A method for extracting bird wing beat frequency based on doppler radial acceleration

Abstract

The present application belongs to the technical field of radar, and particularly relates to a bird wing-flapping frequency extraction method based on Doppler radial acceleration. The present application firstly tracks the instantaneous frequency trajectory of the time-frequency analysis result by using the Viterbi algorithm, dynamically searches for the best path, and thus effectively extracts the signal instantaneous frequency; then performs differential processing, converts the instantaneous Doppler information into Doppler radial acceleration; and finally sets an adaptive threshold, counts the wing-flapping times of the bird, and thus realizes the extraction of the wing-flapping frequency of the bird. The present application proposes a bird wing-flapping frequency extraction method based on Doppler radial acceleration, which provides an effective means for the extraction of the wing-flapping frequency of the bird. Compared with the existing bird wing-flapping frequency extraction method, the present method can greatly improve the success rate and precision of the extraction of the wing-flapping frequency of the bird under the condition of low signal-to-noise ratio, and will be helpful for the bird target recognition of the radar.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of radar, and particularly relates to a bird wing flapping frequency extraction method based on Doppler radial acceleration. BACKGROUND

[0002] Radar identification of birds is very important for bird migration research, environmental impact assessment (such as wind farms), aircraft safety and radar meteorology, and the wing flapping frequency is one of the important parameters for radar to identify bird targets. The typical mode of bird echo is the periodic fluctuation of signal strength, which is caused by the body shape change caused by wing beating. At present, different wing flapping modes reflected by bird echoes can realize the classification of certain bird target groups, such as Passeriformes, songbirds and waders. Each bird has a specific wing flapping frequency when continuously flapping wings, so if the wing flapping frequency of the bird echo is accurately obtained, the bird target recognition rate can be further improved.

[0003] In radar observation, the wing flapping frequency can usually be obtained by spectrum analysis of bird echoes. At present, the main method for wing flapping frequency in the field of biological radar is to perform time-frequency analysis on the echo, extract the maximum value of the instantaneous Doppler frequency and perform Fast Fourier transform (FFT). Due to the complex flight posture and trajectory of birds, the signal-to-noise ratio of the target echo is generally low in actual radar observation, which leads to low accuracy of the wing flapping frequency extraction of the method, and the main reasons are as follows: (1) the time-frequency analysis result of low signal-to-noise ratio often produces spectrum aliasing and path bifurcation problems; (2) the spectrum analysis (FFT) is greatly affected by noise, and the correctness of the wing flapping frequency extraction is not high.

[0004] In summary, if the actual time-frequency path that best fits the actual situation can be dynamically searched, and the wing flapping frequency of the observed bird can be accurately counted, the problems existing in the prior art can be overcome, and the wing flapping frequency extraction accuracy can be improved. SUMMARY

[0005] Therefore, the application provides a bird wing flapping frequency extraction method based on Doppler radial acceleration, which uses edge detection in graph theory as a reference. Firstly, the Viterbi algorithm is used to track the instantaneous frequency trajectory of the time-frequency analysis result, and the best path is dynamically searched to effectively extract the signal instantaneous frequency. Then, difference processing is performed to convert the instantaneous Doppler information into Doppler radial acceleration. Finally, by setting an adaptive threshold, the wing flapping frequency of the bird is extracted.

[0006] The technical solution of the application is as follows:

[0007] A bird wing flapping frequency extraction method based on Doppler radial acceleration, comprising the following steps:

[0008] Step one, time-frequency analysis is performed on the bird radar echo to obtain a bird micro-Doppler spectrogram, and an instantaneous Doppler frequency curve of the bird micro-Doppler spectrogram is extracted based on a Viterbi algorithm;

[0009] Step two, after differential processing and smoothing filtering are performed on the instantaneous Doppler frequency curve extracted in step one, a radial acceleration curve is obtained;

[0010] Step three, the number of peak points of the radial acceleration curve obtained in step two that are above a threshold value is counted, and the number of peak points is taken as the number of bird wing flapping times of the current bird radar echo; in step three, the threshold value is 0.3 times the maximum peak value of the radial acceleration curve obtained in step two;

[0011] Step four, the number of bird wing flapping times of the current bird radar echo obtained in step three is divided by the time length of the current bird radar echo to obtain a bird wing flapping frequency.

[0012] In step one, when the instantaneous Doppler frequency curve of the bird micro-Doppler spectrogram is extracted based on the Viterbi algorithm, the Viterbi algorithm is used to search for an optimal path from the bird micro-Doppler spectrogram to extract the instantaneous Doppler frequency curve, and the optimal path requires that (i) the amplitude of the time-frequency point corresponding to the instantaneous Doppler frequency at each time point is as large as possible; and (ii) the change in the instantaneous Doppler frequency between two adjacent time points is not too large, and the instantaneous Doppler frequency curve is relatively smooth.

[0013] The optimal path f(n) is:

[0014]

[0015] Wherein, n represents a time, n∈[n1,n2], all paths between n1 and n2 are K, k(n) represents a certain path, n1 is a starting time of time-frequency analysis, and n2 is a termination time of time-frequency analysis;

[0016] TF(·) represents time-frequency transformation of the bird radar echo, h(x) is a penalty function corresponding to requirement (i) of the optimal path, and g(x,y) is a penalty function corresponding to requirement (ii) of the optimal path;

[0017] Arranging the time-frequency amplitude TF of the bird micro-Doppler spectrogram at time n into a non-decreasing sequence is:

[0018] TF(n,f1)≥TF(n,f2)≥...≥TF(n,f j )≥...≥TF(n,f Q )

[0019] Wherein, j represents the position in the sequence, Q is the number of signal frequency points, and j = 1, 2,..., Q;

[0020] According to requirement (i), the amplitude of the time-frequency point corresponding to the instantaneous Doppler frequency of each time instant is as large as possible, and therefore the penalty function h(x) is defined as:

[0021] h(TF(n,w j ))=j-1

[0022] According to requirement (ii), the change of the instantaneous Doppler frequency of two adjacent time instants is not too large, and therefore the penalty function g(x, y) is defined as:

[0023]

[0024] Wherein, c is a penalty factor, Δ is the maximum expected value of the change of the instantaneous frequency between adjacent points, i.e., the frequency resolution of the time-frequency transform; x and y are the instantaneous Doppler frequencies of two adjacent time instants;

[0025] The method for obtaining the radial acceleration curve in step two is:

[0026] First, based on the radial velocity and the Doppler frequency, the instantaneous Doppler frequency curve obtained in step one is converted into a radial velocity curve, and the conversion formula of the radial velocity and the Doppler frequency is:

[0027]

[0028] In the formula, f d is the instantaneous Doppler frequency of each time instant, λ is the wavelength of the radar signal, and v is the radial velocity of the target relative to the radar. Then, the radial velocity curve is subjected to a difference processing and divided by the time interval of the time-frequency curve to obtain a radial acceleration curve. Finally, the radial acceleration curve is subjected to a smoothing filtering operation.

[0029] After the radial acceleration curve is obtained in step two, the maximum value of the radial acceleration peak value is counted in step three, and the threshold value is set as 0.3 times the maximum value of the radial acceleration peak value, i.e.:

[0030] Thr=0.3*max(a)

[0031] In the formula, max(a) is the maximum value of the radial acceleration peak value, and Thr is the set threshold value.

[0032] Then, the number of peak points of the radial acceleration curve above the threshold value is counted, and the number of peak points is taken as the bird flapping frequency N of the current bird radar echo.

[0033] The real wing-flapping frequency f of the bird can be calculated by dividing the wing-flapping times N of the current bird radar echo by the time length T of the bird radar echo in step four w :

[0034]

[0035] In step one, the punishment factor c is defined as 2.5, and the Δ is defined as the maximum expected value of the instantaneous frequency change between two continuous points, i.e. the frequency resolution of the time-frequency analysis;

[0036] In step two, the instantaneous Doppler curve is converted into a radial velocity curve, then through differential processing and divided by the time resolution of the time-frequency analysis, the Doppler radial acceleration is obtained

[0037] In step three, the value of the dynamic threshold is set as 0.3 times the maximum value of the Doppler radial acceleration.

[0038] Beneficial effects

[0039] (1) The present application discloses a bird wing-flapping frequency extraction method based on Doppler radial acceleration; the present application can greatly improve the success rate and accuracy of bird wing-flapping frequency extraction under low signal-to-noise ratio conditions, and will help the bird target recognition of radar;

[0040] (2) The present application firstly tracks the instantaneous frequency trajectory of the time-frequency analysis result by using the Viterbi algorithm, dynamically searches the best path to effectively extract the instantaneous frequency of the signal; then through differential processing, the instantaneous Doppler information is converted into Doppler radial acceleration; finally, by setting an adaptive threshold, the wing-flapping times of the bird are counted, so as to realize the extraction of the wing-flapping frequency of the bird.

[0041] (3) The present application proposes a bird wing-flapping frequency extraction method based on Doppler radial acceleration, which provides an effective means for bird wing-flapping frequency extraction.

[0042] (4) Compared with the existing bird wing-flapping frequency extraction method, the present method can greatly improve the success rate and accuracy of bird wing-flapping frequency extraction under low signal-to-noise ratio conditions, and will help the bird target recognition of radar. BRIEF DESCRIPTION OF DRAWINGS

[0043] Figure 1 It is a method flowchart of the present application;

[0044] Figure 2 It is a measured radar picture;

[0045] Figure 3 It is the instantaneous Doppler curve result extracted by the Viterbi algorithm;

[0046] Figure 4Setting a radial acceleration curve and a threshold;

[0047] Figure 5 The present application has higher extraction success rate than traditional algorithms;

[0048] Figure 6 The present application has higher extraction accuracy than traditional algorithms. DETAILED DESCRIPTION

[0049] The present application is described in detail below with reference to the accompanying drawings and examples, and the specific implementation flowchart is shown in Figure 1

[0050] Step one: extracting the instantaneous Doppler curve based on the Viterbi algorithm

[0051] The two-dimensional time-frequency distribution is obtained by time-frequency analysis of the bird echo signal, and the Viterbi algorithm is used to search for the best path from the time-frequency distribution to extract the signal instantaneous Doppler frequency curve. The best path requires: (i) the amplitude of the time-frequency point corresponding to the instantaneous Doppler frequency at each time should be as large as possible; (ii) the change of the instantaneous Doppler frequency between two adjacent time instants should not be too large, and the instantaneous Doppler frequency curve should be relatively smooth. The instantaneous Doppler frequency curve extracted by the Viterbi algorithm can be represented by the path penalty function sum f(n) of the minimum line:

[0052]

[0053] Where n represents the time, n∈[n1,n2], K is the path between n1 and n2, k(n) represents a path, n1 is the starting time of time-frequency analysis, and n2 is the termination time of time-frequency analysis;

[0054] TF(·) represents the time-frequency transformation of the bird radar echo, h(x) is the penalty function corresponding to the best path requirement (i), and g(x,y) is the penalty function corresponding to the best path requirement (ii);

[0055] The time-frequency amplitude TF of the bird micro-Doppler spectrum at time n is arranged in a non-decreasing sequence as follows:

[0056] TF(n,f1)≥TF(n,f2)≥...≥TF(n,f j )≥...≥TF(n,f Q )

[0057] Where j represents the position in the sequence, Q is the number of signal frequency points, and j=1,2,...,Q;

[0058] According to requirement (i), the amplitude of the time-frequency point corresponding to the instantaneous Doppler frequency at each time should be as large as possible, so the penalty function h(x) is defined as: ​

[0059] h(TF(n,w j ))=j-1

[0060] According to requirement (ii), the instantaneous Doppler frequency change between two adjacent time instants should not be too large, so the penalty function g(x,y) is defined as:

[0061]

[0062] where c is the penalty factor, Δ is the maximum expected value of the instantaneous frequency change between adjacent points, i.e., the frequency resolution of the instantaneous frequency transform; x and y are the instantaneous Doppler frequencies at two adjacent time instants.

[0063] Step two: time-frequency curve processing

[0064] After successfully extracting the time-frequency curve, the instantaneous Doppler frequency is converted into the radial velocity according to the following formula to obtain the radial velocity curve.

[0065]

[0066] where f d is the instantaneous Doppler frequency at each time instant, λ is the wavelength of the radar signal, and v is the radial velocity of the target relative to the radar; then, the radial velocity curve is subjected to difference processing and divided by the time interval of the time-frequency curve to obtain the radial acceleration curve; finally, the radial acceleration curve is subjected to smoothing filtering.

[0067] Step three: counting the flapping times

[0068] After obtaining the radial acceleration curve in step two, the maximum value of the radial acceleration peak value is counted. According to a large amount of radar measurement data statistics, the ideal value of the dynamic threshold is about 0.3 times the maximum value of the acceleration. Therefore, the threshold is set according to the following formula:

[0069] Thr=0.3*max(a)

[0070] where max(a) is the maximum value of the radial acceleration peak value, and Thr is the threshold value. The number of peak values of the radial acceleration curve above the threshold value is counted as the flapping times N of the bird.

[0071] Step four: calculating the wing-beat frequency

[0072] The real wing-beat frequency f w of the bird can be calculated by dividing the flapping times N of the bird in the current bird radar echo by the time length T of the bird radar echo.

[0073]

[0074] Embodiment

[0075] To verify the measurement accuracy and correctness of the wing beat frequency extraction method described above, based on the radar measured bird data, a bird wing beat frequency extraction method based on Doppler radial acceleration is used to complete the wing beat frequency measurement of birds, and compared with the traditional method. The radar parameters used in the actual observation are shown in Table 1, and the radar picture is shown in Figure 2 .

[0076] Table 1 Experimental radar system parameters

[0077]

[0078] Taking a group of measured bird data as an example, first, according to step one, the time-frequency analysis is carried out and the instantaneous Doppler frequency curve is extracted based on the Viterbi algorithm, as shown in Figure 3 , the upper graph is the time-frequency analysis result, and the lower graph is the instantaneous Doppler frequency curve extracted based on the Viterbi algorithm.

[0079] Then, according to step two, the radial acceleration curve is obtained, and according to step three, the statistical threshold is calculated, as shown in Figure 4 . The results show that in 3s observation time, the group of data has 38 times of flapping wings, so the wing beat frequency of the bird can be calculated as:

[0080] 38 / 3≈12.67Hz

[0081] Figure 5 and Figure 6 show the wing beat frequency extraction results of the algorithm and the traditional algorithm when the bird echo signal-to-noise ratio is low. The measured data results show that the bird wing beat frequency extraction method based on Doppler radial acceleration proposed in the application can effectively realize the success rate and accuracy of bird wing beat frequency measurement.

[0082] In summary, the above is only a preferred embodiment of the present application, and is not used to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for extracting bird wingbeat frequencies based on Doppler radial acceleration, characterized in that... Includes the following steps: Step 1: Perform time-frequency analysis on the bird radar echo to obtain the bird's micro-Doppler spectrum. Based on the Viterbi algorithm, extract the instantaneous Doppler frequency curve of the bird's micro-Doppler spectrum. Step 2: After differential processing and smoothing filtering of the instantaneous Doppler frequency curve extracted in Step 1, the radial acceleration curve is obtained. Step 3: Count the number of peak points of the radial acceleration curve obtained in Step 2 that are above the threshold value, and use this number of peak points as the number of bird flapping wing counts in the current bird radar echo; Step 4: Divide the number of bird flappings in the current bird radar echo obtained in Step 3 by the duration of the current bird radar echo to obtain the bird flapping frequency. In step one, when extracting the instantaneous Doppler frequency curve of the bird's micro-Doppler spectrum based on the Viterbi algorithm, the Viterbi algorithm is used to search for the optimal path from the bird's micro-Doppler spectrum to extract the instantaneous Doppler frequency curve. The optimal path is the instantaneous Doppler frequency curve of the bird's micro-Doppler spectrum. The optimal path requires that: (i) the amplitude of the time-frequency point corresponding to the instantaneous Doppler frequency at each moment should be as large as possible; and (ii) the instantaneous Doppler frequency curves between two adjacent moments should be smooth. The optimal path for: in, n Indicates time, , and All paths between them are , Indicates a certain path, This is the starting time for time-frequency analysis. This is the termination time of the time-frequency analysis; This indicates that time-frequency transformation is performed on the radar echoes from birds. It is the penalty function corresponding to the optimal path requirement (i). It is the penalty function corresponding to the optimal path requirement (ii); According to requirement (i), the amplitude of the time-frequency point corresponding to the instantaneous Doppler frequency at each moment should be as large as possible, and the penalty function... for: According to requirement (ii), the instantaneous Doppler frequency curves of two adjacent moments are smoothed, and the penalty function is applied. for: in, As a penalty factor, The maximum expected value of the instantaneous frequency change between adjacent points is the frequency resolution of the instantaneous frequency transformation; x, y The instantaneous Doppler frequencies are the frequencies at two adjacent moments. In step two, the method for obtaining the radial acceleration curve is as follows: First, based on the radial velocity and Doppler frequency, the instantaneous Doppler frequency curve obtained in step one is transformed into a radial velocity curve. The conversion formula between radial velocity and Doppler frequency is as follows: In the formula, The instantaneous Doppler frequency at each moment, The wavelength of the radar signal. The radial velocity of the target relative to the radar is then calculated. Next, the radial velocity curve is differentially processed and divided by the time interval of the time-frequency curve to obtain the radial acceleration curve. Finally, the radial acceleration curve is smoothed and filtered. In step three, the threshold value is 0.3 times the maximum value of the radial acceleration curve obtained in step two; In step three, after obtaining the radial acceleration curve in step two, the maximum value of the radial acceleration peak is calculated: In the formula, The maximum value of the radial acceleration peak. The set threshold value; Next, the number of peak points on the radial acceleration curve above the threshold is counted, and this number of peak points is taken as the number of bird flapping wing counts in the current bird radar echo; In step four, the number of bird flapping wing counts from the current bird radar echo is... Divide by the duration of bird radar echoes This allows for the calculation of the bird's true wingbeat frequency. : In step one, the penalty factor Defined as 2.5.

Images