Channelization Detection Method Based on Robust Noise Floor Estimation
By adopting a robust noise floor estimation method in a channelized receiver, the accurate estimation of noise floor characteristic parameters and the generation of adaptive detection thresholds are solved, and the detection performance and sensitivity are improved.
Patent Information
- Application Number
- CN202211217882.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-09-30
- Publication Date
- 2025-06-10
- Estimated Expiration
- 2042-09-30
AI Technical Summary
Traditional channelized receivers are difficult to accurately describe the noise level in complex electromagnetic environments, resulting in inaccurate setting of detection thresholds, false alarms or missed detection, which seriously affects detection performance.
The channelized detection method based on robust noise floor estimation is adopted, and the accurate estimation of noise floor characteristic parameters and the generation of adaptive detection thresholds through steps such as AD sampling, multiphase filtering, time domain energy accumulation and quantile point estimation.
It improves the accuracy of estimating noise floor characteristic parameters in complex electromagnetic environments, realizes the generation of adaptive detection thresholds, improves detection probability and sensitivity, and reduces false alarms and missed detection phenomena.
Smart Images
Figure CN115508788B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of electronic reconnaissance, and particularly relates to a channelized detection method based on robust noise floor estimation. Background Art
[0002] In recent years, with the rapid development of radio technology, the frequency band of electronic countermeasures has become wider and the forms have become more complex and changeable. Traditional electronic reconnaissance receivers can no longer meet the requirements by far. Therefore, digital channelized receivers with large dynamic range, wide instantaneous reception bandwidth, high sensitivity, high resolution, and multi-signal processing capabilities based on the concept of software radio have emerged as the times require. As an important tool for electronic countermeasure equipment to detect target radiation source signals, the accurate setting of the detection threshold of its detector is an important basis for the receiver to achieve high detection sensitivity, and is of great significance for realizing subsequent functions such as radiation source identification, positioning, and tracking.
[0003] In traditional channelized receivers, the noise level of the echo sequence is usually raised by a certain multiple as the detection threshold. Therefore, the accurate description of the statistical characteristics of the noise level is crucial for the setting of the detection threshold. Patent 202118111478.2 proposed "A Pulse Width Adaptive Detection Method", which uses the minimum value of the cumulative noise envelope sequence to describe the noise floor level of the echo sequence, and sets the corresponding uplift multiple according to different cumulative pulse width lengths to obtain a detection threshold adapted to the pulse width. In this method, for a specific scenario, the uplift multiple of the noise floor level is fixed, and the influence of the dynamic fluctuation of the noise variance in a complex environment on the detection threshold is not considered, which is very likely to cause inaccurate setting of the noise floor threshold and result in false alarms or missed detections, seriously affecting the channelized detection performance. Therefore, studying an accurate description method for the noise floor level in the echo sequence and selecting an appropriate detection threshold value are of great significance for adapting to the complex and changeable actual receiving environment and realizing high-sensitivity detection of radiation source signals by electronic countermeasure equipment. Summary of the Invention
[0004] The present invention proposes a channelized detection method based on robust noise floor estimation, which realizes the accurate estimation of the noise floor characteristic parameters in a complex electromagnetic environment; realizes the generation of an adaptive detection threshold and improves the detection probability; by using the quantile estimation method, the parameter estimation of the amplitude sequence with abnormal samples is robust, improves the accuracy of the signal detection threshold, and further improves the sensitivity of radar signal detection.
[0005] The technical solution for realizing the present invention is: a channelized detection method based on robust noise floor estimation, including the following steps:
[0006] Step 1: After performing AD sampling and polyphase filtering on the radar signal, obtain the filtered data and transfer to Step 2.
[0007] Step 2: Accumulate the energy of the filtered data in the time domain to obtain the signal accumulation envelope, sort it in ascending order to generate an ascending signal accumulation envelope sequence, and then proceed to Step 3.
[0008] Step 3: Construct an envelope probability density curve based on the ascending signal accumulation envelope sequence, find the positions of the sequence extreme points, calculate the noise ratio of the sequence according to the positions of the sequence extreme points to obtain the noise accumulation envelope sequence of each channel, and then proceed to Step 4.
[0009] Step 4: Use the expression of the cumulative distribution function of the Gaussian distribution to calculate the estimated value of the noise floor mean and the estimated value of the noise floor variance of the noise accumulation envelope sequence, and then proceed to Step 5.
[0010] Step 5: Use a simulation experiment to determine the reliable false alarm probability, combine the estimated value of the noise floor mean and the estimated value of the noise floor variance to calculate the detection threshold, and then proceed to Step 6.
[0011] Step 6: Compare the signal accumulation envelope with the detection threshold to obtain the detection VP.
[0012] Compared with the prior art, the significant advantages of the present invention are as follows:
[0013] 1) The present invention realizes the accurate estimation of the noise floor characteristic parameters in a complex electromagnetic environment, improving the adaptability to a complex electromagnetic environment.
[0014] 2) It realizes the adaptive generation of a detection threshold under different noise ratio conditions, realizes the optimal matched filtering, and improves the detection probability.
[0015] 3) A quantile estimation method is proposed, which is robust to the parameter estimation of the amplitude sequence with abnormal samples, improves the accuracy of the signal detection threshold, and further improves the sensitivity of radar signal detection. Description of the Drawings
[0016] Figure 1 It is a flowchart of a channelized detection method based on robust noise floor estimation.
[0017] Figure 2 It is a schematic diagram of maximum value holding.
[0018] Figure 3 It is a schematic diagram of the estimation accuracy of the noise floor mean and the noise floor variance.
[0019] Figure 4 It is a schematic diagram of detection sensitivity. Detailed Embodiment
[0020] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.
[0021] Combined with Figure 1 , the channelized detection method based on robust noise floor estimation is as follows:
[0022] Step 1: After performing AD sampling and polyphase filtering on the radar signal in sequence, filtered data is obtained. The filtered data y k (n 1 M) is as follows:
[0023]
[0024] Among them, let the input time variable n 1 correspond to the sampling frequency of 500 MHz before filtering, and the decimation rate M = 32. Then the output sampling frequency time variable n 2 = 500 / 32 = 15.625 MHz; the number of channels Q = 32, ω q (q = 0, 1,..., Q - 1) is the center frequency of each channel. Taking the coefficient length of each channel filter allocation as 8, then h(n 1 ) is a 1024-order low-pass FIR filter, x(n 1 ) is the sampling signal sequence, and j represents the imaginary part.
[0025] Step 2: Perform time-domain energy accumulation on the filtered data input and perform ascending order processing to generate an ascending signal accumulation envelope sequence.
[0026] Step 21: Take the modulus of the filtered data of each channel to obtain the amplitude sequence Amp after filtering for each channel.
[0027] Step 22: According to the pulse width matching accumulation module, mainly when one or several weak signals are submerged or equal to the noise, the number of accumulation points matching the signal width is adopted. The result of noise superposition is almost mutually cancelled, while the weak signal part is retained. The more the data volume of signal superposition, the higher the accumulated energy. Therefore, the maximum signal-to-noise ratio improvement is achieved. Using a length of 1000 ns, perform time-domain energy sliding accumulation on the amplitude sequence Amp after filtering for each channel, and generate a signal accumulation envelope sequence {A 1 , A 2 , …, A L} for each channel, where L is the length of the signal accumulation envelope sequence.
[0028] Step 23: Sort the signal accumulation envelope sequence from small to large to obtain the ascending signal accumulation envelope sequence {A (1) , A (2) , …, A (L)}.
[0029] Step 3: According to the ascending signal accumulation envelope sequence, construct an envelope probability density curve, find the position of the sequence extreme point, calculate the noise ratio of the sequence based on the position of the sequence extreme point, and obtain the noise accumulation envelope sequence of each channel. This step can effectively extract the noise floor information in the signal cumulative envelope sequence and provide a basis for the accurate calculation of the subsequent detection threshold. It specifically includes the following steps:
[0030] Step 31: Use the ascending signal accumulation envelope sequence A = {A (1) , A (2) , …, A (L)} to calculate the envelope probability density distribution sequence and construct an envelope probability density curve.
[0031] Step 31_1: Calculate the maximum value A max and the minimum value A min of the ascending signal accumulation envelope sequence. Divide the interval [A min , A max where the elements of the ascending signal accumulation envelope sequence are located into 200 subintervals. Let i represent the subinterval serial number of the ascending signal accumulation envelope sequence. The subinterval set Ω = {ω 1 , ω 2 , …, ω i , …, ω 200}, where ω i represents the i-th subinterval.
[0032] Step 31_2: Let P = {p 1 , p 2 , …, p i , …, p 200} represent the cumulative number of elements in each subinterval, and initialize P = 0.
[0033] Step 31_3: Determine the subinterval to which the l-th element of the ascending signal accumulation envelope sequence A = {A (1) , A (2) , …, A (L)} belongs. If the subinterval it belongs to is ω i , then let p i = p i +1.
[0034] Step 31_4: Repeat Step 31_3 until l = L ends to obtain the set P composed of the total cumulative number of elements in each subinterval, and satisfy where L is the length of the signal accumulation envelope sequence.
[0035] Step 31_5: Calculate the probability density value prob of the i-th sub-interval using the following formula i :
[0036]
[0037] Obtain the envelope probability density distribution sequence Prob = {prob i , i = 1, 2, …, 200}.
[0038] Step 31_6: Calculate the central value cen of each sub-interval i = (i - 0.5)(A max - A min ) / 200, construct the envelope probability density curve with a curve length of 200, and use the central value of the i-th sub-interval and the probability density value of the i-th sub-interval to form the coordinates of the i-th point on the envelope probability density curve as (cen i , prob i ).
[0039] Step 32: Find the abscissa cen of the first local extreme point of the envelope probability density curve max .
[0040] Step 32_1: Smooth each probability density value prob i in the envelope probability density distribution sequence using the following formula to obtain the corresponding smoothed value prob_smooth i :
[0041]
[0042] Obtain the smoothed envelope probability density distribution sequence Prob_smooth = {prob_smooth i , i = 1, 2, …, 200}.
[0043] Step 32_2: Divide the smoothed envelope probability density distribution sequence into 20 sub-sequences equally, with each sub-sequence containing 10 element values, calculate the maximum value of each sub-sequence, and form the smoothed envelope maximum value sequence local_max k , k = 1, 2, …, 20.
[0044] Step 32_3: Perform differencing on the smoothed envelope maximum value sequence to obtain a difference sequence. Find all elements in the difference sequence with a value of 0, and the corresponding element positions form the zero point position sequence InvariantIndex. Add small positive perturbations to all element values in the smoothed envelope maximum value sequence where the element number satisfies k = InvariantIndex + 1 to obtain the updated smoothed envelope maximum value sequence local_max_mod k , k = 1, 2, …, 20, and the specific implementation method is as follows:
[0045]
[0046] Step 32_4: Calculate the local extreme points of the updated smoothed envelope maximum value sequence, and the corresponding element positions form the extreme point position sequence. Calculate the minimum value MaxIndexIndex of this sequence.
[0047] Step 32_5: Divide the envelope probability density distribution sequence into 20 sub-sequences equally, with each sub-sequence containing 10 element values. Calculate the maximum value position MaxIndex of the MaxIndexIndex-th sub-sequence to obtain the abscissa cen of the first local extreme point of the envelope probability density curve max Satisfy the following expression:
[0048] cen max = MaxIndexIndex * 10 + MaxIndex.
[0049] Step 33: Calculate the area between the first local extreme point of the envelope probability density curve and the horizontal axis on the left side to be Obtain the noise ratio ratio_noise = 2 * area, where prob i represents the probability density value of the i-th sub-interval, and i represents the sub-interval number of the ascending signal accumulation envelope sequence.
[0050] Step 34: Intercept the first cen max element values of the ascending signal cumulative envelope sequence to form the noise accumulation envelope sequence
[0051] Step 4: Use the noise accumulation envelope sequence to calculate the estimated value of the noise floor mean and the estimated value of the noise floor variance according to the cumulative distribution function expression of the Gaussian distribution. Compared with the traditional minimum noise floor calculation method, the estimated values of the noise floor mean and the noise floor variance obtained by this method can better describe the noise floor statistical characteristics of the radar signal. The specific steps are as follows:
[0052] Step 41: Select two cumulative probability values α = 0.25 and β = 0.5, and calculate two sample quantile values of the noise accumulation envelope sequence where [*] represents the integer closest to the serial number *.
[0053] Step 42: According to the cumulative distribution function expression of the Gaussian distribution, the functional relationship between two sample quantile values is obtained as follows:
[0054]
[0055] where erf -1 (*) represents the inverse function of the error function. The noise floor variance estimate is calculated using the above formula
[0056] Step 43: Based on the sample quantile values and the noise floor variance estimate Using any one of the expressions in the following system of equations, the noise floor mean estimate can be obtained
[0057]
[0058] Step 5: Use simulation experiments to determine the reliable false alarm probability and calculate the detection threshold, which specifically includes the following steps:
[0059] Step 51: Simulate AD data with a time length of 2.5 ms, and the ascending signal accumulation envelope sequence length L = 2.5e-3 * 500e6 / 32 = 40000.
[0060] Step 52: Set the false alarm probability P fa = 10 -5 , 10 -6 , 10 -7 , 10 -8 , 10 -9 , 10 -10 , 10 -11 , 10 -12 , 10 -17 , according to the noise floor mean estimate and the noise floor variance estimate, use the following formula to calculate the threshold thr under different false alarm probability conditions:
[0061]
[0062] Step 53: Repeat Step 51 and Step 52 until 1000 Monte Carlo experiments are completed, and statistically obtain the threshold curve under different false alarm probabilities and the sample quantile curve corresponding to the simulation data.
[0063] Step 54: According to the threshold curve under different false alarm probabilities and the sample quantile curve corresponding to the simulation data, the false alarm probability P fa = 10 -8 can achieve the detection requirement of approximately no false alarms, and at the same time, the detection sensitivity will not be lost too much.
[0064] Step 55: If L increases, the fluctuation of the sample quantile becomes smaller, and the selected false alarm probability value can be larger than 10 -8 and the threshold becomes lower; conversely, if L decreases, the threshold needs to be raised.
[0065] Step 6: By comparing the signal accumulation envelope with the detection threshold, the detection VP is obtained, which specifically includes the following steps:
[0066] Step 61: Perform a cross-threshold comparison between the signal accumulation envelope sequence and the threshold value obtained in Step 52 to obtain the rabbit ear detection VP.
[0067] Step 62: Multiply the threshold value by the parameter detection threshold adjustment coefficient to obtain the corresponding parameter threshold, and perform a cross-threshold comparison between the parameter threshold and the accumulation envelope to obtain the parameter detection VP.
[0068] Step 63: Expand the maximum value of the rabbit ear detection VP by 3 us (as Figure 2 shown), corresponding to obtaining the expanded detection, and perform an AND operation between the expanded detection and the parameter detection VP to obtain the original detection VP.
[0069] Through experiments, the estimated value of the noise floor mean and the estimation error of the noise floor variance decrease as the length of the signal accumulation envelope sequence increases (as Figure 3 shown); in addition, by comparing the traditional noise floor detection method and the robust noise floor detection method proposed in the present invention (as Figure 4 shown), the latter can improve the detection sensitivity by about 1 dB.
Claims
1. Channelized detection method based on robust noise floor estimation, characterized in that, it includes the following steps: Step 1: After performing AD sampling and polyphase filtering on the radar signal, obtain the filtered data and transfer to Step 2; Step 2: Perform time-domain energy accumulation on the filtered data to obtain the signal accumulation envelope, perform ascending order processing on it, generate an ascending order signal accumulation envelope sequence, and transfer to Step 3; Step 3: Construct an envelope probability density curve according to the ascending order signal accumulation envelope sequence, find the position of the sequence extreme point, calculate the noise ratio of the sequence according to the position of the sequence extreme point, obtain the noise accumulation envelope sequence of each channel, and transfer to Step 4; Step 4: Use the expression of the cumulative distribution function of the Gaussian distribution to calculate the noise floor mean estimate value and the noise floor variance estimate value of the noise accumulation envelope sequence, and transfer to Step 5; Step 5: Use simulation experiments to determine the reliable false alarm probability, combine the noise floor mean estimate value and the noise floor variance estimate value, calculate the detection threshold, and transfer to Step 6; Step 6: Compare the signal accumulation envelope with the detection threshold to obtain the detection VP.
2. The channelized detection method based on robust noise floor estimation according to claim 1, characterized in that, in Step 3, construct an envelope probability density curve according to the ascending order signal accumulation envelope sequence, find the position of the sequence extreme point, calculate the noise ratio of the sequence according to the position of the sequence extreme point, and obtain the noise accumulation envelope sequence of each channel, specifically as follows: Step 31. Using the ascending signal accumulation envelope sequence A = {A (1) , A (2) , …, A (L)}, calculate the envelope probability density distribution sequence and construct an envelope probability density curve; Step 32: Find the abscissa cen of the first local extreme point of the envelope probability density curve max ; Step 33: Calculate the area between the left side of the first local extreme point of the envelope probability density curve and the horizontal axis as Obtain the noise ratio ratio_noise = 2 * area, where prob i represents the probability density value of the i-th sub-interval, and i represents the sub-interval serial number of the ascending signal accumulation envelope sequence; Step 34: Intercept the first cen max element values of the ascending signal cumulative envelope sequence to form a noise accumulation envelope sequence 3. The channelized detection method based on robust noise floor estimation according to claim 2, characterized in that, in Step 31, use the ascending order signal accumulation envelope sequence to calculate the envelope probability density distribution sequence and construct an envelope probability density curve, specifically including the following steps: Step 31_1: Calculate the maximum value A of the ascending signal accumulation envelope sequence max and the minimum value A min . Divide the interval where the elements of the ascending signal accumulation envelope sequence are located [A min , A max into 200 equal sub-intervals. Let i represent the sub-interval number of the ascending signal accumulation envelope sequence. The sub-interval set Ω = {ω 1 , ω 2 , …, ω i , …, ω 200}, where ω i represents the i-th sub-interval; Step 31_2. Let \(P = \{p 1 , p 2 , \ldots, p i , \ldots, p 200 \}\) represent the cumulative number of elements in each sub - interval, and initialize \(P = 0\); Step 31_3, determine the sub - interval to which the size of the l - th element of the ascending signal accumulation envelope sequence A={A (1) , A (2) , …, A (L)} belongs. If the sub - interval it belongs to is ω i , then let p i = p i +1; Step 31_4. Repeat Step 31_3 until l = L ends, to obtain a set P composed of the total number of accumulated elements in each sub-interval, and it satisfies L is the length of the signal accumulation envelope sequence; Step 31_5: Calculate the probability density value prob of the i-th sub-interval using the following formula i :[[]]END]] Obtain the envelope probability density distribution sequence Prob = {prob i , i = 1, 2, …, 200}; Step 31_6: Calculate the central value cen of each sub-interval i =(i - 0.5)(A max - A min ) / 200, construct an envelope probability density curve with a curve length of 200. The coordinates of the i-th point of the envelope probability density curve are composed of the central value of the i-th sub-interval and the probability density value of the i-th sub-interval, which is (cen i , prob i ).
4. The channelized detection method based on robust noise floor estimation according to claim 3, characterized in that, In step 32, find the abscissa cen of the first local extreme point of the envelope probability density curve max , which specifically includes the following steps: Step 32_1. Smooth each probability density value prob in the envelope probability density distribution sequence using the following formula i to obtain the corresponding smoothed value prob_smooth i : Obtain the smoothed envelope probability density distribution sequence Prob_smooth = {prob_smooth i , i = 1, 2, …, 200}; Step 32_2: Equally divide the smooth envelope probability density distribution sequence into 20 subsequences, with each subsequence containing 10 element values. Calculate the maximum value of each subsequence to form the smooth envelope maximum value sequence local_max k , where k = 1, 2, …, 20; Step 32_3: Perform a difference operation on the smoothed envelope maximum value sequence to obtain a difference sequence. Find all elements with a value of 0 in the difference sequence, and the positions of the corresponding elements form the zero point position sequence InvariantIndex. Add small positive perturbations to all element values in the smoothed envelope maximum value sequence whose element numbers satisfy k = InvariantIndex + 1 to obtain the updated smoothed envelope maximum value sequence local_max_mod k , k = 1, 2, …, 20, and the specific implementation method is as follows: Step 32_4: Calculate the local extreme points of the updated smoothed envelope maximum sequence, the positions of the corresponding elements form an extreme point position sequence, and calculate the minimum value MaxIndexIndex of this sequence; Step 32_5: Divide the envelope probability density distribution sequence into 20 equal sub-sequences, with each sub-sequence containing 10 element values. Calculate the maximum position MaxIndex of the MaxIndexIndex-th sub-sequence to obtain the abscissa cen of the first local extreme point of the envelope probability density curve max Satisfy the following expression: cen max = MaxIndexIndex * 10 + MaxIndex。 5. The channelized detection method based on robust noise floor estimation according to claim 4, characterized in that, in Step 4, use the expression of the cumulative distribution function of the Gaussian distribution to calculate the noise floor mean estimate value and the noise floor variance estimate value of the noise accumulation envelope sequence, specifically including the following steps: Step 41: Select two cumulative probability values α = 0.25 and β = 0.5, and calculate two sample quantile values of the noise accumulation envelope sequence where [*] represents the integer closest to serial number *; Step 42: According to the expression of the cumulative distribution function of the Gaussian distribution, obtain the following functional relationship between two sample quantile values: Among them, erf -1 (*) represents the inverse function of the error function; the noise floor variance estimation value is calculated using the above formula Step 43: Based on the sample quantile value and the estimated value of the noise floor variance Using any one of the expressions in the following equations, the estimated value of the noise floor mean can be obtained 6. The channelized detection method based on robust noise floor estimation according to claim 5, characterized in that, in Step 5, use simulation experiments to determine the reliable false alarm probability, combine the noise floor mean estimate value and the noise floor variance estimate value, and calculate the detection threshold, specifically including the following steps: Step 51: Simulate AD data with a time length of 2.5 ms, and the length L of the ascending order signal accumulation envelope sequence = 2.5e-3 * 500e6 / 32 = 40000; Step 52: Set the false alarm probability P fa = 10 -5 ,10 -6 ,10 -7 ,10 -8 ,10 -9 ,10 -10 ,10 -11 ,10 -12 ,10 -17 ,According to the estimated value of the noise floor mean and the estimated value of the noise floor variance, calculate the threshold thr under different false alarm probability conditions using the following formula: Step 53: Repeat Step 51 and Step 52 until 1000 Monte Carlo experiments are completed, and statistically obtain the threshold curve under different false alarm probabilities and the sample quantile curve corresponding to the simulation data; Step 54. According to the threshold curve under different false alarm probabilities and the sample quantile curve corresponding to the simulation data, the false alarm probability P fa = 10 -8 can meet the detection requirement of approximately no false alarm, and at the same time, the detection sensitivity will not be lost too much; Step 55: If L increases, the fluctuation of the sample quantile becomes smaller, and the selected false alarm probability value can be larger than 10 -8 and the threshold becomes lower; conversely, if L decreases, the threshold needs to be raised.