

4.4: revise the angular frequency parameter:

${\mathrm{\ω}}_n(k+1)={\mathrm{\ω}}_n\left(k\right)+\mathrm{\Δ}{\stackrel{\‾}{\mathrm{\ω}}}_n\left(k\right)$

Return step 2, handle next sampling point.

Angular frequency parameter ω among the present invention _n(k) when carrying out adaptive iteration, as long as the frequency of the N kind single order trapper on each bar branch road in the initial condition is scattered to distribute, just can reduce or avoid a plurality of zero sunken frequency to merge into one phenomenon unreasonably, make zero to fall into frequency points close to actual arrowband strong jamming number.

Compared with prior art the present invention has following advantage:

1) implementation complexity is low: because parallel organization of the present invention is based on the realization of sample interweaving TDM technology, therefore the sampling point speed of each branch road has reduced N-1 doubly, compare with the N road of routine is parallel, its work clock can reduce N-1 doubly, thereby computation complexity reduces significantly, and work clock also can carry out under the operating frequency that FPGA realizes being convenient to.Therefore not only be suitable for the broadband direct sequence spread spectrum communication system, also be suitable for ultra-wideband communication system.

2) trap is effective: adopt sample interweaving time-division way along separate routes, x (k) is divided into the N road, every road signal is equivalent to carry out N: 1 down-sampling extracts, total frequency band that the effect of down-sampling can not make the arrowband disturb broadens, can become narrower because of spectral aliasing on the contrary, therefore total notch bandwidth always is less than or equal to the summation of each arrowband jamming bandwidth, thereby the trap frequency band can be narrower, and is less for the useful signal damage.

3) good stability: because each branch road sample rate has reduced N-1 doubly, the relative bandwidth that the arrowband disturbs broadens the transfer function H of used trapper unit during trap _i(z) limit can be too near unit circle, thereby good stability.

4) can reduce or avoid a plurality of zero falling into to merge into one phenomenon: as long as the frequency ω of the N kind single order trapper on each bar branch road in the initial condition is scattered to distribute, just can reduce or avoid a plurality of zero falling into to merge into one phenomenon unreasonably, make zero to fall into number close to actual arrowband strong jamming number.

Description of drawings

Fig. 1 is the formation schematic diagram that the present invention is based on the parallel organization N rank adaptive notch filter of TDM;

Fig. 2 is single order adaptive notch filter theory diagram of the present invention.

Embodiment

Embodiment 1

With reference to Fig. 1, the present invention is a kind of parallel organization high-order adaptive notch filter based on TDM, comprise string and conversion time-division splitter, and string conversion time division multiplexing device and be connected on splitter and multiplexer between N rank adaptive notch filter; The reception signal x (k) of a communication control processor is input to N rank adaptive notch filter concurrently simultaneously, and after the arrowband strong jamming in the adaptive notch filter filtered signal, remerging is one road signal y (k) output.String of the present invention and conversion time-division splitter carry out splitter along separate routes for adopting sample interweaving TDM technology, and the multiplexer of string conversion time division multiplexing device for adopting the TDM technology to carry out multiple connection, receiving signal x (k) is an intermediate frequency analytic signal, comprise the zero intermediate frequency analytic signal, this signal is the digital sampling point sequence of plural form.

Adaptive notch filter of the present invention adopts N rank (2≤N≤9) parallel organization, and this parallel organization comprises N bar parallel branch, every route N unit cascaded formation of single order adaptive notch filter; Putting in order of trapper is the cyclic shift that trapper puts in order in its adjacent legs in every branch road; Be example with N=8, wherein 8 single order trappers of the 1st branch road put in order be 8,7,6 ..., 2,1; Article 2,8 single order trappers of branch road put in order be 1,8,7,6 ..., 2; The rest may be inferred for remaining; That is to say the down-sampling that the signal of each branch road has carried out 8: 1, so speed reduces 8-1 doubly, namely 7 times, reduced clock rate effectively.In the time of need satisfying certain sampling condition, arrowband wherein disturbs just therefore spectral aliasing can not take place.(1≤i≤N) transfer function of single order adaptive notch filter unit is i in every parallel branch

Angular frequency parameter ω wherein _iBy with output signal energy minimum being the adaptive iteration acquisition of criterion ,-π＜ω _i＜π; Parameter r selects 0.8＜r＜1 according to the bandwidth that hope suppresses.

The sample interweaving TDM technology that is based on parallel organization of the present invention realizes, therefore the sampling point speed of each branch road has reduced N-1 doubly, compare with the N road of routine is parallel, its work clock can reduce N-1 doubly, thereby computation complexity reduces significantly, and work clock also can carry out under the operating frequency that FPGA realizes being convenient to.Therefore not only be suitable for the broadband direct sequence spread spectrum communication system, also be suitable for the application of ultra-wideband communication system.

The process of parallel organization high-order adaptive notch that the present invention is based on TDM is as follows:

Step 1: adopt sample interweaving time-division method along separate routes, the high-speed sampling sampling point sequence x (k) that communication control processor is received is the sampling point sequence branch road of N bar low speed along separate routes, and x (k) is an intermediate frequency analytic signal, referring to Fig. 1;

Step 2: various kinds point sequence branch road carries out cascade connection type high-order adaptive notch respectively, obtains N road error signal, and this N road error signal in fact all is to have suppressed the useful signal that the arrowband disturbs, referring to Fig. 1; Each adaptive notch filter unit is only in the specific angle frequencies omega _iNeighbouring that the arrowband interference trap is extremely zero, angular frequency _iAdaptive iteration, be to adopt the unification of interlock mode to make amendment, referring to Fig. 2;

Step 3: adopting sample interweaving time division multiplexing technology, is one tunnel output with the error signal multiple connection of N road, and the useful signal y (k) that the arrowband disturbs that has been eliminated is referring to Fig. 1.

The high-order adaptive resistance-trap wave process of above-mentioned parallel organization adopts sample interweaving TDM technology, a high-speed sampling sampling point sequence x (k) who receives is the sampling point sequence branch road of N low speed along separate routes, each road signal is equivalent to carry out N: 1 down-sampling, the arrowband disturbs therefore spectral aliasing can or can not take place, if what influence the occurrence frequency aliasing has again, below this is analyzed and give and issued a certificate.

As long as the arrowband analytic signal that [theorem] bandwidth is the B hertz, the sampling rate when it is carried out quadrature sampling be greater than B sampling point/second, and the gained sampled signal just can be used for recovering original signal undistortedly.

Proof procedure is as follows:

1) establishing x (k) is that bandwidth is the sample sequence of the arrowband analytic signal of B hertz, and establishing its sample rate is MB (sampling point/second), wherein integer M＞＞1, its frequency spectrum function is a periodic function in-∞＜f＜∞ scope so, its cycle is MB, namely satisfies relational expression X (f)=X (f+iMB), wherein i=-∞, ...,-1-0,1,2, ..., ∞.These characteristics of noting discrete analytic signal are fundamentally different than discrete real signal, the frequency spectrum of discrete real signal only (∞, 0) or (0, ∞) present in the frequency range periodically;

2) x (k) is carried out N: 1 extraction (being down-sampling), N＜M wherein, M/N is the integer greater than 1; It is that periodicity δ function and { x (k) } of N multiplies each other that this extraction is equivalent to one-period, namely

$\stackrel{\·\·\·}{x}\left(k\right)=\underset{i=-\∞}{\overset{\∞}{\mathrm{\Σ}}}\mathrm{\δ}(k-\mathrm{iN})x\left(k\right)$

Notice that the above-mentioned cycle is that time interval of adjacent two δ pulses of the periodicity δ function of N is N/ (MB), so its frequency spectrum function D (f) is that one-period is the periodicity δ function of MB/N;

3) according to convolution theorem, signal Frequency spectrum

Equal the convolution of frequency spectrum X (f) Yu the D (f) of x (k), namely

$\stackrel{\·\·\·}{X}\left(f\right)=X\left(f\right)*D\left(f\right)=X\left(f\right)*\underset{k=-\∞}{\overset{\∞}{\mathrm{\Σ}}}\mathrm{\δ}(f-\mathrm{kMB}/N)=\underset{k=-\∞}{\overset{\∞}{\mathrm{\Σ}}}X(f-\mathrm{kMB}/N)$

Obviously Also be that one-period is the periodic function of MB/N;

4) above-mentioned convolution process condition that spectral aliasing does not take place is: M/N is for greater than 1 integer; And wherein M can be big arbitrarily, as seen

Sample rate MB/N under the situation of approaching B arbitrarily, can both satisfy the condition of no spectral aliasing sampling;

When 5) condition of spectral aliasing after satisfying above-mentioned down-sampling, not occurring,

Non-null part among non-zero frequency spectrum function in each cycle and the X (f) in the frequency spectrum function in each cycle is accurately to equate, thereby can use

Undistorted ground reconstruct x (k).The theorem card is finished.

The present invention adopts sample interweaving time-division way along separate routes, is the N road along separate routes with x (k); So every road signal is equivalent to carry out N: 1 down-sampling extracts, and sample rate is reduced to the 1/N of former sample rate, referring to Fig. 1.The theorem of crossing according to above-mentioned proof as can be known, as long as satisfy N＜M, and M/N is the integer greater than 1, the signal in each branch road of the present invention can be because of N: spectral aliasing takes place in 1 down-sampling.

Embodiment 2

The present invention is based on the parallel organization high-order adaptive notch filter of TDM and based on the parallel organization high-order adaptive resistance-trap wave method of TDM with embodiment 1.

Again suppose that now x (k) is a wide-band zero intermediate frequency complex signal, its sample rate is B sampling point/second, signal bandwidth B _xSatisfy B _x＜B supposes wherein to contain a very strong arrowband and disturbs J (k), and bandwidth is B _JHertz, and B _J＜＜B _xAbove-mentioned wide-band zero intermediate frequency complex signal x (k) adopts the trap process of parallel organization adaptive notch filter of the present invention may further comprise the steps:

Step 1: adopt sample interweaving time-division way along separate routes, x (k) is divided into the N road; Every road signal is equivalent to carry out N: 1 down-sampling extracts, and sample rate is reduced to the 1/N of former sample rate, referring to Fig. 1.The theorem of crossing according to above-mentioned proof as can be known, as long as MB _J/ N＞B _J, the narrow band signal in each tributary signal just can be because of N: spectral aliasing takes place in 1 down-sampling.

Step 2: every branch road carries out adaptive notch respectively, obtains N road error signal, in fact all is to have suppressed the useful signal that the arrowband disturbs.

Step 3: adopting sample interweaving time division multiplexing technology, is one the tunnel with this N road error signal multiple connection, the useful signal y (k) that the arrowband disturbs that has been eliminated, and its sample rate reverts to former sample rate B sampling point/second.

The present invention adopts sample interweaving time-division way along separate routes, and the wide-band zero intermediate frequency complex signal x (k) that receives is the N road along separate routes; Each branch road carries out adaptive notch respectively, obtains N road error signal, and this N road error signal in fact all is to have suppressed the useful signal that the arrowband disturbs; Adopting the sample interweaving time-division multiplex technology at last, is one the tunnel with this N road error signal multiple connection, obtains y (k), has eliminated the useful signal that the arrowband disturbs exactly.Thereby the present invention not only can realize effective trap for the intermediate frequency analytic signal of plural form, also can realize effective trap for the zero intermediate frequency analytic signal of plural form.

But, have a major issue to further specify.For the situation that exists a plurality of arrowbands to disturb, above-mentioned arrowband sampling thheorem is also inapplicable.For the narrow band signal that has two above frequency bands, even sample rate is higher than wherein wide one or is higher than all narrow band signal bandwidth sums, can not guarantee that all aliasing does not take place frequency spectrum, this convolution relation from the theorem proving of front just can be found out significantly.

Fortunately, the purpose of carrying out trap is not to want these arrowbands of reconstruct to disturb, but will remove them, and also it doesn't matter even therefore there is the spectral aliasing of interference signal, anyway will remove; The down-sampling effect along separate routes of above-mentioned time-division, total frequency band that the arrowband is disturbed broadens, and can become narrower because of spectral aliasing on the contrary, and this just is conducive to keep when trap is eliminated the arrowband interference more wide band useful signal.The arrowband trapper is used as the preliminary treatment before the despreading, and the general interference total bandwidth that requires could effectively improve the jamming margin of system less than 10% of signal total bandwidth; Under such condition, adopt parallel organization of the present invention to realize that many band adaptive notch are feasible.

The total notch bandwidth of the present invention always is less than or equal to the summation of each arrowband jamming bandwidth, thereby the trap frequency band can be narrower, and is less for the useful signal damage, so the trap effect is better.

Embodiment 3

The present invention is based on the parallel organization high-order adaptive notch filter of TDM and based on the parallel organization high-order adaptive resistance-trap wave method of TDM with embodiment 1-2.

It is analytic signal that the present invention requires input signal, and for the two paths of signals that adopts quadrature frequency conversion to obtain, with one the tunnel as real part, another road just constitutes analytic signal as imaginary part.If input signal is not zero intermediate frequency signals, but have the real signal of higher intermediate frequency, as long as so with these signal phase shift 90 degree, obtain its Hilbert transform as imaginary part, original signal just can constitute an analytic signal as real part; But want the frequency range of attention this moment is not the whole frequency range that takes [π ,+π], and is a part wherein.

Embodiment 4

Formation and the signal processing spec of single order adaptive notch filter of the present invention unit

The present invention is based on the parallel organization high-order adaptive notch filter of TDM and based on the parallel organization high-order adaptive resistance-trap wave method of TDM with embodiment 1-3.

With reference to Fig. 2, a kind of parallel organization high-order adaptive notch filter based on TDM proposed by the invention, each single order adaptive notch filter unit by A (z) and A ' (z) two-stage constitute, wherein

The input signal sequence of trapper unit is x (k), with x (k) through the filtered output of prime A (z) as after level A ' input (z), back level A ' output signal e (k) (z) is input in the weights control module as error signal, by the iteration adjustment of weights coefficient h in this module controls front and back stages.

Be limited in the zero point of A (z) on the unit circle and change adaptively; The limit that corresponding radius is introduced at this zero point, namely A ' limit (z) generally is in the unit circle and very close unit circle, because the span of r is 0.8＜r＜1, therefore, each just can suppress an arrowband effectively to zero limit and disturb.Referring to Fig. 2, the parameter ω among the prime A (z) _iWhen constantly revising, the ω of back level A ' in (z) _iAlso change thereupon; Because back level A ' is to the amplitude-frequency characteristic of prime A (z) and the compensation of phase-frequency characteristic (z), whole filter is only at characteristic frequency ω _iNear subside to zero, the amplification quantity of the signal on other frequency nearly all is approximately equal to 1, it is linear that phase shift also is approximately; Therefore has good trap characteristic.

Because be limited in zero point on the unit circle, definition

Then the transfer function of single order adaptive notch filter unit can be expressed as

$H\left(z\right)=A\left(z\right)\·A^{\′}\left(z\right)=\frac{1-e^{\mathrm{j\ω}}\·z^{-1}}{1-r\·e^{\mathrm{j\ω}}\·z^{-1}}=\frac{1-h\·z^{-1}}{1-r\·h\·z^{-1}}$

The input and output signal sequence of single order adaptive notch filter unit is respectively x _n(k) and e _n(k), angular frequency parameter ω wherein _iAdaptive iteration adopt the unification of interlock mode to make amendment, specifically include:

3.1 with transfer function H in the n bar branch road _i(z) angular frequency parameter is made as ω _i(k);

3.2 the adaptive iteration index word that estimates is Δ ω _i(k), Δ ω _i(k)=μ Im[T _i(k)], T wherein _i(k) be ω _i(k) Grad in the adaptive iteration computing, μ are step factor;

3.3 to { ω _i(k), i=1,2 ..., when N} carries out actual modification, adopt its average index word

Simultaneously to the angular frequency parameter ω in the N bar branch road _i(k) make amendment, realize adaptive iteration.

The input signal of single order adaptive notch filter of the present invention unit is x _n(k), output signal is error signal e _n(k), angular frequency parameter ω in its transfer function _n(n=1,2,3 ..., adaptive iteration step N) is:

4.1 initialization: make k=0, e _n(k)=x _n(k)=0; ω _n(k)=π/(n-N/2);

4.2 compute gradient value: make k+1 → k, to input vector X _n=[x _n(k), x _n(k-1) ..., x _n(k-N+1)] ^T

$e_n\left(k\right)=x_n\left(k\right)-e^{j{\mathrm{\ω}}_n\left(k\right)}{e}_n(k-1)$

Output e with single order trapper unit _n(k) as error signal, its mean square error is

J(ω)＝E|e _n(k)| ²＝E[e _n(k)·e _n ^*(k)]

The gradient of mean square error J (ω)

For

$\▿J\left(\mathrm{\ω}\right)=\frac{\∂J\left(\mathrm{\ω}\right)}{\∂{\mathrm{\ω}}_n\left(k\right)}$

$=E\left\{j{e}^{j{\mathrm{\ω}}_n\left(k\right)}{x_n}^{*}\left(k\right)\right[r{e}_n(k-1)-x_n(k-1)]-j{e}^{-j{\mathrm{\ω}}_n\left(k\right)}{x}_n\left(k\right)[r{e_n}^{*}(k-1)-{x_n}^{*}(k-1)\left]\right\}$

Order $T_n\left(k\right)=e^{j{\mathrm{\ω}}_n\left(k\right)}{x}_n^{*}\left(k\right)[r{e}_n(k-1)-x_n(k-1)]$

Average: $\stackrel{\‾}{T}\left(k\right)=\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{T}_n\left(k\right)/N;$

4.3 ask angular frequency parameter ω _nMean value:

$\mathrm{\Δ}{\stackrel{\‾}{\mathrm{\ω}}}_n\left(k\right)=\mathrm{\μIm}\left[\stackrel{\‾}{T}\left(k\right)\right]$ (getting step size mu=0.0001～0.1);

4.4 revise the angular frequency parameter:

${\mathrm{\ω}}_n(k+1)={\mathrm{\ω}}_n\left(k\right)+\mathrm{\Δ}{\stackrel{\‾}{\mathrm{\ω}}}_n\left(k\right)$

Return 4.2, handle next sampling point.

Embodiment 5

The present invention is based on the parallel organization high-order adaptive notch filter of TDM and based on the parallel organization high-order adaptive resistance-trap wave method of TDM with embodiment 1-4.

A ultra-wideband communication system, receiver be with the sampling rate of 1GHz, is that the zero intermediate frequency of 500MHz receives signal x (k) and carries out quadrature sampling to bandwidth; Suppose to exist in the signal band a plurality of arrowbands more much better than than signal of communication, system's despreading processing jamming margin head and shoulders above to disturb, thereby system can't normally receive.Before the receiver despreading, insert the parallel organization 8 rank adaptive notch filters based on TDM of the present invention; Then earlier using sample interweaving TDM technology to received signal, is 8 road parallel signals with it along separate routes, each road parallel signal x _n(k) sampling point speed is reduced to 125MHz, is suitable for FPGA and realizes; The 8 rank trappers that every road forms with the cascade of 8 single order trappers carry out adaptive notch to be handled; Wherein 8 single order trappers of the 1st branch road put in order be 8,7,6 ..., 2,1; Article 2,8 single order trappers of branch road put in order be 1,8,7,6 ..., 2; The rest may be inferred for remaining; After each branch road is finished adaptive notch, adopting the multiple connection of corresponding TDM technology again is one tunnel output signal, the sampling point rate recovery is 1GHz, and the noise cancellation signal that has that this keeps useful signal substantially with regard to obtaining having suppressed the interference of strong arrowband can further carry out despreading demodulation processing.

The parameter r of each single order adaptive notch filter unit is unified among Fig. 1 is taken as 0.98; Parallel adaptive trapper unit can be separate among Fig. 1, and the input signal of n adaptive notch filter unit is x _n(k), output signal is error signal e _n(k), the specific implementation process prescription of weights iterative algorithm is as follows:

(1) initialization: make k=0, e _n(k)=x _n(k)=0; ω _n(k)=π/(n-N/2);

(2) compute gradient value: make k+1 → k, to input vector X _n=[x _n(k), x _n(k-1) ..., x _n(k-N+1)] ^T

$e_n\left(k\right)=x_n\left(k\right)-e^{j{\mathrm{\ω}}_n\left(k\right)}{e}_n(k-1)$

$T_n\left(k\right)=e^{j{\mathrm{\ω}}_n\left(k\right)}{x}_n^{*}\left(k\right)[r{e}_n(k-1)-x_n(k-1)];$

(3) ask angular frequency parameter ω _n(k) mean value: Δ ω _n(k)=μ Im[T _n(k)] (get step size mu=0.0001～0.1)

(4) revise the angular frequency parameter:

ω _n(k+1)＝ω _n(k)+Δω _n(k)

Returned for second step, handle next sampling point.

Embodiment 6

The present invention is based on the parallel organization high-order adaptive notch filter of TDM and based on the parallel organization high-order adaptive resistance-trap wave method of TDM with embodiment 1-4.

A direct sequence spread spectrum communication system, receiver be with the sampling rate of 200MHz, is that the zero intermediate frequency of 100MHz receives signal x (k) and carries out quadrature sampling to bandwidth; Suppose to exist in the signal band a plurality of arrowbands more much better than than signal of communication, system's despreading processing jamming margin head and shoulders above to disturb, thereby system can't normally receive.Before the receiver despreading, insert the parallel organization 4 rank adaptive notch filters based on TDM of the present invention; Then earlier using sample interweaving TDM technology to received signal, is 4 road parallel signals with it along separate routes, each road parallel signal x _n(k) sampling point speed is reduced to 50MHz, is suitable for FPGA and realizes; The 4 rank trappers that every road adopts the cascade of 4 single order trappers to form carry out adaptive notch to be handled; Wherein putting in order of 4 single order trappers of the 1st branch road is H ₄(z), H ₃(z), H ₂(z), H ₁(z); Article 2, putting in order of the trapper of branch road is H ₁(z), H ₄(z), H ₃(z), H ₂(z); Article 3, putting in order of the trapper of branch road is H ₂(z), H ₁(z), H ₄(z), H ₃(z); Article 4, putting in order of the trapper of branch road is H ₃(z), H ₂(z), H ₁(z), H ₄(z).After each branch road is finished adaptive notch, adopting the multiple connection of corresponding TDM technology again is one tunnel output signal, the sampling point rate recovery is 200MHz, and the noise cancellation signal that has that this keeps useful signal substantially with regard to obtaining having suppressed the interference of strong arrowband can further carry out despreading demodulation processing.

The parameter of each parallel adaptive trapper unit, the unified r=0.95 that gets; The input signal of adaptive notch filter unit is x _n(k), output signal is error signal e _n(k), to ω _n(n=1,2, it is 3...N) unified that to carry out the process prescription that adaptive iteration revises as follows:

(1) initialization: make k=0, e _n(k)=x _n(k)=0; ω _n(k)=π/(n-N/2); N=1,2 ..., N;

(2) compute gradient value: make k+1 → k, to input vector X _n=[x _n(k), x _n(k-1) ..., x _n(k-N+1)] ^T

$e_n\left(k\right)=x_n\left(k\right)-e^{j{\mathrm{\ω}}_n\left(k\right)}{e}_n(k-1)$

$T_n\left(k\right)=e^{j{\mathrm{\ω}}_n\left(k\right)}{x}_n^{*}\left(k\right)[r{e}_n(k-1)-x_n(k-1)],n=\mathrm{1,2},...,N$

Average: $\stackrel{\‾}{T}\left(k\right)=\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{T}_n\left(k\right)/N;$

(3) ask the mean value of angular frequency parameter ω (k):

(getting step size mu=0.0001～0.1)

(4) revise the angular frequency parameter:

${\mathrm{\ω}}_n(k+1)={\mathrm{\ω}}_n\left(k\right)+\mathrm{\Δ}{\stackrel{\‾}{\mathrm{\ω}}}_n\left(k\right),n=\mathrm{1,2},...,N$

Returned for second step, handle next sampling point.

Implementation complexity of the present invention is low, trap effect and good stability, can reduce or avoid the phenomenon of zero sunken unreasonable merging, be to make zero to fall into approximate parallel organization high-order adaptive notch filter and the adaptive resistance-trap wave method based on TDM identical with actual arrowband strong jamming number of number.Be applied to antagonism arrowband interference in the digital communication system, especially for the preliminary treatment of receiver in the direct sequence spread spectrum communication system before despreading.

Claims (4)

1. parallel organization adaptive notch filter based on TDM, comprise string and conversion time-division splitter, and string conversion time division multiplexing device and be connected on splitter and multiplexer between N rank adaptive notch filter; The reception signal x (k) of a communication control processor is input to N rank adaptive notch filter concurrently simultaneously, and after the arrowband strong jamming in the adaptive notch filter filtered signal, remerging is one road signal y (k) output; Adaptive notch filter adopts N rank parallel organization, 2≤N≤9, and this parallel organization comprises N bar parallel branch, every route N unit cascaded formation of single order adaptive notch filter; Putting in order of trapper is the cyclic shift that trapper puts in order in its adjacent legs in every branch road; The transfer function of i single order adaptive notch filter unit is in every parallel branch

0≤i≤N-1, wherein angular frequency parameter ω _iBy with output signal energy minimum being the adaptive iteration acquisition of criterion ,-π＜ω _i＜π; Parameter r selects 0.8＜r＜1 according to the bandwidth that hope suppresses; It is characterized in that: described string and conversion time-division splitter carry out splitter along separate routes for adopting sample interweaving TDM technology, the described and multiplexer of string conversion time division multiplexing device for adopting the TDM technology to carry out multiple connection; Described reception signal x (k) is an intermediate frequency analytic signal, comprises the zero intermediate frequency analytic signal, and this signal is the digital sampling point sequence of plural form.

2. parallel organization adaptive resistance-trap wave method based on TDM, it is characterized in that: the adaptive resistance-trap wave process comprises the steps:

Step 1: adopt sample interweaving time-division method along separate routes, the high-speed sampling sampling point sequence x (k) that communication control processor is received is that the sampling point sequence branch road of N bar low speed, x (k) are an intermediate frequency analytic signal along separate routes;

Step 2: various kinds point sequence branch input signal x ₁(k), x ₂(k) ..., x _N(k) carry out cascade connection type high-order adaptive notch respectively, obtain N road error signal e ₁(k), e ₂(k) ..., e _N(k), this N road error signal in fact all is to have suppressed the useful signal that the arrowband disturbs; Each adaptive notch filter unit is only in the specific angle frequencies omega _iNeighbouring that the arrowband interference trap is extremely zero, angular frequency _iAdaptive iteration, be to adopt the unification of interlock mode to make amendment;

Step 3: adopting sample interweaving time division multiplexing technology, is one tunnel output with the error signal multiple connection of N road, and the useful signal y (k) that the arrowband disturbs has been eliminated.

3. the parallel organization adaptive notch filter method based on TDM according to claim 2 is characterized in that: adopt the interlock mode unification realization adaptive iteration of making amendment in the described step 2, specifically include:

3.1 with transfer function H in the n branch road _i(z) angular frequency parameter is made as ω _i(k);

3.2 the adaptive iteration index word that estimates is Δ ω _i(k), Δ ω _i(k)=μ Im[T _i(k)], T wherein _i(k) be ω _i(k) Grad in the adaptive iteration computing, μ are step factor;

3.3 to { ω _i(k), i=1,2 ..., when N} carries out actual modification, adopt its average index word

Simultaneously to the angular frequency parameter ω in the N bar branch road _i(k) make amendment, realize adaptive iteration.

4. the parallel organization high-order adaptive resistance-trap wave method based on TDM according to claim 3, it is characterized in that: the input signal of single order adaptive notch filter unit is x _n(k), output signal is error signal e _n(k), angular frequency parameter ω in its transfer function _n(n=1,2,3 ..., adaptive iteration step N) is:

4.1 initialization: make k=0, e _n(k)=x _n(k)=0; ω _n(k)=π/(n-N/2);

4.2 compute gradient value: make k+1 → k, to input vector X _n=[x _n(k), x _n(k-1) ..., x _n(k-N+1)] ^T

Average:

4.3 ask angular frequency parameter ω _nMean value:

(getting step size mu=0.0001～0.1);

4.4 revise the angular frequency parameter:

Return 4.2, handle next sampling point.

Images