CN109031239A

CN109031239A - Compressed sensing external illuminators-based radar based on information fusion is to object detection method

Info

Publication number: CN109031239A
Application number: CN201810992096.2A
Authority: CN
Inventors: 马俊虎; 安建成; 王爽; 廖红舒; 甘露
Original assignee: University of Electronic Science and Technology of China
Current assignee: University of Electronic Science and Technology of China
Priority date: 2018-08-29
Filing date: 2018-08-29
Publication date: 2018-12-18
Anticipated expiration: 2038-08-29
Also published as: CN109031239B

Abstract

The invention belongs to signal processing technology field, it is related to a kind of compressed sensing based external illuminators-based radar to object detection method.Projection of the present invention first according to sparse signal in dictionary base obtains sparse vector, new calculation matrix is devised then in conjunction with gaussian random matrix and dictionary basic matrix, in the case of different for the time delay of echo under different paths, the different distribution character of compression sampling value is utilized and designs the weight coefficient after one group of optimization, and then proposes compressed sensing based information fusion detection method.Finally, completing the detection to target.This method is directly handled the sampled value of compression, is not necessarily to signal reconstruction, is reduced data operation quantity, can work normally under low signal-to-noise ratio.There is reference and practical application well in detection direction to be detected.

Description

Compressed sensing external illuminators-based radar based on information fusion is to object detection method

Technical field

The invention belongs to signal processing technology field, it is related to a kind of compressed sensing based external illuminators-based radar and target is examined Survey method.

Background technique

In recent years, the application that compressive sensing theory is reached its maturity in Radar Signal Processing, wireless communication, such as Signal is detected, identification and parameter Estimation etc..It is different from traditional Nyquist sampling thheorem, it introduces after compressed sensing, In the case where considerably reducing the analysis to sampled data output, the feature of the front signal of compression is not lost substantially.Radar at present The compressed sensing technology used in system, it usually needs signal is reconstructed using restructing algorithm, and then realizes target detection With parameter Estimation.But the general more demanding signal-to-noise ratio of restructing algorithm, it can just obtain meeting target detection and parameter Estimation Reconstruction signal.For as external illuminators-based radar, since its essence is bistatic non-cooperation radar, target echo power is low, usually not Meet the signal-to-noise ratio requirement of signal reconstruction.Therefore realize that Detection of Weak Signals needs new method using compressed sensing technology.

Signal detection be by receive signal analyze and determine echo signal whether there is, after compressed sensing technology, The mathematical model of target detection is as follows,

Wherein Φ ∈ R^M×NIt is calculation matrix, n ∈ R^N×1It is white Gaussian noise, x ∈ R^N×1It is signal to be detected, it is assumed that H₀It is Signal to be detected is not present, it is assumed that H₁Signal to be detected there are the case where.Presently, there are detection algorithm mainly have it is following several Kind method: (1) Liu Bing in 2010 et al. proposes mean value comparison algorithm, and core concept is that noise is the white noise that mean value is 0, that H₀In the case of E (y)=E (Φ n)=0, H₁In the case of E (y)=E (Φ (x+n))=Φ x, and variance does not change in the case of two kinds Become.So he use method be using actual sample value and its exist two kinds hypothesis in the case of mathematic expectaion deviation as adjudicate Foundation completes detection；(2) 2015 years Alireza Hariri et al. propose maximum likelihood ratio value-based algorithm, and core concept is that do not have There is the likelihood function for obtaining two kinds of situations to signal reconstruction to be detected, then does ratio, it is special by likelihood ratio function probability density Property obtains thresholding formula, makes decisions.But above-mentioned three kinds of detection algorithms are that -5dB or less cannot realize target in signal-to-noise ratio Detection.Which limits the application ranges of detector, in order to detect bigger range, it is necessary to which it is right under more low signal-to-noise ratio to solve The detection of signal to be detected.Property (3) 2017 years Ma Junhu et al. fixed according to sparse vector non-zero position, proposes and is based on The Weak Signal Detecting Arithmetic of sparse domain accumulation.The algorithm solves the problems, such as low signal-to-noise ratio.But this algorithm needs each group of number According to sparse element position be identical.For distributed external illuminators-based radar scene, corresponding to each receiver Time delay and doppler information are different from, then its is sparse for rarefaction representation under dictionary base for the target echo that generates of different receivers Position is not also identical.The above-mentioned method failure based on element position accumulation.Therefore, on the basis of not increasing data volume, into one Step research when sparse vector nonzero element position difference, the detection to signal be very it is necessary to.

Summary of the invention

The technical problem to be solved by the present invention is to study to work as sparse vector nonzero element position not simultaneously to compressed Signal directly detects signal under conditions of no signal reconstruction.

In order to make it easy to understand, the technology used to the present invention is done as described below:

According to formula (1), following signal detection model is established

Wherein y_n∈R^M×1Indicate the signal n=1,2 of n-th compression sampling ..., N_b, n_nFor white Gaussian noise, distribution Are as follows:Sparse signal representation are as follows: x_n=Ψ α_n, Ψ ∈ R^N×NIt is unit dictionary basic matrix.Sparse signal is in dictionary Base projects to obtain sparse vector α_n=Ψ^Hx_n,Nonzero element number much smaller than sparse Vector element number, for different observation frequency α_nIt is different to beTherefore sparse member cannot be used The method of plain position accumulation.X is observed for single_n, when signal determines, sparse vector determines therewith, and non-zero position is corresponding Numerical value is regarded as constant.Therefore, projection of the sparse vector under Gaussian matrix still Gaussian distributed.White Gaussian noise is dilute Thin domain does not have sparsity, projects Gaussian distributed in dictionary base.And projection of the white Gaussian noise under random gaussian matrix Also Gaussian distributed.The present invention combines the calculation matrix of random gaussian matrix and the design of dictionary basic matrix are as follows: In H₀Under signal detection model, y_nDistribution are as follows:In H₁Under detection model, y_nDistribution Are as follows: It is identical to compare its mean value known to both the above Gaussian Profile, variance is different, therefore the detector is set according to the size of variance Meter.But for multiple observed case, due to being set herein by the thought merged using information per signal-to-noise ratio difference all the way Detector is counted, detection probability is effectively improved, specific method is one group of weight coefficient of design to fusion center, is optimized Data afterwards design detector in turn, and concrete scheme is as shown in Figure 1.

In H₀Under assuming that, this is distributed as standardized normal distribution:Then pass through N_bSecondary sight Survey y_nVariance T_n=D (y_n) mean value composition vector are as follows: m₀=[1,1 ..., 1]^T, further analysis obtains T_nVariance are as follows:And in H₁Under detection model:If y_nVariance T_n's The vector of mean value composition are as follows:WhereinAs shown in Figure 1, defining statisticWherein weight coefficientAndWhereinIt is WithFor the diagonal matrix of diagonal entry composition.

To solve the above problems, the technical solution of the present invention is as follows:

A kind of compressed sensing based external illuminators-based radar is used for compressed signal object detection method, this method Directly signal is detected under conditions of no signal reconstruction, which comprises the following steps:

S1, signal detection model is established:

Wherein y_n∈R^M×1Indicate the signal n=1,2 of n-th compression sampling ..., N_b, n_nFor white Gaussian noise, distribution Are as follows:Φ is Gaussian matrix, and Ψ is unit dictionary basic matrix, N_bFor target echo number；

S2, it determines detection threshold, specifically includes:

S21, random gaussian matrix Φ and dictionary basic matrix Ψ is generated, designs calculation matrixN is Chief Signal Boatswain Degree, σ_nFor the received noise power of n-th of receiver；

S22, calculation matrix Φ_nNoise-containing signal compression is sampled to obtain y_n=Φ_n(x_n+n_n), x_nIt is letter to be detected Number；

S23, y is sought_nVariance, be denoted as e_n；

S24, definitionWherein w_nFor weight coefficient, andWherein,Be withFor the diagonal matrix of diagonal entry composition, m₀=[1,1 ..., 1]^TIt is H in step sl₀Detection model It is lower to pass through N_bSecondary observation y_nVariance mean value composition vector,It is H in step sl₁Under detection model y_nVariance mean value composition vector；

S3, according to detector threshold value γ, whether true T > γ is judged, if so, being judged to signal；Otherwise it is judged to signal It is not present, the value of γ is determined according to noise profile.

Beneficial effects of the present invention are to obtain sparse vector in the projection of dictionary base according to sparse signal first, then tie It closes gaussian random matrix and dictionary basic matrix devises new calculation matrix, it is different for the time delay of echo under different paths The case where, the different distribution character of compression sampling value is utilized and designs the weight coefficient after one group of optimization, and then proposes based on pressure Contract the information fusion detection method perceived.Finally, completing the detection to target.This method is directly handled the sampled value of compression, Without signal reconstruction, reduce data operation quantity, can be worked normally under low signal-to-noise ratio.Have in detection direction to be detected fine Reference and practical application.

Detailed description of the invention

Fig. 1 the method for the present invention schematic diagram；

Fig. 2 the method for the present invention is to signal detection flow chart to be detected；

Fig. 3 is M/N=0.5 when compression ratio, the detection performance situation of change of detector when different false-alarm probabilities.

Specific embodiment

Below in conjunction with drawings and examples, technical solution of the present invention is further described.

Embodiment

The detection implementation method of embodiment is as shown in attached drawing 1, attached drawing 2.Shown in specific step is as follows:

Known signal to be detected is single frequency sinusoidal signal x_n=A_n*exp(1j*2πf_n(0:N-1)), wherein For n-th of receiver received signal amplitude, SNR_nFor signal-to-noise ratio, f_n=[0.12 0.25 0.34 0.43 0.52 0.60], n =1 ..., N_b, N_b=6, N=1000.Enable sparse dictionary base Ψ ∈ R^N×NIt is unit Fourier transform matrix, Φ_n∈R^M×NFor with Machine Gaussian matrix, wherein M=500.n_n∈R^N×1For white Gaussian noise, the corresponding signal-to-noise ratio of each receiver is respectively SNR_n= [-5,-7,-9,-8.5,-6,-9.5]。

Step 1: generating random gaussian matrix Φ and dictionary basic matrix Ψ, designs calculation matrixN is signal Length, σ_nFor the received noise power of n-th of receiver；

Step 2: calculation matrix Φ_nNoise-containing signal compression is sampled to obtain y_n=Φ_n(x_n+n_n), x_nIt is to be detected Signal；

Step 3: y is sought_nVariance, be denoted as e_n；

Step 4: definitionWherein w_nFor weight coefficient, andWherein m₀=[1,1 ..., 1]^T,

Step 5: if T > γ, signal has been judged to it；Otherwise signal is judged to be not present, wherein γ is detector threshold value, It is determined according to noise profile.

Fig. 3 is the emulation to embodiment, is illustrated from Fig. 3 in compression ratio M/N=0.5, the inspection under different false-alarm probabilities The detection performance situation for surveying device, as can be seen from Figure 3 when receiver signal-to-noise ratio is all lower, this method still can be with higher general Rate detects target, and is not necessarily to signal reconstruction, therefore, can obtain of the invention with good detection performance.

Claims

1. compressed sensing based external illuminators-based radar is to object detection method, this method is not for having compressed signal Directly signal is detected under conditions of signal reconstruction, which comprises the following steps:

S1, signal detection model is established:

S2, it determines statistic, specifically includes:

S21, random gaussian matrix Φ and dictionary basic matrix Ψ is generated, designs calculation matrixN is signal length, σ_n For the received noise power of n-th of receiver；

S22, calculation matrix Φ_nNoise-containing signal compression is sampled to obtain y_n=Φ_n(x_n+n_n), x_nIt is signal to be detected；

S23, y is sought_nVariance, be denoted as e_n；

S24, statistic is definedWherein w_nFor weight coefficient, andWherein,Be withFor the diagonal matrix of diagonal entry composition, m₀=[1,1 ..., 1]^TIt is H in step sl₀Detection model It is lower to pass through N_bSecondary observation y_nVariance mean value composition vector,It is H in step sl₁Under detection model y_nVariance mean value composition vector；

S3, according to detector threshold value γ, whether true T > γ is judged, if so, being judged to signal；Otherwise signal is judged to not deposit It is determined in the value of, γ according to noise profile.