Adaptive Time-Frequency Analysis Method for Ship Signals Based on Instantaneous Frequency Direction Self-Adjustment

Abstract

This invention discloses an adaptive time-frequency analysis method for ship signals based on instantaneous frequency direction self-adjustment, comprising: acquiring the ship signal to be analyzed and performing a short-time Fourier transform; estimating the instantaneous frequency of the ship signal based on the time-frequency distribution of the short-time Fourier transform and combining it with the time-frequency narrowband characteristics; calculating the smoothed instantaneous frequency and the optimal window length; segmenting the smoothed instantaneous frequency in time, calculating the time-frequency domain direction pre-guidance for each time period, and constructing a Chirp atomic direction overcomplete dictionary; segmenting the ship signal in time, calculating the matching direction of the ship signal in each time period, thereby calculating the improved Chirplet transform time-frequency distribution; and calculating the optimal short-time Fourier transform in combination with the optimal window length to correct the improved Chirplet transform time-frequency distribution. This invention simultaneously utilizes the robustness of the short-time Fourier transform and the anisotropy of the improved Chirplet transform to obtain a time-frequency distribution with high robustness and high time-frequency resolution.

Description

Technical Field

[0001] This invention relates to an adaptive time-frequency analysis method for ship signals based on instantaneous frequency direction self-adjustment, belonging to the field of signal processing technology. Background Technology

[0002] In the field of ship navigation and communication, the time-frequency characteristic analysis of ship signals (such as main engine vibration signals, propeller noise signals, and inter-ship communication signals) is a key technology for realizing equipment condition monitoring, fault diagnosis, and signal identification. Ship signals are usually non-stationary and multi-component, with their frequency components changing dynamically over time (such as frequency drift caused by main engine speed fluctuations and noise modulation under complex sea conditions), and are often accompanied by strong background noise and interference. Therefore, high-precision time-frequency analysis methods are needed to extract effective features.

[0003] Time-frequency analysis methods map one-dimensional time-domain signals to a two-dimensional time-frequency plane, simultaneously describing the signal's distribution characteristics in both the time and frequency domains. This makes them a core tool for processing non-stationary signals. Currently, mainstream linear time-frequency analysis methods include the Short-Time Fourier Transform (STFT) and Chirplet Transform, but these have significant limitations in ship signal analysis.

[0004] The Short-Time Fourier Transform (STFT) performs a local Fourier transform on a signal using a sliding window function of fixed length to achieve time-frequency analysis. Its simple principle and high computational efficiency have led to its widespread application in signal processing. However, due to the fixed window length, STFT cannot simultaneously optimize time and frequency resolution. The Chirplet Transform, an extension of the wavelet transform, introduces a frequency modulation parameter. Its atomic function can be represented as a "linearly frequency-modulated signal modulated by a window function," better matching non-stationary signals with linear frequency modulation characteristics. Compared to STFT, it exhibits higher time-frequency convergence for linearly frequency-modulated signals, effectively capturing linear frequency changes over time. Furthermore, by adjusting parameters such as window length, center frequency, and frequency modulation, it can adapt to the time-frequency characteristics of different signals to some extent. Its drawback is that the parameters of the traditional Chirplet Transform are fixed once set, making it unable to dynamically track the nonlinear changes in instantaneous frequency in ship signals.

[0005] Existing methods for ship signals suffer from several key bottlenecks due to their complex characteristics: the time-frequency resolution cannot be adaptively adjusted based on local signal characteristics, making it difficult to balance transient capture and frequency resolution capabilities; and there is a lack of a mechanism to predict instantaneous frequency change trends, resulting in insufficient tracking accuracy for nonlinear frequency modulation or sudden frequency jumps. Therefore, there is an urgent need for an adaptive time-frequency analysis method that combines instantaneous frequency prediction with dynamic adjustment of analysis parameters to achieve high-precision time-frequency characterization of ship signals, providing a reliable basis for subsequent condition monitoring and fault diagnosis. Summary of the Invention

[0006] The technical problem to be solved by this invention is to provide an adaptive time-frequency analysis method for ship signals based on instantaneous frequency and direction self-adjustment. By performing instantaneous frequency prediction and direction self-adjustment in segments, a more accurate frequency modulation distribution range of the instantaneous frequency of the signal is obtained, and a more accurate adaptive improved Chirplet transform is achieved, resulting in a time-frequency distribution with high robustness and high time-frequency resolution.

[0007] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0008] The adaptive time-frequency analysis method for ship signals based on instantaneous frequency direction self-adjustment includes the following steps:

[0009] Step 1: Acquire the ship signal to be analyzed in time and frequency, and perform short-time Fourier transform on the ship signal to obtain the short-time Fourier transform time-frequency distribution of the ship signal;

[0010] Step 2: Based on the short-time Fourier transform time-frequency distribution of the ship signal, combined with the time-frequency narrowband characteristics, estimate the instantaneous frequency of the ship signal and retain the effective instantaneous frequency.

[0011] Step 3: Perform smoothing calculations on the effective instantaneous frequencies to obtain the smoothed instantaneous frequencies, and calculate the optimal window length;

[0012] Step 4: Divide the smoothed instantaneous frequency into time segments, calculate the time-frequency domain direction pre-guidance for each time segment, and construct an overcomplete dictionary of Chirp atom directions;

[0013] Step 5: Divide the ship signal to be analyzed into time segments. The total number of segments of the ship signal to be analyzed is equal to the total number of segments of the smoothed instantaneous frequency. Calculate the matching direction of the ship signal in each time period based on the Chirp atomic direction overcomplete dictionary. Calculate the improved Chirplet transform time-frequency distribution based on the matching direction of the ship signal in each time period and the optimal window length.

[0014] Step 6: Combine the optimal window length to calculate the optimal short-time Fourier transform, and correct and improve the time-frequency distribution of the Chirplet transform to obtain the final time-frequency distribution.

[0015] Compared with the prior art, the present invention, employing the above technical solution, has the following technical effects:

[0016] 1. This invention utilizes the time-frequency domain energy distribution characteristics of a signal to obtain the direction parameters for adaptive time-frequency analysis. The instantaneous frequency is estimated by calculating the short-time Fourier transform and combining it with time-frequency narrowband characteristics. The time-frequency domain direction pre-guiding information is obtained using the instantaneous frequency estimation result of the signal to be analyzed, and combined with local linearity characteristics, an overcomplete Chirp atomic direction dictionary is obtained. Robust estimation of the instantaneous frequency of the signal to be analyzed can be obtained at a low signal-to-noise ratio, and the computational redundancy of direction matching is reduced based on this instantaneous frequency estimation, suppressing the impact of outliers on the accuracy of direction matching.

[0017] 2. This invention utilizes the characteristics of both the short-time Fourier transform and the improved Chirplet transform to obtain adaptive time-frequency analysis results, achieving robust and high-resolution time-frequency distributions even at low signal-to-noise ratios. By leveraging the difference in cross-correlation between Chirp atoms and the signal under analysis, the normalized cross-correlation function between the signal and Chirp atoms at each time period is calculated to obtain the time-frequency domain direction matching results for each period. The improved Chirplet transform is then calculated using these time-frequency domain direction matching results. Through time-frequency domain direction matching at each time period, a high-resolution anisotropic time-frequency distribution is adaptively obtained, and further corrected using the short-time Fourier transform. This improves the energy focusing and robustness of the time-frequency distribution results for nonlinear frequency-modulated signals under low signal-to-noise ratio conditions. Attached Figure Description

[0018] Figure 1 This is a flowchart of the adaptive time-frequency analysis method for ship signals based on instantaneous frequency direction self-adjustment of the present invention;

[0019] Figure 2 This is the short-time Fourier transform time-frequency distribution of the signal in an embodiment of the present invention;

[0020] Figure 3 This is an improved Chirplet transform time-frequency distribution of the signal in an embodiment of the present invention;

[0021] Figure 4 This is the short-time Fourier transform time-frequency distribution of the signal after interpolation in an embodiment of the present invention;

[0022] Figure 5 This is the final time-frequency distribution of the signal in the embodiment of the present invention. Detailed Implementation

[0023] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0024] like Figure 1As shown, this invention proposes an adaptive time-frequency analysis method for ship signals based on instantaneous frequency direction self-adjustment, comprising the following steps:

[0025] Step 1: Obtain the sampled data sequence of the ship signal to be processed. Calculate the sampled data sequence Short-time Fourier Transform Specifically:

[0026] Step 1.1: Receive data from the sensor The collected data from each sampling point is used as a sampling data sequence of the ship signal to be processed. , ,in, For time-discrete indexes, This represents the total number of sampling points covering the entire duration of the signal. Number of sampling points Need to meet positive integers, The sampling time length, Sampling frequency, and The specific values ​​are determined by the actual application, and generally require that they meet the following requirements. positive real numbers, Positive integers.

[0027] Step 1.2: Initialize all parameters:

[0028] Window length of the short-time Fourier transform window function Initialize to: Positive integers;

[0029] Stepping of the short-time Fourier transform window function Initialize to: Positive integers;

[0030] Time length of short-time Fourier transform Initialized to: satisfy , It is a rounding function;

[0031] Frequency length of short-time Fourier transform Initialized to: satisfy Positive integers.

[0032] Step 1.3: Calculate the sampled data sequence of the ship signal to be processed. Short-time Fourier Transform :

[0033] ,

[0034] ,

[0035] in, For window functions, a rectangular window is used here. The calculated short-time Fourier transform time-frequency distribution Time-dimensional discrete index, , For the frequency dimension discrete index of the time-frequency distribution, , The time length of the time-frequency distribution. The frequency length of the time-frequency distribution. The imaginary unit, .

[0036] Step 2: Estimate the instantaneous frequency of the ship signal using the time-frequency narrowband characteristics. Specifically:

[0037] Step 2.1: Initialize all parameters:

[0038] Invalid frequency-dimensional discrete index set Initialize to: It is an empty set, that is ;

[0039] Calculate the time dimension neighborhood size of narrowband features Initially: Positive integers;

[0040] Calculate the size of the narrowband feature frequency dimension neighborhood Initially: Positive integers;

[0041] Narrowband characteristic coefficients Initialized to: satisfy Positive real numbers;

[0042] Invalid time-frequency domain discrete index upper limit Initialize to: Positive integers;

[0043] instantaneous frequency Initialize to: , .

[0044] Step 2.2: Initialize the time-dimensional discrete index values ​​of the time-frequency distribution. Invalid index cumulative value ;

[0045] Step 2.3, for the time-frequency distribution Time dimension data Select the maximum value of the amplitude and find the frequency index corresponding to the maximum amplitude value. :

[0046] ,

[0047] Wherein, the above formula represents from and Find the one that satisfies the acquisition Maximum frequency dimension discrete index value;

[0048] Step 2.4: Calculate the time-frequency points neighborhood :

[0049] ,

[0050] Wherein, the above formula represents in , Find all pairs of strings that satisfy the conditions within the parentheses. The time and frequency points constitute the neighborhood ;

[0051] Step 2.5: Calculate the average background amplitude in the neighborhood. :

[0052] ,

[0053] in, Let be the average value function; the above formula indicates that the calculation satisfies . and All The mean;

[0054] Step 2.6: Determine the time-frequency point Does it meet the time-frequency narrowband characteristics?

[0055] ,

[0056] If the condition is met, proceed to step 2.8; otherwise, proceed to step 2.7.

[0057] Step 2.7: Update the set of invalid frequency-dimensional discrete indices. And determine whether to continue the search. :

[0058] ,

[0059] ,

[0060] judge Check if the condition is met. If it is met, proceed to step 2.3; otherwise, proceed to step 2.9.

[0061] Step 2.8: Record the time index instantaneous frequency :

[0062] ,

[0063] The instantaneous frequency recorded at this point is the valid instantaneous frequency. After completing this step, let... Let the set be empty, and let the cumulative value of invalid indexes be... Proceed to step 2.10;

[0064] Step 2.9: Record the time index instantaneous frequency :

[0065] ,

[0066] The instantaneous frequency recorded at this point is an invalid instantaneous frequency. After completing this step, let... Let the set be empty, and let the cumulative value of invalid indexes be... Proceed to step 2.10;

[0067] Step 2.10, let Determine whether the following conditions are true:

[0068] ,

[0069] If the above conditions are met, proceed to step 2.3; otherwise, proceed to step 3.

[0070] Step 3: Calculate the smoothed instantaneous frequency And set the optimal window length Specifically:

[0071] Step 3.1: Initialize all parameters:

[0072] Smooth window length Initialize to: Positive integers;

[0073] Highest order of polynomial fitting Initially: satisfies Positive integers;

[0074] Smoothed instantaneous frequency Initialize to: , .

[0075] Step 3.2: Calculate the lower bound of the effective instantaneous frequency index. and the Upper Realm :

[0076] ,

[0077] ,

[0078] in, Indicates from Find the range that satisfies The smallest value, Indicates from Find the range that satisfies The largest The value, calculated using the above formula and combined with the fact that invalid instantaneous frequency values ​​were recorded as -1 in step 2, yields the lower bound of the valid instantaneous frequency index. and the Upper Realm .

[0079] Step 3.3: Initialize the time-dimensional discrete index values ​​of the time-frequency distribution. ;

[0080] Step 3.4: Use polynomials to... The length of the indexed lower window is The instantaneous frequency data is fitted, and the coefficients of the fitting polynomial are solved. :

[0081] Step 3.4.1: Define the lower bound of the current time index window. Upper Realm and fitting error :

[0082] ,

[0083] ,

[0084] .

[0085] Step 3.4.2: Calculate the fitting error Regarding the partial derivatives of the coefficients of each fitted polynomial:

[0086] ;

[0087] For ease of representation, the above system of equations is expressed in matrix form:

[0088] definition for coefficient matrix:

[0089] ,

[0090] in, The coefficient matrix row and column indexes, Represents the coefficient matrix The Line 1 Column elements;

[0091] definition for coefficient vector:

[0092] ,

[0093] Where T is the transpose symbol;

[0094] definition for The constant term vector:

[0095] ,

[0096] in, Vector representing constant terms No. Row elements;

[0097] In summary, the system of equations can be written as:

[0098] ,

[0099] Step 3.4.3: Solve the matrix equation to obtain the coefficient vector. :

[0100] ,

[0101] in, Represents the coefficient matrix Find the reverse. To fit the polynomial coefficient vector, the above equation yields a set of fitting polynomial coefficients that minimize the error. .

[0102] Step 3.5: Calculate the polynomial fitting coefficients. Smoothed instantaneous frequency :

[0103] ,

[0104] Step 3.6, let and determine the index Does it meet the following conditions:

[0105] ,

[0106] If the above conditions are met, proceed to step 3.4; otherwise, proceed to step 3.7.

[0107] Step 3.7: Calculate the optimal window length :

[0108] .

[0109] Step 4: Based on the smoothed instantaneous frequency Calculate the pre-guidance direction for each time period And construct an overcomplete dictionary of Chirp atomic directions. Specifically:

[0110] Step 4.1: Initialize all parameters:

[0111] Total number of segments of the smoothed instantaneous frequency signal Initialized to: satisfy Positive integers;

[0112] Pre-guidance direction for each time period Initialize to: , , Discrete indexes for signal segmentation;

[0113] Number of angle expansion scales Initialized to: satisfy Positive integers;

[0114] Angle expansion factor Initialized to: satisfy The real number.

[0115] Step 4.2: Initialize the signal segment index Lowest Frame Index ;

[0116] Step 4.3: Divide the signal into time segments and obtain the time frame index of each segment. :

[0117] ,

[0118] Step 4.4: Calculate the time-frequency domain directional pre-guidance for each time period. :

[0119] ,

[0120] in, This is the time frame index after signal segmentation. hour, ;

[0121] Step 4.5, let Determine whether the following conditions are true:

[0122] ,

[0123] If the condition is met, proceed to step 4.3; otherwise, proceed to step 4.6.

[0124] Step 4.6: Construct a complete dictionary of Chirp atoms. :

[0125] .

[0126] Step 5: Calculate the matching direction for each time period. And match the direction according to each time period. With optimal window length Calculate the time-frequency distribution of the improved Chirplet transform. Specifically:

[0127] Step 5.1: Initialize all parameters:

[0128] Chirp Atom Complete Dictionary The Middle The first segment of the signal Rotation angle Initialize to: That is, initialized as a Chirp atomic overcomplete dictionary The OK Column elements , , ;

[0129] Gaussian window function parameters of Chirplet transform Initialized to: satisfy positive real numbers, For optimal window length, The attenuation factor of the main lobe is used to satisfy... Positive real numbers;

[0130] Improved frequency length of Chirplet transform time-frequency distribution Initialized to: satisfy Positive integers.

[0131] Step 5.2: Initialize the ship signal segment discrete index. Lowest start time index ;

[0132] Step 5.3: Divide the ship signal to be processed into time segments and obtain the start time index of each segment. :

[0133] ,

[0134] in, For segmented indexes, The total number of segments after segmenting the ship signal is the same as the total number of segments of the smoothed instantaneous frequency signal;

[0135] Step 5.4: Convert the time-domain signal Truncation, obtaining the truncated signal :

[0136] ,

[0137] ,

[0138] in, hour, .

[0139] Step 5.5: Calculate the angle of the chirp atom after windowing and truncation using a complete dictionary. :

[0140] ,

[0141] in, Chirp atoms are represented by an overcomplete dictionary. The Middle OK Column elements;

[0142] Step 5.6: Calculate the truncated signal With dictionary atoms Normalized cross-correlation function :

[0143] ,

[0144] Step 5.7, for the first For segmented signals, select the optimal ordinal number corresponding to the angular dictionary atom that maximizes the cross-correlation function. :

[0145] ,

[0146] The above formula is used to extract from... Find the first one in the middle. Normalized cross-correlation function The ordinal number with the largest absolute value ;

[0147] Step 5.8, Calculate the first... Matching direction of segment signal :

[0148] ,

[0149] The above formula represents the complete dictionary from Chirp atoms. The Find the matching direction in the column ;

[0150] Step 5.9, let Determine whether the following conditions are true:

[0151] ,

[0152] If the condition is met, proceed to step 5.3; otherwise, proceed to step 5.10.

[0153] Step 5.10: Calculate the improved Chirplet time-frequency distribution. :

[0154] ,

[0155] in, To improve the frequency-dimensional discrete index of the Chirplet time-frequency distribution, .

[0156] Step 6: Combine the optimal window length Calculate the optimal short-time Fourier transform And correct and improve the time-frequency distribution of the Chirplet transform. The final time-frequency distribution is obtained. Specifically:

[0157] Step 6.1: Initialize all parameters:

[0158] Short-time Fourier Transform of Optimal Window Length After Interpolation Initialize to: , , ;

[0159] Effective instantaneous frequency time-frequency range tolerance initialization: Positive integers;

[0160] Calculate the scaling factor of the 0-1 time-frequency mask. Initialize to: The real number.

[0161] Step 6.2: Calculate the short-time Fourier transform under the optimal window length. :

[0162] ,

[0163] ,

[0164] in, To employ a window function with the optimal window length, a rectangular window is used here;

[0165] Step 6.3: Initialize the time-frequency distribution discrete index. ;

[0166] Step 6.4: Calculate the time discrete index The optimal window length after interpolation and the short-time Fourier transform time-frequency distribution at that point :

[0167] ,

[0168] For all of the above formulas By performing the above interpolation operation on the frequency index, we can obtain the index. The optimal window length after interpolation and the short-time Fourier transform time-frequency distribution at that point .

[0169] Step 6.5, Calculation Determine whether the following conditions are met:

[0170] ,

[0171] If the condition is met, proceed to step 6.4; otherwise, proceed to step 6.6.

[0172] Step 6.6: Calculate the possible regions where instantaneous frequencies may exist. :

[0173] ,

[0174] The meaning of the above formula is to find in and All time-frequency points within the range that satisfy the conditions in parentheses A set;

[0175] Step 6.7: Calculate the background time-frequency region With the corresponding average amplitude :

[0176] ,

[0177] ,

[0178] The above can be calculated within the background time-frequency region. mean ;

[0179] Step 6.8: Calculate the 0-1 time-frequency mask. :

[0180] ,

[0181] Step 6.9: Calculate the final time-frequency distribution result. :

[0182] .

[0183] The following is an example.

[0184] The simulation signal parameters are set as follows: sampling time length , signal component It is a hyperbolic frequency modulated signal, signal amplitude The start time Duration initial phase Starting frequency Periodic slope sampling frequency Observation data sequence points Signal-to-noise ratio .

[0185] Based on step 1, obtain the sampled data sequence of the ship signal to be processed. , , , , .

[0186] Based on step 2, initialize the window length of the short-time Fourier transform window function. Short-time Fourier transform window function step Frequency length in short-time Fourier transform Invalid frequency-dimensional discrete index set Given an empty set, calculate the size of the narrowband feature's time dimension neighborhood. Calculate the neighborhood size of the narrowband feature frequency dimension. Narrowband characteristic coefficients Invalid frequency dimension discrete index upper limit Instantaneous frequency Initialize to: , Based on the above initialization parameters, calculate the sampled data sequence. Short-time Fourier Transform The time-frequency analysis results are as follows: Figure 2 As shown, the instantaneous frequency of the ship signal is estimated using the time-frequency narrowband characteristics. .

[0187] Based on step 3, initialize the smooth window length. The highest order of polynomial fitting Smoothed instantaneous frequency Based on the initialization parameters described above, calculate the smoothed instantaneous frequency. And the optimal window length is calculated. .

[0188] Based on step 4, initialize the number of signal segments. Pre-guidance direction for each time period Number of angles and scales Angle expansion factor Based on the above initialization parameters, and according to the smoothed instantaneous frequency... Calculate the pre-guidance direction for each time period And construct an overcomplete dictionary of Chirp atomic directions. .

[0189] Based on step 5, initialize the Chirp atomic overcomplete dictionary. The Middle The first segment of the signal Rotation angle The attenuation factor of the main lobe The parameters of the Gaussian window function in the Chirplet transform Improve the frequency length of the time-frequency distribution of the Chirplet transform. Based on the initialization parameters mentioned above, the matching direction for each time period is calculated. And match the direction according to each time period. With optimal window length Calculate the time-frequency distribution of the improved Chirplet transform. ,like Figure 3 As shown.

[0190] Based on step 6, initialize the optimal window length short-time Fourier transform after interpolation. Effective instantaneous frequency time-frequency range tolerance Calculate the scaling factor of the 0-1 time-frequency mask. Based on the above initialization parameters, combined with the optimal window length... Calculate the optimal short-time Fourier transform And perform interpolation, such as Figure 4 As shown, the time-frequency distribution of the Chirplet transform is corrected and improved. The final time-frequency distribution is obtained. ,like Figure 5 As shown.

[0191] Based on the same inventive concept, this application provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the aforementioned adaptive time-frequency analysis method for ship signals based on instantaneous frequency direction self-adjustment.

[0192] Based on the same inventive concept, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the aforementioned adaptive time-frequency analysis method for ship signals based on instantaneous frequency direction self-adjustment.

[0193] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0194] This invention is described with reference to flowchart illustrations of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each step in the flowchart, and combinations of steps in the flowchart, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing device, generate instructions for implementing the steps in the flowchart. Figure 1 A device for a function specified in one or more processes.

[0195] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 The function specified in one or more processes.

[0196] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 Steps of a specified function in one or more processes.

[0197] The above embodiments are merely illustrative of the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solutions based on the technical concept proposed in this invention shall fall within the scope of protection of this invention.

Claims

1. An adaptive time-frequency analysis method for ship signals based on instantaneous frequency direction self-adjustment, characterized in that, The steps include: Step 1: Acquire the ship signal to be analyzed in time and frequency, and perform short-time Fourier transform on the ship signal to obtain the short-time Fourier transform time-frequency distribution of the ship signal; Step 2: Based on the short-time Fourier transform time-frequency distribution of the ship signal, combined with the time-frequency narrowband characteristics, estimate the instantaneous frequency of the ship signal and retain the effective instantaneous frequency. Step 3: Perform smoothing calculations on the effective instantaneous frequencies to obtain the smoothed instantaneous frequencies, and calculate the optimal window length; Step 4: Divide the smoothed instantaneous frequency into time segments, calculate the time-frequency domain direction pre-guidance for each time segment, and construct an overcomplete dictionary of Chirp atom directions; the specific process is as follows: Step 4.1, divide the smoothed instantaneous frequency into time segments. part, Obtain the time frame index for each time period. : in, The time length of the time-frequency distribution. For discrete indexes of instantaneous frequency segments after smoothing, ; Step 4.2: Calculate the time-frequency domain directional pre-guidance for each time period. : in, Sampling frequency, The sampling time length, The frequency length of the time-frequency distribution. and The first and Time frame index for each time period, and The first Hedi The instantaneous frequency after smoothing of the time frame; hour, ; Step 4.3: Construct an overcomplete dictionary of Chirp atom directions : in, To expand the number of scales for the angle, As an angle expansion factor, , ; Step 5: Divide the ship signal to be analyzed into time segments. The total number of segments is equal to the total number of segments of the smoothed instantaneous frequency. Calculate the matching direction of the ship signal in each time period based on the Chirplet atomic direction overcomplete dictionary. Then, calculate the improved Chirplet transform time-frequency distribution based on the matching direction and optimal window length of the ship signal in each time period. The specific process is as follows: Step 5.1: Transfer the ship signals to be analyzed in time and frequency from Step 1. Divided into time Segment, obtain the start time index of each signal segment. : in, For time-discrete indexes, , This represents the total number of sampling points. Discrete index for ship signal segments; Step 5.2, send ship signals Truncation, obtaining the truncated signal : in, hour, ; and The first and The start time index of each segmented signal; Step 5.3: Calculate the dictionary atoms after overcomplete dictionary windowing truncation in the Chirp atom direction. : in, The imaginary unit, An overcomplete dictionary representing the Chirp atom orientation The Middle OK Column elements, , For the smoothed instantaneous frequency Pre-guidance of direction in the time-frequency domain for each time period; Step 5.4, Calculate the truncated signal With dictionary atoms Normalized cross-correlation function : Step 5.5, for the first For segmented ship signals, select the ordinal number corresponding to the dictionary atom that maximizes the absolute value of the normalized cross-correlation function. : Step 5.6, traverse the dictionary from the Chirp atom direction. The Find the first one in the column. Matching direction of segment ship signals : Step 5.7, Calculate the improved Chirplet transform time-frequency distribution. : in, The parameters are the Gaussian window function parameters of the Chirplet transform. , For the optimal window length, The attenuation factor of the main lobe. , To improve the frequency-dimensional discrete index of the Chirplet time-frequency distribution, , To improve the frequency length of the time-frequency distribution in the Chirplet transform, For time-dimensional discrete indexes, Indicates the matching direction; Step 6: Combine the optimal window length to calculate the optimal short-time Fourier transform, and correct and improve the time-frequency distribution of the Chirplet transform to obtain the final time-frequency distribution.

2. The adaptive time-frequency analysis method for ship signals based on instantaneous frequency direction self-adjustment according to claim 1, characterized in that, The specific process of step 1 is as follows: Using time data series This represents the collected ship signals to be analyzed in time and frequency. , , The sampling time length, Sampling frequency, ; The short-time Fourier transform of the collected ship signals is as follows: in, This represents the short-time Fourier transform time-frequency distribution of ship signals. They are respectively Time-dimensional and frequency-dimensional discrete indexes , , , This is the window length of the short-time Fourier transform window function. , For the stepping of the short-time Fourier transform window function, , .

3. The adaptive time-frequency analysis method for ship signals based on instantaneous frequency direction self-adjustment according to claim 1, characterized in that, The specific process of step 2 is as follows: Step 2.1, Initialize the time-dimensional discrete index values ​​of the time-frequency distribution. 、 Invalid index cumulative value and invalid frequency dimension discrete index set Invalid frequency dimension discrete index upper limit , that is to say , , It is an empty set. ; Step 2.2, for the time-frequency distribution... Amplitude in the time dimension Select the maximum value and find the frequency-dimensional discrete index corresponding to the largest amplitude value. : Step 2.3, Calculate the time-frequency points neighborhood : in, They are used to calculate the size of the neighborhood in the time dimension and frequency dimension of the narrowband features, respectively. , , Discrete index for frequency dimension; Step 2.4, in the neighborhood Calculate the average amplitude of the background : in, It is a function of average value. Short-time Fourier transform time-frequency distribution of ship signals The amplitude; Step 2.5, determine the time-frequency point Does it meet the time-frequency narrowband characteristics, i.e., the instantaneous frequency point? amplitude Is it greater than , If the coefficients are narrowband characteristic coefficients, proceed to step 2.7; otherwise, proceed to step 2.

6. Step 2.6, let and Update the set of invalid frequency-dimensional discrete indices and the cumulative value of invalid indices, and determine whether to continue searching. That is, the updated Is it less than If yes, return to step 2.2; otherwise, proceed to step 2.

8. Step 2.7, calculate the discrete index value of the time-frequency distribution in the time dimension. instantaneous frequency : The instantaneous frequency at this point is the effective instantaneous frequency; let Cumulative value of empty set and invalid index Proceed to step 2.9; Step 2.8, let the time dimension discrete index value of the time-frequency distribution be... instantaneous frequency The instantaneous frequency at this point is an invalid instantaneous frequency; let Cumulative value of empty set and invalid index Proceed to step 2.9; Step 2.9, let Update the discrete index value of the time-frequency distribution in the time dimension, and determine the updated value. Is it less than If yes, return to step 2.2; otherwise, proceed to step 3.

4. The adaptive time-frequency analysis method for ship signals based on instantaneous frequency direction self-adjustment according to claim 1, characterized in that, The specific process of step 3 is as follows: Step 3.1, calculate the lower bound of the effective instantaneous frequency index. and the Upper Realm : in, Effective instantaneous frequency; Step 3.2, Initialize the time-dimensional discrete index values ​​of the time-frequency distribution. and smoothed instantaneous frequency , that is to say Smoothed instantaneous frequency The initial value is the effective instantaneous frequency. ; Step 3.3, use a polynomial to adjust the index values The length of the lower smooth window is Fit the effective instantaneous frequency. Solve for the coefficients of the fitted polynomial, i.e.: Step 3.3.1, Define the current index value Window lower bound Upper Realm and fitting error : in, To fit the polynomial coefficients, The highest order of the polynomial fitting is given. ; Step 3.3.2, calculate the fitting error. With respect to the partial derivatives of the coefficients of each fitted polynomial, set the partial derivatives to 0, obtain the matrix equation, and solve it to obtain the coefficients of the fitted polynomial. Step 3.4: Calculate based on the coefficients of the fitted polynomial. Smoothed instantaneous frequency : Step 3.5, let Update the discrete index value of the time-frequency distribution in the time dimension, and determine the updated value. Is it less than or equal to? If yes, return to step 3.3; otherwise, proceed to step 3.

6. Step 3.6, Calculate the optimal window length : in, This is the window length of the short-time Fourier transform window function. and They are respectively and The instantaneous frequency after smoothing at the point.

5. The adaptive time-frequency analysis method for ship signals based on instantaneous frequency direction self-adjustment according to claim 1, characterized in that, The specific process of step 6 is as follows: Step 6.1: Calculate the short-time Fourier transform under the optimal window length. : in, Discrete index for frequency dimension; Step 6.2, calculate the discrete index for each time dimension. Optimal window length after interpolation, short-time Fourier transform time-frequency distribution : Frequency-dimensional discrete index for all improved Chirplet time-frequency distributions Perform the above interpolation operation. To obtain the discrete index for each time dimension Optimal window length after interpolation, short-time Fourier transform time-frequency distribution ; Step 6.3, calculate the possible regions where the instantaneous frequency may exist. : in, and These are the lower and upper bounds of the effective instantaneous frequency index, respectively. For the smoothed instantaneous frequency, Sampling frequency, To provide an effective instantaneous frequency time-frequency range tolerance, ; Step 6.4, Calculate the background time-frequency region With the corresponding average amplitude : Step 6.5, calculate the 0-1 time-frequency mask. : in, This is the proportionality coefficient. ; Step 6.6, calculate the final time-frequency distribution. :

6. A computer device comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the adaptive time-frequency analysis method for ship signals based on instantaneous frequency direction self-adjustment as described in any one of claims 1 to 5.

7. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the adaptive time-frequency analysis method for ship signals based on instantaneous frequency direction self-adjustment as described in any one of claims 1 to 5.

Images