Signal peak detection method and system based on adaptive prominence threshold

Abstract

The application discloses a signal peak value detection method and system based on an adaptive prominence threshold, relates to the technical field of electric digital data processing, and comprises the following steps: smoothing preprocessed data of spectrum data to obtain smooth curve data; performing bottom noise shaping on the smooth curve data based on the preprocessed data to obtain a bottom noise shaping result; performing multi-scale peak value detection on the bottom noise shaping result based on an AMPD algorithm; for each peak value candidate point, calculating the prominence of the peak value candidate point relative to the bottom noise shaping result based on a watershed algorithm to obtain a prominence value list corresponding to all peak value candidate points; calculating an adaptive prominence threshold based on an IQR statistical algorithm according to the prominence value list; and screening the peak value candidate set based on the adaptive prominence threshold to retain peak values with a prominence not less than the adaptive prominence threshold. The application can realize full-adaptive processing, has strong noise resistance, high accuracy, good robustness, high calculation efficiency and low algorithm complexity.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of digital data processing, and particularly relates to a signal peak detection method and system based on an adaptive prominence threshold. BACKGROUND

[0002] In the field of spectral analysis, accurate detection and extraction of spectral peaks are key technologies in signal processing, widely used in wireless communication, acoustic processing, vibration monitoring, medical diagnosis and other fields. The traditional peak detection method has the following limitations: threshold sensitivity: fixed threshold method cannot adapt to different noise environments and signal characteristics;

[0003] Noise interference: false positive detection is prone to occur in low SNR environment; parameter dependence: a large amount of prior knowledge is needed for parameter tuning; poor adaptability: lack of adaptability for spectral data with different characteristics. Although the prior art has certain adaptive ability, it still relies on empirical threshold and cannot be dynamically adjusted according to data characteristics.

[0004] Therefore, the present application develops a signal peak detection method and system based on an adaptive prominence threshold to solve the above problems. SUMMARY

[0005] The present application provides a signal peak detection method and system based on an adaptive prominence threshold to solve the problem of low detection accuracy and strong parameter dependence of existing methods in complex noise environments.

[0006] The present application achieves the above-mentioned purposes through the following technical solutions:

[0007] The present application provides a signal peak detection method based on an adaptive prominence threshold, comprising:

[0008] Obtaining spectral data of a signal;

[0009] Pretreating the spectral data to obtain pretreated data;

[0010] Smoothly processing the pretreated data to obtain smooth curve data;

[0011] Shaping the bottom noise of the smooth curve data based on the pretreated data to obtain a bottom noise shaping result;

[0012] Performing multi-scale peak detection on the bottom noise shaping result based on an AMPD algorithm to obtain a peak candidate set;

[0013] For each peak candidate point, calculating its prominence relative to the bottom noise shaping result based on a watershed algorithm to obtain a list of prominence values corresponding to all peak candidate points;

[0014] According to the saliency value list, an adaptive saliency threshold is calculated based on an IQR statistical algorithm;

[0015] Based on the adaptive saliency threshold, the peak candidate set is screened, and peaks with saliency not less than the adaptive saliency threshold are retained, and then the spectral data index position corresponding to the peak is output.

[0016] Further, the spectral data is preprocessed to obtain preprocessed data, including:

[0017] The spectral data is subjected to minimum-maximum normalization processing to obtain first normalized data:

[0018] The first average value of the first normalized data is calculated, and all spectral line values less than the first average value in the first normalized data are set to the first average value to obtain the preprocessed data.

[0019] Further, the bottom noise of the smoothed curve data is shaped based on the preprocessed data to obtain a bottom noise shaping result, including:

[0020] The preprocessed data is subtracted from the smoothed curve data to obtain an initial bottom noise shaping result;

[0021] The initial bottom noise shaping result is subjected to minimum-maximum normalization processing to obtain second normalized data:

[0022] The second average value of the second normalized data is calculated, and all spectral line values less than the second average value in the second normalized data are set to the second average value to obtain the bottom noise shaping result.

[0023] Further, the bottom noise shaping result is subjected to multi-scale peak detection based on an AMPD algorithm to obtain a peak candidate set, including:

[0024] The number of data points of the spectral data is initialized to obtain a local maximum scale;

[0025] The bottom noise shaping result and the local maximum scale are used to calculate a local maximum value matrix;

[0026] The local maximum value matrix is used to calculate a scale consistency vector;

[0027] The scale consistency vector is used to calculate an optimal detection scale;

[0028] The local maximum value matrix and the optimal detection scale are used to create a consistent peak candidate set to obtain the peak candidate set.

[0029] Further, for each peak candidate point, the saliency of the peak candidate point relative to the bottom noise shaping result is calculated based on a watershed algorithm, including:

[0030] Search in the left and right directions of the current peak candidate point respectively, find the conditional saddle points in the left and right directions respectively, the conditional saddle point is the first point higher than the current peak candidate point or the second average value encountered in the search direction;

[0031] Take the maximum value of the two conditional saddle points in the left and right directions, subtract the maximum value from the height of the peak candidate point to obtain the peak candidate point, and obtain the prominence.

[0032] Further, according to the prominence value list, an adaptive prominence threshold is calculated based on the IQR algorithm, including:

[0033] Determine whether the prominence value list is empty;

[0034] If yes, the preset fixed sensitivity threshold is taken as the adaptive prominence threshold;

[0035] If no, the IQR threshold is calculated based on the quartile distance method, and the minimum value of the IQR threshold and the preset fixed sensitivity threshold is taken as the adaptive prominence threshold.

[0036] Further, the smoothing filtering processing of the preprocessed data includes using a sliding average filter with an adaptive window size to smooth the preprocessed data.

[0037] The application also provides a system for the signal peak detection method based on the adaptive prominence threshold, including:

[0038] An acquisition module is configured to acquire the frequency spectrum data of a signal;

[0039] A preprocessing module is configured to preprocess the frequency spectrum data to obtain preprocessed data;

[0040] A smoothing processing module is configured to smooth the preprocessed data to obtain smoothed curve data;

[0041] A noise shaping module is configured to shape the noise of the smoothed curve data based on the preprocessed data to obtain a noise shaping result;

[0042] A detection module is configured to perform multi-scale peak detection on the noise shaping result based on the AMPD algorithm to obtain a peak candidate set;

[0043] A first calculation module is configured to calculate the prominence of each peak candidate point relative to the noise shaping result based on the watershed algorithm to obtain a prominence value list corresponding to all peak candidate points;

[0044] A second calculation module is configured to calculate an adaptive prominence threshold based on an IQR statistical algorithm according to the list of prominence values;

[0045] A screening module is configured to screen the set of peak candidates based on the adaptive prominence threshold, retain peaks with a prominence not less than the adaptive prominence threshold, and then output the index position of the spectrum data corresponding to the peaks.

[0046] Further, the floor noise shaping module comprises:

[0047] A third calculation module is configured to subtract the preprocessed data from the smooth curve data to obtain an initial floor noise shaping result.

[0048] A normalization module is configured to perform minimum-maximum normalization on the initial floor noise shaping result to obtain second normalized data.

[0049] A fourth calculation module is configured to calculate a second average value of the second normalized data, set all spectral line values less than the second average value in the second normalized data to the second average value, and obtain a floor noise shaping result.

[0050] Further, the first calculation module comprises:

[0051] A search module is configured to search in the left and right directions of the current peak candidate point respectively, and find the conditional saddle points in the left and right directions respectively, wherein the conditional saddle point is the first point higher than the current peak candidate point or the second average value encountered in the search direction.

[0052] A prominence module is configured to take the maximum value of the two conditional saddle points in the left and right directions, subtract the maximum value from the height of the peak candidate point to obtain the peak candidate point, and obtain the prominence.

[0053] The present application has the following advantages:

[0054] The signal peak detection method and system based on the adaptive prominence threshold can realize fully adaptive processing, without manual parameter tuning and automatically adapting to different spectrum characteristics, have strong anti-noise ability, effectively suppress noise interference through multi-stage processing, have high accuracy, ensure accurate peak screening through the IQR adaptive threshold, have good robustness, process various abnormal situations through the double threshold guarantee mechanism, and have high calculation efficiency, low algorithm complexity, and are suitable for real-time processing applications. BRIEF DESCRIPTION OF DRAWINGS

[0055] Figure 1 The present application is based on the signal peak detection method flowchart based on the adaptive prominence threshold.

[0056] Figure 2Flow chart of the AMPD multi-scale peak detection process of the present application;

[0057] Figure 3 Flow chart of the multi-scale local maximum matrix calculation process of the present application;

[0058] Figure 4 Flow chart of the scale consistency vector calculation process of the present application;

[0059] Figure 5 Flow chart of the optimal detection scale calculation process of the present application;

[0060] Figure 6 Flow chart of the consistent peak candidate extraction process of the present application;

[0061] Figure 7 Schematic diagram of the saliency principle of the present application;

[0062] Figure 8 Final peak detection result diagram of Example 1 of the present application;

[0063] Figure 9 Final peak detection result diagram of Example 2 of the present application. DETAILED DESCRIPTION

[0064] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some embodiments but not all embodiments of the present application. The components of the embodiments of the present application described and shown in the drawings can be arranged and designed in various different configurations.

[0065] Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative labor are within the scope of protection of the present application.

[0066] In addition, the terms "first", "second", and the like are only used to distinguish description, and cannot be understood as indicating or implying relative importance.

[0067] The specific embodiments of the present application will be described in detail below with reference to the drawings.

[0068] As Figure 1 shown, a signal peak detection method and system based on adaptive saliency threshold value, comprising:

[0069] Obtaining the frequency spectrum data of the signal;

[0070] The spectrum data is preprocessed to obtain preprocessed data, and the spectrum data is preprocessed to obtain preprocessed data, comprising:

[0071] The spectrum data is preprocessed to obtain preprocessed data, and the spectrum data is preprocessed to obtain preprocessed data, comprising:

[0072] The first average value of the first normalized data is calculated, and all spectral line values less than the first average value in the first normalized data are set to the first average value, to obtain preprocessed data;

[0073] The preprocessed data is smoothed to obtain smoothed curve data, and a sliding average filter with adaptive window size is used to smooth the preprocessed data;

[0074] The bottom noise of the smoothed curve data is shaped based on the preprocessed data to obtain a bottom noise shaping result;

[0075] The bottom noise shaping result is detected by a multi-scale peak value based on the AMPD algorithm to obtain a peak candidate set;

[0076] For each peak candidate point, the prominence of the peak candidate point relative to the bottom noise shaping result is calculated based on the watershed algorithm to obtain a list of prominence values corresponding to all peak candidate points;

[0077] According to the list of prominence values, an adaptive prominence threshold is calculated based on the IQR algorithm;

[0078] The peak candidate set is filtered based on the adaptive prominence threshold, and the peaks with prominence not less than the adaptive prominence threshold are retained, and then the spectral data index position corresponding to the peak is output, and the detection is ended.

[0079] Specifically, it comprises:

[0080] Step S101: spectrum data normalization

[0081] The input original spectrum data x is subjected to minimum-maximum normalization processing to eliminate the dimension effect, and the first normalized data A is obtained:

[0082]

[0083] Wherein, is the spectrum data index, min(x) and max(x) are the minimum and maximum values of the spectrum data, respectively.

[0084] Step S102: mean noise preprocessing

[0085] The first average value μ of the first normalized data A is calculated, and all spectral line values less than the first average value μ in the first normalized data A are set to μ, to obtain preprocessed data B:

[0086]

[0087] wherein:

[0088]

[0089] The average value of the first normalized data A.

[0090] Step S103: Large-scale smoothing filtering processing

[0091] The pre-processed data B is smoothed by a sliding average filter with adaptive window size to obtain smoothed curve data C:

[0092]

[0093] wherein is the adaptive window size, K is the local maximum scale size of the AMPD algorithm, and floor is the floor function.

[0094] Step S104: Base noise shaping processing

[0095] The pre-processed data B is subtracted from the smoothed curve data C to obtain initial base noise shaping result D:

[0096]

[0097] The initial base noise shaping data D is subjected to minimum-maximum normalization processing to stretch the peak value significance to obtain second normalized data E:

[0098]

[0099] The second average value β of the second normalized data E is calculated, and all spectral line values less than the second average value β in the data E are set to β to obtain base noise shaping result X:

[0100]

[0101] wherein:

[0102]

[0103] This step effectively eliminates background noise and baseline drift, enhances peak value significance, and improves the robustness of the subsequent IQR algorithm for extracting an adaptive prominence threshold.

[0104] Step S105: AMPD multi-scale peak detection

[0105] As shown in Figure 2 , the base noise shaping result X is subjected to multi-scale peak detection by the AMPD algorithm to find all possible peak position candidate sets, including the following processes:

[0106] (1) input the one-dimensional data X after the floor noise shaping; the data point number N = size(X);

[0107] (2) initialize the local maximum scale: K = N*0.004, calculate and create a multi-scale local maximum matrix L[N][K], the detailed process is shown in Figure 3 , including: create a two-dimensional multi-scale local maximum matrix L[N][K]; scale loop, loop variable k from 1 to K increment; data point loop variable i from 1 to N increment; check the boundary point condition; if it is a boundary point, mark it as False, and record L[i][k-1]=False; if it is an internal point, execute local maximum detection, compare loop variable j from 1 to k increment; compare X[i] with adjacent points X[i-j] and X[i+j], judge whether X[i] is greater than the adjacent point, if not, mark it as False and jump out of the loop, and record L[i][k-1]=False, if yes, judge whether all j comparisons are completed, if completed, mark it as True, and record L[i][k-1]=True, if not completed, loop variable j is incremented by 1, return to compare X[i] with adjacent points X[i-j] and X[i+j]; judge whether all i point marks are completed, if not, return to check the boundary condition, if yes, judge whether all k scale marks are completed, if not, return to the data point loop from 1 to N, if yes, output the complete L matrix.

[0108] (3) calculate the scale consistency vector γ[N], the detailed process is shown in Figure 4 , including: input the local maximum matrix L[N][K], create a one-dimensional scale consistency vector γ[N], data point loop variable i from 0 to N increment, and then scale loop variable k from 1 to K increment; check whether L[i][k] is True, if yes, γ[i]=γ[i]+1; if not, keep γ[i] unchanged; judge whether all k loops are completed, if not, return to check whether L[i][k] is True, if yes, judge whether all i loops are completed, if yes, output the consistency vector γ, if not, return to the scale loop variable k from 1 to K increment check.

[0109] (4) determine the optimal detection scale, the detailed process is shown in Figure 5As shown, including: input one-dimensional scale consistency vector γ[N], initialization scale consistency vector maximum max_gama is γ[0], initialization optimal detection scale λ is 0; Scale consistency vector point loop variable i from 0 to N increment; Determine whether max_gama is less than γ[i], if yes, max_gama=γ[i], λ=i, if not, keep max_gama unchanged, keep λ unchanged; Determine whether all variable i loop is completed, if not, return to determine whether max_gama is less than γ[i], if yes, output the optimal detection scale mod(λ,K).

[0110] (5) Extracting consistency peak candidate, detailed process as shown in Figure 6 As shown, including input local maximum matrix L[N][K], input optimal detection scale, data point loop variable i from K to N-K increment; Determine whether L[i][λ] is true, if not, go to the next step, if yes, the best scale loop variable k from λ to K increment, determine whether L[i][k] is true, if yes, jump out of the best scale loop and enter the next step, if yes, determine whether all k loop is completed, if not, return to the best scale loop k from λ to K, if yes, add data point index i to the peak candidate set and enter the next step; In the next step, determine whether all i loop is completed, if not, return to determine whether L[i][λ] is true, if yes, output the peak candidate set P.

[0111] Step S106: Prominence calculation

[0112] Prominence refers to the height of a peak relative to its surrounding terrain. It represents the degree to which a peak exists independently, i.e. how much the peak needs to drop to reach higher terrain. As shown in Figure 7 , Figure 7 The coordinates of the peak point p are (20201, 0.370863), the left conditional saddle point coordinates are (20197, 0.0892132), and the right conditional saddle point coordinates are (20204, 0.0892132).

[0113] For a peak point p, the calculation process of its prominence is as follows:

[0114] Assume:

[0115] p is a peak point, index i p , peak X[i p ];

[0116] N is the number of data points;

[0117] left min is the first point found on the left that is higher than p or the average value β in step S104;

[0118] right min the first point found to the right of p that is higher than p or the mean β in step S104;

[0119] Search to the left:

[0120]

[0121] where satisfies:

[0122]

[0123] Search to the right:

[0124]

[0125] where satisfies:

[0126]

[0127] Determine the condition saddle point: in both left and right search directions, find the saddle point left min (0.0892132) and right min (0.0892132);

[0128] Calculate the prominence: the peak value of p minus the maximum peak value of the condition saddle points left min and right min , and the peak point prominence is prominence 20201 = 0.370863 - 0.0892132 = 0.2816498.

[0129] For each peak candidate point, the optimized watershed algorithm is used to calculate its prominence relative to the bottom noise shaping result X:

[0130]

[0131] where, is the peak index, left min and right min are the left and right search directions until a higher point is encountered or the mean value β after bottom noise shaping.

[0132] Step S107: Adaptive prominence threshold calculation

[0133] The IQR statistical method is used to automatically calculate the prominence threshold, including:

[0134] Determine whether the prominence value list is empty;

[0135] If yes, a preset fixed sensitivity threshold is taken as the adaptive prominence threshold.

[0136] If no, an IQR threshold is calculated based on the interquartile range method, and the minimum value between the IQR threshold and the preset fixed sensitivity threshold is taken as the adaptive prominence threshold.

[0137] The adaptive prominence threshold V is calculated according to the formula:

[0138]

[0139] Wherein Q3 is the third quartile of the prominence value, IQR is the interquartile range, and the threshold calibration mechanism is also included: when V > 0.14, it is calibrated to 0.14.

[0140] Step S108: prominence screening and peak value confirmation

[0141] The peak value candidates are screened using the adaptive prominence threshold, and the peak value indexes with a prominence greater than or equal to the adaptive prominence threshold are retained:

[0142]

[0143] Wherein, is the peak value index, is the prominence corresponding to the peak value index, is the peak value index of the mth screening result, is the adaptive prominence threshold.

[0144] Step S109: peak value index output

[0145] The spectral data index position corresponding to the final screened accurate peak value is output, and the spectral peak value detection is completed.

[0146] The application also provides a system for the signal peak value detection method based on the adaptive prominence threshold, comprising:

[0147] An acquisition module is configured to acquire spectral data of a signal.

[0148] A preprocessing module is configured to preprocess the spectral data to obtain preprocessed data.

[0149] A smoothing processing module is configured to perform smoothing processing on the preprocessed data to obtain smoothed curve data.

[0150] A noise shaping module is configured to perform noise shaping on the smoothed curve data based on the preprocessed data to obtain a noise shaping result.

[0151] The detection module is configured to perform multi-scale peak detection on the noise floor shaping result based on an AMPD algorithm to obtain a peak candidate set;

[0152] The first calculation module is configured to calculate, for each peak candidate point, a prominence degree of the peak candidate point relative to the noise floor shaping result based on a watershed algorithm to obtain a list of prominence degree values corresponding to all the peak candidate points;

[0153] The second calculation module is configured to calculate an adaptive prominence threshold based on an IQR statistical algorithm according to the list of prominence degree values.

[0154] The screening module is configured to screen the peak candidate set based on the adaptive prominence threshold to retain peaks with a prominence degree not less than the adaptive prominence threshold, and then output a spectral data index position corresponding to the peaks.

[0155] Further, the noise floor shaping module comprises:

[0156] The third calculation module is configured to subtract the preprocessed data from the smooth curve data to obtain an initial noise floor shaping result.

[0157] The normalization module is configured to perform minimum-maximum normalization processing on the initial noise floor shaping result to obtain second normalized data.

[0158] The fourth calculation module is configured to calculate a second average value of the second normalized data, set all spectral line values less than the second average value in the second normalized data to the second average value, and obtain the noise floor shaping result.

[0159] Further, the first calculation module comprises:

[0160] The search module is configured to search left and right directions of the current peak candidate point respectively to find conditional saddle points in the left and right directions respectively, the conditional saddle point being a first point higher than the current peak candidate point or the second average value encountered in the search direction.

[0161] The prominence degree module is configured to take a maximum value of the two conditional saddle points in the left and right directions, subtract the maximum value from a height of the peak candidate point to obtain the peak candidate point, and obtain the prominence degree.

[0162] The application is further described below through specific embodiments:

[0163] Embodiment 1: Wireless communication spectrum analysis

[0164] The method is applied in spectrum signal analysis to accurately extract channel peak values for signal quality evaluation and interference detection (the fixed sensitivity threshold is 0.14 in the embodiment).

[0165] In the first step, the minimum-maximum normalization processing is performed on the input original spectrum data x to eliminate the dimension influence.

[0166] In the second step, the mean value of the noise floor is preprocessed to weaken the influence of the noise floor on the peak prominence.

[0167] In the third step, the large-scale smoothing filter processing is performed on the preprocessed data by using the sliding average filter with the adaptive window size (the window size is five times the thousandth of the data point number in the embodiment).

[0168] In the fourth step, the noise floor shaping processing is performed to eliminate the background noise and baseline drift, enhance the peak prominence, and improve the robustness of the adaptive prominence threshold extracted by the subsequent IQR algorithm.

[0169] In the fifth step, the AMPD algorithm is used to detect the multi-scale peak values of the mean value processing result after the noise floor shaping to find all possible peak value candidate sets (the window scale of the AMPD algorithm is the thousandth of the data point number).

[0170] In the sixth step, after the peak prominence of each peak value candidate set is calculated, the adaptive prominence threshold is extracted by using the IQR algorithm, combined with the fixed sensitivity threshold, to screen the peak value candidate set, and the peak extraction result is as shown in FIG. 2. Figure 8

[0171] Embodiment 2: Suppression interference analysis of ultra-wideband radar

[0172] In the suppression interference signal of the ultra-wideband radar, the method is used to extract the characteristic frequency peak value to realize the real-time interference detection and adjust the radar parameter (the fixed sensitivity threshold is 0.14 in the embodiment).

[0173] In the first step, the minimum-maximum normalization processing is performed on the input original spectrum data x to eliminate the dimension influence.

[0174] In the second step, the mean value of the noise floor is preprocessed to weaken the influence of the noise floor on the peak prominence.

[0175] In the third step, the large-scale smoothing filter processing is performed on the preprocessed data by using the sliding average filter with the adaptive window size (the window size is five times the thousandth of the data point number in the embodiment).

[0176] In the fourth step, the noise floor shaping processing is performed to eliminate the background noise and baseline drift, enhance the peak prominence, and improve the robustness of the adaptive prominence threshold extracted by the subsequent IQR algorithm.

[0177] ​In the fifth step, the AMPD algorithm is used to detect the multi-scale peak value of the mean value processing result after the noise shaping, and all possible peak position candidate sets (the window scale of the AMPD algorithm is 1 / 1000 of the number of data points) are found.

[0178] In the sixth step, after the peak prominence of each peak value in the peak value candidate set is calculated, the IQR algorithm is used to extract an adaptive prominence threshold, and the peak value candidate set is screened in combination with a fixed sensitivity threshold, and the peak extraction result is as shown in Figure 9

[0179] The signal peak value detection method and system based on the adaptive prominence threshold have the following advantages compared with the prior art:

[0180] (1) The multi-stage processing flow is innovative: normalization, mean value, smoothing filter, noise shaping, and peak detection.

[0181] (2) The noise shaping processing eliminates the influence of noise fluctuation and drift on peak detection, and improves the stability and universality of peak prominence extraction.

[0182] (3) The adaptive threshold mechanism is a double guarantee of the IQR statistical method combined with the sensitivity threshold.

[0183] (4) The fully automatic processing does not require manual parameter optimization and is suitable for various spectrum characteristics.

[0184] (5) The abnormal processing mechanism has perfect threshold calibration and data quantity adaptability adjustment.

[0185] The above description is only the preferred embodiment of the present application, and it should be pointed out that for ordinary skilled in the art, without departing from the technical principles of the present application, a number of improvements and refinements can be made, and these improvements and refinements should be considered as the protection scope of the present application.​

Claims

1. A method for signal peak detection based on adaptive salience threshold, characterized in that, The method comprises the following steps: obtaining original spectrum data of an ultra-wideband radar jamming signal; preprocessing the spectrum data to obtain preprocessed data; performing smoothing processing on the preprocessed data to obtain smoothed curve data; performing bottom noise shaping on the smoothed curve data based on the preprocessed data to obtain a bottom noise shaping result; performing multi-scale peak detection on the bottom noise shaping result based on an AMPD algorithm to obtain a peak candidate set; for each peak candidate point, calculating its prominence relative to the bottom noise shaping result based on a watershed algorithm to obtain a list of prominence values corresponding to all peak candidate points; calculating an adaptive prominence threshold based on an IQR statistical algorithm according to the list of prominence values; performing screening on the peak candidate set based on the adaptive prominence threshold, retaining peaks with a prominence not less than the adaptive prominence threshold, and then outputting the index position of the spectrum data corresponding to the peak, wherein the peak with a prominence not less than the adaptive prominence threshold is the peak of the ultra-wideband radar jamming signal, and finally realizing jamming detection.

2. The method of claim 1, wherein, The preprocessing of the spectrum data to obtain the preprocessed data comprises the following steps: performing minimum-maximum normalization processing on the spectrum data to obtain first normalized data: calculating a first average value of the first normalized data, and setting all spectral line values less than the first average value in the first normalized data to the first average value to obtain the preprocessed data.

3. The method of claim 1, wherein the method is characterized by: The bottom noise shaping on the smoothed curve data based on the preprocessed data to obtain the bottom noise shaping result comprises the following steps: subtracting the preprocessed data from the smoothed curve data to obtain an initial bottom noise shaping result; performing minimum-maximum normalization processing on the initial bottom noise shaping result to obtain second normalized data: calculating a second average value of the second normalized data, and setting all spectral line values less than the second average value in the second normalized data to the second average value to obtain the bottom noise shaping result.

4. The method of claim 3, wherein the method further comprises: The multi-scale peak detection on the bottom noise shaping result based on the AMPD algorithm to obtain the peak candidate set comprises the following steps: initializing a local maximum scale according to the number of data points of the spectrum data; calculating a local maximum value matrix according to the bottom noise shaping result and the local maximum scale; calculating a scale consistency vector according to the local maximum value matrix; calculating an optimal detection scale according to the scale consistency vector; creating a consistent peak candidate set according to the local maximum value matrix and the optimal detection scale to obtain the peak candidate set.

5. The method of claim 4, wherein the method further comprises: For each peak candidate point, the prominence thereof relative to the bottom noise shaping result is calculated based on the watershed algorithm, which comprises the following steps: searching in the left and right directions of the current peak candidate point respectively to find the conditional saddle points in the left and right directions, wherein the conditional saddle point is the first point higher than the current peak candidate point or the second average value encountered in the search direction; taking the maximum value of the two conditional saddle points in the left and right directions, subtracting the maximum value from the height of the peak candidate point to obtain the prominence of the peak candidate point.

6. The method of claim 1, wherein, The adaptive prominence threshold is calculated based on the IQR statistical algorithm according to the list of prominence values, which comprises the following steps: determining whether the list of prominence values is empty; if yes, setting a preset fixed sensitivity threshold as the adaptive prominence threshold; if no, calculating an IQR threshold based on the quartile range method, and taking the minimum value of the IQR threshold and the preset fixed sensitivity threshold as the adaptive prominence threshold.

7. The method of claim 1, wherein the method is characterized by: The smoothing filtering processing on the pretreatment data comprises adopting a sliding average filter with an adaptive window size to smooth the pretreatment data.

8. A system for the adaptive threshold based signal peak detection method according to any one of claims 1 to 7, characterized in that, The method comprises the following steps: An acquisition module is configured to acquire original spectrum data of an ultra-wideband radar jamming signal. A pretreatment module is configured to pretreat the spectrum data to obtain pretreatment data. A smoothing processing module is configured to perform smoothing processing on the pretreatment data to obtain smoothed curve data. A noise floor shaping module is configured to perform noise floor shaping on the smoothed curve data based on the pretreatment data to obtain a noise floor shaping result. A detection module is configured to perform multi-scale peak value detection on the noise floor shaping result based on an AMPD algorithm to obtain a peak value candidate set. A first calculation module is configured to, for each peak value candidate point, calculate its prominence relative to the noise floor shaping result based on a watershed algorithm to obtain a list of prominence values corresponding to all peak value candidate points. A second calculation module is configured to calculate an adaptive prominence threshold based on an IQR statistical algorithm according to the list of prominence values. A screening module is configured to screen the peak value candidate set based on the adaptive prominence threshold, retain peak values with a prominence not less than the adaptive prominence threshold, and then output the spectral data index positions corresponding to the peak values, wherein the peak values with a prominence not less than the adaptive prominence threshold are peak values of the ultra-wideband radar jamming signal, and finally realize jamming detection.

9. The signal peak detection system based on adaptive prominence threshold value according to claim 8, characterized in that, The noise floor shaping module comprises: A third calculation module is configured to subtract the pretreatment data from the smoothed curve data to obtain an initial noise floor shaping result. A normalization module is configured to perform minimum-maximum normalization processing on the initial noise floor shaping result to obtain second normalized data. A fourth calculation module is configured to calculate a second average value of the second normalized data, set all spectral line values less than the second average value in the second normalized data to the second average value, and obtain the noise floor shaping result.

10. The signal peak detection system based on adaptive prominence threshold value according to claim 8, characterized in that, The first calculation module comprises: A search module is configured to search in the left and right directions of the current peak value candidate point respectively to find conditional saddle points in the left and right directions respectively, wherein the conditional saddle point is the first point higher than the current peak value candidate point or the second average value encountered in the search direction. A prominence module is configured to take the maximum value of the two conditional saddle points in the left and right directions, subtract the maximum value from the height of the peak value candidate point to obtain the prominence of the peak value candidate point.

Images