A method for determining the type of ship active signals based on multi-domain feature extraction

Abstract

The present invention discloses a method for determining the type of active ship signals based on multi-domain feature extraction. The method comprises the following steps: first, obtaining the active ship signal to be processed; second, calculating the time-frequency domain distribution modulus; third, calculating the peak feature center window and split window; fourth, calculating the instantaneous frequency index; fifth, calculating the time-frequency domain feature parameters; sixth, calculating the fuzzy domain distribution modulus; seventh, calculating the fuzzy domain feature parameters; and eighth, calculating the multi-domain feature parameters, determining the signal type, and outputting the determined signal type. This method is suitable for low signal-to-noise ratio conditions and can completely and effectively determine the type of active ship signals.

Description

Technical Field

[0001] The present invention belongs to the field of ship signal processing, and in particular relates to a method for determining the type of ship active signals based on multi-domain feature extraction. Background Art

[0002] The type determination of ship active signals is of great significance in the fields of underwater acoustic communication and underwater acoustic countermeasures. Specifically, the types of ship active signals include single frequency, linear frequency modulation and hyperbolic frequency modulation. Since the parameters and corresponding estimation methods of the above three types of signals are different, it is important to make a complete and effective determination of the type, which can provide a basis for the subsequent implementation of effective underwater acoustic countermeasures.

[0003] Signal type judgment is often divided into two modules: (1) type judgment feature extraction, commonly used features include instantaneous features, transform domain features, and statistical features; (2) type judge design, commonly used judges include traditional judges and machine learning judges. However, current research on signal type judgment methods mainly focuses on underwater acoustic communication signals and underwater acoustic targets. Research on ship active signal type judgment methods is relatively limited, and the features used are relatively simple, such as methods for judging single-frequency and frequency-modulated signals based on signal spectrum width and narrowband features, and methods for judging linear and hyperbolic frequency-modulated signals based on linear fitting features of signal instantaneous frequency. In addition, there is no method that focuses on how to make complete judgments on single-frequency, linear frequency-modulated, and hyperbolic frequency-modulated signals. It can be seen that the current type judgment method is difficult to achieve complete and effective judgment of the type of ship active signals and has limitations. Summary of the Invention

[0004] Technical problem, the purpose of the present invention is to provide a method for determining the type of ship active signals based on multi-domain feature extraction, which can determine the type of ship active signals and obtain accurate type determination results even under low signal-to-noise ratio.

[0005] Technical solutions, in order to solve the above technical problems, such as Figure 1 As shown, the present invention proposes a method for determining the type of ship active signal based on multi-domain feature extraction, which includes the following steps:

[0006] (1) Obtain active signals from ships to be processed x ( n );

[0007] (2) Calculation of active signals from ships to be processed x ( n )'s time-frequency domain distribution modulus TF( m , l );

[0008] (3) Calculate the l The peak characteristic center window Ω of the frequency point l,center, local peak feature split window Ω l,local and the global peak feature splitting window Ω l,global ;

[0009] (4) Based on the time-frequency domain distribution modulus TF( m , l ), No. l The peak characteristic center window Ω of the frequency point l,center , local peak feature split window Ω l,local , global peak feature split window Ω l,global , calculate the active signal of the ship to be processed x ( n )'s instantaneous frequency index f ( m );

[0010] (5) Based on the time-frequency domain distribution modulus TF( m , l ) and the instantaneous frequency index f ( m ) Calculate the time-frequency domain characteristic parameters u TF ;

[0011] (6) Calculation of active signals from ships to be processed x ( n ) of the fuzzy domain distribution modulus AM( k , h );

[0012] (7) Based on the fuzzy domain distribution modulus AM ( k , h ) Calculate the characteristic parameters of the fuzzy domain u AM ;

[0013] (8) Based on time-frequency domain characteristic parameters u TF and fuzzy domain characteristic parameters u AM Calculate multi-domain feature parameters u MD , judge the signal type and output the judged signal type type.

[0014] Furthermore, in step (1), the active signal of the ship to be processed is obtained by the following method: x ( n ), specifically including the following steps:

[0015] Received from the ship's hydroacoustic sensor N The real-time data collected at each sampling point is used as the active signal of the ship to be processed x ( n ), n= 0, 1, …, N -1, or extract from memory N The data of the sampling points are used as the active signal of the ship to be processed x ( n ), n = 0, 1, …, N -1, n Active signal for the ship to be processed x ( n ), N Active signal for the ship to be processed x ( n ) corresponds to the number of sampling points of the pulse width, and the value is N =2 κ , k is a positive integer greater than or equal to 8.

[0016] Furthermore, in step (2), the active signal of the ship to be processed is calculated using the following method: x ( n )'s time-frequency domain distribution modulus TF( m , l ), specifically including the following steps:

[0017] (2-1) Initialize the parameters for calculating the time-frequency domain distribution modulus, including the initialization of the following parameters:

[0018] Sliding time window length N w Initialized to: 32≤ N w ≤floor{ N / 5}, where floor{·} is a floor function;

[0019] Sliding time window stepping N s Initialization: 1≤ N s ≤floor{ N w / 4} positive integer;

[0020] Total time frames M Initialized as: M =floor{( N - N w ) / N s}-1;

[0021] Total frequency points L Initialized as: L = Nw / 2+1;

[0022] (2-2) Extracting segmented data sequences y 0( r ), y 1( r ), …, y m ( r ), …, y M-1 ( r ), among which m Segmented data sequence y m ( r )for:

[0023]

[0024] in, m is the time frame index, m =0,1,…, M -1, r is the segment time index;

[0025] (2-3) Calculate the time-frequency domain distribution modulus TF( m , l ):

[0026]

[0027] in, l is the frequency index, j is the imaginary unit, that is, , |·| is the modulo function.

[0028] Furthermore, in step (3), the following method is used to calculate the l The peak characteristic center window Ω of the frequency point l,center , local peak feature split window Ω l,local and the global peak feature splitting window Ω l,global , specifically including the following steps:

[0029] (3-1) Initialize the parameters of the center window and split window for calculating the peak feature, including the initialization of the following parameters:

[0030] Peak feature center window length parameter a 1 Initialized to: 0≤ a 1≤floor{ L / 8-3 / 8}, where L is the total number of frequency points initialized in step (2-1), floor{·} is the floor rounding function;

[0031] Peak feature split window interval frequency points a 2 Initialized to: 1≤ a 2≤floor{ L / 8-3 / 8};

[0032] Local peak feature split window length parameter a 3 Initialized to: 1≤ a 3≤floor{ L / 8-3 / 8};

[0033] No. l The local peak characteristic split window Ω of the frequency point l,local Initialized to an empty set;

[0034] No. l The global peak feature split window Ω of the frequency point l,global Initialized to an empty set;

[0035] (3-2) Update frequency index l :

[0036] ;

[0037] (3-3) Calculate the l The peak characteristic center window Ω of the frequency point l,center :

[0038]

[0039] in, p low,1 ( l )and p high,1 ( l ) are respectively l The lower and upper limits of the center window of the peak characteristic of each frequency point, p low,1 ( l )=max{0, l - a 1}, p high,1 ( l )=min{ L -1, l + a 1}, max{·} and min{·} are the maximum and minimum value functions respectively;

[0040] (3-4) Determine the frequency index l Whether the following peak feature left split window existence conditions are met:

[0041]

[0042] If the condition is met, go to step (3-5); otherwise, go to step (3-7);

[0043] (3-5) Calculate the l The lower limit of the left split window of the local peak feature of the frequency point p low,2 ( l ) and upper limit p high,2 ( l ):

[0044]

[0045] ;

[0046] (3-6) Update l The local peak characteristic split window Ω of the frequency point l,local and the global peak feature splitting window Ω l,global :

[0047]

[0048] ;

[0049] (3-7) Determine the frequency index l Whether the following peak feature right split window existence conditions are met:

[0050]

[0051] If the condition is met, go to step (3-8); otherwise, go to step (3-10);

[0052] (3-8) Calculate the l The lower limit of the right split window of the local peak feature of the frequency point p low,3 ( l ) and upper limit p high,3 ( l ):

[0053]

[0054] ;

[0055] (3-9) Update l The local peak characteristic split window Ω of the frequency point l,local and the global peak feature splitting window Ω l,global :

[0056]

[0057] ;

[0058] (3-10) Determine the frequency index l Whether the following frequency point traversal end conditions are met:

[0059]

[0060] If the condition is met, let the frequency index l = l +1 and return to step (3-3); otherwise, go to step (4).

[0061] Furthermore, in step (4), the active signal of the ship to be processed is calculated using the following method: x ( n )'s instantaneous frequency index f ( m ), specifically including the following steps:

[0062] (4-1) Weight the local peak feature or local Initialized to: 0< or local Positive real numbers < 1;

[0063] (4-2) Update time frame index m :

[0064] ;

[0065] (4-3) Update frequency index l :

[0066] ;

[0067] (4-4) Calculate the m Time frame l The local peak characteristic parameters of each frequency point x local ( m , l ):

[0068]

[0069] in, l 1 is the frequency point traversal index, TF( m , l 1) is the time-frequency domain distribution modulus calculated in steps (2-3), and Respectively m Fixed, taken and Time TF( m, l 1) the mean;

[0070] (4-5) Calculate the m Time frame l The global peak characteristic parameters of each frequency point x global ( m , l ):

[0071]

[0072] in, express m Fixed, taken Time TF( m , l 1) the mean;

[0073] (4-6) Calculate the m Time frame l The peak characteristic weighted result of each frequency point x weight ( m , l ):

[0074] ;

[0075] (4-7) Determine the frequency index l Whether the following frequency point traversal end conditions are met:

[0076]

[0077] If the condition is met, let the frequency index l = l +1, and return to step (4-4); otherwise, go to step (4-8);

[0078] (4-8) Calculate the m The instantaneous frequency index of the time frame f ( m ):

[0079]

[0080] in, Indicates when m When fixed, the peak feature weighted result x weight ( m , l ) to search for the maximum value, and the obtained frequency index l ;

[0081] (4-9) Determine the time frame index mWhether the following time frame traversal end conditions are met:

[0082]

[0083] If the condition is met, let the time frame index m = m +1 and return to step (4-3); otherwise, go to step (5).

[0084] Furthermore, in step (5), the time-frequency domain characteristic parameters are calculated using the following method: u TF , specifically including the following steps:

[0085] (5-1) Initialize the parameters for calculating the time-frequency domain characteristic parameters, including the initialization of the following parameters:

[0086] Time Energy Center Window Length Parameters b 1 Initialized to: 0≤ b 1≤floor{ M / 8-3 / 8}, where floor{·} is a floor function;

[0087] Time energy splitting window interval time frame number b 2 Initialized to: 1≤ b 2≤floor{ M / 8-3 / 8};

[0088] Time energy splitting window length parameter b 3 Initialized to: 1≤ b 3≤floor{ M / 8-3 / 8};

[0089] No. m The time energy splitting window Ψ of the time frame m,div Initialized to an empty set;

[0090] Time energy characteristic parameter threshold m st Initialized as: m st Positive real numbers > 0;

[0091] (5-2) Update time frame index m :

[0092] ;

[0093] (5-3) Calculate the m The time energy center window Ψ of the time frame m,center :

[0094]

[0095] in, q low,1 ( m )and q high,1 ( m ) are respectively m The lower and upper limits of the time energy center window of each time frame, q low,1 ( m )=max{0, m - b 1}, q high,1 ( m )=min{ M -1, m + b 1}, max{·} and min{·} are the maximum and minimum value functions respectively;

[0096] (5-4) Determine the time frame index m Whether the following time energy left splitting window existence conditions are met:

[0097]

[0098] If the condition is met, go to step (5-5); otherwise, go to step (5-7);

[0099] (5-5) Calculate the m The lower limit of the left splitting window of the time energy of the time frame q low,2 ( m ) and upper limit q high,2 ( m ):

[0100]

[0101] ;

[0102] (5-6) Update m The time energy splitting window Ψ of the time frame m,div :

[0103] ;

[0104] (5-7) Determine the time frame index m Whether the following conditions for the existence of the time energy right splitting window are met:

[0105]

[0106] If the condition is met, go to step (5-8); otherwise, go to step (5-10);

[0107] (5-8) Calculate the m The lower limit of the right splitting window of the time energy of the time frame q low,3 ( m ) and upper limit q high,3 ( m ):

[0108]

[0109] ;

[0110] (5-9) Update m The time energy splitting window Ψ of the time frame m,div :

[0111] ;

[0112] (5-10) Calculate the m Time energy characteristic parameters of each time frame r ( m ):

[0113]

[0114] in, f ( m ) is the first m The instantaneous frequency index of the time frame f ( m ), m 1 is the time frame traversal index, TF( m 1, f ( m )) is the time-frequency domain distribution modulus calculated in step (2-3), and Respectively f ( m )Fixed, take and Time TF( m 1, f ( m ))’s mean;

[0115] (5-11) Calculate the m Time energy characterization parameter of a time frame s ( m ):

[0116] ;

[0117] (5-12) Determine the time frame index m Whether the following time frame traversal end conditions are met:

[0118]

[0119] If the condition is met, let the time frame index m = m +1, and return to step (5-3); otherwise, go to step (5-13);

[0120] (5-13) Calculate time-frequency domain characteristic parameters u TF :

[0121] .

[0122] Furthermore, in step (6), the active signal of the ship to be processed is calculated using the following method: x ( n ) of the fuzzy domain distribution modulus AM( k , h ), specifically including the following steps:

[0123] (6-1) Calculation of active signals from ships to be processed x ( n )'s autocorrelation function K z ( n , h ):

[0124]

[0125] in, h is the delay index, h =0,1,…, N -1, express The conjugated form of

[0126] (6-2) Calculate the fuzzy domain distribution modulus AM( k , h ):

[0127]

[0128] in, k is the frequency shift index.

[0129] Furthermore, in step (7), the fuzzy domain characteristic parameters are calculated using the following method: u AM , specifically including the following steps:

[0130] (7-1) Initialize the parameters for calculating the fuzzy domain characteristic parameters, including the initialization of the following parameters:

[0131] Radon transform radial angle quantity E Initialized to: 90 ≤ E A positive integer ≤ 180;

[0132] The Radon transform radial angle set Θ is initialized as:

[0133]

[0134] in, e is the Radon transform radial angle index, e = 0,1, …, E -1, no. e Radon transform radial angle i e Initialized to i e =-π / 2+ e π / ( E -1);

[0135] Fuzzy domain feature parameter mapping parameters m map Initialized as: m map Positive real numbers > 0;

[0136] (7-2) Calculate the fuzzy domain distribution modulus AM( k , h )'s Radon transform result R =[ R ( i 0), R ( i 1), … , R ( i e ), …, R ( i E-1 )], among which, e Radon transform radial angle i e The corresponding Radon transform result R ( i e )for:

[0137]

[0138] Among them, AM( k , h ) is the fuzzy domain distribution modulus calculated in step (6-2), d {·} is the impulse function, that is:

[0139]

[0140] in, w is the independent variable of the impulse function;

[0141] (7-3) Calculate Radon transform results R The mean R mean :

[0142]

[0143] Among them, mean{·} is the mean function;

[0144] (7-4) Calculating fuzzy domain characteristic parameters u AM :

[0145]

[0146] in, R - R mean Indicates that R Each element in is subtracted R mean The resulting vector;

[0147] (7-5) Fuzzy domain feature parameters u AM Mapping to the range 0 to 1:

[0148] .

[0149] Furthermore, in step (8), the following method is used to calculate the multi-domain feature parameters: u MD Determine the signal type and output the determined signal type type, specifically including the following steps:

[0150] (8-1) Initialize the parameters of the judgment signal type, including the initialization of the following parameters:

[0151] Time-frequency domain feature bisection parameter l TF Initialized to: 0< l TF Positive real numbers < 1;

[0152] Fuzzy domain feature dichotomy parameter l AM Initialized to: 0< l AMPositive real numbers < 1;

[0153] (8-2) Determining time-frequency domain characteristic parameters u TF Whether the following threshold conditions are met:

[0154]

[0155] If the conditions are met, then u TF = u TF +1- l AM ;in, u TF The time-frequency domain characteristic parameters calculated in steps (5-13);

[0156] (8-3) Determining the characteristic parameters of the fuzzy domain u AM Whether the following threshold conditions are met:

[0157]

[0158] If the conditions are met, then u AM = u AM - l TF ;in, u AM The fuzzy domain characteristic parameters calculated in step (7-5);

[0159] (8-4) Calculating multi-domain characteristic parameters u MD :

[0160] ;

[0161] (8-5) Based on multi-domain characteristic parameters u MD Determine the signal type:

[0162]

[0163] Among them, type=1 means that the ship's active signal is judged as a single-frequency signal, type=2 means that the ship's active signal is judged as a linear frequency modulation signal, and type=3 means that the ship's active signal is judged as a hyperbolic frequency modulation signal;

[0164] (8-6) Output signal type type.

[0165] Beneficial effects: Compared with the prior art, the technical solution of the present invention has the following beneficial technical effects:

[0166] 1. The present invention makes full use of local and global peak features to effectively extract the instantaneous frequency of the ship's active signal, as shown in steps 3 and 4. This extraction process can reduce the impact of interference such as noise and improve the accuracy of subsequent time-frequency domain feature utilization.

[0167] 2. The present invention makes full use of the time-frequency domain characteristics of single-frequency, linear and hyperbolic frequency-modulated signals and designs time-frequency domain characteristic parameters. As shown in step 5, the designed time-frequency domain characteristic parameters can provide a basis for the calculation of subsequent multi-domain characteristic parameters.

[0168] 3. The present invention makes full use of the fuzzy domain characteristics of single-frequency, linear and hyperbolic FM signals and designs fuzzy domain feature parameters. As shown in step 7, the designed fuzzy domain feature parameters can provide a basis for the subsequent calculation of multi-domain feature parameters.

[0169] 4. The present invention makes full use of the multi-domain characteristics of single-frequency, linear and hyperbolic FM signals, that is, combines the time-frequency domain characteristics and the fuzzy domain characteristics to realize the type judgment of the ship's active signal. As shown in step 8, this judgment process can make a complete and effective type judgment on the ship's active signal. BRIEF DESCRIPTION OF THE DRAWINGS

[0170] Figure 1 Schematic diagram of the process of the present invention;

[0171] Figure 2 is the time-frequency domain distribution modulus of Example 1;

[0172] Figure 3 is the instantaneous frequency of Example 1;

[0173] Figure 4 is the fuzzy domain distribution modulus value of Example 1;

[0174] Figure 5 is the Radon transform result of Example 1;

[0175] Figure 6 is the time-frequency domain distribution modulus value of Example 2;

[0176] Figure 7 is the instantaneous frequency of Example 2;

[0177] Figure 8 is the fuzzy domain distribution modulus value of Example 2;

[0178] Figure 9 is the Radon transform result of Example 2;

[0179] Figure 10 is the time-frequency domain distribution modulus of Example 3;

[0180] Figure 11 is the instantaneous frequency of Example 3;

[0181] Figure 12 is the fuzzy domain distribution modulus value of Example 3;

[0182] Figure 13 This is the Radon transform result of Example 3. DETAILED DESCRIPTION

[0183] In order to better understand the purpose, structure and function of the present invention, the present invention is further described below with reference to the accompanying drawings.

[0184] In the embodiment of the present invention, the received signal model is simulated x ( t )for:

[0185]

[0186] in, A is the signal amplitude, t 0 is the signal start time, t is the signal pulse width, oh ( t ) has a mean of 0 and a variance of s 2 Gaussian white noise with variance s 2 The size depends on the signal-to-noise ratio SNR: SNR=10log 10 ( A 2 / s 2 ), f ( t ) is the instantaneous phase of the signal, expressed as:

[0187]

[0188] in, f 0 is the initial phase of the signal, f 1 is the starting frequency of the signal, m is the linear frequency modulation frequency, defined as: m =( f 2- f 1) / t , f 2 is the signal termination frequency, k 0 is the periodic slope of the hyperbolic FM signal, defined as: k 0=( f 2- f 1) / ( f 1 f 2 t ).

[0189] At sampling frequency f s The signal received by the above simulation x ( t ) to obtain a signal sampling data sequence by discrete sampling x ( n )for:

[0190]

[0191] The signal starting point n 0= round{ t 0 f s}, number of signal sampling points N = round{ τf s}, round{·} is the rounding function, the discrete form of the instantaneous phase of the signal f ( n / f s ) is expressed as:

[0192] .

[0193] Example 1

[0194] The simulation signal is a single-frequency signal, and its parameters are set as: signal amplitude A =1, signal start time t 0=0.01 s, signal pulse width t =2 s, SNR=-10 dB, initial phase of the signal f 0=0; signal starting frequency f 1=1540 Hz, sampling frequency f s =4 kHz.

[0195] The following is the type judgment of the simulation signal:

[0196] According to step (1), the data of 8000 sampling points are extracted from the memory as the ship active signal to be processed x ( n );

[0197] According to step (2), set the sliding time window length N w =512, sliding time window step N s =128, total number of time frames M =57, total number of frequency points L =257; for the active signal of the ship to be processedx ( n ), extract segmented data sequence y 0( r ), y 1( r ),…, y m ( r ), …, y 56 ( r ), and calculate the time-frequency domain distribution modulus TF( m , l ); time-frequency domain distribution modulus TF( m , l )like Figure 2 As shown;

[0198] According to step (3), set the peak feature center window length parameter a 1=0, peak feature split window interval frequency points a 2=2, local peak feature split window length parameter a 3=2, and split the local peak feature into windows Ω l,local and the global peak feature splitting window Ω l,global Initialize to an empty set; calculate the peak feature center window Ω l,center , update the local peak feature split window Ω l,local and the global peak feature splitting window Ω l,global ;

[0199] According to step (4), set the local peak feature weight or local =0.1; calculate local peak characteristic parameters x local ( m , l ), global peak characteristic parameters x global ( m , l ) and peak feature weighted results x weight ( m , l ); Calculate the instantaneous frequency index f ( m ); instantaneous frequency index f ( m ) corresponds to the instantaneous frequency Figure 3 As shown;

[0200] According to step (5), set the time energy center window length parameter b 1=4, the number of time frames between time energy splitting windows b2=5, time energy splitting window length parameter b 3=6, time energy characteristic parameter threshold m st =1, and the time energy splitting window Ψ m,div Initialized to an empty set; calculate the time energy center window Ψ m,center , update the time energy splitting window Ψ m,div ; Calculate time energy characteristic parameters r ( m ) and time energy characterization parameters s ( m );Calculate the time-frequency domain characteristic parameters u TF =0.4035;

[0201] According to step (6), calculate the active signal of the ship to be processed x ( n )'s autocorrelation function K z ( n , h ) and the fuzzy domain distribution modulus AM( k , h ); fuzzy domain distribution modulus AM( k , h )like Figure 4 As shown;

[0202] According to step (7), set the number of Radon transform radial angles E =180, Radon transform radial angle set Θ=[ i 0, i 1,…, i e , …, i 179 ], among which, e Radon transform radial angle i e Initialized as: i e =-π / 2+ e π / 179, fuzzy domain feature parameter mapping parameter m map =100; calculate the fuzzy domain distribution modulus AM( k , h )'s Radon transform result R =[ R ( i 0), R ( i 1), … , R ( i e ), …,R ( i 179 )]; Calculate Radon transform results R The mean R mean =0.0969; calculate the fuzzy domain characteristic parameters u AM =51.6009, and mapping it to the range of 0~1, we get u AM =0.5160; Radon transform result R like Figure 5 As shown;

[0203] According to step (8), set the time-frequency domain feature binary parameters l TF =0.3, fuzzy domain feature dichotomy parameter l AM =0.5; time-frequency domain characteristic parameters u TF Meet the threshold condition u TF =0.4035> l TF =0.3, u TF = u TF +1- l AM =0.9035; fuzzy domain characteristic parameter u AM Threshold conditions not met u AM =0.5160< l AM =0.5; calculate multi-domain characteristic parameters u MD = u TF + u AM =1.4195; Due to the multi-domain characteristic parameters u MD Satisfies 1+ l TF =1.3≤ u MD , signal type type=1, type=1 means that the ship's active signal is judged as a single-frequency signal and the signal type type is output.

[0204] Example 2

[0205] The simulation signal is a linear frequency modulation signal, and its parameters are set as: signal amplitude A =1, signal start time t0=0.01s, signal pulse width t =2 s, SNR=-10 dB, initial phase of the signal f 0=0; signal starting frequency f 1=1540 Hz, signal end frequency f 2=1920 Hz, linear frequency modulation signal frequency m =190 Hz·s -1 , sampling frequency f s =4 kHz.

[0206] The following is the type judgment of the simulation signal:

[0207] According to step (1), the data of 8000 sampling points are extracted from the memory as the ship active signal to be processed x ( n );

[0208] According to step (2), set the sliding time window length N w =512, sliding time window step N s =128, total number of time frames M =57, total number of frequency points L =257; for the active signal of the ship to be processed x ( n ), extract segmented data sequence y 0( r ), y 1( r ),…, y m ( r ), …, y 56 ( r ), and calculate the time-frequency domain distribution modulus TF( m , l ); time-frequency domain distribution modulus TF( m , l )like Figure 6 As shown;

[0209] According to step (3), set the peak feature center window length parameter a 1=0, peak feature split window interval frequency points a 2=2, local peak feature split window length parameter a 3=2, and split the local peak feature into windows Ω l,local and the global peak feature splitting window Ω l,global Initialize to an empty set; calculate the peak feature center window Ω l,center, update the local peak feature split window Ω l,local and the global peak feature splitting window Ω l,global ;

[0210] According to step (4), set the local peak feature weight or local =0.1; calculate local peak characteristic parameters x local ( m , l ), global peak characteristic parameters x global ( m , l ) and peak feature weighted results x weight ( m , l ); Calculate the instantaneous frequency index f ( m ); instantaneous frequency index f ( m ) corresponds to the instantaneous frequency Figure 7 As shown;

[0211] According to step (5), set the time energy center window length parameter b 1=4, the number of time frames between time energy splitting windows b 2=5, time energy splitting window length parameter b 3=6, time energy characteristic parameter threshold m st =1, and the time energy splitting window Ψ m,div Initialized to an empty set; calculate the time energy center window Ψ m,center , update the time energy splitting window Ψ m,div ; Calculate time energy characteristic parameters r ( m ) and time energy characterization parameters s ( m );Calculate the time-frequency domain characteristic parameters u TF =0;

[0212] According to step (6), calculate the active signal of the ship to be processed x ( n )'s autocorrelation function K z ( n , h ) and the fuzzy domain distribution modulus AM( k , h ); fuzzy domain distribution modulus AM( k , h )like Figure 8 As shown;

[0213] According to step (7), set the number of Radon transform radial angles E =180, Radon transform radial angle set Θ=[ i 0, i 1,…, i e , …, i 179 ], among which, e Radon transform radial angle i e Initialized as: i e =-π / 2+ e π / 179, fuzzy domain feature parameter mapping parameter m map =100; calculate the fuzzy domain distribution modulus AM( k , h )'s Radon transform result R =[ R ( i 0), R ( i 1), … , R ( i e ), …, R ( i 179 )]; Calculate Radon transform results R The mean R mean =0.1676; calculate the fuzzy domain characteristic parameters u AM =108.4484, and mapping it to the range of 0~1, we can get u AM =1; Radon transform result R like Figure 9 As shown;

[0214] According to step (8), set the time-frequency domain feature binary parameters l TF =0.3, fuzzy domain feature dichotomy parameter l AM =0.5; time-frequency domain characteristic parameters u TF Threshold conditions not met u TF =0> l TF =0.3; fuzzy domain characteristic parameter u AM Threshold conditions not met uAM =1< l AM =0.5; calculate multi-domain characteristic parameters u MD = u TF + u AM =1; Due to the multi-domain characteristic parameters u MD satisfy l AM =0.5≤ u MD =1<1+ l TF =1.3, signal type type=2, type=2 means that the ship's active signal is judged as a linear frequency modulation signal and the output signal type type.

[0215] Example 3:

[0216] The simulation signal is a hyperbolic frequency modulated signal, and its parameters are set as: signal amplitude A =1, signal start time t 0=0.01s, signal pulse width t =2 s, SNR=-10 dB, initial phase of the signal f 0=0; signal starting frequency f 1=1540 Hz, signal end frequency f 2=1920 Hz, the slope of the hyperbolic FM signal cycle k 0=6.4259×10 -5 , sampling frequency f s =4 kHz.

[0217] The following is the type judgment of the simulation signal:

[0218] According to step (1), the data of 8000 sampling points are extracted from the memory as the ship active signal to be processed x ( n );

[0219] According to step (2), set the sliding time window length N w =512, sliding time window step N s =128, total number of time frames M =57, total number of frequency points L =257; for the active signal of the ship to be processed x ( n ), extract segmented data sequence y 0(r ), y 1( r ),…, y m ( r ), …, y 56 ( r ), and calculate the time-frequency domain distribution modulus TF( m , l ); time-frequency domain distribution modulus TF( m , l )like Figure 10 As shown;

[0220] According to step (3), set the peak feature center window length parameter a 1=0, peak feature split window interval frequency points a 2=2, local peak feature split window length parameter a 3=2, and split the local peak feature into windows Ω l,local and the global peak feature splitting window Ω l,global Initialize to an empty set; calculate the peak feature center window Ω l,center , update the local peak feature split window Ω l,local and the global peak feature splitting window Ω l,global ;

[0221] According to step (4), set the local peak feature weight or local =0.1; calculate local peak characteristic parameters x local ( m , l ), global peak characteristic parameters x global ( m , l ) and peak feature weighted results x weight ( m , l ); Calculate the instantaneous frequency index f ( m ); instantaneous frequency index f ( m ) corresponds to the instantaneous frequency Figure 11 As shown;

[0222] According to step (5), set the time energy center window length parameter b 1=4, the number of time frames between time energy splitting windows b 2=5, time energy splitting window length parameter b 3=6, time energy characteristic parameter threshold m st=1, and the time energy splitting window Ψ m,div Initialized to an empty set; calculate the time energy center window Ψ m,center , update the time energy splitting window Ψ m,div ; Calculate time energy characteristic parameters r ( m ) and time energy characterization parameters s ( m );Calculate the time-frequency domain characteristic parameters u TF =0;

[0223] According to step (6), calculate the active signal of the ship to be processed x ( n )'s autocorrelation function K z ( n , h ) and the fuzzy domain distribution modulus AM( k , h ); fuzzy domain distribution modulus AM( k , h )like Figure 12 As shown;

[0224] According to step (7), set the number of Radon transform radial angles E =180, Radon transform radial angle set Θ=[ i 0, i 1,…, i e , …, i 179 ], among which, e Radon transform radial angle i e Initialized as: i e =-π / 2+ e π / 179, fuzzy domain feature parameter mapping parameter m map =100; calculate the fuzzy domain distribution modulus AM( k , h )'s Radon transform result R =[ R ( i 0), R ( i 1), … , R ( i e ), …, R ( i 179 )]; Calculate Radon transform results R The meanR mean =0.1721; calculate the fuzzy domain characteristic parameters u AM =35.4842, and mapping it to the range of 0~1, we can get u AM =0.3548; Radon transform result R like Figure 13 As shown;

[0225] According to step (8), set the time-frequency domain feature binary parameters l TF =0.3, fuzzy domain feature dichotomy parameter l AM =0.5; time-frequency domain characteristic parameters u TF Threshold conditions not met u TF =0> l TF =0.3; fuzzy domain characteristic parameter u AM Meet the threshold condition u AM =0.3548< l AM =0.5, u AM = u AM - l TF =0.0548; Calculate multi-domain characteristic parameters u MD = u TF + u AM =0.0548; Due to the multi-domain characteristic parameters u MD satisfy u MD =0.0548< l AM =0.5, signal type type=3, type=3 means that the ship's active signal is judged as a hyperbolic FM signal and the output signal type type.

[0226] It will be appreciated that the present invention is described with reference to certain embodiments, and those skilled in the art will appreciate that various changes or equivalent substitutions may be made to these features and embodiments without departing from the spirit and scope of the present invention. Furthermore, under the teachings of the present invention, these features and embodiments may be modified to adapt to specific circumstances and materials without departing from the spirit and scope of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are intended to be protected by the present invention.

Claims

1. A method for determining the type of active ship signals based on multi-domain feature extraction, characterized in that: The method comprises the following steps: (1) Obtain active signals from ships to be processed x ( n ); (2) Calculation of active signals from ships to be processed x ( n )'s time-frequency domain distribution modulus TF( m , l ); (3) Calculate the l The peak characteristic center window Ω of the frequency point l,center , local peak feature split window Ω l,local and the global peak feature splitting window Ω l,global ; (4) Based on the time-frequency domain distribution modulus TF( m , l ), No. l The peak characteristic center window Ω of the frequency point l,center , local peak feature split window Ω l,local , global peak feature split window Ω l,global , calculate the active signal of the ship to be processed x ( n )'s instantaneous frequency index f ( m ); (5) Based on the time-frequency domain distribution modulus TF( m , l ) and the instantaneous frequency index f ( m ) Calculate the time-frequency domain characteristic parameters υ TF ; (6) Calculation of active signals from ships to be processed x ( n ) of the fuzzy domain distribution modulus AM( k , h ); (7) Based on the fuzzy domain distribution modulus AM ( k , h ) Calculate the characteristic parameters of the fuzzy domain υ AM ; (8) Based on time-frequency domain characteristic parameters υ TF and fuzzy domain characteristic parameters υ AM Calculate multi-domain feature parameters υ MD , judge the signal type and output the judged signal type type; In step (2), the active signal of the ship to be processed is calculated using the following method: x ( n )'s time-frequency domain distribution modulus TF( m , l ), specifically including the following steps: (2-1) Initialize the parameters for calculating the time-frequency domain distribution modulus, including the initialization of the following parameters: Sliding time window length N w Initialized to 32≤ N w ≤floor{ N / 5}, where floor{·} is a floor function; Sliding time window stepping N s Initialized to 1≤ N s ≤floor{ N w / 4} positive integer; Total time frames M Initialized to M =floor{( N - N w ) / N s }-1; Total frequency points L Initialized to L = N w / 2+1; (2-2) Extracting segmented data sequences y 0( r ), y 1( r ),…, y m ( r ),…, y M-1 ( r ), among which m Segmented data sequence y m ( r )for: in, m is the time frame index, m =0,1,…, M -1, r is the segment time index; (2-3) Calculate the time-frequency domain distribution modulus TF( m , l ): in, l is the frequency index, j is the imaginary unit, that is, , |·| is the modulo function.

2. The method for determining the type of active ship signal based on multi-domain feature extraction according to claim 1 is characterized in that: In step (1), the following method is used to obtain the active signal of the ship to be processed x ( n ), specifically including the following steps: Received from the ship's hydroacoustic sensor N The real-time data collected at each sampling point is used as the active signal of the ship to be processed x ( n ), n = 0,1,…, N -1, or extract from memory N The data of the sampling points are used as the active signal of the ship to be processed x ( n ), n = 0, 1, …, N -1, n Active signal for the ship to be processed x ( n ), N Active signal for the ship to be processed x ( n ) corresponds to the number of sampling points of the pulse width, and the value is N =2 κ , κ is a positive integer greater than or equal to 8.

3. The method for determining the type of active ship signals based on multi-domain feature extraction according to claim 1 is characterized in that: In step (3), the following method is used to calculate the l The peak characteristic center window Ω of the frequency point l,center , local peak feature split window Ω l,local and the global peak feature splitting window Ω l,global , specifically including the following steps: (3-1) Initialize the parameters of the center window and split window for calculating the peak feature, including the initialization of the following parameters: Peak feature center window length parameter a 1 initialized to 0≤ a 1≤floor{ L / 8-3 / 8}, where L is the total number of frequency points initialized in step (2-1); Peak feature split window interval frequency points a 2 Initialized to 1≤ a 2≤floor{ L / 8-3 / 8}; Local peak feature split window length parameter a 3 initialized to 1≤ a 3≤floor{ L / 8-3 / 8}; No. l The local peak characteristic split window Ω of the frequency point l,local Initialized to an empty set; No. l The global peak feature split window Ω of the frequency point l,global Initialized to an empty set; (3-2) Update frequency index l : ； (3-3) Calculate the l The peak characteristic center window Ω of the frequency point l,center : in, p low,1 ( l )and p high,1 ( l ) are respectively l The lower and upper limits of the center window of the peak characteristic of each frequency point, p low,1 ( l )=max{0, l - a 1}, p high,1 ( l )=min{ L -1, l + a 1}, max{·} and min{·} are the maximum and minimum value functions respectively; (3-4) Determine the frequency index l Whether the following peak feature left split window existence conditions are met: If the condition is met, go to step (3-5); otherwise, go to step (3-7); (3-5) Calculate the l The lower limit of the left split window of the local peak feature of the frequency point p low,2 ( l ) and upper limit p high,2 ( l ): ； (3-6) Update l The local peak characteristic split window Ω of the frequency point l,local and the global peak feature splitting window Ω l,global : ； (3-7) Determine the frequency index l Whether the following peak feature right split window existence conditions are met: If the condition is met, go to step (3-8); otherwise, go to step (3-10); (3-8) Calculate the l The lower limit of the right split window of the local peak feature of the frequency point p low,3 ( l ) and upper limit p high,3 ( l ): ； (3-9) Update l The local peak characteristic split window Ω of the frequency point l,local and the global peak feature splitting window Ω l,global : ； (3-10) Determine the frequency index l Whether the following frequency point traversal end conditions are met: If the condition is met, let the frequency index l = l +1 and return to step (3-3); otherwise, go to step (4).

4. The method for determining the type of active ship signal based on multi-domain feature extraction according to claim 3 is characterized in that: In step (4), the active signal of the ship to be processed is calculated using the following method: x ( n )'s instantaneous frequency index f ( m ), specifically including the following steps: (4-1) Weight the local peak feature η local Initialized to 0< η local Positive real numbers < 1; (4-2) Update time frame index m : ； (4-3) Update frequency index l : ； (4-4) Calculate the m Time frame l The local peak characteristic parameters of each frequency point χ local ( m , l ): in, l 1 is the frequency point traversal index, TF( m , l 1) is the time-frequency domain distribution modulus calculated in steps (2-3), and Respectively m Fixed, taken and Time TF( m , l 1) the mean; (4-5) Calculate the m Time frame l The global peak characteristic parameters of each frequency point χ global ( m , l ): in, express m Fixed, taken Time TF( m , l 1) the mean; (4-6) Calculate the m Time frame l The peak characteristic weighted result of each frequency point χ weight ( m , l ): ； (4-7) Determine the frequency index l Whether the following frequency point traversal end conditions are met: If the condition is met, let the frequency index l = l +1, and return to step (4-4); otherwise, go to step (4-8); (4-8) Calculate the m The instantaneous frequency index of the time frame f ( m ): in, Indicates when m When fixed, the peak feature weighted result χ weight ( m , l ) to search for the maximum value and obtain the frequency index l ; (4-9) Determine the time frame index m Whether the following time frame traversal end conditions are met: If the condition is met, let the time frame index m = m +1 and return to step (4-3); otherwise, go to step (5).

5. The method for determining the type of active ship signal based on multi-domain feature extraction according to claim 4 is characterized in that: In step (5), the time-frequency domain characteristic parameters are calculated using the following method: υ TF , specifically including the following steps: (5-1) Initialize the parameters for calculating the time-frequency domain characteristic parameters, including the initialization of the following parameters: Time Energy Center Window Length Parameters b 1 initialized to 0≤ b 1≤floor{ M / 8-3 / 8} non-negative integer; Time energy splitting window interval time frame number b 2 Initialized to 1≤ b 2≤floor{ M / 8-3 / 8}; Time energy splitting window length parameter b 3 initialized to 1≤ b 3≤floor{ M / 8-3 / 8}; No. m The time energy splitting window Ψ of the time frame m,div Initialized to an empty set; Time energy characteristic parameter threshold μ st Initialized to μ st Positive real numbers > 0; (5-2) Update time frame index m : ； (5-3) Calculate the m The time energy center window Ψ of the time frame m,center : in, q low,1 ( m )and q high,1 ( m ) are respectively m The lower and upper limits of the time energy center window of each time frame, q low,1 ( m )=max{0, m - b 1}, q high,1 ( m )=min{ M -1, m + b 1}, max{·} and min{·} are the maximum and minimum value functions respectively; (5-4) Determine the time frame index m Whether the following time energy left splitting window existence conditions are met: If the condition is met, go to step (5-5); otherwise, go to step (5-7); (5-5) Calculate the m The lower limit of the left splitting window of the time energy of the time frame q low,2 ( m ) and upper limit q high,2 ( m ): ； (5-6) Update m The time energy splitting window Ψ of the time frame m,div : ； (5-7) Determine the time frame index m Whether the following conditions for the existence of the time energy right splitting window are met: If the condition is met, go to step (5-8); otherwise, go to step (5-10); (5-8) Calculate the m The lower limit of the right splitting window of the time energy of the time frame q low,3 ( m ) and upper limit q high,3 ( m ): ； (5-9) Update m The time energy splitting window Ψ of the time frame m,div : ； (5-10) Calculate the m Time energy characteristic parameters of each time frame ρ ( m ): in, f ( m ) is the first m The instantaneous frequency index of the time frame f ( m ), m 1 is the time frame traversal index, TF( m 1, f ( m )) is the time-frequency domain distribution modulus calculated in step (2-3), and Respectively f ( m )Fixed, take and Time TF( m 1, f ( m ))’s mean; (5-11) Calculate the m Time energy characterization parameter of a time frame σ ( m ): ； (5-12) Determine the time frame index m Whether the following time frame traversal end conditions are met: If the condition is met, let the time frame index m = m +1, and return to step (5-3); otherwise, go to step (5-13); (5-13) Calculate time-frequency domain characteristic parameters υ TF : 。 6. The method for determining the type of active ship signals based on multi-domain feature extraction according to claim 5 is characterized in that: In step (6), the active signal of the ship to be processed is calculated using the following method: x ( n ) of the fuzzy domain distribution modulus AM( k , h ), specifically including the following steps: (6-1) Calculation of active signals from ships to be processed x ( n )'s autocorrelation function K z ( n , h ): in, h is the delay index, h =0,1,…, N -1, express The conjugated form of (6-2) Calculate the fuzzy domain distribution modulus AM( k , h ): in, k is the frequency shift index.

7. The method for determining the type of active ship signals based on multi-domain feature extraction according to claim 6 is characterized in that: In step (7), the fuzzy domain characteristic parameters are calculated using the following method: υ AM , specifically including the following steps: (7-1) Initialize the parameters for calculating the fuzzy domain characteristic parameters, including the initialization of the following parameters: Radon transform radial angle quantity E Initialized to 90 ≤ E A positive integer ≤180; The Radon transform radial angle set Θ is initialized as: in, e is the Radon transform radial angle index, e = 0,1, …, E -1, no. e Radon transform radial angle θ e Initialized to θ e =-π / 2+ e π / ( E -1); Fuzzy domain feature parameter mapping parameters μ map Initialized to μ map Positive real numbers > 0; (7-2) Calculate the fuzzy domain distribution modulus AM( k , h )'s Radon transform result R =[ R ( θ 0), R ( θ 1), … , R ( θ e ),…, R ( θ E-1 )], among which, e Radon transform radial angle θ e The corresponding Radon transform result R ( θ e )for: Among them, AM( k , h ) is the fuzzy domain distribution modulus calculated in step (6-2), round{·} is the rounding function, δ {·} is the impulse function, that is: in, w is the independent variable of the impulse function; (7-3) Calculate Radon transform results R The mean R mean : Among them, mean{·} is the mean function; (7-4) Calculating fuzzy domain characteristic parameters υ AM : in, R - R mean Indicates that R Each element in is subtracted R mean The resulting vector; (7-5) Fuzzy domain feature parameters υ AM Mapping to the range 0 to 1: 。 8. The method for determining the type of active ship signals based on multi-domain feature extraction according to claim 7 is characterized in that: In step (8), the following method is used to calculate the multi-domain feature parameters υ MD Determine the signal type and output the determined signal type type, specifically including the following steps: (8-1) Initialize the parameters of the judgment signal type, including the initialization of the following parameters: Time-frequency domain feature bisection parameter λ TF Initialized to 0< λ TF Positive real numbers < 1; Fuzzy domain feature dichotomy parameter λ AM Initialized to 0< λ AM Positive real numbers < 1; (8-2) Determining time-frequency domain characteristic parameters υ TF Whether the following threshold conditions are met: If the conditions are met, then υ TF = υ TF +1- λ AM ;in, υ TF The time-frequency domain characteristic parameters calculated in steps (5-13); (8-3) Determining the characteristic parameters of the fuzzy domain υ AM Whether the following threshold conditions are met: If the conditions are met, then υ AM = υ AM - λ TF ;in, υ AM The fuzzy domain characteristic parameters calculated in step (7-5); (8-4) Calculating multi-domain characteristic parameters υ MD : ； (8-5) Based on multi-domain characteristic parameters υ MD Determine the signal type: Among them, type=1 means that the ship's active signal is judged as a single-frequency signal, type=2 means that the ship's active signal is judged as a linear frequency modulation signal, and type=3 means that the ship's active signal is judged as a hyperbolic frequency modulation signal; (8-6) Output signal type type.

Images