Signal detection method and Laplacian correlation detector solving device and method

Abstract

The invention discloses a signal detection method and a Laplacian correlation detector solving device and method. The signal detection method comprises the following steps: acquiring to-be-detected signals under two channels; solving a preset Laplacian correlation detector according to the to-be-detected signal to obtain a detection statistic; determining a detection result according to a comparison result of the detection statistical magnitude and a preset threshold value; wherein the preset threshold value is calculated through a preset false alarm probability. The signal detection efficiency can be improved.

Description

Technical Field

[0001] The present invention relates to the field of signal processing, and in particular to a signal detection method and a Laplace correlation detector solving device and method. Background Art

[0002] In fields such as signal processing and statistics, the correlation coefficient is an important tool for measuring the strength of the statistical relationship between random variables. Among them, the Pearson's Product Moment Correlation Coefficient (PPMCC) proposed by Karl Pearson, the founder of mathematical statistics, the Spearman's rho (SR) proposed by statistician Spearman, and the Kendall's tau (KT) proposed by psychologist Kendall are the three most commonly used correlation coefficients. PPMCC is the optimal asymptotically unbiased estimate of the parent correlation coefficient when the sample follows a bivariate normal distribution, and its statistical properties are very mature. Due to its excellent detection performance, PPMCC occupies a dominant position in both theoretical research and practical applications. However, in practical applications, data is often interfered by noise, especially impulse noise, such as low-frequency atmospheric noise, electromagnetic interference generated by natural lightning and thunderstorms, man-made noise, and interference noise generated by electronic equipment. PPMCC is very sensitive to noise in the sample, and its performance may drop sharply or even fail when the sample is affected by impulse noise. In contrast, SR and KT only consider the ordering information of samples, that is, the rank information, so they are less sensitive to non-normal distribution and have better robustness to impulse noise. This enables SR and KT to provide more reliable correlation analysis results when processing data containing non-normal distribution noise.

[0003] However, SR and KT still have the problem of low robustness and detection performance. In the prior art, compared with the three classical correlation coefficients mentioned above, the research and application of the Laplace kernel correlation coefficient are relatively less. The characteristics of the Laplace kernel function include strong locality, unbounded support, parameter dependence and computational cost. It is very sensitive to neighboring points, and the influence away from the center point weakens rapidly. Although the kernel function decays rapidly with increasing distance, it theoretically has a non-zero response to all inputs.

[0004] Therefore, it is of practical research significance to apply the correlation detector based on Laplace kernel correlation coefficient to signal detection. Summary of the invention

[0005] In order to overcome the defects of the above-mentioned prior art, the present invention provides a signal detection method and a Laplace correlation detector solving device and method. The signal detection method can improve the efficiency of signal detection, and the Laplace correlation detector solving device and method can improve the solving efficiency of the Laplace correlation detector.

[0006] An embodiment of the present invention provides a signal detection method, comprising the following steps:

[0007] The signals to be detected under two channels are collected;

[0008] Solve a preset Laplace correlation detector according to the signal to be detected to obtain a detection statistic;

[0009] The detection result is determined according to the comparison result between the detection statistic and the preset threshold value; wherein the preset threshold value is calculated by a preset false alarm probability.

[0010] Furthermore, the preset Laplace detection model is specifically:

[0011]

[0012] Among them, X i , Y i are the signals to be detected for the two channels, V X 、V Y X i , Y i Corresponding to the median of the signal sequence to be detected, n X , η Y The preset kernel widths for the two channels respectively.

[0013] Another embodiment of the present invention provides a solution device based on the preset Laplace correlation detector described in any one of the above embodiments, comprising: an input unit, a calculation unit and a register; the input unit, the calculation unit and the register are connected in sequence and are respectively composed of a plurality of FPGA integrated circuits;

[0014] The input unit is used to preprocess the signal to be detected and input the sequence of the signal to be detected obtained after the preprocessing into the calculation unit;

[0015] The calculation unit is used to solve the preset Laplace detection model according to the signal sequence to be detected through the plurality of FPGA integrated circuits using a pipeline processing method, and input the solution result into the register;

[0016] The register is used to store the solution result.

[0017] Further, the input unit includes a row storage block, a column storage block and a sorting network module;

[0018] The row storage block and the column storage block are used to serially store the to-be-detected signal sequence, the output end of the row storage block is connected to the calculation unit, and the output end of the column storage block is connected to the sorting network module;

[0019] The sorting network module is used to select the median of the signal sequence to be detected and output the median to the calculation unit.

[0020] Further, the calculation unit includes a comparison operation module, a subtractor, a first multiplier, a second multiplier, an adder and a second divider;

[0021] The input end of the comparison operation module is connected to the input unit, and the output end is connected to the first multiplier for calculating and The result of i , Y i are the signals to be detected for the two channels, V X 、V Y are the medians of the signal sequences to be detected of the two channels, η X , η Y They are the preset kernel widths of external inputs respectively;

[0022] The input end of the subtractor is connected to the input unit, and the output end is connected to the first multiplier, respectively, for calculating X i -V X , Y i -V Y Result;

[0023] The output terminal of the first multiplier is connected to the second multiplier for calculating and Result;

[0024] The output terminal of the second multiplier is connected to the adder for calculating Result;

[0025] The output end of the adder is connected to the second divider for calculating Results

[0026] The output terminal of the second divider is connected to the register for calculating result.

[0027] Preferably, the comparison operation module includes a comparator, a first divider and an exponential operation circuit;

[0028] The output terminal of the comparator is connected to the first divider for calculating |X i -V X |、|Y i -V Y |Results;

[0029] The output terminal of the first divider is connected to the exponential operation circuit for calculating and Result;

[0030] The output end of the exponential operation circuit is connected to the first multiplier for calculating and result.

[0031] Furthermore, the solution device also includes an assembly line;

[0032] The pipeline is arranged between the input unit and the calculation unit for caching the signal sequence to be detected, and the output end of the pipeline is respectively connected to the comparison operation module and the subtractor.

[0033] Preferably, the solving device further comprises a control unit, and the control unit is connected to the pipeline and is used to control the output end of the pipeline.

[0034] Another embodiment of the present invention provides a method for solving a Laplace correlation detector based on the solving device described in any one of the above embodiments, comprising the following steps:

[0035] Preprocessing the signal to be detected by the input unit to obtain a signal sequence to be detected;

[0036] According to the signal sequence to be detected, the preset Laplace detection model is solved by the calculation unit, and the solution result is input into the register.

[0037] Furthermore, after obtaining the signal sequence to be detected, the method further includes:

[0038] The median of the signal sequence to be detected is extracted through a sorting network module.

[0039] Compared with the prior art, the beneficial effects of this application are:

[0040] Based on the Laplace kernel correlation coefficient, a Laplace correlation detector under dual channels is proposed. Compared with the existing detectors, the Laplace correlation detector provided by the present invention can eliminate the influence of abnormal signal points on the detector performance to the maximum extent, and has better stability and detection performance than other non-parametric detectors and parametric detectors when dealing with pulse noise.

[0041] At the same time, by utilizing the high-performance parallel processing capabilities of FPGA integrated circuits, the calculation time of the Laplace correlation detector related operations is shortened, and the solution efficiency of the Laplace correlation detector is improved, thereby enabling real-time and fast signal correlation processing based on the Laplace correlation detector. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A schematic flow chart of a signal detection method provided by an embodiment of the present invention.

[0043] Figure 2 A schematic diagram of a model evaluation index result provided by an embodiment of the present invention.

[0044] Figure 3 A schematic diagram of the results of a Monte Carlo comparison test provided by an embodiment of the present invention.

[0045] Figure 4 A schematic diagram of the structure of a Laplace correlation detector solving device provided by an embodiment of the present invention.

[0046] Figure 5 A schematic structural diagram of a preferred embodiment of a Laplace correlation detector solution device provided by an embodiment of the present invention.

[0047] Figure 6 A schematic flow chart of a method for solving a Laplace correlation detector provided by another embodiment of the present invention. DETAILED DESCRIPTION

[0048] The drawings are for illustrative purposes only and should not be construed as limiting the present patent;

[0049] It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0050] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0051] Reference Figure 1 , is a flow chart of a signal detection method provided by an embodiment of the present invention, comprising the following steps:

[0052] S1: Collect the signals to be detected under two channels;

[0053] S2: solving a preset Laplace correlation detector according to the signal to be detected to obtain a detection statistic;

[0054] S3: Determine a detection result according to a comparison result between the detection statistic and a preset threshold value; wherein the preset threshold value is calculated by a preset false alarm probability.

[0055] For step S2, specifically, the preset Laplace detection model is:

[0056]

[0057] Among them, X i , Y i are the signals to be detected for the two channels, V X 、V Y X i , Y i Corresponding to the median of the signal sequence to be detected, n X , η Y The preset kernel widths for the two channels respectively.

[0058] In a preferred embodiment, in order to effectively illustrate that the preset Laplace correlation detector has excellent robustness and detection performance in combating impulse noise, this preferred embodiment will intuitively demonstrate the advantages of the Laplace correlation detector over several other traditional detectors through comparative experiments.

[0059] Before the comparative experiment, in order to help users select a more appropriate kernel width (η X ,η Y ), we need to first design an experiment to select the appropriate kernel width. The noise model and parameters of the experiment are designed as follows:

[0060] The input signal is known and follows the standard normal distribution model N(0,1), where n = 1000. The selected impulse noise includes two types of noise, namely the mixed Gaussian noise model (Contaminated Gaussian model, CGM) and the Laplace noise model (Laplacian noise model, LNM), so Z i and W i Two noise models were constructed, and represents the noise model under mixed Gaussian noise, and Represents the noise model under Laplace noise, then:

[0061]

[0062] Among them, ε = 0.05 represents the probability of pulse component occurrence in a mixed Gaussian noise environment, represents the variance of the pulse component of the mixed Gaussian noise, σ 31 =σ 32 =1 indicates the standard deviation of Laplace noise.

[0063] The experiment of the optimal kernel width is based on the model evaluation index (area under the curve, AUC). Figure 2 , is a schematic diagram of a model evaluation index result provided by an embodiment of the present invention.

[0064] Depend on Figure 2 It can be seen that AUC is the ordinate and the abscissa is the kernel width of the X channel from 0 to 20 (i.e. η X ∈[0-20]), the kernel width of the Y channel is fixed to 1 for the accuracy of the experiment, that is, η Y =1. Figure 2 At the same time, the influence of five kinds of attenuation coefficient kernel width on AUC under mixed Gaussian noise model and Laplace noise model was made. When the noise is mixed Gaussian noise model, the maximum value of AUC under five kinds of attenuation coefficients is all at η X =2; when the noise is the Laplace noise model, the AUC reaches its maximum value under the five attenuation coefficients (θ1, θ2) at η X = around 0.8.

[0065] It can be seen that selecting an appropriate kernel width according to the specific environmental noise conditions will greatly improve the detection performance of the Laplace correlation detector. According to the experimental results, when the noise is mixed Gaussian noise, select η X =2 is optimal, when the noise is Laplace noise, select η X =0.8 is optimal.

[0066] In order to illustrate the excellent robustness and detection performance of the Laplace correlation detector against impulse noise, this preferred embodiment compares the performance of the Laplace correlation detector proposed in the present invention with the traditional Spearman rank correlation coefficient detector, the Kendall rank correlation coefficient and the detection probability (Probability of detection, PD) of the matched filter through a Monte Carlo experiment.

[0067] The experimental settings are as follows: the attenuation coefficients of the two channels are set equal (that is, θ1 = θ2 = θ), and the false alarm probability is set to 5%. By looking up the table, the detection threshold (ξ) corresponding to the area of ​​the upper α quantile of the standard Gaussian distribution of 5% can be obtained. The detection probability can be calculated according to the following formula:

[0068]

[0069] Among them, μ 1 and σ 1 The mean and standard deviation of the received signal under the assumption that is 1,

[0070] Reference Figure 3 , is a schematic diagram of the results of a Monte Carlo comparison test provided by an embodiment of the present invention. Figure 3 It can be seen that when the background noise is mixed Gaussian noise or Laplace noise, the matched filter is basically ineffective. Since KT and SR only consider the ordering information of the samples, that is, the rank information, they are less sensitive to non-normal distribution and have better robustness to impulse noise. The Laplace correlation detector constructed using the median of the received signal proposed in the present invention surpasses KT and SR in detection performance. Since the construction of the Laplace correlation detector does not require operations between two random variables, it has a faster solution speed in the solution process. By utilizing the high-performance parallel processing capability of FPGA, the calculation time can be further shortened, and real-time and fast LC-based signal correlation processing can be realized, which can achieve real-time calculation solution speed while maintaining excellent detection performance.

[0071] Reference Figure 4 , is a schematic diagram of a Laplace correlation detector solving device provided by an embodiment of the present invention, comprising: an input unit, a calculation unit and a register; the input unit, the calculation unit and the register are connected in sequence and are respectively composed of a plurality of FPGA integrated circuits;

[0072] The input unit is used to preprocess the signal to be detected and input the sequence of the signal to be detected obtained after the preprocessing into the calculation unit;

[0073] The calculation is used to solve the preset Laplace detection model by using the plurality of FPGA integrated circuits according to the sequence of signals to be detected using a pipeline processing method, and input the solution result into the register;

[0074] The register is used to store the solution result.

[0075] Further, the input unit includes a row storage block, a column storage block and a sorting network module;

[0076] The row storage block and the column storage block are used to serially store the to-be-detected signal sequence, the output end of the row storage block is connected to the calculation unit, and the output end of the column storage block is connected to the sorting network module;

[0077] The sorting network module is used to select the median of the signal sequence to be detected and output the median to the calculation unit.

[0078] In a preferred embodiment, in practical applications, many signal processing tasks require detecting signals under noise interference. The signal detection problem under dual-channel impulse noise is usually modeled as follows:

[0079]

[0080] Among them, X i and Y i are the receiving signals of the two channels, θ1 and θ2 represent the attenuation coefficients of the two channels with (θ1θ2≠0) or without signal (θ1θ2=0), s i is a known transmitted signal without noise interference, Z i and W i These are the impulse noises of the two channels respectively.

[0081] make and Represents the system model The received signal of length n is represents the known transmitted signal undisturbed by noise, and Respectively represent the impulse noise of the two receiving channels.

[0082] Reference Figure 5 , which is a schematic diagram of a preferred embodiment of a Laplace correlation detector solution device provided by an embodiment of the present invention. Figure 5 It can be seen that the signal to be detected X i and Y i are input into the row storage block and the column storage block respectively, and two groups of signal sequences to be detected arranged in ascending order of amplitude are obtained, X1,...X i ,...X n and Y1,...Y i ,...Y n , completing the first level cache.

[0083] After completing the first-level cache, one set of signal sequences to be detected is directly output to the calculation unit, and the other set of signal sequences to be detected is sent to the sorting network module to extract the corresponding median V X and V Y , then V X and V Y is also output to the computing unit.

[0084] Further, the calculation unit includes a comparison operation module, a subtractor, a first multiplier, a second multiplier, an adder and a second divider;

[0085] The input end of the comparison operation module is connected to the input unit, and the output end is connected to the first multiplier for calculating and The result of i , Y i are the signals to be detected for the two channels, V X 、V y are the medians of the signal sequences to be detected of the two channels, η X , η Y They are the preset kernel widths of external inputs respectively;

[0086] The input end of the subtractor is connected to the input unit, and the output end is connected to the first multiplier, respectively, for calculating X i -V X , Y i -V Y Result;

[0087] The output terminal of the first multiplier is connected to the second multiplier for calculating and Result;

[0088] The output terminal of the second multiplier is connected to the adder for calculating Result;

[0089] The output end of the adder is connected to the second divider for calculating Result;

[0090] The output terminal of the second divider is connected to the register for calculating result.

[0091] Preferably, the comparison operation module includes a comparator, a first divider and an exponential operation circuit;

[0092] The output terminal of the comparator is connected to the first divider for calculating |X i -V X |、|Y i -V Y |Results;

[0093] The output terminal of the first divider is connected to the exponential operation circuit for calculating and Result;

[0094] The output end of the exponential operation circuit is connected to the first multiplier for calculating and result.

[0095] In a preferred embodiment, referring to Figure 5 , in the calculation unit, the signal sequence to be detected X i , Y i The corresponding median V X 、V Y are input to the comparator and subtractor at the same time, and are completed in two comparators and two subtractors respectively. i -V X |、|Y i -V Y | and X i -V X , Y i -V Y Calculation.

[0096] Then, the |X calculated by the two comparators i -V X |、|Y i -V Y | and the preset kernel width η of the external input X , η Y Input into the two first dividers respectively to complete and Calculation.

[0097] Then, the two first dividers are calculated and Input into two exponential operation circuits respectively to complete and Calculation.

[0098] Then, the two exponential operation circuits calculate and the X calculated by two subtractors i -V X , Y i -V Y Input into the two first multipliers respectively to complete and Calculation.

[0099] Then, the two first multipliers calculated and At the same time, it is input into the second multiplier to complete Calculation.

[0100] Then, the second multiplier calculates Input to the adder to complete Calculation.

[0101] Finally, the adder calculates Input to the second divider to complete Thus, the solution process ends, and the second divider inputs the final solution result into the register.

[0102] Furthermore, the solution device also includes an assembly line;

[0103] The pipeline is arranged between the input unit and the calculation unit for caching the signal sequence to be detected, and the output end of the pipeline is respectively connected to the comparison operation module and the subtractor.

[0104] Preferably, the solving device further comprises a control unit, and the control unit is connected to the pipeline and is used to control the output end of the pipeline.

[0105] In a preferred embodiment, referring to Figure 5 The solving device is also provided with a pipeline and a control unit. The pipeline is used to temporarily store intermediate calculation results. In this preferred embodiment, the pipeline is used to perform secondary caching on the signal sequence to be detected output by the row storage module and the median output by the sorting network module.

[0106] The control unit is used to control the signal sequence to be detected and the corresponding median cached in the pipeline to be sent to the first comparator and the subtractor respectively at the same time, so as to realize multi-line simultaneous calculation and improve the calculation efficiency of the calculation unit.

[0107] In summary, the Laplace correlation detector solution device provided by the embodiment of the present invention is specifically an FPGA operation circuit built according to the above-mentioned module components. After the construction is completed, it is necessary to define the interface of each module component and perform behavioral level modeling on each module component. Figure 2 As shown in the figure, each module is connected and verified by behavioral level simulation. Finally, each module is modeled at register transfer level to realize various specific functions.

[0108] Therefore, the embodiment of the present invention utilizes the high-performance parallel processing capability of the FPGA integrated circuit to shorten the calculation time of the Laplace correlation detector related operations and improve the solution efficiency of the Laplace correlation detector.

[0109] Reference Figure 6 , is a flow chart of a method for solving a Laplace correlation detector provided by another embodiment of the present invention, comprising the following steps:

[0110] S4: preprocessing the signal to be detected through the input unit to obtain a signal sequence to be detected;

[0111] S5: According to the signal sequence to be detected, the preset Laplace detection model is solved by the calculation unit, and the solution result is input into the register.

[0112] Furthermore, after obtaining the signal sequence to be detected, the method further includes:

[0113] The median of the signal sequence to be detected is extracted through a sorting network module.

[0114] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the embodiments here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. A signal detection method, characterized in that: The following steps are involved: The signals to be detected under two channels are collected; Solve a preset Laplace correlation detector according to the signal to be detected to obtain a detection statistic; The detection result is determined according to the comparison result between the detection statistic and the preset threshold value; wherein the preset threshold value is calculated by a preset false alarm probability.

2. The signal detection method according to claim 1, characterized in that: The preset Laplace detection model is specifically: Among them, X i , Y i are the signals to be detected for the two channels, V X 、V Y X i , Y i Corresponding to the median of the signal sequence to be detected, n X , η Y The preset kernel widths for the two channels respectively.

3. A solution device based on the preset Laplace correlation detector according to any one of claims 1-2, characterized in that: include: An input unit, a calculation unit and a register; the input unit, the calculation unit and the register are connected in sequence and are respectively composed of a plurality of FPGA integrated circuits; The input unit is used to preprocess the signal to be detected and input the sequence of the signal to be detected obtained after the preprocessing into the calculation unit; The calculation unit is used to solve the preset Laplace detection model according to the signal sequence to be detected through the plurality of FPGA integrated circuits using a pipeline processing method, and input the solution result into the register; The register is used to store the solution result.

4. The Laplace correlation detector solving device according to claim 3, characterized in that: The input unit includes a row storage block, a column storage block and a sorting network module; The row storage block and the column storage block are used to serially store the to-be-detected signal sequence, the output end of the row storage block is connected to the calculation unit, and the output end of the column storage block is connected to the sorting network module; The sorting network module is used to select the median of the signal sequence to be detected and output the median to the calculation unit.

5. The Laplace correlation detector solving device according to claim 4, characterized in that: The calculation unit includes a comparison operation module, a subtractor, a first multiplier, a second multiplier, an adder and a second divider; The input end of the comparison operation module is connected to the input unit, and the output end is connected to the first multiplier for calculating and The result of i , Y i are the signals to be detected for the two channels, V X 、V Y are the medians of the signal sequences to be detected of the two channels, η X , η Y They are the preset kernel widths of external inputs respectively; The input end of the subtractor is connected to the input unit, and the output end is connected to the first multiplier, respectively, for calculating X i -V X , Y i -V Y Result; The output terminal of the first multiplier is connected to the second multiplier for calculating (X i - and Result; The output terminal of the second multiplier is connected to the adder for calculating Result; The output end of the adder is connected to the second divider for calculating Results The output terminal of the second divider is connected to the register for calculating result.

6. The Laplace correlation detector solving device according to claim 5, characterized in that: The comparison operation module includes a comparator, a first divider and an exponential operation circuit; The output terminal of the comparator is connected to the first divider for calculating |X i -V X |、|Y i -V Y |Results; The output terminal of the first divider is connected to the exponential operation circuit for calculating and Result; The output end of the exponential operation circuit is connected to the first multiplier for calculating and result.

7. The Laplace correlation detector solving device according to claim 5, characterized in that: The solution device also includes an assembly line; The pipeline is arranged between the input unit and the calculation unit for caching the signal sequence to be detected, and the output end of the pipeline is respectively connected to the comparison operation module and the subtractor.

8. The Laplace correlation detector solving device according to claim 7, characterized in that: The solving device further comprises a control unit, which is connected to the pipeline and is used to control an output end of the pipeline.

9. A method for solving a Laplace correlation detector based on the solving device according to any one of claims 3 to 8, characterized in that: The following steps are involved: Preprocessing the signal to be detected by the input unit to obtain a signal sequence to be detected; According to the signal sequence to be detected, the preset Laplace detection model is solved by the calculation unit, and the solution result is input into the register.

10. The method for solving the Laplace correlation detector according to claim 9, characterized in that: After obtaining the signal sequence to be detected, it also includes: The median of the signal sequence to be detected is extracted through a sorting network module.

Images