Signal source direction positioning method based on machine learning

Abstract

The invention discloses a signal source direction positioning method based on machine learning, and belongs to the technical field of signal source positioning, and the method comprises the following steps: determining peak points of X, Y and Z channel data according to a signal-to-noise ratio (SNR), selecting effective pulse signals in antenna array signals, selecting pulse signals of three channels, carrying out data cleaning, feature dimension reduction and feature extraction, and carrying out feature extraction; the random forest model training speed and the model prediction accuracy are improved, and the signal feature extraction in the same time window meets the time synchronism. Training a random forest model to predict the signal source direction, predicting the signal source direction of a single antenna, finally calculating an array antenna prediction result by adopting a voting method, and outputting the array antenna prediction result. According to the method, the defect that a traditional direction finding method is insufficient in adaptive capacity to various errors is overcome, near-real-time estimation is achieved, the low signal-to-noise ratio adaptive capacity and the space angle resolution capacity are enhanced, and meanwhile the method has the advantages of being resistant to environment interference and low signal-to-noise ratio environment and the like.

Description

Technical Field

[0001] The invention relates to a signal source direction positioning method based on machine learning, and belongs to the technical field of signal source positioning. Background Art

[0002] Signal processing is widely used in many fields, among which array signal processing is a key component. It plays an important role in wireless communications, radar sonar, aerospace and other fields, and promotes the development of these fields. The basic principle of array signal processing is to arrange a number of sensors according to a specific spatial geometry (such as linear, L-shaped, circular, etc.) to form a sensor array. When there is a signal to be received in space, the sensor array collaborates to collect, transmit and store data to obtain various information about the spatial signal source for subsequent analysis and processing.

[0003] The signal source direction positioning in signal processing has important research significance. In the civilian field, signal source positioning is widely used in meteorological radar, earthquake detection, fishing vessel navigation and biomedicine, and plays an important role. In the military field, signal source positioning is the basis for target positioning, radiation source identification, threat level determination and effective implementation of interference, which is of great significance to modern electronic warfare.

[0004] Array signal processing is an important branch of modern signal processing and is widely used in mobile communications, radar, sonar, radio astronomy and other fields. At present, the classic array signal source positioning method belongs to the model-driven method, such as the MUSIC algorithm, the ESPRIT algorithm and its various improved methods. However, the model-driven method always faces severe challenges in actual engineering applications, such as complex environments such as array errors and low signal-to-noise ratios, and the inability to predict the direction of the signal source in real time. In recent years, machine learning has been introduced into array signal processing as a data-driven tool to improve adaptability in complex environments. Through machine learning, the waveform characteristics of the signal source can be extracted from the observed data, and a nonlinear mapping relationship between the characteristics and the signal angle can be established. Using machine learning to achieve high-precision, real-time and intelligent signal source direction positioning has become a key issue to be solved.

[0005] Signals are easily affected by the environment. For some signals with low signal-to-noise ratio, the traditional model-driven method is ineffective in predicting the direction of the signal source. The traditional method requires multiple antennas and high signal-to-noise ratio data to locate the direction of the signal source. In reality, there are many signals with low energy, low signal-to-noise ratio, and weak signals received by array antennas. In this case, it is difficult for traditional methods to accurately locate the direction of the signal source. This paper proposes a signal source direction positioning method based on machine learning. Summary of the invention

[0006] Signal source positioning is widely used in radar, wireless communication, astronomy and other fields. Traditional methods mainly include beamforming, subspace, maximum likelihood and sparse reconstruction. However, these methods are often difficult to adapt to complex and changeable signals and application environments in practical applications, and cannot meet the needs of fast (real-time) positioning of signal sources. In the case of low signal-to-noise ratio, weak signal, and few signal receivers (very few signal receivers receive signal data), traditional signal source positioning methods will fail. In response to the above problems, the present invention proposes a signal source direction positioning method based on machine learning.

[0007] The technical solution adopted by the present invention is: a signal source direction positioning method based on machine learning, the main idea is: obtain antenna array signal data for data cleaning, feature extraction, normalization, and train a random forest model to predict the source direction of the signal in the antenna array. The main steps are as follows:

[0008] Step 1: Get the antenna array signal data set data = {TRACEX, TRACEY, TRACEZ, AZIMUTH, ZENITH}, where TRACEX represents the direction x channel data set, and the i-th direction x channel data TRACEX i ={x1,x2,…,x n}, where x1,x2,x n Indicates the x-channel data TRACEX in the i-th direction i The first element, the second element, the nth element. TRACEY represents the y-channel data set, where the i-th y-channel data TRACEY i ={y1,y2,…,y n}, where y1,y2,y n Indicates the y channel data TRACEY in the i-th direction i The first element, the second element, the nth element. TRACEZ represents the z-channel data set, where the i-th z-channel data TRACEZ i ={z1,z2,…,z n}, where z1,z2,z n Indicates the z channel data TRACEX in the i-th direction iThe first element, the second element, and the nth element. AZIMUTH represents azimuth angle data, and ZENITH represents zenith angle data. Define the signal-to-noise ratio threshold snrthreshold for filtering data, feature matrix X, true value matrix Y, signal-to-noise ratio SNRX of direction x channel data set TRACEX, signal-to-noise ratio SNRY of direction y channel data set TRACEY, signal-to-noise ratio SNRZ of direction z channel data set TRACEZ, pulse signal data POSWINDOWX of direction x channel data set TRACEX, pulse signal data POSWINDOWY of direction y channel data set TRACEY, pulse signal data POSWINDOWZ of direction z channel data set TRACEZ, pulse signal data POSWINDOW composed of pulse signal data POSWINDOW of direction x channel data set TRACEX, direction y channel data set TRACEY, and direction z channel data set TRACEZ, azimuth angle set AZIMUTHSET is the set of azimuth angle data AZIMUTH, and zenith angle set ZENITHSET is the set of zenith angle data ZENITH The set of H, the Cartesian product number set LABEL of the azimuth angle set AZIMUTHSET and the zenith angle set ZENITHSET, the pulse signal windows windows of the direction x channel data set TRACEX, the direction y channel data set TRACEY, and the direction z channel data set TRACEZ, the pulse signal transition window windowb of the direction x channel data set TRACEX, the direction y channel data set TRACEY, and the direction z channel data set TRACEZ, the peak point POSX of the direction x channel data set TRACEX, the peak point POSY of the direction y channel data set TRACEY, and the peak point POSZ of the direction z channel data set TRACEZ, the peak point POS generated according to the direction x channel data set TRACEX, the direction y channel data set TRACEY, and the direction z channel data set TRACEZ, the prediction result set YPRED, and the prediction result category set YCOUNT.

[0009] Step 2: Traverse the antenna array signal data set data and calculate the x-channel data TRACEX in the i-th direction according to the following formula i Peak Point POSX i

[0010] POSX i = argmax(|TRACEX i |)

[0011] Among them POSX i Indicates the x-channel data TRACEX in the i-th direction i Peak point, TRACEX iIndicates the ith data of the direction x channel data set TRACEX. Define the square average value set MSX of the pulse signal of the direction x channel data set TRACEX, and calculate the ith data of the direction x channel data TRACEX according to the following formula i Pulse signal square average MSX i ,

[0012]

[0013] Among them MSX i Indicates the x-channel data TRACEX in the i-th direction i Square average, n represents the x channel data TRACEX in the i-th direction i Length, windows represents the x-channel data TRACEX in the i-th direction i Pulse signal window. TRACEX i,j Indicates the x-channel data TRACEX in the i-th direction i The jth element defines the background signal square average value set MSBX of the direction x channel data set TRACEX. The background signal ith direction x channel data TRACEX is calculated according to the following formula i Square mean MSBX i .

[0014]

[0015] Among them MSBX i Indicates the x-channel data TRACEX in the i-th direction i The average square of the background signal, n represents the data of the x channel in the i-th direction TRACEX i Length, windowb represents the x-channel data TRACEX in the i-th direction i Pulse signal transition window, TRACEX i,j Indicates the x-channel data TRACEX in the i-th direction i The jth element. Define the signal-to-noise ratio set SNRX of the direction x channel data set TRACEX, and calculate the i-th direction x channel data TRACEX according to the following formula i Signal-to-Noise Ratio SNRX i

[0016]

[0017] Among them SNRX i Indicates the i-th x channel data TRACEX i Signal-to-Noise Ratio, MSX i Indicates the x-channel data TRACEX in the i-th direction i Pulse signal data square average, MSBXi Indicates the x-channel data TRACEX in the i-th direction i The squared average of the background signal.

[0018] Step 3: Calculate the signal-to-noise ratio SNRY of the y-direction channel data set TRACEY and the signal-to-noise ratio SNRZ of the z-direction channel data set TRACEZ according to step 2 above.

[0019] Step4: Traverse the antenna array signal data set data, select the direction x channel data set TRACEX signal-to-noise ratio SNRX greater than the signal-to-noise ratio threshold snrthreshold or the direction y channel data set TRACEY signal-to-noise ratio SNRY greater than the signal-to-noise ratio threshold snrthreshold or the direction z channel data set TRACEZ signal-to-noise ratio SNRZ greater than the signal-to-noise ratio threshold snrthreshold to generate a valid data set subdata = {TRACEX, TRACEY, TRACEZ, SNRX, SNRY, SNRZ, AZIMUTH, ZENITH}. Among them, TRACEX represents the direction x channel data set, TRACEY represents the direction y channel data set, TRACEZ represents the direction z channel data set, SNRX represents the direction x channel data set TRACEX signal-to-noise ratio, SNRY represents the direction y channel data set TRACEY signal-to-noise ratio, SNRZ represents the direction z channel data set TRACEZ signal-to-noise ratio. AXIMUTH represents the azimuth data, and ZENITH represents the zenith angle data.

[0020] Step 5: Traverse the valid data set subdata and calculate the peak point POSXi of the x-channel data TRACEXi in the i-th direction according to the following formula

[0021] POSX i = argmax(|TRACEX i |)

[0022] Among them POSX i Indicates the x-channel data TRACEX in the i-th direction i Peak point, TRACEX i Indicates the x-channel data in the ith direction. Calculate the y-channel data TRACEY in the ith direction according to the following formula i Peak point POSY i

[0023] POSY i =argmax(|TRACEY i |)

[0024] Among them POSY i Indicates the y channel data TRACEY in the i-th directioni Peak point, TRACEY i Indicates the y channel data in the i-th direction. Calculate the z channel data in the i-th direction TRACEZ according to the following formula i Peak point POSZ i

[0025] POSZ i =argmax(|TRACEZ i |)

[0026] POSZ i Indicates the z channel data TRACEZ in the i-th direction i Peak point, TRACEZ i Represents the data of the z-channel in the i-th direction. Find the peak point POS from the TRACEX peak point POSX of the x-channel data set, the TRACEY peak point POSY of the y-channel data set, and the TRACEZ peak point POSZ of the z-channel data set, and calculate the peak point POS of the i-th data according to the following formula i ,

[0027]

[0028] Among them, POS i Indicates the peak point of the i-th data, POSX i Indicates the x-channel data TRACEX in the i-th direction i Peak point, POSY i Indicates the y channel data TRACEY in the i-th direction i Peak point, POSZ i Indicates the z channel data TRACEZ in the i-th direction i Peak point, SNRX i SNRY i 、SNRZ i Respectively represent TRACEX i TRACEY i TRACEZ i Signal-to-noise ratio.

[0029] Step 6: Traverse the valid data set subdata and calculate the x channel data TRACEX in the i-th direction according to the following formula i Pulse signal data POSWINDOWX i

[0030]

[0031] POSWINDOWX i Indicates the x-channel data TRACEX in the i-th direction iPulse signal data, n represents the x-channel data TRACEX in the i-th direction i Length, windows represents the x-channel data TRACEX in the i-th direction i Pulse signal window, POS i Indicates the peak point of the i-th data. TRACEX i,j Indicates the x-channel data TRACEX in the i-th direction i The jth element.

[0032] Calculate the y channel data TRACEY in the i-th direction according to the following formula i Pulse signal data POSWINDOWY i ,

[0033]

[0034] POSWINDOWY i Indicates the y channel data TRACEY in the i-th direction i Pulse signal data, n represents the y channel data TRACEY in the i-th direction i Length, windows represents the y channel data TRACEY in the i-th direction i Pulse signal window, POS i Indicates the peak point of the i-th data, TRACEY i,j Indicates the y channel data TRACEY in the i-th direction i The jth element.

[0035] Calculate the z channel data TRACEZ in the i-th direction according to the following formula i Pulse signal data POSWINDOWZ i ,

[0036]

[0037] POSWINDOWZ i Indicates the z channel data TRACEZ in the i-th direction i Pulse signal data, n represents the z channel data TRACEZ in the i-th direction i Length, windows represents the z channel data TRACEZ in the i-th direction i Pulse signal window, POS i Indicates the peak point of the i-th data, TRACEZ i,j Indicates the z channel data TRACEZ in the i-th direction i The jth element.

[0038] Step 7: Traverse the valid data set subdata and generate the input data set datainput = {(POSWINDOWX, POSWINDOWY, POSWINDOWZ, AZIMUTH, ZENITH)}, where POSWINDOWX represents the direction x channel data set TRACEX pulse signal data, where the i-th direction x channel data set TRACEX pulse signal data POSWINDOWX i ={x1,x2,…,x n}, where x1,x2,x n Indicates the pulse signal data POSWINDOWX of the x-channel data set TRACEX in the i-th direction i The first element, the second element, the nth element. POSWINDOWY represents the direction y channel data set TRACEY pulse signal data set, where the i-th direction y channel data set TRACEY pulse signal data POSWINDOWY i ={y1,y2,…,y n}, where y1,y2,y n Indicates the y-channel data set TRACEY pulse signal data POSWINDOWY in the i-th direction i The first element, the second element, the nth element, POSWINDOWZ represents the z-channel data set TRACEZ pulse signal data set, where the i-th z-channel data set TRACEZ pulse signal data POSWINDOWZ i ={z1,z2,…,z n}, where z1,z2,z n Indicates the pulse signal data POSWINDOWZ of the z channel data set TRACEZ in the i-th direction i The first element, the second element, the nth element. AZIMUTH represents the azimuth angle data, and ZENITH represents the zenith angle data. Define the pulse signal data set POSWINDOW = {(POSWINDOWX1, POSWINDOWY1, POSWINDOWZ1), (POSWINDOWX2, POSWINDOWY2, POSWINDOWZ2),

[0039] …,

[0040] (POSWINDOWX m ,POSWINDOWY m ,POSWINDOWZ m )}

[0041] Where m represents the size of the pulse signal data set, POSWINDOWX1, POSWINDOWX2, POSWINDOWX m Indicates the first element, second element, and mth element of the pulse signal data set POSWINDOWX in the direction x channel data set TRACEX. POSWINDOWY1, POSWINDOWY2, POSWINDOWY m Indicates the first, second, and mth elements of the pulse signal data set POSWINDOWY in the direction y channel data set TRACEY. POSWINDOWZ1, POSWINDOWZ2, POSWINDOWZ m Indicates the first, second, and mth elements of the pulse signal data set POSWINDOWZ in the z-channel data set TRACEZ. Define the minimum value of the i-th pulse signal data to be POSWINDOW i,min , the maximum value of the i-th pulse signal data is POSWINDOW i,max According to the following formula, the POSWINDOW of the i-th pulse signal data i The jth element POSWINDOW i,j Normalize it,

[0042] POSWINDOW i,min = min(POSWINDOW i )

[0043] POSWINDOW i,max =max(POSWINDOW i )

[0044]

[0045] POSWINDOW i,min Indicates the i-th pulse signal data POSWINDOW i Minimum, POSWINDOW i,max Indicates the i-th pulse signal data POSWINDOW i Maximum value, POSWINDOW i,j Represents the jth element of the i-th pulse signal data.

[0046] Step 8: Define the Cartesian product number set LABEL of the azimuth set AZIMUTHSET and the zenith angle set ZENITHSET, and calculate the Cartesian product number set LABEL of the azimuth set AZIMUTHSET and the zenith angle set ZENITHSET according to the following formula

[0047] LABEL=AZIMUTHSET×ZENITHSET

[0048] ={LABEL 0,0 ,LABEL 0,1 ,…,LABEL m,n}

[0049] in,

[0050] LABEL i,j =(AZIMUTHSET i ,ZENITHSET j ),i∈(0,m),j∈(0,n)

[0051] Where LABEL i,j It represents the Cartesian product number of the i-th element of the azimuth angle set AZIMUTHSET and the j-th element of the zenith angle set ZENITHSET. m represents the size of the azimuth angle set AZIMUTHSET, and n represents the size of the zenith angle set ZENITHSET. Construct the input feature data matrix X = {X1, X2, X3, ..., X n}, where X i ={POSWINDOW i}, where POSWINDOW i represents the i-th pulse signal data. Select the real value matrix Y = {Y1, Y2, Y3, ..., Y n}, where Y i =(AZIMUTH i ,ZENITH i )=LABEL i,i AZIMUTH i Indicates the azimuth AZIMUTH data of the i-th data, ZENITH i Indicates the i-th zenith angle ZENITH data, LABEL i,i Indicates the Cartesian product number of the i-th azimuth angle data AZIMUTH and the i-th zenith angle data ZENITH.

[0052] Step 9: Select the first 70% of the feature matrix X and the true value matrix Y as the training data set train_set, and the remaining 30% as the validation set valid_set. Initialize the random forest model parameters, and train the random forest model with the training data set train_set and the validation data set valid_set. The trained random forest model inputs the feature matrix X and returns the prediction result set YPRED.

[0053] Step 10: Define the predicted value ytrue that appears the most times in the prediction result set YPRED. Count the number of occurrences of elements in the prediction result set YPRED to generate the prediction result category set YCOUNT, traverse the prediction result set YPRED, and calculate the number of occurrences of the i-th predicted result set YPRED element YCOUNT according to the following formula i ,

[0054] YCOUNT i =YCOUNT i +1,i=YPRED i

[0055] According to the following formula, calculate the element ytrue that appears most times in the prediction result set:

[0056] ytrue=argmax(YCOUNT)

[0057] Among them, ytrue represents the element with the most occurrences in the prediction result YPRED, and YCOUNT represents the number of occurrences of the elements in the prediction result YPRED. According to the element ytrue with the most occurrences in the prediction result, find the corresponding zenith angle and azimuth angle, that is, ytrue=LABLE i,j =(AZIIMUTHSET i ,ZENITHSET j ), where ytrue represents the element with the most occurrences in the prediction result set YPRED, LABEL i,j Indicates the Cartesian product number of the i-th element of the azimuth set AZIMUTHSET and the j-th element of the zenith angle set ZENITHSET, AZIMUTHSET i Represents the i-th element of the azimuth set, ZENITHSET j Represents the jth element of the zenith angle set, outputting the azimuth angle AZIMUTHSET i ,ZENITHSET j .

[0058] Beneficial effects:

[0059] 1. The present invention extracts features related to the incident angle of the signal source from the array observation signal, and establishes a nonlinear mapping relationship between the features and the signal angle parameters through a machine learning model. This solves the defect of the traditional direction finding method that it is not adaptable to various errors, realizes near real-time estimation, and enhances the adaptability to low signal-to-noise ratio and spatial angle resolution.

[0060] 2. The present invention accelerates model training through feature extraction. To address the problem that deep neural networks have large training data and are prone to overfitting, the random forest model is used to adopt an integrated model of multiple decision trees to reduce the data volume required for training the model and shorten the training time, thereby reducing the risk of overfitting and improving the model effect.

[0061] 3. The present invention has the advantages of being resistant to environmental interference and low signal-to-noise ratio environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 It is a flow chart of the present invention.

[0063] Figure 2 It is a schematic diagram of the system application scenario of the present invention. DETAILED DESCRIPTION

[0064] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0065] like Figure 1 As shown, a signal source direction positioning method based on machine learning includes the following steps:

[0066] Step 1: Get the antenna array signal data set data = {TRACEX, TRACEY, TRACEZ, AZIMUTH, ZENITH}, where TRACEX represents the direction x channel data set, and the i-th direction x channel data TRACEX i ={x1,x2,…,x n}, where x1,x2,x n Indicates the x-channel data TRACEX in the i-th direction i The first element, the second element, the nth element. TRACEY represents the y-channel data set, where the i-th y-channel data TRACEY i ={y1,y2,…,y n}, where y1,y2,y n Indicates the y channel data TRACEY in the i-th direction i The first element, the second element, the nth element. TRACEZ represents the z-channel data set, where the i-th z-channel data TRACEZ i ={z1,z2,…,z n}, where z1,z2,z n Indicates the z channel data TRACEZ in the i-th direction iThe first element, the second element, and the nth element. AZIMUTH represents azimuth angle data, and ZENITH represents zenith angle data. Define the signal-to-noise ratio threshold snrthreshold for filtering data, feature matrix X, true value matrix Y, signal-to-noise ratio SNRX of direction x channel data set TRACEX, signal-to-noise ratio SNRY of direction y channel data set TRACEY, signal-to-noise ratio SNRZ of direction z channel data set TRACEZ, pulse signal data POSWINDOWX of direction x channel data set TRACEX, pulse signal data POSWINDOWY of direction y channel data set TRACEY, pulse signal data POSWINDOWZ of direction z channel data set TRACEZ, pulse signal data POSWINDOW composed of pulse signal data POSWINDOW of direction x channel data set TRACEX, direction y channel data set TRACEY, and direction z channel data set TRACEZ, azimuth angle set AZIMUTHSET is the set of azimuth angle data AZIMUTH, and zenith angle set ZENITHSET is the set of zenith angle data ZENITH The set of H, the Cartesian product number set LABEL of the azimuth angle set AZIMUTHSET and the zenith angle set ZENITHSET, the pulse signal windows windows of the direction x channel data set TRACEX, the direction y channel data set TRACEY, and the direction z channel data set TRACEZ, the pulse signal transition window windowb of the direction x channel data set TRACEX, the direction y channel data set TRACEY, and the direction z channel data set TRACEZ, the peak point POSX of the direction x channel data set TRACEX, the peak point POSY of the direction y channel data set TRACEY, and the peak point POSZ of the direction z channel data set TRACEZ, the peak point POS generated according to the direction x channel data set TRACEX, the direction y channel data set TRACEY, and the direction z channel data set TRACEZ, the prediction result set YPRED, and the prediction result category set YCOUNT.

[0067] Step 2: Traverse the antenna array signal data set data and calculate the x-channel data TRACEX in the i-th direction according to the following formula i Peak Point POSX i

[0068] POSX i = argmax(|TRACEX i |)

[0069] Among them POSX i Indicates the x-channel data TRACEX in the i-th direction i Peak point, TRACEX iIndicates the ith data of the direction x channel data set TRACEX. Define the square average value set MSX of the pulse signal of the direction x channel data set TRACEX, and calculate the ith data of the direction x channel data TRACEX according to the following formula i Pulse signal square average MSX i ,

[0070]

[0071] Among them MSX i Indicates the x-channel data TRACEX in the i-th direction i Square average, n represents the x channel data TRACEX in the i-th direction i Length, windows represents the x-channel data TRACEX in the i-th direction i Pulse signal window. TRACEX i,j Indicates the x-channel data TRACEX in the i-th direction i The jth element defines the background signal square average value set MSBX of the direction x channel data set TRACEX. The background signal ith direction x channel data TRACEX is calculated according to the following formula i Square mean MSBX i .

[0072]

[0073] Among them MSBX i Indicates the x-channel data TRACEX in the i-th direction i The average square of the background signal, n represents the data of the x channel in the i-th direction TRACEX i Length, windowb represents the x-channel data TRACEX in the i-th direction i Pulse signal transition window, TRACEX i,j Indicates the x-channel data TRACEX in the i-th direction i The jth element. Define the signal-to-noise ratio set SNRX of the direction x channel data set TRACEX, and calculate the i-th direction x channel data TRACEX according to the following formula i Signal-to-Noise Ratio SNRX i

[0074]

[0075] Among them SNRX i Indicates the i-th x-channel data TRAXEX i Signal-to-Noise Ratio, MSX i Indicates the x-channel data TRACEX in the i-th direction i Pulse signal data square average, MSBXi Indicates the x-channel data TRACEX in the i-th direction i The squared average of the background signal.

[0076] Step 3: Calculate the signal-to-noise ratio SNRY of the y-direction channel data set TRACEY and the signal-to-noise ratio SNRZ of the z-direction channel data set TRACEZ according to step 2 above.

[0077] Step 4: Traverse the antenna array signal data set data, select the direction x channel data set TRACEX signal-to-noise ratio SNRX greater than the signal-to-noise ratio threshold snrthreshold or the direction y channel data set TRACEY signal-to-noise ratio SNRY greater than the signal-to-noise ratio threshold snrthreshold or the direction z channel data set TRACEZ signal-to-noise ratio SNRZ greater than the signal-to-noise ratio threshold snrthreshold to generate a valid data set subdata = {TRACEX, TRACEY, TRACEZ, SNRX, SNRY, SNRZ, AZIMUTH, ZENITH}. Among them, TRACEY represents the direction x channel data set, TRACEY represents the direction y channel data set, TRACEZ represents the direction z channel data set, SNRX represents the direction x channel data set TRACEX signal-to-noise ratio, SNRY represents the direction y channel data set TRACEY signal-to-noise ratio, and SNRZ represents the direction z channel data set TRACEZ signal-to-noise ratio. AZIMUTH represents the azimuth data, and ZENITH represents the zenith angle data.

[0078] Step 5: Traverse the valid data set subdata and calculate the x channel data TRACEX in the i-th direction according to the following formula i Peak Point POSX i

[0079] POSX i = argmax(|TRACEX i |)

[0080] Among them POSX i Indicates the x-channel data TRACEX in the i-th direction i Peak point, TRACEY i Indicates the x-channel data in the ith direction. Calculate the y-channel data TRACEY in the ith direction according to the following formula i Peak point POSY i

[0081] POSY i =argmax(|TRACEY i |)

[0082] Among them POSY iIndicates the y channel data in the i-th direction ARACEY i Peak point, TRACEY i Indicates the y channel data in the i-th direction. Calculate the z channel data in the i-th direction TRACEZ according to the following formula i Peak point POSZ i

[0083] POSZ i =argmax(|TRACEZ i |)

[0084] POSZ i Indicates the z channel data TRACEZ in the i-th direction i Peak point, TRACEZ i Represents the data of the z-channel in the i-th direction. Find the peak point POS from the TRACEX peak point POSX of the x-channel data set, the TRACEY peak point POSY of the y-channel data set, and the TRACEZ peak point POSZ of the z-channel data set, and calculate the peak point POS of the i-th data according to the following formula i ,

[0085]

[0086] Among them, POS i Indicates the peak point of the i-th data, POSX i Indicates the x-channel data TRACEX in the i-th direction i Peak point, POSY i Indicates the y channel data TRACEY in the i-th direction i Peak point, POSZ i Indicates the z channel data TRACEZ in the i-th direction i Peak point, SNRX i SNRY i 、SNRZ i Respectively represent TRACEX i TRACEY i TRACEZ i Signal-to-noise ratio.

[0087] Step 6: Traverse the valid data set subdata and calculate the x channel data TRACEX in the i-th direction according to the following formula i Pulse signal data POSWINDOWX i

[0088]

[0089] POSWINDOWX iIndicates the x-channel data TRACEX in the i-th direction i Pulse signal data, n represents the x-channel data TRACEX in the i-th direction i Length, windows represents the x-channel data TRACEX in the i-th direction i Pulse signal window, POS i Indicates the peak point of the i-th data. TRACEX i,j Indicates the x-channel data TRACEX in the i-th direction i The jth element. Calculate the y channel data TRACEY in the i-th direction according to the following formula i Pulse signal data POSWINDOWY i ,

[0090]

[0091] POSWINDOWY i Indicates the y channel data TRACEY in the i-th direction i Pulse signal data, n represents the y channel data TRACEY in the i-th direction i Length, windows represents the y channel data TRACEY in the i-th direction i Pulse signal window, POS i Indicates the peak point of the i-th data, TRACEY i,j Indicates the y channel data TRACEY in the i-th direction i The jth element. Calculate the i-th direction channel data TRACEZ according to the following formula i Pulse signal data POSWINDOWZ i ,

[0092]

[0093] POSWINDOWZ i Indicates the z channel data TRACEZ in the i-th direction i Pulse signal data, n represents the z channel data TRACEZ in the i-th direction i Length, windows represents the z channel data TRACEZ in the i-th direction i Pulse signal window, POS i Indicates the peak point of the i-th data, TRACEZ i,j Indicates the z channel data TRACEZ in the i-th direction i The jth element.

[0094] Step 7: Traverse the valid data set subdata and generate the input data set datainput = {(POSWINDOWX, POSWINDOWY, POSWINDOWZ, AZIMUTH, ZENITH)}, where POSWINDOWX represents the direction x channel data set TRACEX pulse signal data, where the i-th direction x channel data set TRACEX pulse signal data POSWINDOWX i ={x1,x2,…,x n}, where x1,x2,x n Indicates the pulse signal data POSWINDOWX of the x-channel data set TRACEX in the i-th direction i The first element, the second element, the nth element. POSWINDOWY represents the direction y channel data set TRACEY pulse signal data set, where the i-th direction y channel data set TRACEY pulse signal data POSWINDOWY i ={y1,y2,…,y n}, where y1,y2,y n Indicates the y-channel data set TRACEY pulse signal data POSWINDOWY in the i-th direction i The first element, the second element, the nth element, POSWINDOWZ represents the z-channel data set TRACEZ pulse signal data set, where the i-th z-channel data set TRACEZ pulse signal data POSWINDOWZ i ={z1,z2,…,z n}, where z1,z2,z n Indicates the pulse signal data POSWINDOWZ of the z channel data set TRACEZ in the i-th direction i The first element, the second element, the nth element. AZIMUTH represents the azimuth angle data, and ZENITH represents the zenith angle data. Define the pulse signal data set POSWINDOW = {(POSWINDOWX1, POSWINDOWY1, POSWINDOWZ1), (POSWINDOWX2, POSWINDOWY2, POSWINDOWZ2),

[0095] ...,(POSWINDOWX m ,POSWINDOWY m ,POSWINDOWZ m}

[0096] Where m represents the size of the pulse signal data set, POSWINDOWX1, POSWINDOWX2, POSWINDOWX mIndicates the first element, second element, and mth element of the pulse signal data set POSWINDOWX in the direction x channel data set TRACEX. POSWINDOWY1, POSWINDOWY2, POSWINDOWY m Indicates the first, second, and mth elements of the pulse signal data set POSWINDOWY in the direction y channel data set TRACEY. POSWINDOWZ1, POSWINDOWZ2, POSWINDOWZ m Indicates the first, second, and mth elements of the pulse signal data set POSWINDOWZ in the z-channel data set TRACEZ. Define the minimum value of the i-th pulse signal data to be POSWINDOW i,min , the maximum value of the i-th pulse signal data is POSWINDOW i,max According to the following formula, the POSWINDOW of the i-th pulse signal data i The jth element POSWINDOW i,j Normalize it,

[0097] POSWINDOW i,min = min(POSWINDOW i )

[0098] POSWINDOW i,max =max(POSWINDOW i )

[0099]

[0100] POSWINDOW i,min Indicates the i-th pulse signal data POSWINDOW i Minimum, POSWINDOW i,max Indicates the i-th pulse signal data POSWINDOW i Maximum value, POSWINDOW i,j Represents the jth element of the i-th pulse signal data.

[0101] Step 8: Define the Cartesian product number set LABEL of the azimuth set AZIMUTHSET and the zenith angle set ZENITHSET, and calculate the Cartesian product number set LABEL of the azimuth set AZIMUTHSET and the zenith angle set ZENITHSET according to the following formula

[0102] LABEL=AZIMUTHSET×ZENITHSET

[0103] ={LABEL 0,0,LABEL 0,1 ,…,LABEL m,n}

[0104] in,

[0105] LABEL i,j =(AZIMUTHSET i ,ZENITHSET j ),i∈(0,m),j∈(0,n)

[0106] Where LABEL i,j It represents the Cartesian product number of the i-th element of the azimuth angle set AZIMUTHSET and the j-th element of the zenith angle set ZENITHSET. m represents the size of the azimuth angle set AZIMUTHSET, and n represents the size of the zenith angle set ZENITHSET. Construct the input feature data matrix X = {X1, X2, X3, ..., X n}, where X = {POSWINDOW i}, where POSWINDOW i represents the i-th pulse signal data. Select the real value matrix Y = {Y1, Y2, Y3, ..., Y n}, where Y i =(AZIMUTH i ,ZENITH i )=LABEL i,i AZIMUTH i Indicates the azimuth AZIMUTH data of the i-th data, ZENITH i Indicates the i-th zenith angle ZENITH data, LABEL i,i Indicates the Cartesian product number of the i-th azimuth angle data AZIMUTH and the i-th zenith angle data ZENITH.

[0107] Step 9: Select the first 70% of the feature matrix X and the true value matrix Y as the training data set train_set, and the remaining 30% as the validation set valid_set. Initialize the random forest model parameters, and train the random forest model with the training data set train_set and the validation data set valid_set. The trained random forest model inputs the feature matrix X and returns the prediction result set YPRED.

[0108] Step 10: Define the predicted value ytrue that appears the most times in the prediction result set YPRED. Count the number of occurrences of elements in the prediction result set YPRED to generate the prediction result category set YCOUNT, traverse the prediction result set YPRED, and calculate the number of occurrences of the i-th predicted result set YPRED element YCOUNT according to the following formula i ,

[0109] YCOUNT i =YCOUNT i +1,i=YPRED i

[0110] According to the following formula, calculate the element ytrue that appears most times in the prediction result set:

[0111] ytrue=argmax(YCOUNT)

[0112] Among them, ytrue represents the element with the most occurrences in the prediction result YPRED, and YCOUNT represents the number of occurrences of the elements in the prediction result YPRED. According to the element ytrue with the most occurrences in the prediction result, find the corresponding zenith angle and azimuth angle, that is, ytrue=LABLE i,j =(AZIMUTHSET i ,ZENITHSET j ), where ytrue represents the element with the most occurrences in the prediction result set YPRED, LABEL i,j Indicates the Cartesian product number of the i-th element of the azimuth set AZIMUTHSET and the j-th element of the zenith angle set ZENITHSEYT, AZIMUTHSET i Represents the i-th element of the azimuth set, ZENITHSET j Represents the jth element of the zenith angle set, outputting the azimuth angle AZIMUTHSET i ,ZENITHSET j .

[0113] Figure 2 This is a schematic diagram of the system application scenario of the present invention. The system of the present invention is applied in an array signal processing system. When an electromagnetic wave encounters an antenna, the antenna will induce an electromotive force, and the antenna will convert the received electromagnetic wave signal into a current signal. The signal receiving and processing unit converts the electromagnetic wave received by the antenna into an electrical signal to generate signal data and generate an array signal data set. The data preprocessing unit performs operations such as deduplication of array signal data and removal of outliers. The feature extraction unit determines the peak points of the X, Y, and Z channel data by calculating the signal-to-noise ratio SNR, selects the effective pulse signal in the antenna array signal, and constructs a feature matrix X based on the effective pulse signal in the antenna array signal. The random forest model unit uses the feature matrix X as an input value to output the predicted zenith angle and azimuth angle.

[0114] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A signal source direction positioning method based on machine learning, characterized in that: The steps include: Step 1: Get the antenna array signal data set data; Step 2: Traverse the antenna array signal data set data and calculate the x-channel data TRACEX in the i-th direction according to the following formula i Peak Point POSX i POSX i =argmax(|TRACEX i |) Among them POSX i Indicates the x-channel data TRACEX in the i-th direction i Peak point, TRACEX i Indicates the i-th data of the direction x channel data set TRACEX, Define the direction x channel data set TRACEX pulse signal square average value set MSX, and calculate the i-th direction x channel data TRACEX according to the following formula i Pulse signal square average MSX i , Among them MSX i Indicates the x-channel data TRACEX in the i-th direction i Square average, n represents the x channel data TRACEX in the i-th direction i Length, windows represents the x-channel data TRACEX in the i-th direction i Pulse signal window, TRACEX i，j Indicates the x-channel data TRACEX in the i-th direction i The jth element, Define the background signal square mean value set MSBX of the direction x channel data set TRACEX, and calculate the background signal of the ith direction x channel data TRACEX according to the following formula i Square mean MSBX i : Among them MSBX i Indicates the x-channel data TRACEX in the i-th direction i The average square of the background signal, n represents the data of the x channel in the i-th direction TRACEX i Length, windowb represents the x-channel data TRACEX in the i-th direction i Pulse signal transition window, TRACEX i，j Indicates the x-channel data TRACEX in the i-th direction i The jth element, Define the signal-to-noise ratio set SNRX of the direction x channel data set TRACEX, and calculate the i-th direction x channel data TRACEX according to the following formula i Signal-to-Noise Ratio SNRX i Among them SNRX i Indicates the i-th x channel data TRACEX i Signal-to-Noise Ratio, MSX i Indicates the x-channel data TRACEX in the i-th direction i Pulse signal data square average, MSBX i Indicates the x-channel data TRACEX in the i-th direction i The square mean of the background signal; Step 3: Calculate the signal-to-noise ratio SNRY of the y-channel data set TRACEY and the signal-to-noise ratio SNRZ of the z-channel data set TRACEZ according to step 2 above; Step 4: Traverse the antenna array signal data set data, select the direction x channel data set TRACEX whose signal-to-noise ratio SNRX is greater than the signal-to-noise ratio threshold snrthreshold, or the direction y channel data set TRACEY whose signal-to-noise ratio SNRY is greater than the signal-to-noise ratio threshold snrthreshold, or the direction z channel data set TRACEZ whose signal-to-noise ratio SNRZ is greater than the signal-to-noise ratio threshold snrthreshold to generate a valid data set subdata; Step 5: Traverse the valid data set subdata and calculate the x channel data TRACEX in the i-th direction i Peak Point POSX i , y channel data TRACEY in the i-th direction i Peak point POSY i and the z channel data TRACEZ in the i-th direction i Peak point POSZ i Then find the peak point POS from the peak point POSX of the direction x channel data set TRACEX, the peak point POSY of the direction y channel data set TRACEY, and the peak point POSZ of the direction z channel data set TRACEZ, and calculate the peak point POS of the i-th data. i ; Step 6: Traverse the valid data set subdata and calculate the x channel data TRACEX in the i-th direction i Pulse signal data POSWINDOWX i , y channel data TRACEY in the i-th direction i Pulse signal data POSWINDOWY i , z channel data TRACEZ in the i-th direction i Pulse signal data POSWINDOWZ i ; Step7: Traverse the valid data set subdata and generate the input data set datainput; Step 8: Define the Cartesian product number set LABEL of the azimuth angle set AZIMUTHSET and the zenith angle set ZENITHSET, and calculate the Cartesian product number set LABEL of the azimuth angle set AZIMUTHSET and the zenith angle set ZENITHSET; Step 9: Select the first 70% of the feature matrix X and the true value matrix Y as the training data set train_set, and the remaining 30% as the validation set valid_set, initialize the random forest model parameters, and train the random forest model with the training data set train_set and the validation data set valid_set. After training, the random forest model inputs the feature matrix X to return the prediction result set YPRED; Step 10: Define the predicted value with the highest number of occurrences in the prediction result set YPRED as ytrue, count the number of occurrences of elements in the prediction result set YPRED to generate the prediction result category set YCOUNT, traverse the prediction result set YPRED, and calculate the number of occurrences of the i-th predicted result set YPRED element YCOUNT according to the following formula i , YCOUNT i =YCOUNT i +1,i=YPRED i According to the following formula, calculate the element ytrue that appears most times in the prediction result set: ytrue=argmax(YCOUNT) Among them, ytrue represents the element with the largest number of occurrences in the prediction result YPRED, and YCOUNT represents the number of occurrences of the elements in the prediction result YPRED. According to the element ytrue with the largest number of occurrences in the prediction result, the corresponding zenith angle and azimuth angle are found, that is, ytrue=LABLE i，j =(AZIMUTHSET i ,ZENITHSET j ), where ytrue represents the element with the most occurrences in the prediction result set YPRED, LABEL i，j Indicates the Cartesian product number of the i-th element of the azimuth set AZIMUTHSET and the j-th element of the zenith angle set ZENITHSET, AZIMUTHSET i Represents the i-th element of the azimuth set, ZENITHSET j Represents the jth element of the zenith angle set, outputting the azimuth angle AZIMUTHSET i , zenith angle ZENITHSET j .

2. The signal source direction positioning method based on machine learning according to claim 1, characterized in that: In the Step 1, the antenna array signal data set data = {TRACEX, TRACEY, TRACEZ, AZIMUTH, ZENITH}, where TRACEX represents the direction x channel data set, and the i-th direction x channel data TRACEX i ={x1, x2, ..., x n }, where x1, x2, x n Indicates the x-channel data TRACEX in the i-th direction i The first element, the second element, the nth element, TRACEY represents the y-channel data set, where the i-th y-channel data TRACEY i ={y1, y2, ..., y n }, where y1, y2, y n Indicates the y channel data TRACEY in the i-th direction i The first element, the second element, the nth element, TRACEZ represents the z-channel data set, where the i-th z-channel data TRACEZ i ={z1, z2, ..., z n }, where z1, z2, z n Indicates the z channel data TRACEZ in the i-th direction i The first element, the second element, the nth element, AZIMUTH represents azimuth angle data, ZENITH represents zenith angle data, defines the signal-to-noise ratio threshold snrthreshold for filtering data, feature matrix X, true value matrix Y, signal-to-noise ratio SNRX of direction x channel data set TRACEX, signal-to-noise ratio SNRY of direction y channel data set TRACEY, signal-to-noise ratio SNRZ of direction z channel data set TRACEZ, pulse signal data POSWINDOWX of direction x channel data set TRACEY, pulse signal data POSWINDOWY of direction y channel data set TRACEY, pulse signal data POSWINDOWZ of direction z channel data set TRACEZ, pulse signal data POSWINDOW composed of pulse signal data POSWINDOW of direction x channel data set TRACEX, direction y channel data set TRACEY and direction z channel data set TRACEZ, azimuth angle set AZIMUTHSET is the set of azimuth angle data AZIMUTH, zenith angle set The angle set ZENITHSET is a set of zenith angle data ZENITH, the Cartesian product number set LABEL of the azimuth angle set AZIMUTHSET and the zenith angle set ZENITHSET, the pulse signal windows windows of the direction x channel data set TRACEX, the direction y channel data set TRACEY, and the direction z channel data set TRACEZ, the pulse signal transition window windowb of the direction x channel data set TRACEX, the direction y channel data set TRACEY, and the direction z channel data set TRACEZ, the peak point POSX of the direction x channel data set TRACEX, the peak point POSY of the direction y channel data set TRACEY, and the peak point POSZ of the direction z channel data set TRACEZ, the peak point POS is generated according to the direction x channel data set TRACEX, the direction y channel data set TRACEY, and the direction z channel data set TRACEZ, the prediction result set YPRED, and the prediction result category set YCOUNT.

3. The signal source direction positioning method based on machine learning according to claim 1, characterized in that: The valid data set subdata in Step 4 is {TRACEX, TRACEY, TRACEZ, SNRX, SNRY, SNRZ, AZIMUTH, ZENITH}, where TRACEX represents the direction x channel dataset, TRACEY represents the direction y channel dataset, TRACEZ represents the direction z channel dataset, SNRX represents the TRACEX signal-to-noise ratio of the direction x channel dataset, SNRY represents the TRACEY signal-to-noise ratio of the direction y channel dataset, SNRZ represents the TRACEZ signal-to-noise ratio of the direction z channel dataset, AZIMUTH represents the azimuth data, and ZENITH represents the zenith angle data.

4. The signal source direction positioning method based on machine learning according to claim 1, characterized in that: The x-channel data TRACEX in the i-th direction in Step 5 i Peak Point POSX i The calculation formula is as follows: POSX i =argmax(|TRACEX i |) Among them POSX i Indicates the x-channel data TRACEX in the i-th direction i Peak point, TRACEX i Represents the x-channel data in the i-th direction; The y channel data TRACEY in the i-th direction i Peak point POSY i The calculation formula is as follows: POSITION i =argmax(|TRACEY i |) Among them POSY i Indicates the y channel data TRACEY in the i-th direction i Peak point, TRACEY i Represents the y channel data in the i-th direction; The z channel data TRACEZ in the i-th direction i Peak point POSZ i The calculation formula is as follows: POSZ i =argmax(|TRACEZ i |) POSZ i Indicates the z channel data TRACEZ in the i-th direction i Peak point, TRACEZ i Represents the z channel data in the i-th direction; The peak point POS of the i-th data i The calculation formula is as follows: Among them, POS i Indicates the peak point of the i-th data, POSX i Indicates the x-channel data TRACEX in the i-th direction i Peak point, POSY i Indicates the y channel data TRACEY in the i-th direction i Peak point, POSZ i Indicates the peak point of the z-channel data TRACEZi in the i-th direction, SNRX i SNRY i 、SNRZ i Respectively represent TRACEX i TRACEY i TRACEZ i Signal-to-noise ratio.

5. The signal source direction positioning method based on machine learning according to claim 1, characterized in that: The x-channel data TRACEX in the i-th direction in Step 6 i Pulse signal data POSWINDOWX i The formula is as follows: POSWINDOWX i Indicates the x-channel data TRACEX in the i-th direction i Pulse signal data, n represents the x-channel data TRACEX in the i-th direction i Length, windows represents the x-channel data TRACEX in the i-th direction i Pulse signal window, POS i Indicates the peak point of the i-th data, TRACEX i，j Indicates the x-channel data TRACEX in the i-th direction i The jth element; The y channel data TRACEY in the i-th direction i Pulse signal data POSWINDOWY i The announcement is as follows: POSWINDOWY i Indicates the y channel data TRACEY in the i-th direction i Pulse signal data, n represents the y channel data TRACEY in the i-th direction i Length, windows represents the y channel data TRACEY in the i-th direction i Pulse signal window, POS i Indicates the peak point of the i-th data, TRACEY i，j Indicates the y channel data TRACEY in the i-th direction i The jth element; The z channel data TRACEZ in the i-th direction i Pulse signal data POSWINDOWZ i The formula is as follows: POSWINDOWZ i Indicates the z channel data TRACEZ in the i-th direction i Pulse signal data, n represents the z channel data TRACEZ in the i-th direction i Length, windows represents the z channel data TRACEZ in the i-th direction i Pulse signal window, POS i Indicates the peak point of the i-th data, TRACEZ i，j Indicates the z channel data TRACEZ in the i-th direction i The jth element.

6. The signal source direction positioning method based on machine learning according to claim 1, characterized in that: The input data set datainput in Step 7 = {(POSWINDOWX, POSWINDOWY, POSWINDOWZ, AZIMUTH, ZENITH)}, where POSWINDOWX represents the direction x channel data set TRACEX pulse signal data, where the i-th direction x channel data set TRACEX pulse signal data POSWINDOWX i ={x1, x2, ..., x n }, where x1, x2, x n Indicates the pulse signal data POSWINDOWX of the x-channel data set TRACEX in the i-th direction i The first element, the second element, the nth element, POSWINDOWY represents the direction y channel data set TRACEY pulse signal data set, where the i-th direction y channel data set TRACEY pulse signal data POSWINDOWY i ={y1, y2, ..., y n }, where y1, y2, y n Indicates the y-channel data set TRACEY pulse signal data POSWINDOWY in the i-th direction i The first element, the second element, the nth element, POSWINDOWZ represents the z-channel data set TRACEZ pulse signal data set, where the i-th z-channel data set TRACEZ pulse signal data POSWINDOWZ i ={z1, z2, ..., z n }, where z1, z2, z n Indicates the pulse signal data POSWINDOWZ of the z channel data set TRACEZ in the i-th direction i The first element, the second element, the nth element, AZIMUTH represents the azimuth angle data, ZENITH represents the zenith angle data, and the pulse signal data set POSWINDOW is defined as POSWINDOW = {(POSWINDOWX1, POSWINDOWY1, POSWINDOWZ1), (POSWINDOWX2, POSWINDOWY2, POSWINDOWZ2), ..., (POSWINDOWX m ,POSWINDOWY m ,POSWINDOWZ m )} Where m represents the size of the pulse signal data set, POSWINDOWX1, POSWINDOWX2, POSWINDOWX m Indicates the first element, second element, and mth element of the pulse signal data set POSWINDOWX in the direction x channel data set TRACEX, POSWINDOWY1, POSWINDOWY2, and POSWINDOWY m Indicates the first element, second element, and mth element of the pulse signal data set POSWINDOWY in the direction y channel, POSWINDOWZ1, POSWINDOWZ2, and POSWINDOWZ m Indicates the first element, second element, and mth element of the pulse signal data set POSWINDOWZ of the z-channel data set TRACEZ, and defines the minimum value of the i-th pulse signal data as POSWINDOW i，min , the maximum value of the i-th pulse signal data is POSWINDOW i，max , according to the following formula, the jth element POSWINDOW of the i-th pulse signal data POSWINDOWi is i，j Normalize it, POSWINDOW i，min =min(POSWINDOW i ) POSWINDOW i，max =max(POSWINDOW i ) POSWINDOW i，min Indicates the i-th pulse signal data POSWINDOW i Minimum, POSWINDOW i，max Indicates the i-th pulse signal data POSWINDOW i Maximum value, POSWINDOW i，j Represents the jth element of the i-th pulse signal data.

7. The signal source direction positioning method based on machine learning according to claim 1, characterized in that: The calculation formula of the Cartesian product number set LABEL of the azimuth angle set AZIMUTHSET and the zenith angle set ZENITHSET in Step 8 is as follows: LABEL=AZIMUTHSET×ZENITHSET ={LABEL 0，0 ,LABEL 0，1 ,...,LABEL m，n } in, LABEL i，j =(AZIMUTHSET i ,ZENITHSET j ),i∈(0,m),j∈(0,n) Where LABEL i，j represents the Cartesian product number of the i-th element of the azimuth angle set AZIMUTHSET and the j-th element of the zenith angle set ZENITHSET, m represents the size of the azimuth angle set AZIMUTHSET, n represents the size of the zenith angle set ZENITHSET, and constructs the input feature data matrix X = {X1, X2, X3, ..., X n }, where X i ={POSWINDOW i }, where POSWINDOW i Represents the i-th pulse signal data, select the real value matrix Y = {Y1, Y2, Y3, ..., Y n }, where Y i =(AZIMUTH i ,ZENITH i )=LABEL i，i AZIMUTH i Indicates the azimuth AZIMUTH data of the i-th data, ZENITH i Indicates the i-th zenith angle ZENITH data, LABEL i，i Indicates the Cartesian product number of the i-th azimuth angle data AZIMUTH and the i-th zenith angle data ZENITH.

Images