Auditory nerve feature extraction method for ship radiation noise identification

Abstract

The invention belongs to the technical field of signal processing. The invention provides an auditory nerve feature extraction method for ship radiation noise recognition. The method comprises the following steps: acquiring electroencephalogram data induced by ship radiation noise stimulation, and preprocessing electroencephalogram signals induced by different ship radiation noise stimulation to obtain electroencephalogram segment data; performing electroencephalogram feature statistical analysis on the electroencephalogram segmented data to obtain an electroencephalogram data set; performing feature extraction and dimension reduction operation on the electroencephalogram data set to obtain electroencephalogram features; training and testing an SVM (Support Vector Machine) model by utilizing the electroencephalogram characteristics to obtain a trained SVM model; and inputting a to-be-tested data set into the trained SVM model to obtain auditory nerve features. According to judgment strategies adopted by different types of ships, a basis is provided for designing a more effective recognition mechanism, and particularly, a personalized recognition method can be adopted during ship classification, so that the adaptability and accuracy of the system are improved.

Description

Technical Field

[0001] The disclosed embodiments relate to the field of signal processing technology, and in particular to an auditory nerve feature extraction method for ship radiated noise identification. Background Art

[0002] Ship radiated noise is the noise generated by the mechanical operation and movement of the ship and radiated into the water. It is one of the main sources of marine environmental noise. Using ship radiated noise to identify and track different types of ships has important application value in the fields of marine traffic safety and marine environmental monitoring. However, the marine acoustic environment is complex and changeable. The interaction of multiple natural and human factors has affected the propagation of sound waves in the water in many ways, which brings great challenges to the identification of ship radiated noise.

[0003] Early identification of ship radiated noise mainly relied on traditional signal processing technology, which identified the ship type by analyzing the spectrum and time domain characteristics of the radiated noise. With the rapid development of machine learning and pattern recognition technology, researchers began to explore the use of feature extraction and classification algorithms to improve the recognition accuracy of ship radiated noise. In recent years, the application of deep learning technology has provided new solutions in this field. Deep learning models such as convolutional neural networks (CNN) and recurrent neural networks (RNN) have been widely used to improve recognition efficiency and accuracy. At the same time, researchers have also constructed a public underwater acoustic target recognition dataset Deepship, and based on this dataset, a large number of studies on the recognition of different types of ships have been carried out. Although many achievements have been made, in actual scenarios, the existing deep learning recognition models cannot directly replace humans. The reason is that a good deep learning recognition model requires a large amount of data set training, the model trained with a small data set has poor generalization performance, and the deep learning model has poor interpretability.

[0004] In comparison, humans have advantages that machines do not have in some special fields. For example, humans can make correct decisions with small samples based on experience and intuition. Therefore, exploring human recognition mechanisms has important research value for inspiring machine recognition algorithms. Electroencephalogram (EEG) signals can reflect the activities of the human brain in the perception, understanding and cognition of acoustic signals. By extracting and analyzing EEG features, human cognitive abilities can be integrated into machine detection systems to improve the performance of ship radiated noise identification. In the field of computer vision, some scholars have decoded EEG signals during image detection and applied them to machine detection systems. In the field of human-computer voice interaction, the team proposed an intention recognition method that integrates EEG, text and voice features for speech interaction. In the field of acoustic target detection, a team has also proposed a sound target detection method based on EEG signals. These studies have achieved good results.

[0005] However, there is no relevant research on the key neural response feature extraction methods for ship radiated noise type identification.

[0006] Therefore, it is necessary to improve one or more problems existing in the above-mentioned related technical solutions.

[0007] It should be noted that this section is intended to provide background or context for the technical solutions of the present disclosure stated in the claims. The description herein is not admitted to be prior art by virtue of being included in this section. Summary of the invention

[0008] The purpose of the embodiments of the present disclosure is to provide an auditory nerve feature extraction method for ship radiated noise identification, thereby overcoming one or more problems caused by the limitations and defects of related technologies at least to a certain extent.

[0009] According to an embodiment of the present disclosure, a method for extracting auditory nerve features for identifying ship radiated noise is provided, the method comprising: Acquiring electroencephalogram (EEG) data induced by ship radiation noise stimulation, and preprocessing the EEG data to obtain EEG segmentation data; Performing EEG feature statistical analysis on the EEG segmented data to obtain an EEG data set; Performing feature extraction and dimensionality reduction operations on the EEG data set to obtain EEG features; Using the EEG features to train and test the SVM model to obtain the trained SVM model; The data set to be tested is input into the trained SVM model to obtain auditory nerve features.

[0010] Furthermore, the step of preprocessing the EEG data to obtain EEG segmented data includes: The EEG data are sequentially downsampled, low-pass filtered and high-pass filtered, mains eliminated, artifact eliminated, re-referenced, independent component analyzed, independent component selected, dipfit source localized and segmented to obtain the EEG segmented data; wherein the EEG segmented data includes S short group, Y short group, S long group, Y long group, correct short group, incorrect short group, correct long group and incorrect long group.

[0011] Furthermore, the downsampling operation is: reducing the sampling rate from 1000 Hz to 256 Hz; the low-pass filtering and high-pass filtering are 60 Hz low-pass filtering and 1 Hz high-pass filtering; The mains frequency of the mains power elimination is 50Hz; The segmentation operation is: according to the ship type discrimination task, to obtain the EEG segmentation data.

[0012] Furthermore, the step of performing EEG feature statistical analysis on the EEG segmented data to obtain an EEG data set includes: The EEG segmented data are grouped to obtain an S-short group-Y-short group, an S-long group-Y-long group, a correct short group-an incorrect short group, and a correct long group-an incorrect long group; Pre-calculating the EEG features of statistical analysis according to the S-short group-Y-short group, the S-long group-Y-long group, the correct short group-incorrect short group and the correct long group-incorrect long group, clustering independent components using ERSP and dipole positions as clustering features; Set statistical parameters for permutation testing and FDR correction; The drawing parameters are set and statistics are performed to obtain the EEG data set.

[0013] Furthermore, the EEG features include Spectrum features and ERSP features.

[0014] Furthermore, the step of performing feature extraction and dimensionality reduction operations on the EEG data set to obtain EEG features includes: Extracting time-frequency features of the EEG data set; The PCA algorithm is used to reduce the dimension of the time-frequency features to obtain the auditory nerve features, and the number of feature dimensions with principal component cumulative contribution rates of 0.5, 0.6, 0.7, 0.8, and 0.9 are selected as the final number of feature vectors.

[0015] Furthermore, the step of training and testing the SVM model using the EEG features to obtain the trained SVM model includes: The EEG features are divided into 5 parts by using 5-fold cross validation, and 4 of them are used as training sets and 1 is used as a test set in turn; wherein the sample ratio of S ship and Y ship in the EEG features is 3:1, and stratified sampling is used when dividing the training set and the test set to ensure that the category ratio in each fold is the same as the category ratio in the original data set; The SVM model is trained and tested using the training set and the test set to obtain the trained SVM model; wherein, during the training process of the SVM model, the misclassification cost of the S ship and the Y ship is adjusted to 1:3.

[0016] Furthermore, the SVM model is implemented by the fitcsvm() function in MATLAB, KernelFunction is the Gaussian kernel rbf, BoxConstraint is set to 5, KernelScale is set to auto, the misclassification cost Cost is set to cost_matrix = [0 3; 1 0], and Standardize is set to true.

[0017] The technical solution provided by the embodiments of the present disclosure may have the following beneficial effects: In the embodiment of the present disclosure, through the above-mentioned auditory nerve feature extraction method for ship radiation noise identification, on the one hand, the EEG data induced by ship radiation noise stimulation is collected, and the EEG signals induced by different ship radiation noise stimulations are preprocessed to obtain EEG segmented data; EEG feature statistical analysis is performed on the EEG segmented data to obtain an EEG data set; feature extraction and dimensionality reduction operations are performed on the EEG data set to obtain EEG features; the SVM model is trained and tested using the EEG features to obtain a trained SVM model; the test data set is input into the trained SVM model to obtain auditory nerve features. On the other hand, according to the discrimination strategies adopted by different types of ships, a basis is provided for designing a more effective recognition mechanism, especially when classifying ships, a personalized recognition method can be adopted to improve the adaptability and accuracy of the system. This method is used to explore the key neural representations of auditory perception when people are discriminating the type of ship radiation noise, in order to provide a basis for the subsequent human-machine fusion ship radiation noise type recognition and the improvement of ship radiation noise recognition methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The accompanying drawings herein are incorporated into the specification and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the specification are used to explain the principles of the present disclosure. Obviously, the accompanying drawings described below are only some embodiments of the present disclosure, and for ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without creative work.

[0019] Figure 1 A step diagram showing an auditory nerve feature extraction method for ship radiated noise identification in an exemplary embodiment of the present disclosure; Figure 2 Showing the types of errors in hypothesis testing in an exemplary embodiment of the present disclosure; Figure 3 A schematic diagram showing the principle of the SVM linear separable model in an exemplary embodiment of the present disclosure; Figure 4 A schematic diagram of EEG data processing grouping in an exemplary embodiment of the present disclosure is shown; FIG5( a ) shows the average reaction time under different ship noise stimuli in an exemplary embodiment of the present disclosure; FIG5( b ) shows the correct rates under different ship noise stimuli in an exemplary embodiment of the present disclosure; FIG5( c ) shows the average reaction time under correct decision and wrong decision in an exemplary embodiment of the present disclosure; Figure 6 A cluster diagram of short-group dipoles without adding noise in an exemplary embodiment of the present disclosure is shown; FIG7 (a) shows a cluster diagram showing significant differences in the spectra of the Cls4 Spectrum under stimulation of two types of ships in the short group without adding noise in an exemplary embodiment of the present disclosure; FIG7( b ) shows a cluster diagram showing significant differences in the spectra of the Cls7 Spectrum under the stimulation of two types of ships in the short group without adding noise in an exemplary embodiment of the present disclosure; FIG7 (c) shows a cluster diagram showing significant differences in the spectra of the Cls9 Spectrum under the stimulation of two types of ships in the short group without adding noise in an exemplary embodiment of the present disclosure; FIG7( d ) shows a cluster diagram showing significant differences in the spectra of the two types of ships under stimulation without adding noise when the Cls10 spectra are used in an exemplary embodiment of the present disclosure; FIG7 (e) shows a cluster diagram showing significant differences in the spectra of the Cls11 Spectrum under stimulation of two types of ships in the short group without adding noise in an exemplary embodiment of the present disclosure; FIG8 (a) shows a cluster diagram showing significant differences in ERSP under two types of ship stimulation in the short group without adding noise when Cls4 ERSP in an exemplary embodiment of the present disclosure; FIG8( b ) shows a cluster diagram showing significant differences in ERSP under two types of ship stimulation in the Cls9 ERSP in an exemplary embodiment of the present disclosure; FIG8( c ) shows a cluster diagram showing significant differences in ERSP under two types of ship stimulation in the short group without adding noise when ERSP is Cls10 in an exemplary embodiment of the present disclosure; FIG8( d ) shows a cluster diagram showing significant differences in ERSP under two types of ship stimulation in the Cls11 ERSP group without adding noise in an exemplary embodiment of the present disclosure. Fig. 9 A long group dipole cluster diagram without adding noise in an exemplary embodiment of the present disclosure is shown; Fig.10 It shows that the Spectrum of the two types of ships in the long group without adding noise in the exemplary embodiment of the present disclosure has significantly different clusters; FIG. 11 ( a ) shows the clustering of ERSPs with significant differences under the stimulation of two types of ships in the long group without adding noise when Cls4 ERSP in an exemplary embodiment of the present disclosure; FIG. 11( b ) shows the clustering of ERSPs with significant differences under the stimulation of two types of ships in the long group without adding noise at Cls6 ERSP in an exemplary embodiment of the present disclosure; Fig.12 A noisy short group dipole cluster diagram in an exemplary embodiment of the present disclosure is shown; FIG. 13 (a) shows the clustering of the spectra with significant differences under the stimulation of two types of ships in the Cls4 Spectrum in the exemplary embodiment of the present disclosure, with the noise short group; FIG. 13 ( b ) shows the clustering of the spectra with significant differences under the stimulation of two types of ships in the Cls7 Spectrum in the exemplary embodiment of the present disclosure, with the noise short group; Fig.14 A cluster diagram of short-group dipoles without adding noise in an exemplary embodiment of the present disclosure is shown; FIG. 15 (a) shows the clustering of the Cls3 Spectrum under the correct and wrong decisions of the short group without adding noise in the exemplary embodiment of the present disclosure; FIG. 15( b ) shows the clustering of the Cls6 Spectrum in the exemplary embodiment of the present disclosure, in which the Spectrum is significantly different between the correct and incorrect decisions of the short group without adding noise; FIG. 15 ( c ) shows the clustering of the Cls7 Spectrum in the exemplary embodiment of the present disclosure, in which the Spectrum is significantly different between the correct and incorrect decisions of the short group without adding noise; FIG. 15 ( d ) shows the clustering of the spectra with significant differences between the correct and incorrect decisions of the short group without adding noise when the Cls10 spectra are used in the exemplary embodiment of the present disclosure; FIG. 15 ( e ) shows the clustering of the spectra with significant differences between the correct and incorrect decisions of the short group without adding noise when the Cls12 spectra are used in an exemplary embodiment of the present disclosure; FIG. 16 ( a ) shows the clustering of ERSPs with significant differences between correct and incorrect decisions of the short group without adding noise when the ERSPs are Cls7 in an exemplary embodiment of the present disclosure; FIG. 16 ( b ) shows the clustering of ERSPs with significant differences between correct and incorrect decisions of the short group without adding noise when the ERSP is Cls12 in an exemplary embodiment of the present disclosure; Fig.17 A long group dipole cluster diagram without adding noise in an exemplary embodiment of the present disclosure is shown; Fig.18It shows that the Spectrum of the correct and incorrect decisions of the long group without adding noise in the exemplary embodiment of the present disclosure has a significant difference in clustering; Fig.19 It shows that the clustering of ERSP under correct and incorrect decisions of the long group without adding noise in the exemplary embodiment of the present disclosure has significant differences; Fig. 20 A noisy short group dipole cluster diagram in an exemplary embodiment of the present disclosure is shown; Fig.21 It shows the clustering of Spectrum with significant difference under correct and wrong decision of short group with noise in the exemplary embodiment of the present disclosure; FIG. 22 ( a ) shows the clustering of ERSPs with significant differences between correct and incorrect decisions of the noisy short group when the ERSP is Cls4 in an exemplary embodiment of the present disclosure; FIG. 22( b ) shows the clustering of ERSPs with significant differences between correct and incorrect decisions of the noisy short group at Cls7 ERSP in an exemplary embodiment of the present disclosure. DETAILED DESCRIPTION

[0020] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that the disclosure will be more comprehensive and complete and to fully convey the concepts of the example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0021] In addition, the accompanying drawings are only schematic illustrations of the embodiments of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated descriptions will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities.

[0022] This example embodiment provides an auditory neural feature extraction method for ship radiated noise identification. Figure 1 As shown in , the auditory nerve feature extraction method for ship radiated noise identification may include: steps S101 to S105.

[0023] Step S101: acquiring EEG data induced by ship radiation noise stimulation, and preprocessing the EEG data to obtain EEG segmentation data; Step S102: performing EEG feature statistical analysis on the EEG segmented data to obtain an EEG data set; Step S103: performing feature extraction and dimensionality reduction operations on the EEG data set to obtain EEG features; Step S104: training and testing the SVM model using the EEG features to obtain the trained SVM model; Step S105: input the test data set into the trained SVM model to obtain auditory nerve features.

[0024] Through the above-mentioned auditory nerve feature extraction method for ship radiation noise identification, on the one hand, the EEG data induced by ship radiation noise stimulation is collected, and the EEG signals induced by different ship radiation noise stimulation are preprocessed to obtain EEG segmented data; EEG feature statistical analysis is performed on the EEG segmented data to obtain an EEG data set; feature extraction and dimensionality reduction operations are performed on the EEG data set to obtain EEG features; the SVM model is trained and tested using the EEG features to obtain a trained SVM model; the test data set is input into the trained SVM model to obtain auditory nerve features. On the other hand, according to the discrimination strategies adopted by different types of ships, a basis is provided for designing a more effective recognition mechanism, especially when classifying ships, a personalized recognition method can be adopted to improve the adaptability and accuracy of the system. This method is used to explore the key neural representations of auditory perception when people are discriminating the type of ship radiation noise, in order to provide a basis for the subsequent human-machine fusion ship radiation noise type recognition and the improvement of ship radiation noise recognition methods.

[0025] Next, we will refer to Figure 1 to Figure 2 2(b) describes in more detail each step of the above-mentioned auditory nerve feature extraction method for ship radiated noise identification in this example embodiment.

[0026] In step S101, EEG data induced by ship radiation noise stimulation is obtained, and the EEG data is preprocessed to obtain EEG segmented data.

[0027] Specifically, EEG signals are very weak, usually only tens of microvolts, and are easily contaminated by irrelevant noise, thus forming various artifacts. Therefore, it is necessary to preprocess and reduce noise on the collected raw EEG data to minimize or eliminate the impact of these artifacts and retain the original real EEG information. The specific steps include downsampling, filtering, re-referencing, segmentation, baseline correction, independent component analysis (ICA), and artifact subspace reconstruction (ASR).

[0028] In step S102 and step S103, EEG feature statistical analysis is performed on the EEG segmented data to obtain an EEG data set; and feature extraction and dimensionality reduction operations are performed on the EEG data set to obtain EEG features.

[0029] Specifically, spectrum is the distribution diagram of the signal in frequency, which can describe the energy distribution of the signal at different frequencies. Spectrum analysis can transform the signal from the time domain to the frequency domain in order to more intuitively analyze the frequency distribution characteristics of the signal. The basis of spectrum analysis is Fourier transform. Fourier transform converts the time domain signal into the frequency domain signal by expressing the time series signal as the sum of a series of sine functions and cosine functions.

[0030] In reality, the EEG signals collected by EEG equipment need to be processed by analog-to-digital converters, so the EEG signals actually collected and analyzed are all sampled at a series of discrete time points. In addition, since computers can only store and process data with a limited number of bits, the time domain, frequency domain and amplitude of EEG signals are all discretized.

[0031] EEG signals are highly non-stationary, which means that the statistical characteristics and spectral density of the signal will change over time, resulting in some time-varying spectral features. Spectral analysis cannot identify the time-varying spectral features of non-stationary EEG signals, which will cause the important information conveyed by these features to be ignored. Therefore, time-frequency analysis technology is needed to provide the distribution of signal power at each time point and each frequency point in order to observe how the spectrum changes over time.

[0032] Commonly used time-frequency analysis methods are based on a general sliding window analysis approach. The sliding window method assumes that non-stationary signals can be divided into a series of internally stationary short data segments, that is, although the spectrum of the overall signal is time-varying, the spectrum of any short data segment is fixed. Then, conventional spectrum analysis methods can be performed on each short data segment. Finally, the spectrum estimates of all short data segments are stacked together to form a spectrum power distribution diagram in the joint time-frequency domain.

[0033] Time-frequency analysis is usually used to detect and analyze ERS (Event-Related Synchronization) and ERD (Event-Related Desynchronization) in EEG signals. ERS and ERD can be detected from ERSP (Event-Related Spectral Perturbation). ERSP is used to describe the change of EEG signal power at different frequencies over time when an event or stimulus occurs, that is, event-related EEG spectrum changes, which are relative to the baseline spectrum of spontaneous EEG. The increase in spectral power in a specific frequency band is called ERS, and the decrease is called ERD.

[0034] Permutation test is a nonparametric test method in hypothesis testing. Unlike traditional parametric tests (such as t-test and analysis of variance), permutation test does not rely on specific distribution assumptions of data. It is still effective when the sample size is small or the data does not meet the normal distribution assumption. Therefore, permutation test is particularly ideal for statistical analysis of EEG data. The principle of permutation test is to repeatedly randomly rearrange data labels to generate new sample combinations, calculate the test statistic for each combination, and thus construct a permutation distribution of the test statistic. Finally, by comparing the position of the original test statistic in the permutation distribution, the probability p value of its occurrence in the distribution is calculated and compared with the preset significance level α to evaluate whether the differences observed in the original data are significant.

[0035] Significance check: The false positive problem is affected by the significance level control. Figure 2 The error types in hypothesis testing are shown in Figure 1. If a set of hypothesis tests are performed simultaneously, the probability of one or more false positives is called the overall type I error rate. If only one comparison is performed, the probability of a false positive is If m independent comparisons are made, the probability of at least one false positive, i.e. the overall type I error rate, is When m takes a larger value, even if there is no significant difference between the samples under the two conditions, the probability of detecting one or more false positives is very high, and it cannot be guaranteed that the significance level This is the so-called multiple comparisons problem, so a multiple comparison correction is needed to control the overall type I error rate.

[0036] False Discovery Rate (FDR) correction is a multiple comparison correction method that attempts to achieve a balance between false positives and false negatives, controlling the ratio of false positives to true positives within a certain range, and has strong statistical power. FDR is defined as the expected value of the ratio of the number of false positives to the number of all rejected null hypotheses. FDR correction corrects the errors caused by multiple comparisons by controlling FDR. When rejecting multiple null hypotheses H0, the FDR correction method can control the possibility of false positives to find a suitable combination of results.

[0037] Feature extraction The features of EEG signals are mainly divided into three categories: time domain features, frequency domain features, and time-frequency domain features. Time domain features reflect the information of EEG signals in the time domain, frequency domain features reflect the information of EEG signals in the frequency domain, and time-frequency domain features are a combination of time domain features and frequency domain features. EEG signals are usually divided into delta band (1-4Hz), theta band (4-8Hz), alpha band (8-13Hz), beta band (13-30Hz), and gamma band (30-50Hz) according to frequency. This study extracted the frequency domain features and time-frequency domain features of these five frequency bands for subsequent learning and classification.

[0038] Power spectral density (PSD) is a commonly used frequency domain feature to describe the distribution of the power of a random signal along the frequency. It is the result of square normalization of the spectrum.

[0039] PCA Dimensionality Reduction The extracted EEG features usually have extremely high data dimensions, but the number of samples is very limited. Directly inputting them into machine learning algorithms for training often causes overfitting problems and reduces the training and prediction rates. Through dimensionality reduction algorithms, high-dimensional features can be reduced to low dimensions while retaining most of the original data information to reduce data complexity and computational costs, thereby reducing the risk of overfitting in machine learning and speeding up machine learning training and prediction rates.

[0040] Principal Component Analysis (PCA) is a mainstream, simple, unsupervised linear dimensionality reduction algorithm. It aims to "minimize reconstruction error" or "maximize scatter points" and uses orthogonal transformation to linearly transform the observed values ​​of a series of possibly related variables, thereby projecting them into a series of linearly unrelated variable values, which are called principal components.

[0041] Assume there are m n-dimensional data, , where each x is an n-dimensional column vector. The main calculation process of the PCA algorithm is as follows: (1) Decentralize the original data: (1) (2) Calculate the covariance matrix: (2) (3) Perform eigenvalue decomposition on the covariance matrix to obtain the characteristic matrix (arranged from large to small eigenvalues), and take the first k columns to form the matrix , P is equivalent to a new coordinate system, each column of which is a coordinate axis.

[0042] (4) Project the original data into the P coordinate system to obtain the reduced-dimensional data: (3) In step S104 and step S105, the SVM model is trained and tested using the EEG features to obtain the trained SVM model; the data set to be tested is input into the trained SVM model to obtain the auditory nerve features.

[0043] Specifically, support vector machines (SVMs) and their related learning algorithms and recognition patterns are widely used in classification and regression tasks as common supervised learning models. The basic principle is to find an optimal hyperplane to divide the data space so that sample points of different categories can be separated by as large an interval as possible. This hyperplane is determined by some of the data points closest to it, which are called support vectors.

[0044] The most basic support vector machine is a non-probabilistic binary linear classifier. Given a set of linearly separable training instances, each labeled as belonging to one or the other of two categories, the SVM training algorithm builds a model that assigns new instances to one of the two categories. The SVM model represents instances as points in space, so that the mapping makes instances of separate categories separated by as wide a clear interval as possible. Then, new instances are mapped to the same space and predicted to belong to a category based on which side of the interval they fall. Figure 3 This is a schematic diagram of the SVM classifier when it is linearly separable.

[0045] In practical applications, most problems are linearly inseparable. In this case, using a linear classifier alone cannot achieve correct classification. When faced with such nonlinear classification problems, SVM uses a nonlinear kernel function to map the nonlinear classification problem in the original low-dimensional space to a high-dimensional space, thereby converting it into a linear classification problem.

[0046] In a specific embodiment, the method for constructing a neural representation model of the human brain's auditory perception of ship radiated noise proposed in this application has the following specific process: Figure 1 The collected data set is shown in Table 1. This application uses two methods to group and analyze EEG data. The specific methods and groupings are as follows: Figure 4 shown.

[0047] Table 1 Data composition of each subject

[0048] like Figure 1 As shown, the technical solution of this application mainly involves three aspects: EEG data preprocessing, EEG feature statistical analysis and machine learning classification.

[0049] First, the collected EEG data induced by ship radiation noise stimulation need to be preprocessed, including: reducing the sampling rate from 1000Hz to 256Hz, 60Hz low-pass filtering and 1Hz high-pass filtering, eliminating the mains (mains frequency is 50Hz), ASR, re-reference (average), independent component analysis, dipfit source localization and segmentation according to grouping standards (grouping such as Figure 4 As shown, the duration of the short group without noise is 5s, the duration of the long group without noise is 25s, the duration of the short group with noise is 8s, and the duration of the long group with noise is 35s).

[0050] Then, the preprocessed EEG segmentation data was statistically analyzed. The specific steps were as follows: select and import paired EEG segmentation data under the same grouping standard (S ship short group-Y ship short group, S ship long group-Y ship long group, correct short group-wrong short group, correct long group-wrong long group), pre-calculate the EEG features (Spectrum and ERSP) that you want to perform statistical analysis on, cluster the independent components using ERSP and dipole position as clustering features (select the k-means method, and set the number of clusters for the non-noise group and the noise group to 10 and 6, respectively), set statistical parameters (permutation test and FDR correction) and drawing parameters (time domain range is grouping time, frequency band range is 1-60Hz), and obtain statistical results. Among them, ships are divided into S ships and Y ships according to their differences in radiated noise.

[0051] In addition, behavioral analysis was performed on the original data. Based on the results of behavioral analysis and EEG feature statistical analysis, as well as considering the observability of the sample size, the EEG data for ship type discrimination without environmental noise was selected to construct a machine learning data set. According to the results of behavioral analysis, it can be seen that the short group data of the three subjects has a higher confidence level, and the EEG feature statistical analysis results also show that the difference in the short group is more significant, so the preprocessed EEG signal is divided into a data segment of [-1s, 5s] relative to the start of the target ship noise stimulus. By observing the data, it is found that only the EEG data of the first 122 channels are valid. Since there is a lot of overlap and redundancy in the 128-channel data, removing the EEG data of 6 channels will not have a great impact on the final machine learning results.

[0052] This application adopts two methods to analyze the EEG data of the classification task. One is to divide the EEG data into S ship noise stimulation group and Y ship noise stimulation group according to the ship type to which the current ship radiation noise stimulus belongs. The other is to divide the EEG data into correct decision group and incorrect decision group according to whether the subject's decision is correct.

[0053] The first classification method is to further divide each group of data into S-ship short reaction time group, Y-ship short reaction time group, S-ship long reaction time group, and Y-ship long reaction time group according to the subject's key reaction time. Calculate the average reaction time and accuracy of each data group, and plot the results into a bar graph. Then, a paired sample t-test was performed on the average reaction time and accuracy rate under S-ship noise stimulation and Y-ship noise stimulation in the short group and the long group respectively (performed in Excel, a paired sample t-test was performed in the short group. Because the sample sizes of S-ship and Y-ship were different, the sample size with the smaller sample size was used as the benchmark to sample the sample size with the larger sample size. For example, the sample size of the S-ship short group was 147, and the sample size of the Y-ship short group was 32. Then, when performing the paired sample t-test, all the samples of the Y-ship short group were included, and then 32 samples were extracted from the S-ship short group using the sampling method in Excel's data-data analysis. Then, a paired sample t-test was performed using the "t-test: paired two-sample analysis of mean" method in Excel's data-data analysis. A paired sample t-test was performed in the long group using the same operation), the significance level α was set to 0.05, the p-value was calculated, and α and p-value were compared. If p<α, it means that there is a significant difference in the average reaction time and accuracy rate under S-ship noise stimulation and Y-ship noise stimulation in the short group and the long group.

[0054] The second classification method is to further divide each group of data into a correct decision short group, an incorrect decision short group, a correct decision long group, and an incorrect decision long group according to the reaction time of the subjects. Calculate the average reaction time of each data group and plot the results into a bar graph. Then perform a paired sample t test on the average reaction time under correct and incorrect decisions in the short group and the long group (the test steps are the same as above), set the significance level α=0.05, calculate the p value, compare α and p value, p<α means that there is a significant difference in the average reaction time under correct and incorrect decisions in the short group and the long group.

[0055] The goal of machine learning is to identify ship types based on EEG features. The EEG data for ship type identification without environmental noise not only has a considerable sample size, but also is selected to construct a machine learning dataset based on the results of behavioral analysis and EEG feature statistical analysis.

[0056] After processing, we finally obtained 681 EEG data samples under S-ship noise stimulation and 223 EEG data samples under Y-ship noise stimulation. The size of each data sample is [122, 1536], where 122 is the number of effective channels and 1536 is the sampling point, which is equal to the product of the sampling rate (256 Hz) and the time length of the sample data segment (6 s).

[0057] For all samples, the frequency domain feature PSD and time-frequency features are extracted using the Welch method and short-time Fourier transform, respectively. The frequency domain information obtained from each channel is a one-dimensional vector containing all frequency domain feature points, and the time-frequency domain information is a matrix with F rows and T columns (F represents frequency points and T represents time points). Then, the frequency domain information (size is [F, 122]) and time-frequency information (size is [F, T, 122]) of the 122 channels are rearranged into a one-dimensional vector as the original frequency domain feature vector and time-frequency domain feature vector of the sample.

[0058] Specifically, when the Welch method is used to extract the frequency domain feature PSD, the Welch method is implemented through the pwelch() function in MATLAB, using the Hamming window as the window function, the window size is set to 256 (1 second), and the window moving step is set to 128. The Welch method is used to calculate each channel of each trial to obtain the spectral density pxx, which is a one-dimensional vector with a size of [F, 1], where F is the number of frequency points. Then all pxx are stored in the corresponding frequency positions of the corresponding channels of a three-dimensional matrix with a shape of [F, channels, trials] to obtain the frequency domain features of all trials and channels. Then, the reshape method of MATLAB is used to convert the three-dimensional matrix into a two-dimensional matrix with a shape of [F*channels, trials], that is, the frequency domain features of all channels of each trial are flattened into a one-dimensional vector. Finally, the zscore method is used to normalize the eigenvectors of the two-dimensional matrix, and then the PCA method is used to calculate the principal components of the normalized data.

[0059] STFT is implemented using the Spectrogram() function in MATLAB, using a Hamming window as the window function, with the window size set to 256 (1 second) and the window moving step set to 10.

[0060] The above-mentioned Welch method and short-time Fourier transform are used to extract the frequency domain features PSD and time-frequency features of the five frequency bands, and after PCA dimensionality reduction, they are saved as .mat files.

[0061] From the statistical analysis results of EEG features, it can be seen that the delta frequency band (1-4Hz), theta frequency band (4-8Hz), alpha frequency band (8-13Hz), beta frequency band (13-30Hz) and gamma frequency band (30-50Hz) all contribute to the discrimination of ship types, so features are extracted for these frequency bands respectively.

[0062] Finally, the zscore method is used to normalize the eigenvectors of the two-dimensional matrix, and then the pca method is used to calculate the principal components of the normalized data. The PCA algorithm (implemented by MATLAB's pca() function) is used to reduce the dimensionality of the eigenvectors, and the number of feature dimensions with cumulative contributions of the principal components of 0.5, 0.6, 0.7, 0.8, and 0.9 are selected as the final number of eigenvectors.

[0063] Support vector machine (SVM) was selected as the algorithm for sample classification. The EEG features extracted above were used as input to train the ship type discrimination model, and the performance of the model was evaluated by 5-fold cross validation. The 5-fold cross validation divided the data set into 5 parts, and 4 of them were used as training data and 1 as test data for training and testing in turn. The category distribution of S ship and Y ship in the data set was not balanced (the sample ratio of S ship and Y ship was 3:1). In order to ensure the stability and representativeness of cross validation, stratified sampling was used when dividing the training set and the test set to ensure that the category ratio of each fold was the same as that in the original data set. At the same time, during the SVM training process, the misclassification cost of S ship and Y ship was adjusted to 1:3 to avoid the model being biased towards one class during training. SVM was implemented by the fitcsvm() function in MATLAB, KernelFunction selected Gaussian kernel rbf, BoxConstraint was set to 5, KernelScale was set to auto, the misclassification cost Cost was set to cost_matrix = [0 3; 1 0], and Standardize was set to true.

[0064] Specifically, in the SVM stage, firstly, according to the return value of the PCA method (score: data after PCA dimension reduction; latent: variance of each principal component, indicating the importance of each principal component), the number of feature dimensions with cumulative contribution rate of principal components (calculated according to variance latent) of 0.5, 0.6, 0.7, 0.8, and 0.9 is selected as the final number of feature vectors, and the data is randomly shuffled and saved again. Then, these processed EEG features are used as input to train the SVM model for ship type discrimination, and the performance of the model is evaluated by 5-fold cross validation. The 5-fold cross validation divides the data set into 5 parts, and 4 of them are used as training data and 1 as test data for training and testing in turn. The category distribution of S ship and Y ship in the data set is not balanced (the sample ratio of S ship and Y ship is 3:1). In order to ensure the stability and representativeness of cross validation, stratified sampling is used when dividing the training set and test set to ensure that the category ratio of each fold is the same as that of the original data set. At the same time, during the SVM training process, the misclassification cost of S ship and Y ship was adjusted to 1:3 to avoid the model being biased towards one class during the training process. SVM was implemented using the fitcsvm() function in MATLAB, with the Gaussian kernel rbf selected as KernelFunction, BoxConstraint set to 5, KernelScale set to auto, the misclassification cost Cost set to cost_matrix = [0 3; 1 0], and Standardize set to true.

[0065] In a specific embodiment, 1. Behavioral data analysis: Table 2 t-test results of behavioral data (* indicates p<0.05, the first row numbers correspond to the group numbers in Figure 5)

[0066] Figures 5 (a) to 5 (c) and Table 2 are all the results of behavioral data statistical analysis; Figures 5 (a) to 5 (c) are the results of behavioral data analysis of subjects in different groups (* indicates groups with significant differences, see Table 2 for specific information). Among them, Figure 5 (a) is the average reaction time under different ship noise stimuli; Figure 5 (b) is the accuracy rate under different ship noise stimuli; Figure 5 (c) is the average reaction time under correct and incorrect decisions.

[0067] The results show that in terms of behavior, for the same type of ships, some samples of the same type of ships require a long reaction time, while others require a short reaction time, indicating that there are large differences within the samples. Comparing the average reaction time and accuracy of different types of ships, it is found that regardless of whether the ocean environmental noise is added, the shorter the reaction time of the S ship, the higher the accuracy, while the longer the reaction time of the Y ship, the higher the accuracy. This shows that the discrimination strategies required for the S ship and the Y ship are different. Comparing the behavioral responses of the two types of ships, it can also be found that the accuracy of the Y ship is low, the discrimination is more difficult, and it takes longer to make a comprehensive decision based on the stimulus information and previous knowledge and experience.

[0068] 2. Statistical Analysis of EEG Features 2.1 EEG characteristics under different ship noise stimulation Here are the statistical analysis results of the EEG characteristics induced by ship noise stimulation. By comparing the statistical characteristics of the EEG signals induced by S-ship noise stimulation and Y-ship noise stimulation, the differences in the brain responses of the subjects when processing different ship noise stimuli are studied. The EEG data under different ship noise stimulations are divided into a short group without noise, a long group without noise, a short group with noise and a long group with noise. Then, the EEG characteristics under S-ship noise stimulation and Y-ship noise stimulation in each group are statistically analyzed to obtain the dipole aggregation position, Spectrum and ERSP of different clusters in each group. The results show that the Spectrum and ERSP of the EEG signals induced by the two types of ship stimulations in the short group without noise, the long group without noise and the short group with noise have different degrees of differences, and there are no significant differences in the Spectrum and ERSP of the long group with noise.

[0069] (1) Short group without noise Table 3 shows the brain region information of the short group clustering without noise.

[0070] like Figure 6 As shown, this is the dipole clustering diagram of the short group without noise.

[0071] As shown in Figures 7(a), 7(b), 7(c), 7(d) and 7(e), there are cluster diagrams showing significant differences in Spectrum under the stimulation of two types of ships in the short group without noise.

[0072] As shown in Figures 8(a), 8(b), 8(c) and 8(d), there are significant differences in the clustering of ERSP under the stimulation of two types of ships in the short group without noise.

[0073] Table 3 Brain region information of short group clustering without noise (* indicates brain regions with significant differences in Spectrum, # indicates brain regions with significant differences in ERSP)

[0074] (2) No noise long group Table 4 shows the brain region information of the clustered long group without noise.

[0075] like Fig. 9 As shown, this is a long group dipole clustering diagram without noise.

[0076] like Fig.10 As shown, there are clusters with significant differences in Spectrum under the stimulation of two types of ships in the no-noise group.

[0077] As shown in Figure 11 (a) and Figure 11 (b), there are significant differences in the clustering of ERSP under the stimulation of two types of ships in the long group without noise.

[0078] Table 4 Brain region information of clustering of the long group without noise (* indicates brain regions with significant differences in Spectrum, # indicates brain regions with significant differences in ERSP)

[0079] (3) Noise-added short group Table 5 shows the brain region information of the short group clustering with noise.

[0080] like Fig.12 As shown, it is the noisy short group dipole clustering diagram.

[0081] As shown in Figures 13(a) and 13(b), the Spectra of the two types of ships in the noise-added short group have significant differences in clustering.

[0082] Table 5 Brain region information of the short group with noise (* indicates brain regions with significant differences in Spectrum)

[0083] Figure 6-Figure 1The results of Figure 3 (b) and Tables 3-5 show that for the ship type discrimination task, the spectral characteristics and time-frequency characteristics of the occipital, frontal, parietal and temporal lobes of the short group without noise showed significant differences in the two types of ship type discrimination, and there were many different areas. These differences can be found within 1000ms of the stimulus onset, and the theta (4-8Hz), alpha (8-13Hz) and beta (13-30Hz) frequency bands are the most significant. In contrast, only the EEG characteristics of the occipital and frontal lobes in the long group without noise have significant differences, and the time-frequency characteristics are more significant. In the short group with noise, only the occipital, parietal and temporal lobes have a small number of spectral feature differences. Combining the results of the short group without noise and the long group without noise, it can be found that the occipital and frontal lobes have high stability and importance in ship type discrimination. We speculate that the frontal lobe is mainly responsible for high-level cognitive functions such as attention control and cognitive decision-making in this task, while the occipital lobe may be involved in the association and memory of ship radiated noise. The results of the short group without noise and the short group with noise showed that the addition of marine environmental noise increased the difficulty of auditory perception and cognitive discrimination, significantly interfering with the differences in EEG characteristics, but the occipital lobe, parietal lobe and temporal lobe still showed significantly different spectral characteristics, so it is speculated that these three brain regions play a key role in coping with noise interference. In various cases, the EEG characteristics of the occipital lobe showed certain significant differences, indicating that it plays an extremely critical role in distinguishing ship types.

[0084] 2.2 EEG characteristics of correct and incorrect decisions Here, the statistical characteristics of the EEG signals of the subjects when making correct and incorrect decisions are compared in order to find the difference in brain activity under correct and incorrect decisions. The EEG data are still divided into the short group without noise, the long group without noise, the short group with noise, and the long group with noise. Among them, the Spectrum and ERSP of the EEG signals under different decisions in the short group without noise, the long group without noise, and the short group with noise show different degrees of difference.

[0085] (1) Short group without noise like Fig.14 As shown, this is the dipole clustering diagram of the short group without noise.

[0086] Table 6 shows the brain region information of the short group clustering without noise.

[0087] Table 6 Brain region information of clustering of short group without noise (* indicates brain regions with significant differences in Spectrum, # indicates brain regions with significant differences in ERSP)

[0088] As shown in Figures 15(a), 15(b), 15(c), 15(d) and 15(e), there are clusters with significant differences in Spectrum under correct and incorrect decisions of the short group without noise.

[0089] As shown in Figure 16 (a) and Figure 16 (b), there are clusters with significant differences in ERSP under correct and incorrect decisions in the short group without noise.

[0090] (2) No noise long group like Fig.17 As shown, this is a long group dipole clustering diagram without noise.

[0091] Table 7 shows the brain region information of the clustered long group without noise.

[0092] Table 7 Brain region information of clustering of the long group without noise (* indicates brain regions with significant differences in Spectrum, # indicates brain regions with significant differences in ERSP)

[0093] like Fig.18 As shown in Figure 1, there are significant differences in the Spectrum of the correct and incorrect decisions of the long group without adding noise.

[0094] like Fig.19 As shown, there are clusters with significant differences in ERSP under correct and incorrect decisions in the long group without noise.

[0095] (3) Noise-added short group like Fig. 20 As shown, it is the noisy short group dipole clustering diagram.

[0096] Table 8 shows the brain region information of the short group clustering with noise.

[0097] Table 8 Brain region information of the short noise group clustering (* indicates brain regions with significant differences in Spectrum, # indicates brain regions with significant differences in ERSP)

[0098] like Fig.21 As shown in Figure 3, the Spectrum of the noisy short group has significant differences under correct and incorrect decisions.

[0099] As shown in Figure 22 (a) and Figure 22 (b), there are clusters with significant differences in ERSP under correct and incorrect decisions in the noisy short group.

[0100] Figure 14-Figure 2 The results in Tables 2(b) and 6-8 show that the frontal lobe plays a key role in the decision-making process of the ship type discrimination task. The frontal lobe has a high-level cognitive function to regulate decision-making. Therefore, in the ship type discrimination task, the active state of the frontal lobe may contribute to more accurate judgment and classification.

[0101] 3. Machine learning classification results Table 9 SVM classification results of frequency domain features (The best average classification effect under the current contribution rate is marked in bold)

[0102] Table 10 SVM classification results of time-frequency domain features (The best average classification effect under the current contribution rate is marked in bold)

[0103] Tables 9 and 10 show the SVM classification accuracy of frequency domain features and time-frequency domain features in different frequency bands at different principal component cumulative contribution rates. The analysis of machine learning classification results shows that in the frequency domain features, the classification effect of the beta band is better, and in the time-frequency features, the classification effects of the delta, theta and beta bands are better, which is consistent with the statistical analysis results of the EEG features of the ship type discrimination task. And the classification effect of the time-frequency features is better than that of the frequency domain features. All the results are significantly higher than the chance level (50%), indicating the effectiveness of the EEG feature extraction method proposed in this application for ship type discrimination, as well as the feasibility and application potential of combining EEG features for ship radiated noise identification.

[0104] The above details the construction principle, process and results of the neural representation model of the human brain's auditory perception of ship radiated noise proposed in this application, from which it can be seen that this application has the following beneficial effects: (1) Behavioral analysis found that different types of ships require different discrimination strategies, which can provide a basis for designing more effective identification mechanisms, especially when classifying ships, personalized identification methods can be used to improve the adaptability and accuracy of the system.

[0105] (2) The EEG characteristics of theta (4-8 Hz), alpha (8-13 Hz) and beta (13-30 Hz) frequency bands are particularly different and can be quickly detected within 1000 ms after stimulation, indicating that these low-frequency bands play a key role in ship type discrimination. This provides theoretical support for the development of ship noise recognition models based on these specific frequency bands.

[0106] (3) The frontal lobe plays an important role in the decision-making process of the ship type discrimination task. This finding provides a direction for further research on the function of this brain region and how to enhance its performance in the corresponding task.

[0107] (4) The results of machine learning analysis show that the classification effect of ship type discrimination based on EEG features is excellent in delta, theta and beta frequency bands, especially the classification effect of time-frequency features is significantly better than that of frequency domain features. This proves the discriminative power of EEG signals in complex tasks and its practical application potential in ship noise identification.

[0108] Through the above-mentioned auditory nerve feature extraction method for ship radiation noise identification, on the one hand, the EEG data induced by ship radiation noise stimulation is collected, and the EEG signals induced by different ship radiation noise stimulation are preprocessed to obtain EEG segmented data; EEG feature statistical analysis is performed on the EEG segmented data to obtain an EEG data set; feature extraction and dimensionality reduction operations are performed on the EEG data set to obtain EEG features; the SVM model is trained and tested using the EEG features to obtain a trained SVM model; the test data set is input into the trained SVM model to obtain auditory nerve features. On the other hand, according to the discrimination strategies adopted by different types of ships, a basis is provided for designing a more effective recognition mechanism, especially when classifying ships, a personalized recognition method can be adopted to improve the adaptability and accuracy of the system. This method is used to explore the key neural representations of auditory perception when people are discriminating the type of ship radiation noise, in order to provide a basis for the subsequent human-machine fusion ship radiation noise type recognition and the improvement of ship radiation noise recognition methods.

[0109] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example" or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present disclosure. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification.

[0110] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any modification, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present disclosure are indicated by the appended claims.

Claims

1. An auditory nerve feature extraction method for ship radiated noise identification, characterized in that: The method includes: Acquiring electroencephalogram (EEG) data induced by ship radiation noise stimulation, and preprocessing the EEG data to obtain EEG segmentation data; Performing EEG feature statistical analysis on the EEG segmented data to obtain an EEG data set; Performing feature extraction and dimensionality reduction operations on the EEG data set to obtain EEG features; Using the EEG features to train and test the SVM model to obtain the trained SVM model; The data set to be tested is input into the trained SVM model to obtain auditory nerve features.

2. The auditory nerve feature extraction method for ship radiated noise identification according to claim 1 is characterized in that: The step of preprocessing the EEG data to obtain EEG segmented data includes: The EEG data are sequentially downsampled, low-pass filtered and high-pass filtered, mains eliminated, artifact eliminated, re-referenced, independent component analyzed, independent component selected, dipfit source localized and segmented to obtain the EEG segmented data; wherein the EEG segmented data includes S short group, Y short group, S long group, Y long group, correct short group, incorrect short group, correct long group and incorrect long group.

3. The auditory nerve feature extraction method for ship radiated noise identification according to claim 2 is characterized in that: The downsampling operation is: reducing the sampling rate from 1000 Hz to 256 Hz; the low-pass filtering and high-pass filtering are 60 Hz low-pass filtering and 1 Hz high-pass filtering; The mains frequency of the mains power elimination is 50Hz; The segmentation operation is: according to the ship type discrimination task, to obtain the EEG segmentation data.

4. The auditory nerve feature extraction method for ship radiated noise identification according to claim 2 is characterized in that: The step of performing EEG feature statistical analysis on the EEG segmented data to obtain an EEG data set includes: The EEG segmented data are grouped to obtain an S-short group-Y-short group, an S-long group-Y-long group, a correct short group-an incorrect short group, and a correct long group-an incorrect long group; Pre-calculating the EEG features of statistical analysis according to the S-short group-Y-short group, the S-long group-Y-long group, the correct short group-incorrect short group and the correct long group-incorrect long group, clustering independent components using ERSP and dipole positions as clustering features; Set statistical parameters for permutation testing and FDR correction; The drawing parameters are set and statistics are performed to obtain the EEG data set.

5. The auditory nerve feature extraction method for ship radiated noise identification according to claim 4 is characterized in that: The EEG features include Spectrum features and ERSP features.

6. The auditory nerve feature extraction method for ship radiated noise identification according to claim 5 is characterized in that: The step of performing feature extraction and dimensionality reduction operations on the EEG data set to obtain EEG features includes: Extracting time-frequency features of the EEG data set; The PCA algorithm is used to reduce the dimension of the time-frequency features to obtain the auditory nerve features, and the number of feature dimensions with principal component cumulative contribution rates of 0.5, 0.6, 0.7, 0.8, and 0.9 are selected as the final number of feature vectors.

7. The auditory nerve feature extraction method for ship radiated noise identification according to claim 6 is characterized in that: The step of training and testing the SVM model using the EEG features to obtain the trained SVM model includes: The EEG features are divided into 5 parts by using 5-fold cross validation, and 4 of them are used as training sets and 1 is used as a test set in turn; wherein the sample ratio of S ship and Y ship in the EEG features is 3:1, and stratified sampling is used when dividing the training set and the test set to ensure that the category ratio in each fold is the same as the category ratio in the original data set; The SVM model is trained and tested using the training set and the test set to obtain the trained SVM model; wherein, during the training process of the SVM model, the misclassification cost of the S ship and the Y ship is adjusted to 1:

3.

8. The auditory nerve feature extraction method for ship radiated noise identification according to claim 7 is characterized in that: The SVM model is implemented by the fitcsvm() function in MATLAB, KernelFunction is the Gaussian kernel rbf, BoxConstraint is set to 5, KernelScale is set to auto, the misclassification cost Cost is set to cost_matrix = [0 3; 1 0], and Standardize is set to true.

Images