CN104901708A - Compressive sampling broadband digital receiver and signal processing method thereof - Google Patents

Abstract

The present invention discloses a compressive sampling broadband digital receiver and a signal processing method thereof. The broadband digital receiver includes an analog-to-digital converter, a mixer, a lowpass pre-filter, a decimator, a uniform channelization filter and a channel selection module. The analog-to-digital converter collects a signal x(n) according to a sampling frequency f<NYQ>. The mixer mixes the received signal with a pseudorandom sequence p(n) to obtain a mixed sequence x(n). The lowpass pre-filter filters the mixed sequence to obtain a filtered sequence x<D>(n). The decimator performs decimation by an integer factor R<D> on the filtered sequence to obtain a sequence x<p>(n). The uniform channelization filter processes the received sequence to obtain M output signals y<i><d>(n). The channel selection module selects, from the received signals, first R signals to obtain the final R output signals. The receiver of the present invention has a simple structure, can lower the system complexity, and can achieve the Nyquist sampling.

Description

A kind of wideband digital receiver of compression sampling and signal processing method thereof

Technical field

The invention belongs to data acquisition process field, particularly relate to a kind of wideband digital receiver of compression sampling and signal processing method thereof of the sub-nyquist sampling for communication or radar signal.

Background technology

Wideband digital receiver has important application in frequency spectrum perception, spacing wave collection and identification field.Along with the development of Analog-digital Converter (ADC) acquisition technique, digital receiver is more and more close to antenna, it is to the all-digital acquisition of signal and ex-post analysis and process, improve receiver system sensitivity, the aspects such as the restructural of system, relative to traditional analog receiver, there is advantage significantly.What current wideband digital receiver adopted mostly is uniform channel structure, as WOLA structure, structure based on FFT, based on the structure of CEM.Due to based on general principle be utilize equiband digital filter bank to construct uniform sub-channel structure, therefore its sample rate wanting system total is still Nyquist (Nyquist) speed, and considers the setting of bandpass sampling and the setting of mixing during design.

Summary of the invention

The object of this invention is to provide the wideband digital receiver that a kind of structure is simply applied to the compression sampling of sub-nyquist sampling, object of the present invention also comprises provides a kind of system complexity that can reduce, the signal processing method of the wideband digital receiver of compression sampling.

A wideband digital receiver for compression sampling, comprises analog-digital converter, frequency mixer, lowpass pre-filter, withdrawal device, uniform channel filter and channel selection block,

Analog-digital converter is by sample frequency f _nYQcollection signal x (n) sends frequency mixer to;

Frequency mixer is by signal x (n) of reception and pseudo random sequence mixing, obtains sequence after mixing send lowpass pre-filter to;

Lowpass pre-filter is to sequence after mixing carry out filtering, obtain sequence after filtering send withdrawal device to;

Withdrawal device is to after filtered sequence carry out R _dintegral multiple doubly obtains sequence after extracting send uniform channel filter to;

The sequence received processes by uniform channel filter, obtains M road output signal send channel selection block to;

Channel selection block is from the signal received in select front R road signal, obtain final R road output signal y _j(n), j=0,1,2...R-1.

A signal processing method for the wideband digital receiver of compression sampling, comprises the following steps:

Step one: analog-digital converter is by sample frequency f _nYQcollection signal x (n) sends frequency mixer to;

Step 2: frequency mixer is by signal x (n) and pseudo random sequence frequency mixing processing, r ∈ Ζ, Ζ is integer, M _p=T _pf _nYQ, T _pbe the time cycle of pseudo random sequence, make frequency b is input signal bandwidth;

Step 3: lowpass pre-filter is to sequence after mixing carry out filtering, obtain sequence after filtering after filtering, the frequency domain value of sequence is:

${\tilde{x}}_D\left(e^{\mathrm{j\&omega;}}\right)=\frac{1}{M_p}\underset{k=0}{\overset{M_p-1}{\mathrm{\&Sigma;}}}P\left(k\right)X(\mathrm{\&omega;}-k\frac{2\mathrm{\&pi;}}{M_p}),-\frac{\mathrm{\&pi;}}{R_D}\&le;\mathrm{\&omega;}\&le;\frac{\mathrm{\&pi;}}{R_D}$

Wherein R _dget and be not more than f _nYQ/ (2 (Rf _p+ f _t)) maximum integer;

Step 4: filtered sequence after withdrawal device, every R _dindividual sampled point is sampled once, obtains sequence

Step 5: the sequence received processes by uniform channel filter, obtains M road output signal

$y_i^d\left(n\right),i=\mathrm{0,1,2}...M-1;$

Step 6: channel selection block is from the signal received in select front R road signal, obtain final R road output signal y _jn (), j=0,1,2...R-1, the sample frequency on each road is f _s=f _nYQ/ (R _dk), K is the extraction yield of M road uniform channel filter, f _s=f _p, and

Beneficial effect:

(1) MWC the Theory Construction wideband digital receiver is adopted.MWC theory is mainly used in the application of analog information conversion (AIC).The present invention applies it in the realization of wideband digital receiver, the sampling realizing sub-Nyquist can be ensured on the one hand, ensureing, under the prerequisite that input signal information can be always completely recovered, to adopt less data to store, thus save the storage resources of receiver.On the other hand, because it adopts pseudo-random signal to carry out mixed modulated, under the prerequisite ensureing input signal nyquist sampling, just can use structure of the present invention, and the setting of bandpass sampling when need not consider that Conventional wide band digital received designs and the offering question of mixing.

(2) lowpass pre-filter and withdrawal device.The present invention is provided with lowpass pre-filter and withdrawal device in the middle of frequency mixer and channelized receiver, and object is the way reducing the setting of uniform channel receiver, to reduce the complexity of system.Signal after the effect of withdrawal device is through low-pass filtering can realize first time reduction of speed sampling.

(3) the MWC structure of single channel mixing combines with uniform channel receiver structure.The MWC structure that the present invention adopts the mixing of single channel to design, considers the frequency spectrum characteristic of random signal modulation, i.e. its mixed signal dTFT conversion be input signal x (n) DTFT conversion x (e ^{j ω}) with f _pfor the linear combination of cycle shift.As shown in Figure 4, its frequency spectrum is with f _pfor interval is equally distributed.Therefore uniform channel receiver structure can be adopted further.Make like this to extract the front end that the factor is put into channelized receiver, reduce the operating frequency of system, thus reduce the complexity of system further.

The present invention is under the recoverable condition of guarantee signal, can realize the wideband digital receiver design of sub-Nyquist compression sampling.The signal of one or more known or the unknown that what wideband digital receiver generally received within a period of time is.The mode of this Received signal strength is consistent with multi-band signal model, i.e. frequency-domain sparse signal.Therefore can according to compressed sensing principle, for spectrum sparse signal, Analog-digital Converter is expanded to analog information conversion (AIC), thus remove redundant information, can compression sampling data further.(MWC) model is changed in modulation broadband based on compression sampling, can well process multi-band signal, a kind of successfully AIC compression sampling structure.MWC theoretical demand signal and the mixing of multichannel pseudorandom (PN) sequence.Choosing of pseudo random sequence has larger impact to the reconstruct of signal.For reducing the complexity of PN sequence selection, the present invention adopts single channel mixing model, and the signal relying on pseudo-random signal modulation to produce in the cycle shift of frequency domain, thus can build multiple sub-band filter filter, realizes the compression sampling of signal.This single channel mixing model is applicable to building the uniform channelized receiver in M road and realizes very much, thus from M road, select R road as the observing matrix of compression sampling again, and CTF algorithm finally can be adopted from R circuit-switched data to recover primary signal.

Accompanying drawing explanation

Fig. 1 is the theory diagram of wideband digital receiver of the present invention;

Fig. 2 is based on heterogeneous structure uniform channel Filter Principle block diagram;

Fig. 3 is the spectrogram signal of multi-band signal (N=2) and pseudo random sequence;

Fig. 4 is that after pseudo random sequence and input signal mixing, spectrogram is illustrated;

Fig. 5 is linear frequency modulation (LFM) signal 6 road compression sampling data;

Fig. 6 is the reconstruction signal of LFM signal and the primary signal comparison diagram at frequency domain;

Fig. 7 is the reconstruction signal of LFM signal and the primary signal comparison diagram in time domain;

The MSE performance comparison of Fig. 8 wideband digital receiver of the present invention and standard MWC algorithm.

Embodiment

Below in conjunction with accompanying drawing, the present invention is described in further details.

The present invention is applied to the digital received systems such as communication, radar, and object is the wideband digital receiver structure providing a kind of effective sub-nyquist sampling, and can ensure accurately to recover primary signal.

The object of the present invention is achieved like this: through analog-digital converter with sample frequency f _nYQfirst signal x (n) gathered is T with the cycle _ppseudo random sequence mixing, obtains sequence after mixing then be [0, f by cut-off frequency _nYQ/ 2R _d] h _dthe filtering of (n) lowpass pre-filter, after through R _dintegral multiple doubly obtains sequence after extracting sequence enters M road uniform channel receiver; Before finally selecting from M road exports, R road is as the observation data of compression sampling.

The condition that the present invention is suitable for is:

(1) input signal can be the radar signal (as LFM signal) in multi-band signal or broadband, supposes that it has N number of maximum bandwidth to be no more than the signal of B, then requires last selection way R >=4N.

(2) x (n) sample sequence will meet nyquist sampling theorem, if its sample frequency is f _nYQ.

(3) be one-period be M _pcycle pseudo random sequence, namely r ∈ Ζ, wherein Ζ is integer.M _p=T _pf _nYQ, T _pthe time cycle of pseudo random sequence, main value sequence be that p (n), p (n) can select two-value ± 1Bernoulli random sequence.Requirement

(4) the R road sampled data y after compression sampling ₁(n), y ₂(n) ..., y _r(n) ..., y _rn (), the sample frequency on each road is f _s=f _nYQ/ (R _dk), K is the extraction yield of M road uniform channel receiver.Require f _s>=f _p, in the present invention, get f _s=f _p.

We obtain thus, and the total sample rate of system is get sub-nyquist sampling system can be obtained.

The present invention is the MWC principle according to compression sampling, carries out the uniform channel filtering of M road then, therefrom choose R road as compression sampling data to through the sequence corresponding with the linear combination frequency spectrum of the cycle shift that pseudo random sequence mixing obtains.Simultaneously for reducing the way of channelizing, reducing the complexity of system, between frequency mixer and channelizing filtering, adding one-level prepositive low frequency filter and extraction operation again.Below in conjunction with accompanying drawing and example, the present invention is described in detail.

Composition graphs 1, gives system principle diagram of the present invention, and system is by digital mixing module, and low pass filter blocks, decimator module, channel filter module, channel selection block forms, and wherein channel filter module specific implementation as shown in Figure 2.E in Fig. 2 _lz () is l multinomial component of channelizing design mesarcs filter, IDFT represents inverse discrete Fourier transform.The handling process of our analytic signal successively below.

The discrete time Fourier transform (DTFT) of input signal x (n) is

$x\left(n\right)-\frac{1}{2\mathrm{\&pi;}}{\&Integral;}_{-\mathrm{\&pi;}}^{\mathrm{\&pi;}}X\left(e^{\mathrm{j\&omega;}}\right)e^{\mathrm{jn\&omega;}}\mathrm{d\&omega;}---\left(1\right)$

Here, ω=2 π fT, f _nYQ=1/T is Nyquist sample frequency.Pseudorandom is for getting { the Bernoulli sequence of-1 ,+1} two-value discrete Fourier series be

$\tilde{p}\left(n\right)=\frac{1}{M_p}\underset{k=0}{\overset{M_p-1}{\mathrm{\&Sigma;}}}P\left(k\right)e^{j\frac{2\mathrm{\&pi;}}{N}\mathrm{nk}}---\left(2\right)$

Here P (k) is pseudo random sequence the Fourier transform of main value sequence p (n).Therefore the output signal after frequency mixer $\tilde{x}\left(n\right)=x\left(n\right)\tilde{p}\left(n\right)$ DTFT be transformed to

$\tilde{x}\left(e^{\mathrm{j\&omega;}}\right)=\underset{n=-\&infin;}{\overset{\&infin;}{\mathrm{\&Sigma;}}}x\left(n\right)\&CenterDot;\tilde{p}\left(n\right)e^{-\mathrm{jn\&omega;}}=\frac{1}{M_p}\underset{k=0}{\overset{M_p-1}{\mathrm{\&Sigma;}}}P\left(k\right)X(\mathrm{\&omega;}-k\frac{2\mathrm{\&pi;}}{M_p})---\left(3\right)$

Here the present invention gives the signal of the DTFT conversion of input multi-band signal (N=2) and the Fourier series of pseudo random sequence for N=2, Fig. 3; Spectrogram after Fig. 4 gives pseudo random sequence and inputs multi-band signal mixing is illustrated.

By dTFT transformation results formula (3) and Fig. 4 known, its DTFT convert the DTFT conversion x (e of x (n) ^{j ω}) with frequency f _pthe linear combination of cycle shift, its frequency spectrum is with f _pfor interval is equally distributed.Therefore optional R subband wherein forms the observation data of compression sampling.In order to realize the filtering to R subband, the present invention selects the even channel filter receiver based on heterogeneous structure shown in Fig. 2.If but directly adopting Fig. 2 structure, system can be along with the increase of number of sub-bands, the number of channel also can increase thereupon, and therefore system can become complicated, higher requirement is had to the resource occupation of hardware, therefore invention introduces lowpass pre-filter to arrange, the R subband required for filtering, and then carry out channelizing filtering.The pass band width of lowpass pre-filter is Rf _p, the bandwidth of each subband of channelizing can be f _p.Simultaneously in order to reduce the difficulty of lowpass pre-filter design, reduce design exponent number, the transition band of low pass filter is optionally wider (is set to f _t), then the total bandwidth of channelizing is designed to Rf _p+ f _t.Therefore be equivalent to through R H _r(f) subfilter, H _rf () is defined as:

$H_r\left(f\right)=\left\{\begin{array}{cc}1& (2r-3)\frac{f_s}{2}\&le;f\&le;(2r-1)\frac{f_s}{2},1\&le;r\&le;R\\ 0& \mathrm{otherwise}\end{array}\right.---\left(4\right)$

In sum, then after low-pass filtering module, signal is then its frequency domain exports and is:

${\tilde{x}}_D\left(e^{\mathrm{j\&omega;}}\right)=\frac{1}{M_p}\underset{k=0}{\overset{M_p-1}{\mathrm{\&Sigma;}}}P\left(k\right)X(\mathrm{\&omega;}-k\frac{2\mathrm{\&pi;}}{M_p}),-\frac{\mathrm{\&pi;}}{R_D}\&le;\mathrm{\&omega;}\&le;\frac{\mathrm{\&pi;}}{R_D}---\left(5\right)$

Here R _dget and be not more than f _nYQ/ (2 (Rf _p+ f _t)) maximum integer, signal through R _dafter times decimator module, i.e. time-domain signal every R _dindividual sampled point is sampled and once obtains signal then frequency spectrum is extended to-π≤ω≤π.Again through channel filter module, the m sub-channels of system exports y ^d _mn the DTFT of () signal converts Y ^d _m(ω) be:

${Y^d}_m\left(\mathrm{\&omega;}\right)=\underset{l=-L_0}{\overset{L_0}{\mathrm{\&Sigma;}}}{P}^{\&prime;}\left(l\right)X(\mathrm{\&omega;}-(l-m)\frac{2\mathrm{\&pi;}}{M_p}),\mathrm{\&omega;}\&Element;[-\frac{2\mathrm{\&pi;}}{M_p},\frac{2\mathrm{\&pi;}}{M_p}]---\left(6\right)$

Module is selected finally by R path channels, namely from channel filter arrive m road export in choose the output of front R road as compression sampling, consider the selected branch road of all R, then being write as matrix form is

Y(ω)＝Φ z(ω) (7)

Wherein $\underset{\&OverBar;}{z}\left(\mathrm{\&omega;}\right)={\left[\begin{array}{cccccc}X(\mathrm{\&omega;}+L_0\frac{2\mathrm{\&pi;}}{M_p})& X(\mathrm{\&omega;}+(L_0-1)\frac{2\mathrm{\&pi;}}{M_p})& ...& 0& ...& X(\mathrm{\&omega;}-L_0\frac{2\mathrm{\&pi;}}{M_p})\end{array}\right]}^T$ For length is 2L ₀the column vector of+1, it is sparse solution to be asked. $\mathrm{\&Phi;}={\left[\begin{array}{cccc}{\underset{\&OverBar;}{\mathrm{\&phi;}}}_1^T& {\underset{\&OverBar;}{\mathrm{\&phi;}}}_2^T& ...& {\underset{\&OverBar;}{\mathrm{\&phi;}}}_R^T\end{array}\right]}^T$ Be size be R × (2L ₀+ 1) compressed sensing matrix, wherein ${\underset{\&OverBar;}{\mathrm{\&phi;}}}_r^T=\left[\begin{array}{cccc}{\underset{\&OverBar;}{0}}_{r-1}^T& P^{\&prime;}(-L_0)& ...& P^{\&prime;}\end{array},(L_0-r+1)\right],$ L ₀elect the minimum integer that can comprise all nonzero values of X (ω) frequency spectrum as. it is a column vector having m-1 neutral element.

$P^{\&prime;}\left(l\right)=\left\{\begin{array}{cc}\frac{1}{M_p}P\left(l\right)& (0\&le;l\&le;L_0)\\ \frac{1}{M_p}P(M+l)& (-L_0\&le;l<0)\end{array}\right.---\left(8\right)$

Matrix y(ω)=Φ z(ω) be solved to the Solve problems of sparse theory, consider that processing signals is multi-band model, then Solve problems is IMV problem simultaneously.Therefore solve and can adopt CTF algorithm, IMV is converted to MMV problem.

We obtain thus, and the total sample rate of system is get sub-nyquist sampling system can be obtained.

Provide typical simulation result below, to verify feasibility of the present invention.The employing frequency of input signal is f _nYQ=960MHz, sampling number is 4096 points; f _s=f _p=30MHz; R=6; Band-pass filter design cut-off frequency is 240MHz.It is 6 road compression sampling data after the compression sampling of linear frequency modulation (LFM) signal that Fig. 5 gives input signal.Fig. 6 and Fig. 7 sets forth the frequency-domain and time-domain comparison diagram of reconstruction signal and primary signal after the compressed sampling of LFM signal.The bandwidth of LFM signal is 10MHz, and signal to noise ratio snr is 20dB.Sample rate after compression sampling is Rf _s=180MHz, is less than Nyquist sampling frequency=960MHz.Fig. 7 gives the MSE performance of restoring signal under the different signal to noise ratio of 0dB-30dB, and contrasts with standard MWC algorithm.Its MSE computing formula is

$\mathrm{MSE}=\frac{{\left|\right|\hat{x}\left(n\right)-x\left(n\right)\left|\right|}^2}{{\left|\right|x\left(n\right)\left|\right|}^2}---\left(9\right)$

Wherein, reconstruction signal and primary signal is respectively with x (n).Because the present invention only adopts a road mixing, therefore its MSE performance is better than standard MWC algorithm.As can be seen from emulation, as shown in Figure 8, structure of the present invention can realize the reconstruct to signals such as the LFM of broadband reception under the condition of sub-nyquist sampling.

Claims (2)

1. a wideband digital receiver for compression sampling, is characterized in that: comprise analog-digital converter, frequency mixer, lowpass pre-filter, withdrawal device, uniform channel filter and channel selection block,

Analog-digital converter is by sample frequency f _nYQcollection signal x (n) sends frequency mixer to;

Frequency mixer is by signal x (n) of reception and pseudo random sequence mixing, obtains sequence after mixing send lowpass pre-filter to;

Lowpass pre-filter is to sequence after mixing carry out filtering, obtain sequence after filtering send withdrawal device to;

Withdrawal device is to after filtered sequence carry out R _dintegral multiple doubly obtains sequence after extracting send uniform channel filter to;

The sequence received processes by uniform channel filter, obtains M road output signal i=0,1,2...M-1, send channel selection block to;

Channel selection block is from the signal received in select front R road signal, obtain final R road output signal y _j(n), j=0,1,2...R-1.

2., based on the signal processing method of the wideband digital receiver of a kind of compression sampling according to claim 1, it is characterized in that, comprise the following steps:

Step one: analog-digital converter is by sample frequency f _nYQcollection signal x (n) sends frequency mixer to;

Step 2: frequency mixer is by signal x (n) and pseudo random sequence frequency mixing processing, r ∈ Ζ, Ζ is integer, M _p=T _pf _nYQ, T _pbe the time cycle of pseudo random sequence, make frequency b is input signal bandwidth;

Step 3: lowpass pre-filter is to sequence after mixing carry out filtering, obtain sequence after filtering after filtering, the frequency domain value of sequence is:

${\tilde{x}}_D\left(e^{\mathrm{j\&omega;}}\right)=\frac{1}{M_p}\underset{k=0}{\overset{M_p-1}{\mathrm{\&Sigma;}}}P\left(k\right)X(\mathrm{\&omega;}-k\frac{2\mathrm{\&pi;}}{M_p}),-\frac{\mathrm{\&pi;}}{R_D}\&le;\mathrm{\&omega;}\&le;\frac{\mathrm{\&pi;}}{R_D}$

Wherein R _dget and be not more than f _nYQ/ (2 (Rf _p+ f _t)) maximum integer;

Step 4: filtered sequence after withdrawal device, every R _dindividual sampled point is sampled once, obtains sequence

Step 5: the sequence received processes by uniform channel filter, obtains M road output signal

i＝0,1,2...M-1；

Step 6: channel selection block is from the signal received in select front R road signal, obtain final R road output signal y _jn (), j=0,1,2...R-1, the sample frequency on each road is f _s=f _nYQ/ (R _dk), K is the extraction yield of M road uniform channel filter, f _s=f _p, and R meets