Multi-carrier frequency point extraction method for high-speed frequency hopping signal

Abstract

The invention discloses a multi-carrier frequency point extraction method for a high-speed frequency hopping signal, and relates to the field of signal processing. The method comprises the following steps: calculating a time-frequency power spectrum of a received signal; generating a power local extreme value matrix and a corresponding frequency point information matrix based on the time-frequency power spectrum; generating a local strongest power vector and a corresponding frequency vector; differentiating the frequency vector, and positioning a frequency abrupt change position; estimating a frequency hopping interval and generating a serial number corresponding to a real frequency hopping moment; performing straight line fitting based on the frequency abrupt change position and the serial number corresponding to the real frequency hopping moment, and calculating a frequency hopping moment estimation result; and averaging the frequency points of the same carrier at multiple moments in the frequency point information matrix based on the frequency hopping moment estimation result to obtain the extracted multi-carrier frequency points. According to the method, a more accurate frequency hopping time estimation result can be obtained, extraction of the multi-carrier frequency points of the high-speed frequency hopping signal can be realized based on the power local extreme value matrix and the frequency hopping time estimation result, and errors are small.

Description

Technical Field

[0001] The present invention relates to the field of signal processing, and particularly to a method for extracting multi-carrier frequency points of high-speed frequency-hopping signals. Background Art

[0002] Frequency-hopping communication is a typical modern communication technology. When conducting research on frequency-hopping communication parameter estimation, extracting the characteristic parameters of intercepted frequency-hopping signals is the key to obtaining useful intelligence and achieving interference.

[0003] Generally speaking, frequency-hopping communication is divided into high-speed frequency-hopping and low-speed frequency-hopping according to the frequency-hopping rate per unit time. The present invention focuses on high-speed frequency-hopping with a frequency-hopping rate above 1000 hops per second, and its conclusions are equally applicable to low-speed frequency-hopping. In addition, there are often multiple users in frequency-hopping communication, that is, information is transmitted through multiple carrier frequency points within a unit time. The paper titled "A Blind Source Separation Algorithm for Multiple Frequency-Hopping Signals Based on Time-Frequency Analysis" improves the extraction of the time-frequency ridge line, and then obtains the minimum hop period by detecting mutation points through wavelet transform to separate multiple frequency-hopping signals one by one and achieve parameter estimation, but does not further estimate each carrier frequency point. The paper titled "Blind Estimation Algorithm of Subcarrier Frequencies in Multi-Carrier CDMA and Its Performance Analysis" proposes a method for estimating the subcarrier frequencies of multi-carrier CDMA signals using high-order cyclic cumulants, but has requirements for the signal length and cannot process signals in the ms - us level. The paper titled "Research on an Improved MSK Carrier Estimation Method Based on Generalized Cyclic Correlation Entropy" uses the Sigmoid function to improve the generalized Gaussian kernel function in the generalized cyclic correlation entropy algorithm, derives the SGCCE function of MSK signals, and retrieves the position of the maximum spectral peak in the positive half-axis to achieve carrier frequency estimation. However, the algorithm is extremely complex, not conducive to engineering implementation, and is only applicable to MSK signals. Summary of the Invention

[0004] In view of this, the present invention proposes a method for extracting multi-carrier frequency points of high-speed frequency-hopping signals. This method can obtain a more accurate estimation result of the frequency-hopping moment. Based on the power local maximum matrix and the frequency-hopping moment estimation result, the extraction of multi-carrier frequency points of high-speed frequency-hopping signals can be realized with small errors.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0006] A method for extracting multi-carrier frequency points of high-speed frequency-hopping signals, comprising the following steps:

[0007] Step 1, calculating the time-frequency power spectrum of the received signal;

[0008] Step 2, generating a power local maximum matrix and its corresponding frequency point information matrix based on the time-frequency power spectrum of the received signal;

[0009] Step 3: Generate the local strongest power vector and the corresponding frequency vector;

[0010] Step 4: Differentiate the frequency vector and set a certain threshold to locate the frequency mutation position;

[0011] Step 5: Estimate the frequency hopping interval and generate the sequence numbers corresponding to the true frequency hopping moments;

[0012] Step 6: Perform linear fitting based on the frequency mutation position and the sequence numbers corresponding to the true frequency hopping moments, and calculate the estimated results of the frequency hopping moments;

[0013] Step 7: Based on the estimated results of the frequency hopping moments, average the frequency points at multiple moments of the same carrier in the frequency point information matrix obtained in Step 2. The average value is the extracted multi-carrier frequency points, and the extraction of the multi-carrier frequency points of the high-speed frequency hopping signal is completed.

[0014] Furthermore, the specific method of Step 1 is as follows:

[0015] Step 101: Down-convert, filter, and down-sample the frequency hopping signal to obtain the digital signal sequence x(k), where k = 0, 1, 2......, K-1, and K is the length of the digital signal sequence x(k); Use the Blackman window b(·) for x(k) to obtain a narrower main lobe width and a higher main side lobe ratio, and process it through the short-time Fourier transform algorithm to obtain X(m,n), where n indicates the time-domain sampling points and m indicates the frequency sampling points. Specifically:

[0016]

[0017] where m = 0, 1, 2,......, M-1; n = 0, 1, 2,......, N-1; M is the window length of the Blackman window b(·); N is the total number of time-domain sampling points, represents rounding down, and L is the window sliding interval of the Blackman window b(·); F d is the frequency-domain sampling interval, and F d = F s / M, and F s is the sampling rate of the digital signal sequence x(k); j is the imaginary unit; the superscript * represents conjugate;

[0018] Step 102: Calculate the corresponding time-frequency power spectrum Y(m,n) according to the short-time Fourier transform result X(m,n):

[0019]

[0020] Furthermore, the specific method of Step 2 is as follows:

[0021] Step 201, set the threshold P th , roughly estimate the bandwidth B, the power local maximum matrix P, and the frequency point information matrix F; where the sizes of both P and F are G×N, that is, the number of user carriers × the total number of time-domain sampling points, and their initial values are set to all 0;

[0022] Step 202, set n 0 to represent the time-domain pointer, and let n 0 = 0;

[0023] Step 203, set g 0 to represent the user carrier pointer, and let g 0 = 0;

[0024] Step 204, calculate the power maximum value 0 of the user carrier g 0 and the corresponding frequency sampling point at time n If let and then execute Step 205; otherwise, let P(g 0 , n 0 ) = 0, F(g 0 , n 0 ) = 0, and let g 0 = G - 1 and then execute Step 205;

[0025] Step 205, determine whether g 0 = G - 1 holds. If g 0 = G - 1 holds, then continue to determine whether n 0 = N - 1 holds. If n 0 = N - 1 holds, then execute Step 3. If n 0 = N - 1 does not hold, then let n 0 = n 0 + 1 and then execute Step 203; if g 0 = G - 1 does not hold, then set the time-frequency power spectrum within the roughly estimated bandwidth B centered on the frequency sampling point to 0, and let g 0 = g 0 + 1, and execute Step 204.

[0026] Furthermore, the specific method of the said Step 3 is:

[0027] Step 301, let n 1 = 1, set the local strongest power vector P m with a length of N, set the local strongest frequency vector F m with a length of N, and let P m (0) = P(0, 0), F m(0) = F(0, 0), set the power difference threshold as dP th , the frequency difference threshold as dF th , the maximum power difference threshold as dP mth ;

[0028] Step 302, determine whether the maximum value of the elements in the (n 1 + 1)-th column of the power local maximum matrix P is 0. If it is 0, then replace the element values in the n 1 -th column of P with the element values in the (n 1 + 1)-th column of P, and execute Step 303; otherwise, directly execute Step 303;

[0029] Step 303, calculate the power difference ΔP = |P(0, n 1 ) - P m (n 1 - 1)|,

[0030] the frequency difference ΔF = |F(0, n 1 ) - F m (n 1 - 1)|, where |*| represents taking the absolute value. If both the power difference and the frequency difference are within the threshold range, that is, satisfying ΔP < dP th , and ΔF < dF th , then P m (n 1 ) = P(0, n 1 ), F m (n 1 ) = F(0, n 1 ), and execute Step 306; otherwise, execute Step 304;

[0031] Step 304, let the user carrier pointer g 1 = 1;

[0032] Step 305, calculate the power difference ΔP = |P(g 1 , n 1 ) - P m (n 1 - 1)|,

[0033] the frequency difference ΔF = |F(g 1 , n 1 ) - F m (n 1 - 1)|;

[0034] Let ΔP m = |P(g 1 , n 1 ) - P(0, n 1 - 1)|, and determine whether both ΔP < dP th, and ΔF < dF th , and ΔP m < dP mth , if the above 3 inequalities are satisfied simultaneously, then a real jump has not occurred at this time. Let P m (n 1 ) = P(g 1 , n 1 ), F m (n 1 ) = F(g 1 , n 1 ), and execute step 306;

[0035] If the above 3 inequalities are not satisfied simultaneously, determine whether g 1 = G - 1 holds. If g 1 = G - 1 holds, re - store the maximum value in the local maximum vector, that is, P m (n 1 ) = P(0, n 1 ), F m (n 1 ) = F(0, n 1 ), and execute step 306;

[0036] If g 1 = G - 1 does not hold, let g 1 = g 1 + 1, and execute step 305;

[0037] Step 306, determine whether n 1 = N - 1 holds. If n 1 = N - 1 does not hold, let n 1 = n 1 + 1, and repeat step 302; otherwise, complete the calculation and execute step 4.

[0038] Further, the specific manner of step 4 is as follows:

[0039] Step 401, calculate the first - order difference result ΔF of F m , that is, ΔF(n 2 ) = |F m (n 2 + 1)-F m (n 2 )|, n 2 = 0, 1, 2, …, N - 2; and set the results in ΔF that are less than dF th to 0;

[0040] Step 402, extract all the time - domain pointer numbers corresponding to the non - zero values in ΔF and store them in the time - domain pointer number vector N n0 with a length of J, where J represents the number of non - zero values in ΔF; the Nn0 The time-domain pointer number therein is the frequency mutation position;

[0041] Step 403: Perform a first-order difference operation on the time-domain pointer number vector N n0 to obtain its difference result ΔN n0 :

[0042] ΔN n0 (j) = |N n0 (j + 1) - N n0 (j)|, where j = 0, 1, 2, …, J - 2.

[0043] Further, the specific manner of the said Step 5 is as follows:

[0044] Step 501: Sort the elements in the difference result ΔN n0 in ascending order to obtain the sorted median. Calculate the average value of the elements in ΔN n0 whose difference from the sorted median is within ±10%, and denote the average value as the frequency hopping interval ΔN;

[0045] Step 502: Calculate the sequence number Z of the actual frequency hopping moment: Set the initial iteration value Z(0) = 0, and the value of each subsequent iteration is:

[0046] Z(j 1 ) = <(N n0 (j 1 ) - N n0 (j 1 - 1)) / ΔN> + Z(j 1 - 1), where j 1 = 1, 2, 3, …, J - 1;

[0047] where <·> is rounding to the nearest integer.

[0048] Further, the specific manner of the said Step 6 is as follows:

[0049] Step 601: Use the least squares fitting algorithm to fit the functional relationship between the sequence number Z of the actual frequency hopping moment and the time-domain pointer number vector N n0 into a linear function;

[0050] Step 602: Substitute the sequence number Z of the actual frequency hopping moment into the fitted linear function in turn to obtain the estimation result N Jn of each frequency hopping moment.

[0051] Further, the specific manner of the said Step 7 is as follows:

[0052] Step 701: First, let the frequency hopping moment index j 2 = 0, and the user carrier pointer g 2 = 0;

[0053] Step 702: Extract all the frequency points F(g 2 ,n 3 ) of the current user's carrier before frequency hopping in the frequency point information matrix F, and calculate their average value FG(g 2 ,n 3 );

[0054] When j 2 = 0, n 3 = 0, 1, 2,..., N Jn (Z(0));

[0055]

[0056] When j 2 = 1, 2, 3,…, J - 1, n 3 = N Jn (Z(j 2 )) - 1), N Jn (Z(j 2 )) - 1) + 1, N Jn (Z(j 2 )) - 1) + 2,..., N Jn (Z(j 2 ));

[0057]

[0058] F G (g 2 ,n 3 ) is the frequency point of the user's carrier g 2 in the (j 2 + 1)-th frequency hopping signal segment;

[0059] Step 703: Judge whether g 2 = G - 1 holds. If g 2 = G - 1 holds, then set g 2 = 0 and execute Step 703; if g 2 = G - 1 does not hold, then set g 2 = g 2 + 1 and return to execute Step 702;

[0060] Step 703: Judge whether j 2 = J - 1 holds. If j 2 = J - 1 holds, then complete the extraction of the multi - carrier frequency points of the high - speed frequency hopping signal; if j 2 = J - 1 does not hold, then set j 2 = j 2 + 1 and return to execute Step 702.

[0061] Due to the adoption of the above technical solution, the beneficial effects of the present invention compared with the prior art are as follows:

[0062] Based on the time-frequency power spectrum calculated by the short-time Fourier transform (STFT), the present invention calculates the power local extreme value matrix and the frequency local extreme value matrix through an improved local maximum estimation algorithm, and further calculates the local strongest power vector and the local strongest frequency vector. The time-domain mutation moment, that is, the frequency hopping moment, is obtained through the first-order difference processing of the local strongest frequency vector. After correcting its order, fitting is performed, and then a more accurate frequency hopping moment estimation result is obtained. Based on the power local extreme value matrix and the frequency hopping moment estimation result, the extraction of multi-carrier frequency points of high-speed frequency hopping signals can be realized with small errors. Brief Description of the Drawings

[0063] Figure 1 It is a flowchart of a method for extracting multi-carrier frequency points of a high-speed frequency hopping signal in an embodiment of the present invention.

[0064] Figure 2 It is the time-frequency power spectrum in a method for extracting multi-carrier frequency points of a high-speed frequency hopping signal in an embodiment of the present invention.

[0065] Figure 3 It is a schematic diagram of the power local extreme value matrix of a method for extracting multi-carrier frequency points of a high-speed frequency hopping signal in an embodiment of the present invention.

[0066] Figure 4 It is a schematic diagram of the frequency information matrix of a method for extracting multi-carrier frequency points of a high-speed frequency hopping signal in an embodiment of the present invention.

[0067] Figure 5 It is a plot of the local strongest power vector of a method for extracting multi-carrier frequency points of a high-speed frequency hopping signal in an embodiment of the present invention.

[0068] Figure 6 It is a plot of the local strongest frequency vector of a method for extracting multi-carrier frequency points of a high-speed frequency hopping signal in an embodiment of the present invention.

[0069] Figure 7 It is a plot of the first-order difference of the local strongest frequency vector of a method for extracting multi-carrier frequency points of a high-speed frequency hopping signal in an embodiment of the present invention.

[0070] Figure 8 It is a plot of the non-zero time-domain pointer numbers of a method for extracting multi-carrier frequency points of a high-speed frequency hopping signal in an embodiment of the present invention.

[0071] Figure 9 It is a plot of the corrected non-zero time-domain pointer numbers of a method for extracting multi-carrier frequency points of a high-speed frequency hopping signal in an embodiment of the present invention.

[0072] Figure 10The straight-line fitting result of a method for extracting multi-carrier frequency points of a high-speed frequency-hopping signal and the plot of the time-domain pointer at each frequency-hopping moment in the embodiment of the present invention.

[0073] Figure 11 The estimation error statistics of 200 multi-carrier frequency point estimation experiments in the embodiment of the present invention. Detailed implementation manners

[0074] The following further describes the content of the present invention in conjunction with the accompanying drawings and specific embodiments.

[0075] A method for extracting multi-carrier frequency points of a high-speed frequency-hopping signal, as Figure 1 shown, includes the following steps:

[0076] Step 1, calculate the time-frequency power spectrum of the received signal;

[0077] Step 2, based on the time-frequency power spectrum of the received signal, generate a power local maximum matrix and its corresponding frequency point information matrix;

[0078] Step 3, generate a local strongest power vector and its corresponding frequency vector;

[0079] Step 4, perform a difference on the frequency vector and set a certain threshold to locate the frequency mutation position;

[0080] Step 5, estimate the frequency-hopping interval and generate the sequence numbers corresponding to the true frequency-hopping moments;

[0081] Step 6, perform a straight-line fitting based on the frequency mutation position and its corresponding sequence numbers of the true frequency-hopping moments, and calculate the estimated result of the frequency-hopping moment;

[0082] Step 7, based on the estimated result of the frequency-hopping moment, calculate the average of the frequency points of the same carrier at multiple moments in the frequency point information matrix obtained in Step 2, and the average value is the extracted multi-carrier frequency point, thus completing the extraction of the multi-carrier frequency points of the high-speed frequency-hopping signal.

[0083] Further, the specific manner of Step 1 is:

[0084] Step 101, perform down-conversion, filtering, and down-sampling on the frequency-hopping signal to obtain a digital signal sequence x(k), where k = 0, 1, 2......, K - 1, and K is the length of the digital signal sequence x(k); use the Blackman window b(·) for x(k) to obtain a narrower main lobe width and a higher main-to-side lobe ratio, and process it through the short-time Fourier transform algorithm to obtain X(m, n), where n indicates the time-domain sampling points and m indicates the frequency sampling points. Specifically:

[0085]

[0086] where m = 0, 1, 2,......, M - 1; n = 0, 1, 2,......, N - 1; M is the window length of the Blackman window b(·); N is the total number of time-domain sampling points, denotes rounding downwards, L is the window sliding interval of the Blackman window b(·); F d is the frequency-domain sampling interval, F d = F s / M, F s is the sampling rate of the digital signal sequence x(k); j is the imaginary unit; the superscript * denotes conjugate;

[0087] Specifically, in this embodiment, the hopping speed of the frequency-hopping signal is 16000 hops / s, the signal bandwidth is 10 MHz, the sampling rate is F s = 2.4 GHz, K = 6000000, M = 4096, L = 128, F = F s / M = 58.59 kHz,

[0088] Step 102, as Figure 2 shown, calculate the corresponding time-frequency power spectrum Y(m, n) according to the short-time Fourier transform result X(m, n):

[0089]

[0090] Furthermore, as Figure 3 shown, as Figure 4 shown, the specific manner of step 2 is:

[0091] Step 201, set the threshold P th , the rough estimated bandwidth B, the power local maximum value matrix P and the frequency point information matrix F; where the sizes of P and F are both G×N, that is, the number of user carriers × the total number of time-domain sampling points, and their initial values are set to all 0;

[0092] Step 202, set n 0 to represent the time-domain pointer, and let n 0 = 0;

[0093] Step 203, set g 0 to represent the user carrier pointer, and let g 0 = 0;

[0094] Step 204, calculate the power maximum value 0 at the time of user carrier g 0 and n in the time-frequency power spectrum and its corresponding frequency sampling point If Let After that, step 205 is executed; otherwise, let P(g 0 , n 0 ) = 0, F(g 0 , n 0 ) = 0, and let g 0 = G - 1, then step 205 is executed;

[0095] Step 205: Determine whether g 0 = G - 1 holds. If g 0 = G - 1 holds, then continue to determine whether n 0 = N - 1 holds. If n 0 = N - 1 holds, then step 3 is executed. If n 0 = N - 1 does not hold, then let n 0 = n 0 + 1, and then step 203 is executed. If g 0 = G - 1 does not hold, then set the time - frequency power spectrum within the coarse - estimation bandwidth B centered on the frequency sampling point to 0, then let g 0 = g 0 + 1, and step 204 is executed.

[0096] Furthermore, as shown in Figure 5 , as shown in Figure 6 , the specific manner of the said step 3 is as follows:

[0097] Step 301: Let n 1 = 1, set the local strongest power vector P m of length N, set the local strongest frequency vector F m of length N, and let P m (0) = P(0, 0), F m (0) = F(0, 0), set the power - difference threshold to dP th , the frequency - difference threshold to dF th , and the highest power - difference threshold to dP mth ;

[0098] Specifically, in this embodiment, set the power - difference threshold dP th = 1.06, the frequency - difference threshold dF th = 17, and the highest power - difference threshold dP mth = 20;

[0099] Step 302: Determine whether the maximum value of the elements in the (n 1 + 1)-th column of the power local - maximum matrix P is 0. If it is 0, then replace the element values in the n 1 -th column of P with the element values in the n 1 + 1)-th column of P, and execute step 303; otherwise, directly execute step 303;

[0100] Step 303, calculate the power difference ΔP = |P(0, n 1 ) - P m (n 1 - 1)|,

[0101] and the frequency difference ΔF = |F(0, n 1 ) - F m (n 1 - 1)|, where |*| represents taking the absolute value. If both the power difference and the frequency difference are within the threshold range, that is, satisfying ΔP < dP th , and ΔF < dF th , then P m (n 1 ) = P(0, n 1 ), F m (n 1 ) = F(0, n 1 ), and execute Step 306; otherwise, execute Step 304;

[0102] Step 304, set the user carrier pointer g 1 = 1;

[0103] Step 305, calculate the power difference ΔP = |P(g 1 , n 1 ) - P m (n 1 - 1)|,

[0104] and the frequency difference ΔF = |F(g 1 , n 1 ) - F m (n 1 - 1)|;

[0105] Let ΔP m = |P(g 1 , n 1 ) - P(0, n 1 - 1)|, and determine whether it simultaneously satisfies ΔP < dP th , and ΔF < dF th , and ΔP m < dP mth . If all three of the above inequalities are satisfied simultaneously, then no actual jump has occurred at this time. Let P m (n 1 ) = P(g 1 , n 1 ), F m (n 1 ) = F(g 1 , n 1 ), and execute Step 306;

[0106] If the above three inequalities are not satisfied simultaneously, determine whether g 1 = G - 1 holds. If g 1 = G - 1 holds, store the maximum value in the local maximum vector again, that is, P m (n 1 ) = P(0, n 1 ), F m (n 1 ) = F(0, n 1 ), and execute step 306;

[0107] If g 1 = G - 1 does not hold, let g 1 = g 1 + 1, and execute step 305;

[0108] Step 306, determine whether n 1 = N - 1 holds. If n 1 = N - 1 does not hold, let n 1 = n 1 + 1, and repeat step 302; otherwise, complete the calculation and execute step 4.

[0109] Furthermore, the specific manner of step 4 is as follows:

[0110] Step 401, as Figure 7 shown, calculate the first-order difference result ΔF of F m , that is, ΔF(n 2 ) = |F m (n 2 + 1) - F m (n 2 )|, n 2 = 0, 1, 2,..., N - 2; and set the results in ΔF that are less than dF th to 0;

[0111] Step 402, as Figure 8 shown, extract all the time-domain pointer numbers corresponding to the non-zero values in ΔF and store them in the time-domain pointer number vector N n0 with a length of J, where J represents the number of non-zero values in ΔF; the time-domain pointer numbers in N n0 are the frequency mutation positions;

[0112] Step 403, perform a first-order difference operation on the time-domain pointer number vector N n0 to obtain its difference result ΔN n0 :

[0113] ΔN n0 (j) = |N n0 (j + 1) - N n0(j)|, where j = 0, 1, 2, …, J - 2.

[0114] Furthermore, the specific manner of step 5 is as follows:

[0115] Step 501: Sort the elements in the difference result ΔN n0 in ascending order to obtain the sorted median. Calculate the average value of the elements in ΔN n0 whose difference from the sorted median is within ±10%, and denote the average value as the hopping frequency interval ΔN.

[0116] Step 502: Calculate the sequence number Z of the actual hopping frequency moment: Set the initial iteration value Z(0) = 0, and the subsequent iteration value is:

[0117] Z(j 1 ) = <(N n0 (j 1 ) - N n0 (j 1 - 1)) / ΔN> + Z(j 1 - 1), where j 1 = 1, 2, 3, …, J - 1;

[0118] where <·> represents rounding to the nearest integer. As Figure 9 shown, it is the functional relationship between the sequence number Z of the actual hopping frequency moment and the time domain pointer number vector N n0 .

[0119] Furthermore, the specific manner of step 6 is as follows:

[0120] Step 601: Use the least squares fitting algorithm to fit the functional relationship between the sequence number Z of the actual hopping frequency moment and the time domain pointer number vector N n0 into a linear function.

[0121] Step 602: As Figure 10 shown, substitute the sequence number Z of the actual hopping frequency moment into the fitted linear function in turn to obtain the estimated result N Jn of each hopping frequency moment.

[0122] Furthermore, the specific manner of step 7 is as follows:

[0123] Step 701: First, let the hopping frequency moment index j 2 = 0, and the user carrier pointer g 2 = 0;

[0124] Step 702: Extract all the frequency points F(g 2 , n 3 ) of the current user carrier before hopping occurs from the frequency point information matrix F, and calculate their average value F G (g2 , n 3 );

[0125] When j 2 = 0, n 3 = 0, 1, 2,..., N Jn (Z(0));

[0126]

[0127] When j 2 = 1, 2, 3,…, J - 1, n 3 = N Jn (Z(j 2 ) - 1), N Jn (Z(j 2 ) - 1) + 1, N Jn (Z(j 2 ) - 1) + 2,..., N Jn (Z(j 2 ));

[0128]

[0129] F G (g 2 , n 3 ) is the frequency point of the user carrier g 2 in the (j 2 + 1)-th hopping signal segment;

[0130] Step 703, determine whether g 2 = G - 1 holds. If g 2 = G - 1 holds, then set g 2 = 0 and execute Step 703; if g 2 = G - 1 does not hold, then set g 2 = g 2 + 1 and return to execute Step 702;

[0131] Step 703, determine whether j 2 = J - 1 holds. If j 2 = J - 1 holds, then complete the extraction of the multi - carrier frequency points of the high - speed hopping signal; if j 2 = J - 1 does not hold, then set j 2 = j 2 + 1 and return to execute Step 702.

[0132] Specifically, in this embodiment, 200 multi - carrier frequency point estimation experiments are repeated, and the average value of the frequency point estimation of each carrier in each hopping period is statistically calculated, as Figure 11As shown, the average error of 200 times is only 0.0131 MHz, which is 0.0131 of the signal bandwidth. The test proves that the frequency estimation error of this method is small.

[0133] In summary, the present invention obtains a more accurate estimation result of the frequency hopping moment. Based on the power local maximum matrix and the frequency hopping moment estimation result, the extraction of multi-carrier frequency points of high-speed frequency hopping signals can be realized with small errors.

[0134] Those skilled in the art will realize that the described embodiments are to help readers understand the principles of the present invention, and it should be understood that the protection scope of the present invention is not limited to the described embodiments. For those skilled in the art, various changes and modifications can be made to the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the scope of the claims of the present invention.

Claims

1. A method for extracting multi-carrier frequencies of high-speed frequency hopping signals, characterized in that: The following steps are involved: Step 1, calculating the time-frequency power spectrum of the received signal; Step 2, based on the time-frequency power spectrum of the received signal, generate a power local maximum value matrix and its corresponding frequency point information matrix; Step 3, generating the local strongest power vector and the corresponding frequency vector; Step 4: Differentiate the frequency vector and set a certain threshold to locate the frequency mutation position; Step 5, estimating the frequency hopping interval and generating a sequence number corresponding to the actual frequency hopping moment; Step 6, performing straight line fitting based on the frequency mutation position and the sequence number of the corresponding real frequency hopping moment, and calculating the frequency hopping moment estimation result; Step 7, based on the frequency hopping time estimation result, average the frequencies of the same carrier at multiple times in the frequency information matrix obtained in step 2, and the average value is the extracted multi-carrier frequency, thereby completing the multi-carrier frequency extraction of the high-speed frequency hopping signal.

2. The method for extracting multi-carrier frequencies of a high-speed frequency hopping signal according to claim 1, characterized in that: The specific method of step 1 is: Step 101, down-convert, filter and down-sample the frequency hopping signal to obtain a digital signal sequence x(k), k=0, 1, 2..., K-1, K is the length of the digital signal sequence x(k); use a Blackman window b(·) on x(k) to obtain a narrower main lobe width and a higher main-side lobe ratio, and process it through a short-time Fourier transform algorithm to obtain X(m,n), where n indicates a time domain sampling point and m indicates a frequency sampling point. Specifically: Where, m=0,1,2,......,M-1; n=0,1,2,......,N-1; M is the window length of the Blackman window b(·); N is the total number of time domain sampling points, represents rounding down, L is the window sliding interval of the Blackman window b(·); F d is the frequency domain sampling interval, F d =F s / M,F s is the sampling rate of the digital signal sequence x(k); j is the imaginary unit; the superscript * denotes conjugation; Step 102, calculate the corresponding time-frequency power spectrum Y(m,n) according to the short-time Fourier transform result X(m,n):

3. The method for extracting multi-carrier frequencies of a high-speed frequency hopping signal according to claim 2, characterized in that: The specific method of step 2 is: Step 201, setting threshold P th , rough estimated bandwidth B, power local maximum value matrix P and frequency information matrix F; the size of P and F are both G×N, that is, the number of user carriers × the total number of time domain sampling points, and their initial values ​​are set to all 0; Step 202, set n0 to represent the time domain pointer, and set n0=0; Step 203, setting g0 to represent the user carrier pointer, setting g0 = 0; Step 204: Calculate the maximum power of the user carrier at the time g0 and n0 in the time-frequency power spectrum. and its corresponding frequency sampling points like make Then execute step 205; otherwise, set P(g0,n0)=0, F(g0,n0)=0, and set g0=G-1 and then execute step 205; Step 205, determine whether g0=G-1 is true, if g0=G-1 is true, then continue to determine whether n0=N-1 is true, if n0=N-1 is true, then execute step 3, if n0=N-1 is not true, then set n0=n0+1 and then execute step 203; if g0=G-1 is not true, then the frequency sampling point After the time-frequency power spectrum within the rough estimation bandwidth B centered at is set to 0, g0=g0+1 is set, and step 204 is executed.

4. The method for extracting multi-carrier frequencies of a high-speed frequency hopping signal according to claim 3, characterized in that: The specific method of step 3 is: Step 301, let n1 = 1, set the local strongest power vector P with a length of N m , set the local strongest frequency vector F with length N m , and let P m (0)=P(0,0),F m (0) = F(0,0), set the power difference threshold to dP th , the frequency difference threshold is dF th , the maximum power difference threshold is dP mth ; Step 302, determine whether the maximum value of the element in the n1+1th column of the power local maximum value matrix P is 0. If it is 0, replace the element value in the n1th column of P with the element value in the n1th column of P, and execute step 303; Otherwise, directly execute step 303; Step 303, calculate the power difference ΔP = |P(0,n1)-P m (n1-1)|, Frequency difference ΔF=|F(0,n1)-F m (n1-1)|, where |*| represents the absolute value. If the power difference and frequency difference are both within the threshold range, ΔP is satisfied. <dP th , and ΔF <dF th , then P m (n1)=P(0,n1),F m (n1)=F(0,n1), and execute step 306; Otherwise, execute step 304; Step 304, set the user carrier pointer g1=1; Step 305, calculate the power difference ΔP = |P(g1,n1)-P m (n1-1)|, Frequency difference ΔF=|F(g1,n1)-F m (n1-1)|; Let ΔP m =|P(g1,n1)-P(0,n1-1)|, determine whether ΔP is satisfied at the same time <dP th , and ΔF <dF th , and ΔP m <dP mth , if the above three inequalities are satisfied at the same time, then no real jump occurs at this time, let P m (n1)=P(g1,n1),F m (n1)=F(g1,n1), and execute step 306; If the above three inequalities are not satisfied at the same time, determine whether g1=G-1 is true. If g1=G-1 is true, store the maximum value in the local maximum vector again, that is, P m (n1)=P(0,n1),F m (n1)=F(0,n1), and execute step 306; If g1=G-1 does not hold, set g1=g1+1 and execute step 305; Step 306, determine whether n1=N-1 is true. If n1=N-1 is not true, set n1=n1+1 and repeat step 302; otherwise, complete the calculation and execute step 4.

5. The method for extracting multi-carrier frequencies of a high-speed frequency hopping signal according to claim 4, characterized in that: The specific method of step 4 is: Step 401, calculate F m The first-order difference result ΔF, that is, ΔF(n2)=|F m (n2+1)-F m (n2)|, n2=0,1,2,…,N-2; And let ΔF be less than dF th The result is set to 0; Step 402: extract all time domain pointer numbers corresponding to non-zero values ​​in ΔF and store them in a time domain pointer number vector N with a length of J. n0 , J represents the number of non-zero values ​​in ΔF; n0 The time domain pointer number in is the frequency mutation position; Step 403: number the time domain pointer vector N n0 Perform a first-order difference operation to obtain the difference result ΔN n0 : ΔN n0 (j)=|N n0 (j+1)-N n0 (j)|,j=0,1,2,…,J-2。 6. The method for extracting multi-carrier frequencies of a high-speed frequency hopping signal according to claim 5, characterized in that: The specific method of step 5 is: Step 501: Subtract the difference result ΔN n0 Sort the elements in the order of size, get the median of the sort, and convert ΔN n0 The elements whose difference with the sorted median is within ±10% are averaged, and the average value is recorded as the frequency hopping interval ΔN; Step 502, calculate the sequence number Z of the real frequency hopping moment: set the iteration initial value Z(0)=0, and the value of each subsequent iteration is: Z(j1)=<(N n0 (j1)-N n0 (j1-1)) / ΔN>+Z(j1-1),j1=1,2,3,…,J-1; Among them, <·> is rounded to the nearest integer.

7. The multi-carrier frequency extraction method based on a high-speed frequency hopping signal according to claim 6, characterized in that: The specific method of step 6 is: Step 601: Use the least squares fitting algorithm to transform the sequence number Z of the real frequency hopping moment and the time domain pointer number vector N n0 The functional relationship of is fitted into a linear function; Step 602: Substitute the sequence number Z of the real frequency hopping time into the fitted linear function in sequence to obtain the estimated result N of each frequency hopping time. Jn .

8. The multi-carrier frequency extraction method based on a high-speed frequency hopping signal according to claim 7, characterized in that: The specific method of step 7 is: Step 701, first set the frequency hopping time index j2=0, and the user carrier pointer g2=0; Step 702: Take out all the frequency points F(g2,n3) of the current user carrier before the frequency hopping occurs from the frequency point information matrix F, and calculate their average value F G (g2,n3); When j2=0, n3=0,1,2,...,N Jn (Z(0)); When \(j2 = 1, 2, 3, \ldots, J - 1\), \(n3 = N\) Jn (Z(j2)-1),N Jn (Z(j2)-1)+1,N Jn (Z(j2)-1)+2,...,N Jn (Z(j2)); F G (g2, n3) is the frequency of user carrier g2 in the j2+1th frequency hopping signal; Step 703, determine whether g2=G-1 is true. If g2=G-1 is true, set g2=0 and execute step 703; if g2=G-1 is not true, set g2=g2+1 and return to execute step 702; Step 703, determine whether j2=J-1 is established. If j2=J-1 is established, the multi-carrier frequency point extraction of the high-speed frequency hopping signal is completed; if j2=J-1 is not established, set j2=j2+1 and return to execute step 702.

Images