A broadband signal detection method based on scale iteration and spectrum compensation

Abstract

The invention belongs to the field of wireless communication technology, and in particular to a broadband signal detection method based on scale iteration and spectrum compensation. Aiming at the problems that the existing broadband signal detection algorithm is easily affected by noise, has poor detection effect under low signal-to-noise ratio, has high complexity, takes a long detection time, and requires a large amount of prior information, the following scheme is proposed, comprising the following steps: Step A: A received signal is sampled by AD to obtain a broadband sampling signal, and the broadband signal is divided into multiple signal segments of appropriate bandwidths through digital channelization; the purpose of the invention is to make the noise floor estimation more accurate and avoid false detection caused by uneven noise floor for broadband signals by utilizing the characteristics of morphological filtering bandwidth screening; utilize smoothing iteration to overcome the error caused by strong pulses and white noise in the signal to signal parameter estimation, improve the detection performance of frequency band aliasing signals by spectrum compensation, realize blind detection of broadband signals, and improve detection accuracy.

Description

Technical Field

[0001] The present invention belongs to the technical field of wireless communications, and in particular relates to a broadband signal detection method based on scale iteration and spectrum compensation. Background Art

[0002] The purpose of broadband signal detection is to quickly find signals on the broadband spectrum and estimate the center frequency and bandwidth of each signal to provide prior information for subsequent signal analysis and processing. It has a large number of applications in various civil and military communication systems, such as satellite spectrum detection, radar signal scanning, and battlefield information perception, all of which require broadband spectrum detection.

[0003] However, in non-cooperative communications, spectrum detection is usually completed under conditions such as lack of prior information, complex electromagnetic environment, and the presence of an uneven colored noise floor. This requires that the broadband signal detection algorithm be able to achieve blind detection without prior information under low signal-to-noise ratio and have the ability to resist an uneven noise floor.

[0004] There are usually several detection algorithms for broadband signals: First, signal energy detection method, such as the energy detection method for unknown deterministic signals proposed by Urkowitz (1967), and the improved multi-antenna signal energy detection scheme proposed by Li Yan (2017). These methods have the advantage of anti-frequency deviation, but are easily affected by noise, and the detection effect is poor under low signal-to-noise ratio; second, cyclostationary detection method, such as the algorithm based on second-order cyclostationary characteristics proposed by A. Tani (2016), which has the advantage of good detection performance, but high complexity and long detection time; third, matched filter detection method, such as the detection method combining parallel matched filter and segmented filter proposed by Z. Zhang (2010). This method is an optimal signal detection method in theory, but requires a lot of prior information such as carrier frequency, bandwidth, modulation mode, etc., which is difficult to implement in practice and is not suitable for non-cooperative communication. In addition, some new algorithms have been proposed in recent years, and the signal detection performance has been greatly improved compared with the classical algorithms. The gradient double threshold algorithm proposed by Zhang Yang (2016) calculates the gradient of the broadband power spectrum segmentally, and then performs adaptive double threshold detection. The power spectrum cancellation method proposed by Qi Peihan (2014) uses the ratio of the sum of the intensities of some spectral lines in the segmented frequency band to the sum of the intensities of all spectral lines as the test statistic to judge the existence of the signal. Summary of the invention

[0005] Aiming at the problems that the existing detection algorithms of broadband signals are easily affected by noise, have poor detection effect under low signal-to-noise ratio, high complexity, long detection time and require a lot of prior information, a broadband signal detection method based on scale iteration and spectrum compensation is proposed.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] A broadband signal detection method based on scale iteration and spectrum compensation comprises the following steps:

[0008] Step A: The received signal is sampled by AD to obtain a broadband sampling signal, and the broadband signal is divided into multiple signals of appropriate bandwidth through digital channelization;

[0009] Step B: Take several segments of broadband signals and calculate the maximum power spectrum according to the power spectrum density of each segment of the signal;

[0010] Step C: using the scale iterative morphological filtering algorithm to estimate and remove the colored noise floor, and calculate the white noise power threshold;

[0011] Step D: Detect the signal with the highest amplitude in the power spectrum, estimate its center frequency, bandwidth, and the start and end positions of the signal frequency band, and eliminate the signal and complete the compensation according to the compensation algorithm. After each signal is detected, if the difference in residual signal energy before and after the two signal detections is higher than the preset energy threshold, step D is performed again with the compensated power spectrum, otherwise the signal detection is terminated.

[0012] Preferably, the bandwidth screening characteristics of morphological filtering are utilized to perform morphological filtering iterations on the power spectrum to make the noise floor estimation more accurate, smoothing iterations are utilized to overcome the errors in signal parameter estimation caused by strong pulses and white noise in the signal, and spectrum compensation is used to improve the detection performance of frequency band aliasing signals.

[0013] Preferably, the step A specifically comprises the following steps:

[0014] A1: Calculate the center frequency of the real signal channelization subchannel D is the data extraction multiple, and the entire frequency band is divided into D symmetrical sub-bands of real signals;

[0015] A2: Anti-aliasing filter h LP (n) is a low-pass FIR filter, K is the number of channels, D is the decimation factor, and N and D have an integer multiple relationship, that is, K = FD, then the output of the kth subchannel in the traditional channelization structure is:

[0016]

[0017] A3: Rewrite the above formula into a polyphase filter structure, and get the output expression of the kth subchannel:

[0018]

[0019] A4: Set the center frequency ω k Substituting into the formula, the output expression of the kth subchannel is:

[0020]

[0021] Preferably, in step B, the received data is segmented m times, each segment is L in length, and the interval between segments is K, and the extracted data can be represented in matrix form:

[0022]

[0023] For each segment of data, the power spectrum P of each segment of data is estimated by the periodogram method. i (f), namely

[0024]

[0025] Among them, w(n) is the selected window function. After finding the power spectrum of each segmented signal, the maximum value of the power spectrum is found at each frequency point, that is,

[0026]

[0027] Preferably, the step C specifically comprises the following steps:

[0028] B1: Scale B bi Minimum detection bandwidth B min To perform a morphological filter PN1(f);

[0029] B2: Scale B bi Take the value of the last filtering scale B bi-1 A morphological filter is performed once if the scale is greater than the maximum detection bandwidth B. max , then B max A morphological filter is performed for the scale, and the result of the i-th filter is PN i (f);

[0030] B3: Difference the filtering results of the i-th and i-1-th adjacent filtering times, and the difference result is ΔPN i , compare the two differential results frequency by frequency point, if ΔPN i (f)>ΔPN i-1 (f), then the scale at frequency point f is updated to B bi , otherwise the scale remains unchanged;

[0031] B4: If B bi =B max , then stop the iteration;

[0032] B5: The scale B determined iteratively at each frequency point f bi (f) Perform a morphological filter to estimate the colored noise floor;

[0033] B6: Subtract B5 from the original power spectrum to estimate the noise floor and obtain a power spectrum with a flat noise floor;

[0034] B7: Convert the power spectrum into a logarithmic power spectrum, define an appropriate number of power bands from 0 to the maximum value on the power spectrum and count the number of frequency points falling in each band. The more power bands there are, the more accurate the white noise threshold estimation is. The number of frequency points in adjacent power bands is differentiated respectively, and the upper limit of the power band with the largest differential result is used as the white noise power threshold.

[0035] Preferably, the step D specifically includes the following steps:

[0036] C1: Find the maximum position f in the power spectrum max As the center frequency f Ctr , at the center frequency amplitude P(f Ctr )-3dB to calculate the 3dB bandwidth B 3dB , the left edge of the bandwidth is f 3dBL , the right boundary is f 3dBR ;

[0037] C2: The power spectrum is [f 3dBL ,f 3dBR ] with scale λ sm =0.1*B 3dB After smoothing and filtering, it returns to C1 and enters C3 after five cycles;

[0038] C3: After five cycles of smoothing, the bandwidth obtained is the 3dB bandwidth. 3dBL ,f 3dBR ] range according to the following formula to calculate the center frequency;

[0039]

[0040] C4: in f 3dBL To the left, f 3dBR The first minimum or zero point of the power spectrum is found on the right as the main lobe boundary of the signal, and the left boundary is f LS , the right boundary is f RS , in [f LS ,f RS ] range, the signal is reset to zero and removed, and the out-of-band slope k is calculated according to the following formula LS , k RS :

[0041]

[0042]

[0043] C5: in f LS , f RS The slope k LS , k RS To the estimated signal band [f LS ,fRS ] until the two lines intersect, and the two extended lines replace the in-band spectrum lines of the estimated signal to reduce the damage to the unestimated signal;

[0044] C6: If the energy estimated by C4 is lower than the minimum judgment energy or the bandwidth is smaller than the minimum detection bandwidth, it will be discarded. If the maximum amplitude of the residual power spectrum is not higher than the white noise threshold, the signal detection is completed.

[0045] The beneficial effects of the present invention are as follows: for broadband signals, the characteristics of morphological filtering bandwidth screening are utilized to perform morphological filtering iterations on the power spectrum, so that the noise floor estimation is more accurate and false detection caused by an uneven noise floor is avoided; smoothing iteration is utilized to overcome the errors in signal parameter estimation caused by strong pulses and white noise in the signal, and the detection performance of frequency band aliasing signals is improved through spectrum compensation, so as to realize blind detection of broadband signals and improve detection accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 A flowchart of a method for detecting a broadband signal based on scale iteration and spectrum compensation is provided;

[0047] Figure 2 It is a digital channelized structure;

[0048] Figure 3 It is a schematic diagram of the spectrum of broadband signals;

[0049] Figure 4 It is the signal detection process;

[0050] Figure 5 This is a schematic diagram of the colored noise floor estimation results;

[0051] Figure 6 It is a schematic diagram of signal detection results;

[0052] Figure 7 is the probability of the signal being correct under different signal-to-noise ratios. DETAILED DESCRIPTION

[0053] The technical solutions in the embodiments of the present invention will be described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0054] Embodiment 1

[0055] like Figure 1-7 As shown, a broadband signal detection method based on scale iteration and spectrum compensation includes the following steps:

[0056] Step A: The received signal is sampled by AD to obtain a broadband sampling signal, and the broadband signal is divided into multiple signals of appropriate bandwidth through digital channelization;

[0057] Step B: Take several segments of broadband signals and calculate the maximum power spectrum according to the power spectrum density of each segment of the signal;

[0058] Step C: using the scale iterative morphological filtering algorithm to estimate and remove the colored noise floor, and calculate the white noise power threshold;

[0059] Step D: Detect the signal with the highest amplitude in the power spectrum, estimate its center frequency, bandwidth, and the start and end positions of the signal frequency band, and eliminate the signal and complete the compensation according to the compensation algorithm. After each signal is detected, if the difference in residual signal energy before and after the two signal detections is higher than the preset energy threshold, step D is performed again with the compensated power spectrum, otherwise the signal detection is terminated.

[0060] In this embodiment, step A specifically includes the following steps:

[0061] A1: Calculate the center frequency of the real signal channelization subchannel D is the data extraction multiple, and the entire frequency band is divided into D symmetrical sub-bands of real signals;

[0062] A2: Anti-aliasing filter h LP (n) is a low-pass FIR filter, K is the number of channels, D is the decimation factor, and N and D have an integer multiple relationship, that is, K = FD, then the output of the kth subchannel in the traditional channelization structure is:

[0063]

[0064] A3: Rewrite the above formula into a polyphase filter structure, and get the output expression of the kth subchannel:

[0065]

[0066] A4: Set the center frequency ω k Substituting into the formula, the output expression of the kth subchannel is:

[0067]

[0068] In this embodiment, in step B, the received data is segmented m times, each segment is L in length, and the interval between segments is K. The extracted data can be represented in matrix form:

[0069]

[0070] For each segment of data, the power spectrum P of each segment of data is estimated by the periodogram method. i (f), namely

[0071]

[0072] Among them, w(n) is the selected window function. After finding the power spectrum of each segmented signal, the maximum value of the power spectrum is found at each frequency point, that is,

[0073]

[0074] In this embodiment, step C specifically includes the following steps:

[0075] B1: Scale B bi Minimum detection bandwidth B min To perform a morphological filter PN1(f);

[0076] B2: Scale B bi Take the value of the last filtering scale B bi-1 A morphological filter is performed once if the scale is greater than the maximum detection bandwidth B. max , then B max A morphological filter is performed for the scale, and the result of the i-th filter is PN i (f);

[0077] B3: Difference the filtering results of the i-th and i-1-th adjacent filtering times, and the difference result is ΔPN i , compare the two differential results frequency by frequency point, if ΔPN i (f)>ΔPN i-1 (f), then the scale at frequency point f is updated to B bi , otherwise the scale remains unchanged;

[0078] B4: If B bi =B max , then stop the iteration;

[0079] B5: The scale B determined iteratively at each frequency point f bi (f) Perform a morphological filter to estimate the colored noise floor;

[0080] B6: Subtract B5 from the original power spectrum to estimate the noise floor and obtain a power spectrum with a flat noise floor;

[0081] B7: Convert the power spectrum into a logarithmic power spectrum, define an appropriate number of power bands from 0 to the maximum value on the power spectrum and count the number of frequency points falling in each band. The more power bands there are, the more accurate the white noise threshold estimation is. The number of frequency points in adjacent power bands is differentiated respectively, and the upper limit of the power band with the largest differential result is used as the white noise power threshold.

[0082] In this embodiment, the step D specifically includes the following steps:

[0083] C1: Find the maximum position f in the power spectrum max As the center frequency f Ctr , at the center frequency amplitude P(fCtr )-3dB to calculate the 3dB bandwidth B 3dB , the left edge of the bandwidth is f 3dBL , the right boundary is f 3dBR ;

[0084] C2: The power spectrum is [f 3dBL ,f 3dBR ] with scale λ sm =0.1*B 3dB After smoothing and filtering, it returns to C1 and enters C3 after five cycles;

[0085] C3: After five cycles of smoothing, the bandwidth obtained is the 3dB bandwidth. 3dBL ,f 3dBR ] range according to the following formula to calculate the center frequency;

[0086]

[0087] C4: in f 3dBL To the left, f 3dBR The first minimum or zero point of the power spectrum is found on the right as the main lobe boundary of the signal, and the left boundary is f LS , the right boundary is f RS , in [f LS ,f RS ] range, the signal is reset to zero and removed, and the out-of-band slope k is calculated according to the following formula LS , k RS :

[0088]

[0089]

[0090] C5: in f LS , f RS The slope k LS , k RS To the estimated signal band [f LS ,f RS ] until the two lines intersect, and the two extended lines replace the in-band spectrum lines of the estimated signal to reduce the damage to the unestimated signal;

[0091] C6: If the energy estimated by C4 is lower than the minimum judgment energy or the bandwidth is smaller than the minimum detection bandwidth, it will be discarded. If the maximum amplitude of the residual power spectrum is not higher than the white noise threshold, the signal detection is completed.

[0092] Embodiment 2

[0093] like Figure 1-7 As shown, a broadband signal detection method based on scale iteration and spectrum compensation includes the following steps:

[0094] Step A: The received signal is sampled by AD to obtain a broadband sampling signal, and the broadband signal is divided into multiple signals of appropriate bandwidth through digital channelization;

[0095] Step B: Take several segments of broadband signals and calculate the maximum power spectrum according to the power spectrum density of each segment of the signal;

[0096] Step C: using the scale iterative morphological filtering algorithm to estimate and remove the colored noise floor, and calculate the white noise power threshold;

[0097] Step D: Detect the signal with the highest amplitude in the power spectrum, and remove the signal and complete the compensation according to the compensation algorithm. After each signal is detected, if the difference in residual signal energy before and after the two signal detections is higher than the preset energy threshold, step D is performed again with the compensated power spectrum, otherwise the signal detection ends.

[0098] In this embodiment, step A specifically includes the following steps:

[0099] A1: Calculate the center frequency of the real signal channelization subchannel D is the data extraction multiple, and the entire frequency band is divided into D symmetrical sub-bands of real signals;

[0100] A2: Anti-aliasing filter h LP (n) is a low-pass FIR filter, K is the number of channels, D is the decimation factor, and N and D have an integer multiple relationship, that is, K = FD, then the output of the kth subchannel in the traditional channelization structure is:

[0101]

[0102] A3: Rewrite the above formula into a polyphase filter structure, and get the output expression of the kth subchannel:

[0103]

[0104] A4: Set the center frequency ω k Substituting into the formula, the output expression of the kth subchannel is:

[0105]

[0106] In this embodiment, in step B, the received data is segmented m times, each segment is L in length, and the interval between segments is K. The extracted data can be represented in matrix form:

[0107]

[0108] For each segment of data, the power spectrum P of each segment of data is estimated by the periodogram method. i (f), namely

[0109]

[0110] Among them, w(n) is the selected window function. After finding the power spectrum of each segmented signal, the maximum value of the power spectrum is found at each frequency point, that is,

[0111]

[0112] In this embodiment, step C specifically includes the following steps:

[0113] B1: Scale B bi Minimum detection bandwidth B min To perform a morphological filter PN1(f);

[0114] B2: Scale B bi Take the value of the last filtering scale B bi-1 The morphological filtering is performed twice, and the result of the i-th filtering is PN i (f);

[0115] B3: Difference the filtering results of the i-th and i-1-th adjacent filtering times, and the difference result is ΔPN i , if ΔPN i (f)>ΔPN i-1 (f), then the scale at frequency point f is updated to B bi , otherwise the scale remains unchanged;

[0116] B4: If B bi =B max , then stop the iteration;

[0117] B5: The scale B determined iteratively at each frequency point f bi (f) Perform a morphological filter to estimate the colored noise floor;

[0118] B6: Subtract B5 from the original power spectrum to estimate the noise floor and obtain a power spectrum with a flat noise floor;

[0119] B7: Convert the power spectrum into a logarithmic power spectrum, define an appropriate number of power bands from 0 to the maximum value on the power spectrum and count the number of frequency points falling in each band. The more power bands there are, the more accurate the white noise threshold estimation is. The number of frequency points in adjacent power bands is differentiated respectively, and the upper limit of the power band with the largest differential result is used as the white noise power threshold.

[0120] In this embodiment, the step D specifically includes the following steps:

[0121] C1: Find the maximum position f in the power spectrum max As the center frequency f Ctr , at the center frequency amplitude P(fCtr )-3dB to calculate the 3dB bandwidth B 3dB , the left edge of the bandwidth is f 3dBL , the right boundary is f 3dBR ;

[0122] C2: The power spectrum is [f 3dBL ,f 3dBR ] with scale λ sm =0.1*B 3dB After smoothing and filtering, it returns to C1 and enters C3 after five cycles;

[0123] C3: After five cycles of smoothing, the bandwidth obtained is the 3dB bandwidth. 3dBL ,f 3dBR ] range according to the following formula to calculate the center frequency;

[0124]

[0125] C4: in f 3dBL To the left, f 3dBR Find the first minimum or zero point of the power spectrum on the right as the main lobe boundary of the signal. LS ,f RS ] range, the signal is reset to zero and removed, and the out-of-band slope k is calculated according to the following formula LS , k RS :

[0126]

[0127]

[0128] C5: in f LS , f RS The slope k LS , k RS To the estimated signal band [f LS ,f RS ] until the two lines intersect, reducing the damage to the unestimated signal;

[0129] C6: If the energy estimated by C4 is lower than the minimum judgment energy or the bandwidth is smaller than the minimum detection bandwidth, it will be discarded. If the maximum amplitude of the residual power spectrum is not higher than the white noise threshold, the signal detection is completed.

[0130] Embodiment 3

[0131] like Figure 1-7 As shown, a broadband signal detection method based on scale iteration and spectrum compensation includes the following steps:

[0132] Step A: The received signal is sampled by AD to obtain a broadband sampling signal, and the broadband signal is divided into multiple signals of appropriate bandwidth through digital channelization;

[0133] Step B: Take several segments of broadband signals and calculate the maximum power spectrum according to the power spectrum density of each segment of the signal;

[0134] Step C: using the scale iterative morphological filtering algorithm to estimate and remove the colored noise floor, and calculate the white noise power threshold;

[0135] Step D: Detect the signal with the highest amplitude in the power spectrum. If the residual signal energy difference before and after two signal detections is higher than a preset energy threshold, perform step D again with the compensated power spectrum. Otherwise, the signal detection ends.

[0136] In this embodiment, step A specifically includes the following steps:

[0137] A1: Calculate the center frequency of the real signal channelization subchannel D is the data extraction multiple, and the entire frequency band is divided into D symmetrical sub-bands of real signals;

[0138] A2: Anti-aliasing filter h LP (n) is a low-pass FIR filter, K is the number of channels, D is the decimation factor, and N and D have an integer multiple relationship, that is, K = FD, then the output of the kth subchannel in the traditional channelization structure is:

[0139]

[0140] A4: Set the center frequency ω k Substituting into the formula, the output expression of the kth subchannel is:

[0141]

[0142] In this embodiment, in step B, the received data is segmented m times, each segment is L in length, and the interval between segments is K. The extracted data can be represented in matrix form:

[0143]

[0144] For each segment of data, the power spectrum P of each segment of data is estimated by the periodogram method. i (f), namely

[0145]

[0146] Among them, w(n) is the selected window function. After finding the power spectrum of each segmented signal, the maximum value of the power spectrum is found at each frequency point, that is,

[0147]

[0148] In this embodiment, step C specifically includes the following steps:

[0149] B1: Scale B bi Minimum detection bandwidth B min To perform a morphological filter PN1(f);

[0150] B2: Scale B bi Take the value of the last filtering scale B bi-1 A morphological filter is performed once if the scale is greater than the maximum detection bandwidth B. max , then B max A morphological filter is performed for the scale, and the result of the i-th filter is PN i (f);

[0151] B3: Difference the filtering results of the i-th and i-1-th adjacent filtering times, and the difference result is ΔPN i , compare the two differential results frequency by frequency point, if ΔPN i (f)>ΔPN i-1 (f), then the scale at frequency point f is updated to B bi , otherwise the scale remains unchanged;

[0152] B4: The scale B determined iteratively at each frequency point f bi (f) Perform a morphological filter to estimate the colored noise floor;

[0153] B5: The noise floor is estimated by subtracting B5 from the original power spectrum to obtain a power spectrum with a flat noise floor;

[0154] B6: Convert the power spectrum into a logarithmic power spectrum, define an appropriate number of power bands from 0 to the maximum value on the power spectrum and count the number of frequency points falling in each band, and use the upper limit of the power band with the largest difference result as the white noise power threshold.

[0155] In this embodiment, the step D specifically includes the following steps:

[0156] C1: Find the maximum position f in the power spectrum max As the center frequency f Ctr , at the center frequency amplitude P(f Ctr )-3dB to calculate the 3dB bandwidth B 3dB , the left edge of the bandwidth is f 3dBL , the right boundary is f 3dBR ;

[0157] C2: The power spectrum is [f 3dBL ,f 3dBR ] with scale λ sm =0.1*B 3dBAfter smoothing and filtering, it returns to C1 and enters C3 after five cycles;

[0158] C3: The resulting bandwidth is the 3dB bandwidth. 3dBL ,f 3dBR ] range according to the following formula to calculate the center frequency;

[0159]

[0160] C4: in f 3dBL To the left, f 3dBR The first minimum or zero point of the power spectrum is found on the right as the main lobe boundary of the signal, and the left boundary is f LS , the right boundary is f RS , in [f LS ,f RS ] range, the signal is reset to zero and removed, and the out-of-band slope k is calculated according to the following formula LS , k RS :

[0161]

[0162]

[0163] C5: in f LS , f RS The slope k LS , k RS To the estimated signal band [f LS ,f RS ] until the two lines intersect, and the two extended lines replace the in-band spectrum lines of the estimated signal to reduce the damage to the unestimated signal;

[0164] C6: If the energy estimated by C4 is lower than the minimum judgment energy or the bandwidth is smaller than the minimum detection bandwidth, it will be discarded. If the maximum amplitude of the residual power spectrum is not higher than the white noise threshold, the signal detection is completed.

[0165] Embodiment 4

[0166] like Figure 1-7 As shown, a broadband signal detection method based on scale iteration and spectrum compensation includes the following steps:

[0167] Step A: The received signal is sampled by AD to obtain a broadband sampling signal;

[0168] Step B: Take several segments of broadband signals and calculate the maximum power spectrum according to the power spectrum density of each segment of the signal;

[0169] Step C: using the scale iterative morphological filtering algorithm to estimate and remove the colored noise floor, and calculate the white noise power threshold;

[0170] Step D: Detect the signal with the highest amplitude in the power spectrum, estimate its center frequency, bandwidth, and the start and end positions of the signal frequency band, and eliminate the signal and complete the compensation according to the compensation algorithm. After each signal is detected, if the difference in residual signal energy before and after the two signal detections is higher than the preset energy threshold, step D is performed again with the compensated power spectrum, otherwise the signal detection is terminated.

[0171] In this embodiment, step A specifically includes the following steps:

[0172] A1: Calculate the center frequency of the real signal channelization subchannel D is the data extraction multiple, and the entire frequency band is divided into D symmetrical sub-bands of real signals;

[0173] A2: Anti-aliasing filter h LP (n) is a low-pass FIR filter, K is the number of channels, D is the decimation factor, and N and D have an integer multiple relationship, that is, K = FD, then the output of the kth subchannel in the traditional channelization structure is:

[0174]

[0175] A3: Rewrite the above formula into a polyphase filter structure, and get the output expression of the kth subchannel:

[0176]

[0177] A4: Set the center frequency ω k Substituting into the formula, the output expression of the kth subchannel is:

[0178]

[0179] In this embodiment, in step B, the received data is segmented m times, each segment is L in length, and the interval between segments is K. The extracted data can be represented in matrix form:

[0180]

[0181] For each segment of data, the power spectrum P of each segment of data is estimated by the periodogram method. i (f), namely

[0182]

[0183] Among them, w(n) is the selected window function. After finding the power spectrum of each segmented signal, the maximum value of the power spectrum is found at each frequency point, that is,

[0184]

[0185] In this embodiment, step C specifically includes the following steps:

[0186] B1: Scale B bi Minimum detection bandwidth B min To perform a morphological filter PN1(f);

[0187] B2: Scale B bi Take the value of the last filtering scale B bi-1 The morphological filtering is performed twice, and the result of the i-th filtering is PN i (f);

[0188] B3: Difference the results of the two adjacent filtering times, the i-th and the i-1-th, if ΔPN i (f)>ΔPN i-1 (f), then the scale at frequency point f is updated to B bi , otherwise the scale remains unchanged;

[0189] B4: If B bi =B max , then stop the iteration;

[0190] B5: The scale B determined iteratively at each frequency point f bi (f) Perform a morphological filter to estimate the colored noise floor;

[0191] B6: Subtract B5 from the original power spectrum to estimate the noise floor and obtain a power spectrum with a flat noise floor;

[0192] B7: Convert the power spectrum into a logarithmic power spectrum. The more power bands there are, the more accurate the white noise threshold estimation is. The frequency points of adjacent power bands are differentiated respectively, and the upper limit of the power band with the largest differential result is used as the white noise power threshold.

[0193] In this embodiment, the step D specifically includes the following steps:

[0194] C1: Find the maximum position f in the power spectrum max As the center frequency f Ctr , at the center frequency amplitude P(f Ctr )-3dB to calculate the 3dB bandwidth B 3dB , the left edge of the bandwidth is f 3dBL , the right boundary is f 3dBR ;

[0195] C2: The power spectrum is [f 3dBL ,f 3dBR ] with scale λ sm =0.1*B 3dB After smoothing and filtering, it returns to C1 and enters C3 after five cycles;

[0196] C3: After five cycles of smoothing, the bandwidth obtained is the 3dB bandwidth. 3dBL ,f 3dBR ] range according to the following formula to calculate the center frequency;

[0197]

[0198] C4: in f 3dBL To the left, f 3dBR The first minimum or zero point of the power spectrum is found on the right as the main lobe boundary of the signal, and the left boundary is f LS , the right boundary is f RS , in [f LS ,f RS ] range, the signal is reset to zero and removed, and the out-of-band slope k is calculated according to the following formula LS , k RS :

[0199]

[0200]

[0201] C5: in f LS , f RS The slope k LS , k RS To the estimated signal band [f LS ,f RS ] until the two lines intersect, and the two extended lines replace the in-band spectrum lines of the estimated signal to reduce the damage to the unestimated signal;

[0202] C6: If the energy estimated by C4 is lower than the minimum judgment energy or the bandwidth is smaller than the minimum detection bandwidth, it will be discarded. If the maximum amplitude of the residual power spectrum is not higher than the white noise threshold, the signal detection is completed.

[0203] Embodiment 5

[0204] like Figure 1-7 As shown, a broadband signal detection method based on scale iteration and spectrum compensation includes the following steps:

[0205] Step A: The received signal is sampled by AD to obtain a broadband sampling signal, and the broadband signal is divided into multiple signals of appropriate bandwidth through digital channelization;

[0206] Step B: Take several segments of broadband signals and calculate the maximum power spectrum according to the power spectrum density of each segment of the signal;

[0207] Step C: using the scale iterative morphological filtering algorithm to estimate and remove the colored noise floor, and calculate the white noise power threshold;

[0208] Step D: Detect the signal with the highest amplitude in the power spectrum, estimate its center frequency, bandwidth, and the start and end positions of the signal frequency band. If the residual signal energy difference before and after the two signal detections is higher than the preset energy threshold, perform step D again with the compensated power spectrum, otherwise the signal detection ends.

[0209] In this embodiment, step A specifically includes the following steps:

[0210] A1: Calculate the center frequency of the real signal channelization subchannel The entire frequency band is divided into D symmetrical subbands of real signals;

[0211] A2: Anti-aliasing filter h LP (n) is a low-pass FIR filter, N and D have an integer multiple relationship, that is, K = FD, then the output of the kth subchannel in the traditional channelization structure is:

[0212]

[0213] A3: Rewrite the above formula into a polyphase filter structure, and get the output expression of the kth subchannel:

[0214]

[0215] A4: Set the center frequency ω k Substituting into the formula, the output expression of the kth subchannel is:

[0216]

[0217] In this embodiment, in step B, the received data is segmented m times, each segment is L in length, and the interval between segments is K. The extracted data can be represented in matrix form:

[0218]

[0219] For each segment of data, the power spectrum P of each segment of data is estimated by the periodogram method. i (f), namely

[0220]

[0221] Among them, w(n) is the selected window function. After finding the power spectrum of each segmented signal, the maximum value of the power spectrum is found at each frequency point, that is,

[0222]

[0223] In this embodiment, step C specifically includes the following steps:

[0224] B1: Scale B bi Minimum detection bandwidth B minTo perform a morphological filter PN1(f);

[0225] B2: Scale B bi Take the value of the last filtering scale B bi-1 A morphological filter is performed twice as fast as B max A morphological filter is performed for the scale, and the result of the i-th filter is PN i (f);

[0226] B3: Difference the results of the two adjacent filtering times, the i-th and the i-1-th, if ΔPN i (f)>ΔPN i-1 (f), then the scale at frequency point f is updated to B bi , otherwise the scale remains unchanged;

[0227] B4: If B bi =B max , then stop the iteration;

[0228] B5: The scale B determined iteratively at each frequency point f bi (f) Perform a morphological filter to estimate the colored noise floor;

[0229] B6: Subtract B5 from the original power spectrum to estimate the noise floor and obtain a power spectrum with a flat noise floor;

[0230] B7: Define an appropriate number of power bands from 0 to the maximum value on the power spectrum and count the frequency points falling in each band. The more accurate the white noise threshold estimation is, the upper limit of the power band with the largest differential result is used as the white noise power threshold.

[0231] In this embodiment, the step D specifically includes the following steps:

[0232] C1: Find the maximum position f in the power spectrum max As the center frequency f Ctr , at the center frequency amplitude P(f Ctr )-3dB to calculate the 3dB bandwidth B 3dB , the left edge of the bandwidth is f 3dBL , the right boundary is f 3dBR ;

[0233] C2: The power spectrum is [f 3dBL ,f 3dBR ] with scale λ sm =0.1*B 3dB After smoothing and filtering, it returns to C1 and enters C3 after five cycles;

[0234] C3: After five cycles of smoothing, the bandwidth obtained is the 3dB bandwidth. 3dBL ,f 3dBR] range according to the following formula to calculate the center frequency;

[0235]

[0236] C4: in f 3dBL To the left, f 3dBR The first minimum or zero point of the power spectrum is found on the right as the main lobe boundary of the signal, and the left boundary is f LS , the right boundary is f RS , in [f LS ,f RS ] range, the signal is reset to zero and removed, and the out-of-band slope k is calculated according to the following formula LS , k RS :

[0237]

[0238]

[0239] C5: in f LS , f RS The slope k LS , k RS To the estimated signal band [f LS ,f RS ] until the two lines intersect, reducing the damage to the unestimated signal;

[0240] C6: If the energy estimated by C4 is lower than the minimum judgment energy or the bandwidth is smaller than the minimum detection bandwidth, it will be discarded. If the maximum amplitude of the residual power spectrum is not higher than the white noise threshold, the signal detection is completed.

[0241] Therefore, the signal detection process is as follows Figure 4 shown.

[0242] The algorithm performance simulation results are as follows Figure 5-Figure 7 As shown in the figure, the simulation parameters are set as follows: the signal sampling rate is 8MHz, the symbol rate is 2MBaud, and the modulation mode is QPSK. The sampling time is set to 25ms, there are 3 TDMA signals, and the burst duration of each signal is 0.36ms. The channel environment is additive white Gaussian noise.

[0243] To verify the accuracy of the colored noise floor estimation of this patent, Figure 5 The results of the estimation of the colored noise floor of a broadband power spectrum and the effect after removing the noise floor are given. The signal bandwidth is 37.5MHz, the center frequency is 963MHz, and the sampling rate is 75MHz. It can be seen that the noise floor fits closely with the signal fluctuations. After removing the colored noise, the power spectrum noise floor is flat, only white noise remains, and the number of signals does not change, indicating that the algorithm has good noise floor estimation performance and is suitable for use under fluctuating noise floor conditions.

[0244] To verify the multi-signal detection performance of this patent, Figure 6 The detection results of a broadband power spectrum signal are given. It can be seen that both strong and weak signals can be detected, and when the signals overlap, the strong signal does not affect the detection of the weak signal.

[0245] To further verify the stability of the algorithm performance in this paper, the experimental conditions are set as follows: there are 25 sub-signals in the broadband signal, the modulation modes include BPSK, QPSK, 8PSK, 8QAM, 16QAM, the signal bandwidth range is 2k~5MHz, the Gaussian white noise channel, and the signal-to-noise ratio is -14~6dB. Since the energy detection algorithm, the gradient double threshold algorithm and the power spectrum cancellation method are all greatly affected by the uneven noise floor, the power spectrum noise floor is a flat noise floor. Figure 7 It can be seen that the correct detection probability of the algorithm in this paper is always much higher than that of the classic energy detection algorithm. When the signal-to-noise ratio is extremely low, it is not much different from the gradient double threshold algorithm and the power spectrum cancellation method. It is higher than the three algorithms starting at -10dB, and the detection probability is higher than 90% when SNR≥0dB.

[0246] The specific calculation process of the broadband signal detection and measurement method system based on scale iteration and spectrum compensation can be found in the above embodiment, and the embodiment of the present invention will not be described in detail here.

[0247] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein, but these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A broadband signal detection and measurement method based on scale iteration and spectrum compensation, characterized in that: The steps include: Step A: The received signal is sampled by AD to obtain a broadband sampling signal, and the broadband signal is divided into multiple signals of appropriate bandwidth by digital channelization; Step B: Take several segments of broadband signals and calculate the maximum power spectrum according to the power spectrum density of each segment of the signal; Step C: using the scale iterative morphological filtering algorithm to estimate and remove the colored noise floor, and calculate the white noise power threshold; Step D: Detect the signal with the highest amplitude in the power spectrum, estimate its center frequency, bandwidth, and the start and end positions of the signal band, and remove the signal and complete the compensation according to the compensation algorithm. After each signal is detected, if the residual signal energy difference before and after the two signal detections is higher than the preset energy threshold, step D is performed again with the compensated power spectrum, otherwise the signal detection ends; The step C specifically comprises the following steps: B1: Scale B bi Minimum detection bandwidth B min To perform a morphological filter PN1(f); B2: Scale B bi Take the value of the last filtering scale B bi-1 A morphological filter is performed once if the scale is greater than the maximum detection bandwidth B. max , then B max A morphological filter is performed for the scale, and the result of the i-th filter is PN i (f); B3: Difference the filtering results of the i-th and i-1-th adjacent filtering times, and the difference result is ΔPN i , compare the two differential results frequency by frequency point, if ΔPN i (f)>ΔPN i-1 (f), then the scale at frequency point f is updated to B bi , otherwise the scale remains unchanged; B4: If B bi =B max , then stop the iteration; B5: The scale B determined iteratively at each frequency point f bi (f) Perform a morphological filter to estimate the colored noise floor; B6: Subtract B5 from the original power spectrum to estimate the noise floor and obtain a power spectrum with a flat noise floor; B7: Convert the power spectrum into a logarithmic power spectrum, define an appropriate number of power bands from 0 to the maximum value on the power spectrum and count the number of frequency points falling in each band. The more power bands there are, the more accurate the white noise threshold estimation is. The number of frequency points of adjacent power bands is differentiated respectively, and the upper limit of the power band with the largest differential result is used as the white noise power threshold; The step D specifically includes the following steps: C1: Find the maximum position f in the power spectrum max As the center frequency f Ctr , at the center frequency amplitude P(f Ctr )-3dB to calculate the 3dB bandwidth B 3dB , the left edge of the bandwidth is f 3dBL , the right boundary is f 3dBR ; C2: The power spectrum is [f 3dBL ,f 3dBR ] with scale λ sm =0.1*B 3dB After smoothing and filtering, it returns to C1 and enters C3 after five cycles; C3: After five cycles of smoothing, the bandwidth obtained is the 3dB bandwidth. 3dBL ,f 3dBR ] range according to the following formula to calculate the center frequency; C4: in f 3dBL To the left, f 3dBR The first minimum or zero point of the power spectrum is found on the right as the main lobe boundary of the signal, and the left boundary is f LS , the right boundary is f RS , in [f LS ,f RS ] range, the signal is reset to zero and removed, and the out-of-band slope k is calculated according to the following formula LS , k RS : C5: in f LS , f RS The slope k LS , k RS To the estimated signal band [f LS ,f RS ] until the two lines intersect, and the two extended lines replace the in-band spectrum lines of the estimated signal to reduce the damage to the unestimated signal; C6: If the energy estimated by C4 is lower than the minimum judgment energy or the bandwidth is smaller than the minimum detection bandwidth, it will be discarded. If the maximum amplitude of the residual power spectrum is not higher than the white noise threshold, the signal detection is completed.

2. A broadband signal detection and measurement method based on scale iteration and spectrum compensation according to claim 1, characterized in that: By utilizing the bandwidth screening characteristics of morphological filtering, morphological filtering iteration is performed on the power spectrum to make the noise floor estimation more accurate. Smoothing iteration is used to overcome the error in signal parameter estimation caused by strong pulses and white noise in the signal. Spectral compensation is used to improve the detection performance of frequency band aliasing signals.

3. The method for detecting and measuring a broadband signal based on scale iteration and spectrum compensation according to claim 2, characterized in that: The step A specifically comprises the following steps: A1: Calculate the center frequency of the real signal channelization subchannel D is the data extraction multiple, and the entire frequency band is divided into D symmetrical sub-bands of real signals; A2: Anti-aliasing filter h LP (n) is a low-pass FIR filter, K is the number of channels, D is the decimation factor, and N and D have an integer multiple relationship, that is, K = FD, then the output of the kth subchannel in the traditional channelization structure is: A3: Rewrite the above formula into a polyphase filter structure, and get the output expression of the kth subchannel: A4: Set the center frequency ω k Substituting into the formula, the output expression of the kth subchannel is:

4. The method for detecting and measuring a broadband signal based on scale iteration and spectrum compensation according to claim 3, characterized in that: In step B, the received data is segmented m times, each segment is L in length, and the interval between segments is K. The extracted data can be represented in matrix form: For each segment of data, the power spectrum P of each segment of data is estimated by the periodogram method. i (f), namely Among them, w(n) is the selected window function. After finding the power spectrum of each segmented signal, the maximum value of the power spectrum is found at each frequency point, that is,

Images