Marine low-power-consumption weak electromagnetic signal identification and acquisition method and system

Abstract

The invention discloses an ocean low-power-consumption weak electromagnetic signal identification and acquisition method, which comprises the following steps of: removing noise and interference in an ocean weak electromagnetic signal through adaptive filtering, amplifying the preprocessed signal by using a low-power-consumption amplifier, and carrying out secondary filtering on the amplified signal through a digital filter to eliminate residual noise; wavelet transform and Fourier transform are adopted to extract frequency, amplitude and phase features of the signals to construct feature vectors, and the feature vectors are classified and identified based on a machine learning algorithm; the signal acquisition frequency is controlled through the low-power-consumption microcontroller, and the identified data is compressed and encrypted; the signals processed in the fourth step are transmitted to a remote terminal through a low-power-consumption layered transmission method based on an energy cost function; the method improves the recognition precision, reduces the energy consumption, enhances the practicability and response speed of the system, and is suitable for multiple fields of marine environment monitoring, underwater communication and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of marine electromagnetic signal recognition, and in particular to a method and system for recognizing and collecting marine low-power weak electromagnetic signals. Background Art

[0002] With current technological advancements, the identification and acquisition of weak ocean electromagnetic signals remains a challenge in marine scientific research and technological applications. With the continuous development of marine resources and the increasing complexity of the marine environment, the need to monitor weak ocean electromagnetic signals is becoming increasingly urgent. However, due to the unique characteristics of the marine environment, such as deep-sea pressure and seawater conductivity, electromagnetic signals are susceptible to interference and attenuation during propagation, making them weak and difficult to capture. Therefore, how to efficiently and accurately identify and acquire these weak signals has become a pressing issue in marine scientific research and technology.

[0003] Existing technologies for the identification and collection of weak marine electromagnetic signals face the following major problems and technical deficiencies. On the one hand, traditional signal processing methods are limited in their effectiveness in removing noise and enhancing signals, making them inadequate for high-precision identification. On the other hand, existing amplification and filtering technologies often consume high amounts of energy, making them unsuitable for long-term, continuous monitoring. Furthermore, the accuracy and efficiency of feature extraction and intelligent recognition technologies need to be improved to cope with the complex and changing marine environment. These issues and technical deficiencies limit the application and development of existing technologies for the identification and collection of weak marine electromagnetic signals. Summary of the Invention

[0004] The present invention aims to solve the problems of low recognition accuracy, high energy consumption and low efficiency in the prior art of recognition and collection of weak ocean electromagnetic signals.

[0005] The present invention adopts the following scheme: a method for identifying and collecting low-power marine weak electromagnetic signals, comprising the following steps:

[0006] Step 1: Adaptive filtering is used to remove noise and interference from weak ocean electromagnetic signals, and gain control and phase adjustment techniques are used to enhance signal strength and stability.

[0007] Step 2: Amplify the pre-processed signal using a low-power amplifier and perform secondary filtering on the amplified signal using a digital filter to eliminate residual noise;

[0008] Step 3: Use wavelet transform and Fourier transform to extract the frequency, amplitude and phase characteristics of the signal to construct feature vectors, and classify and identify the feature vectors based on machine learning algorithms;

[0009] Step 4: Control the signal acquisition frequency through a low-power microcontroller and compress and encrypt the identified data; the method for controlling the signal acquisition frequency includes: constructing an electromagnetic field gradient threshold model based on Kalman prediction, when the real-time signal S(t) meets When η is the dynamic threshold coefficient, σ noise is the standard deviation of background bioelectric noise, To monitor the signal gradient, a high-precision sampling mode (e.g., sampling rate fs = 1kHz) is started. When the signal becomes stable, the system switches to a low-power monitoring mode (e.g., fs = 10Hz) and uses a time slot allocation algorithm to avoid sampling conflicts with neighboring nodes.

[0010] Step 5: The signal processed in step 4 is transmitted to the remote terminal using a low-power hierarchical transmission method based on an energy cost function, wherein the data acquisition frequency control and the hierarchical transmission method are used to reduce the overall power consumption of the system;

[0011] Furthermore, the specific steps of step 1 are as follows: in step 1, the adaptive filtering adopts a variable step size LMS algorithm to dynamically adjust the filter parameters, specifically including:

[0012] The error between the filtered output signal and the expected signal is calculated, the filter coefficients are iteratively adjusted until the error reaches a preset threshold, and the quality of the filtered signal is evaluated based on the signal-to-noise ratio (SNR).

[0013] Furthermore, the gain control in step 1 adopts an automatic gain control (AGC) system, which maintains the signal amplitude stable by dynamically adjusting the amplifier gain value, calculates the gain coefficient in real time according to the input signal amplitude, and controls the output signal amplitude through negative feedback.

[0014] Furthermore, the digital filter in step 2 is a finite impulse response (FIR) filter or an infinite impulse response (IIR) filter, whose order and cutoff frequency are designed according to the frequency domain characteristics of the ocean electromagnetic signal, and the computational complexity is reduced by optimizing the filter algorithm.

[0015] Furthermore, the machine learning algorithm in step three includes a support vector machine (SVM) and a deep learning model, wherein the support vector machine uses a Gaussian kernel function to classify feature vectors, and the deep learning model is a recurrent neural network (RNN) for processing time-related electromagnetic signal features.

[0016] Furthermore, the method for compressing the identified data in step 4 includes:

[0017] (1) Wavelet transform: Multi-scale wavelet transform is performed on the collected weak ocean electromagnetic signals to decompose the signals into wavelet coefficients at different frequencies and time scales. The wavelet transform formula is:

[0018]

[0019] Among them, s(n) is the original signal, ψ j,k (n) is the wavelet basis function, j and k represent the scale and translation parameters respectively, W j,k is the wavelet coefficient;

[0020] (2) Adaptive quantization: Adaptive quantization strategy is used to quantize the wavelet coefficients according to their amplitudes. The quantization step length Δ is calculated as follows:

[0021]

[0022] Among them, σ is the standard deviation of the signal, β is the quantization coefficient, which is used to control the quantization accuracy; the quantization step size Δ is dynamically adjusted with the amplitude of the wavelet coefficient. For coefficients with larger amplitudes, a smaller quantization step size is used to retain more details, while for coefficients with smaller amplitudes, a larger quantization step size is used to reduce the amount of data;

[0023] (3) Data encoding: Encode the quantized data using Huffman coding or other efficient lossless coding algorithms to further compress the data volume; the encoded data is transmitted to the remote receiving end through a low-power wireless communication module.

[0024] Furthermore, the layered transmission protocol in step five includes: designing an ocean underwater acoustic-electromagnetic hybrid communication protocol and defining a transmission energy cost function:

[0025]

[0026] Among them C CSI Adaptive modulation is channel state information, D is the compressed data packet size, and PTX is the transmission power; when the node is located in an area with strong current disturbance, short burst forward error correction (FEC) coding transmission is preferred; during the silent period, it switches to the low duty cycle underwater acoustic beacon synchronization mode.

[0027] Another invention of the present invention is a marine low-power weak electromagnetic signal identification and collection system, comprising:

[0028] The pre-processing module removes noise and interference from weak ocean electromagnetic signals through adaptive filtering, and uses gain control and phase adjustment technology to enhance signal strength and stability;

[0029] The low-power amplification and filtering module uses a low-power amplifier to amplify the pre-processed signal and performs secondary filtering on the amplified signal through a digital filter to eliminate residual noise;

[0030] The intelligent recognition module uses wavelet transform and Fourier transform to extract the frequency, amplitude and phase characteristics of the signal to construct feature vectors, and classifies and identifies the feature vectors based on machine learning algorithms;

[0031] The data acquisition and transmission module controls the signal acquisition frequency through a low-power microcontroller and compresses and encrypts the identified data.

[0032] As another invention of the present invention, it also relates to a computer storage medium, on which a computer program is stored. When the computer program is executed by a processor, the above-mentioned frequency conversion synchronization method is implemented.

[0033] In general, the above technical solutions conceived by the present invention can achieve the following beneficial effects compared with the prior art:

[0034] (1) The method for identifying and collecting low-power marine weak electromagnetic signals of the present invention improves the recognition accuracy, reduces energy consumption, and enhances the practicality and response speed of the system. It is suitable for multiple fields such as marine environmental monitoring and underwater communication, and shows broad application prospects and market demand.

[0035] (2) The present invention's method for identifying and collecting low-power, weak electromagnetic signals from the ocean optimizes the data collection and transmission process by precisely controlling the data collection frequency and transmission protocol, thereby reducing system power consumption. The data compression module uses advanced compression algorithms to reduce data volume and improve transmission efficiency; the encryption module ensures the security of data transmission; and the low-power wireless communication module uses optimized transmission protocols and communication frequency bands to achieve real-time, reliable data transmission. This system significantly reduces overall system energy consumption and extends its service life. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 is a logic principle diagram of an embodiment of the present invention;

[0037] Figure 2 is a flow chart of the adaptive filtering technology according to an embodiment of the present invention;

[0038] Figure 3 It is the amplitude-based feedback digital AGC system framework of an embodiment of the present invention;

[0039] Figure 4 is a flow chart of a digital filtering technology according to an embodiment of the present invention;

[0040] Figure 5 is a wavelet basis phase plane diagram of an embodiment of the present invention;

[0041] Figure 6 This is a flow chart of a signal recognition technology based on support vector machine and deep learning according to an embodiment of the present invention;

[0042] Figure 7 This is a diagram of a low-power data acquisition and transmission system adopted in an embodiment of the present invention. DETAILED DESCRIPTION

[0043] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only intended to illustrate the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

[0044] Please refer to Figure 1 The present invention relates to a method for identifying and collecting low-power marine weak electromagnetic signals, comprising the following steps:

[0045] Step 1: Adaptive filtering is used to remove noise and interference from weak ocean electromagnetic signals, and gain control and phase adjustment techniques are used to enhance signal strength and stability.

[0046] Step 2: Amplify the pre-processed signal using a low-power amplifier and perform secondary filtering on the amplified signal using a digital filter to eliminate residual noise;

[0047] Step 3: Use wavelet transform and Fourier transform to extract the frequency, amplitude and phase characteristics of the signal to construct feature vectors, and classify and identify the feature vectors based on machine learning algorithms;

[0048] Step 4: Control the signal acquisition frequency through a low-power microcontroller and compress and encrypt the identified data; the method for controlling the signal acquisition frequency includes: constructing an electromagnetic field gradient threshold model based on Kalman prediction, when the real-time signal S(t) meets When η is the dynamic threshold coefficient, σ noise is the standard deviation of background bioelectric noise, To monitor the signal gradient, a high-precision sampling mode is started; when the signal is stable, it switches to a low-power monitoring mode and avoids sampling conflicts with neighboring nodes through a time slot allocation algorithm;

[0049] Step 5: The signal processed in step 4 is transmitted to the remote terminal using a low-power hierarchical transmission method based on an energy cost function, wherein the data acquisition frequency control and the hierarchical transmission method are used to reduce the overall power consumption of the system.

[0050] The S101 marine electromagnetic signal receiver receives weak electromagnetic signals from the ocean environment, which often contain significant amounts of noise and interference. To extract useful signal components, the device employs adaptive filtering technology. Based on the signal's frequency and noise characteristics, the filter parameters are designed and adjusted to effectively remove noise and interference. Furthermore, signal enhancement techniques such as gain control and phase adjustment are employed to improve signal strength and stability, providing high-quality input for subsequent signal processing.

[0051] To reduce power consumption, the S102 uses a low-power amplifier during the signal amplification stage, featuring low power consumption, high amplification factor, and excellent stability. To further improve signal quality, a digital filter is installed after the amplifier, utilizing digital signal processing technology to further filter the signal and remove residual noise and interference. The digital filter design fully considers the characteristics of marine electromagnetic signals, ensuring effective filtering and reducing power consumption.

[0052] During the feature extraction phase, S103 uses wavelet transform and Fourier transform signal processing techniques to extract features from the amplified and filtered signal. By extracting key characteristic parameters such as the signal's frequency, amplitude, and phase, it constructs a feature vector, providing basic data for subsequent intelligent recognition.

[0053] In the intelligent recognition stage (S104), the present invention uses a support vector machine (SVM) and deep learning machine learning algorithms to build a training model to learn and identify the extracted feature parameters. By using a large amount of training data and an optimization algorithm, it can accurately classify and identify weak marine electromagnetic signals, improving the accuracy and efficiency of signal recognition.

[0054] S105 During the data acquisition and transmission phase, the present invention designs a low-power data acquisition and transmission system. This system includes a low-power microcontroller, a data compression module, an encryption module, and a low-power wireless communication module. The low-power microcontroller is responsible for controlling the operation of the entire system, including signal acquisition, processing, and transmission. The data compression module uses advanced compression algorithms to compress the collected data to reduce the data volume and improve transmission efficiency. The encryption module uses encryption technology to ensure the security of data transmission and prevent data from being stolen or tampered with during transmission. The low-power wireless communication module is responsible for transmitting the compressed and encrypted data to the remote receiving end, realizing real-time data transmission and remote monitoring. By optimizing the data acquisition frequency and transmission protocol, power consumption is further reduced, ensuring the long-term stable operation of the system.

[0055] The specific implementation of each step is as follows:

[0056] The weak electromagnetic signals received by the S101 marine electromagnetic signal receiving device in the marine environment often contain a large amount of noise and interference signals. In order to extract useful signal components, the present invention adopts adaptive filtering technology to design and adjust the filter parameters according to the frequency characteristics and noise characteristics of the signal to effectively remove noise and interference. At the same time, through signal enhancement technologies such as gain control and phase adjustment, the strength and stability of the signal are improved, providing high-quality input for subsequent signal processing. The specific implementation steps are as follows ( Figure 2 ):

[0057] The S201 uses a professional marine electromagnetic signal receiving device to accurately capture weak electromagnetic signals from the deep ocean. However, these raw signals are often accompanied by a large amount of noise and interference. To improve signal quality, preliminary signal processing such as signal amplification and DC component removal is performed. Based on the preliminary signal characteristics, the filter type and order are carefully selected, and the initial filter coefficients are preset.

[0058] S202 enters the adaptive filtering stage, where a certain error will exist between the filter's output signal and the desired signal. The filter parameters are adjusted in real time based on the error signal, continuously reducing the output error so that the filter's output signal gradually approaches the desired signal. Each signal iteration and comparison fine-tunes the filter parameters based on the new error signal until the filter's output error reaches a preset threshold or the number of iterations reaches an upper limit. The signal-to-noise ratio (SNR) is calculated to assess the quality of the filtered signal to ensure that the filtering effect meets the requirements. Adaptive filtering uses the least mean square (LMS) algorithm.

[0059] The step size parameter of the LMS algorithm affects its convergence speed and steady-state error. Larger values for μ result in faster convergence, but also larger steady-state errors. This is because a large step size can cause the algorithm to oscillate back and forth near the optimal solution, preventing it from fully reaching the optimal solution, resulting in larger errors. Conversely, smaller step sizes reduce the steady-state error, but also slow convergence. An iterative and comparative approach is employed to compare the optimal solution ranges for different step sizes and select a step size with minimal fluctuations.

[0060] The S203 utilizes gain control technology to enhance weak signals. It accurately measures and analyzes the received signal to determine its current amplitude level, enhancing signal strength without introducing additional noise or nonlinear distortion. Automatic Gain Control (AGC) dynamically adjusts the amplifier's gain based on real-time changes in signal strength to maintain stable signal amplitude.

[0061] Automatic gain control is a method of automatically adjusting the signal level amplitude, which can control the output signal within a certain dynamic range. Traditional digital AGC systems generally use the difference signal between the best steady-state signal of the demodulation and the input signal as the decision factor for gain control. The control process is as follows: Figure 3 shown.

[0062] To design an amplitude-based feedforward AGC system, the amplitude data of the input signal is first estimated and calculated (generally the average value of the amplitude is used).

[0063] Then, the optimal demodulation steady-state level of the receiver is used as the target threshold and the difference between it and the amplitude estimate is made. The difference signal from the previous stage is used to calculate the gain coefficient. The calculation formula is:

[0064] A(n+1)=A(n)+α[R-|A(n)x(n)|]

[0065] Where: A(n) is the gain control coefficient of the AGC system, which is a delay of A(n+1) per unit time, and the initial gain coefficient value is set to 1; n is the discrete time series; α is the weighting factor, which controls the overall speed performance of the system; R is the target steady-state value; and x(n) is the input signal.

[0066] Finally, the negative feedback control method is used to gradually adjust the input signal, and the adjusted signal is used for the next judgment adjustment. Let Y(n) be the output signal. The output signal of the amplitude-based feedback digital AGC is:

[0067] Y(n)=A(n)x(n)

[0068] During signal processing, S204's phase information carries important information, such as the signal's arrival time and frequency characteristics. The signal phase is precisely measured and, based on the measurement results, compensated using a phase adjuster to ensure accurate and stable phase adjustment and avoid introducing additional phase errors. Furthermore, the signal's frequency characteristics must be considered to ensure that the phase remains consistent within the signal bandwidth to maintain overall signal stability.

[0069] After gain control and phase adjustment, the signal's strength and stability are significantly improved, providing high-quality input for subsequent signal processing.

[0070] In order to reduce power consumption, S102 uses a low-power amplifier with low power consumption, high amplification factor and good stability in the signal amplification stage. In order to further improve the signal quality, a digital filter is set at the end of the amplifier, and the signal is further filtered using digital signal processing technology to remove residual noise and interference. The design of the digital filter fully considers the characteristics of marine electromagnetic signals, ensuring the filtering effect and reducing power consumption. The specific implementation steps are as follows ( Figure 4 ):

[0071] S301 determines the requirements for key parameters such as signal amplification factor, signal frequency range, input and output impedance requirements, and power consumption limit. After determining the requirements, select a low-power amplifier with characteristics such as low power consumption, high amplification factor and good stability. Factors such as the amplifier's packaging form, heat dissipation performance, and compatibility with existing circuits need to be considered.

[0072] S302 designs the input and output circuits of the amplifier to ensure that the signal is properly transmitted and amplified. At the same time, pay attention to the stability of the circuit to avoid problems such as self-oscillation. Reduce the static power consumption of the circuit by selecting appropriate components such as resistors and capacitors.

[0073] After completing the circuit design in S303, simulation and testing are required to verify the circuit's performance. Use professional circuit simulation software to simulate the circuit to predict its actual performance. During the simulation process, pay attention to key parameters such as the circuit's amplification factor, frequency response, and power consumption. After the simulation passes, build the actual circuit for testing to verify the circuit's actual performance and stability.

[0074] S304 optimizes and improves the circuit based on the simulation and test results. If the circuit performance does not meet expectations, circuit parameters can be adjusted or more appropriate components can be selected to improve the circuit performance. If the circuit power consumption is too high, the circuit design can be optimized by improving the circuit layout and wiring to reduce power consumption. Through continuous optimization and improvement, the low-power amplifier can achieve optimal performance during the signal amplification stage while meeting power consumption requirements.

[0075] To further improve the quality of marine electromagnetic signals, the S305 incorporates a digital filter after the amplifier. Depending on the signal, either an FIR (Finite Impulse Response) or IIR (Infinite Impulse Response) filter is selected, and key parameters such as the filter order and cutoff frequency are determined. By employing a fixed-coefficient filter and optimizing the filter implementation algorithm, the filter's computational complexity is moderate, resulting in a low-power filter design.

[0076] The system function of the FIR filter is as follows:

[0077]

[0078] Where: N is the length of the FIR filter unit impulse response h(n). H(z) has N-1 zeros on the Z plane and an N-1 multiplicity pole at the origin, so H(z) is always stable. Assuming that the amplitude-frequency characteristic of the ideal low-pass digital filter in the passband is |H d (e jω )|=1, phase-frequency characteristics Cutoff frequency ω c =2πf c , with linear phase and group delay of α, the corresponding time domain filter function is h d (n), the frequency response is:

[0079]

[0080] The FIR digital filter requires a finite-length unit impulse response h(n) to approximate the infinite-length unit impulse response h of the ideal filter. d (n). Currently, the commonly used method is to use a finite-length window function sequence ω(n) to intercept h d The main part of (n) h(n) = h d (n)·ω(n), n=0, 1, 2, ..., N-1, the frequency response e obtained in this way -jω Approximately the ideal filter frequency response H d (e -jω ). The linear phase filter requires that h(n) must be even symmetric, with the center of symmetry being 1 / 2 of its length, i.e. h(n) = h(N-1-n), and α = (N-1) / 2, so the window function ω(n) must also be even symmetric about the center, i.e. ω(n) = ω(N-1-n).

[0081] In the design of FIR low-pass filter based on window function, the cutoff frequency of the filter, filter length and selection of window function are key factors.

[0082] S306 implements and tests the filter to ensure it can accurately remove noise while maintaining signal integrity in practical applications. Test metrics include the filter's frequency response, phase response, power consumption, and filtering effectiveness. Based on the test results, the filter is optimized and adjusted. If the filter's performance does not meet expectations, the filter's order, cutoff frequency, or coefficients can be adjusted to adjust the filter's parameters. Furthermore, more advanced filtering algorithms or techniques, such as adaptive filtering and multi-band filtering, can be used to further improve filter performance.

[0083] In the feature extraction stage, S103 uses wavelet transform and Fourier transform signal processing technology to extract features from the amplified and filtered signal. By extracting key characteristic parameters such as the signal's frequency, amplitude, and phase, a feature vector is constructed to provide basic data for subsequent intelligent recognition. The specific implementation steps are as follows:

[0084] Signal features are extracted through wavelet transform features. Wavelet transform, with its multi-resolution characteristics, demonstrates a powerful ability to characterize local signal features in both the time and frequency domains. Its core idea is to decompose the signal into a superposition of a series of wavelet functions. Wavelet functions are a type of waveform with a finite support set that oscillates between positive and negative. Their characteristics enable wavelet transform to provide localized information in both the time and frequency domains, making it more suitable for processing non-stationary signals. The specific application steps are as follows:

[0085] 1. Select appropriate wavelet function and scale parameters: Different wavelet functions and scale parameters will have different effects on feature extraction results. Therefore, in practical applications, it is necessary to select appropriate wavelet functions and scale parameters based on specific tasks and data characteristics.

[0086] 2. Perform wavelet transform on the original signal: By calculating the inner product of the signal and the wavelet function, the wavelet coefficients at different scales are obtained.

[0087] 3. Extract wavelet coefficients as features: Depending on the task requirements, some or all wavelet coefficients can be selected as features. These features can be used for subsequent machine learning model training.

[0088] 4. Further processing of features: In order to improve the performance of the model, the extracted wavelet coefficients can be further processed, such as dimensionality reduction and normalization.

[0089] The wavelet transform obtains different wavelet coefficients by scaling and translating the mother wavelet. Generally, translation is used to obtain the temporal information of the data, while scaling is used to obtain the frequency information of the data. These coefficients represent the relationship between local data. Large wavelet coefficients indicate a high correlation and similarity between the wavelet and the data at that location, and these wavelet coefficients are retained for reconstruction. Small wavelet coefficients are considered to indicate a stationary signal, while large wavelet coefficients are considered to indicate fluctuations in the non-stationary region.

[0090] After wavelet transformation, the data f(t) can be processed and analyzed in the time-frequency domain to distinguish the characteristics of effective signals and noise, thereby effectively removing noise. In the time-frequency domain, the principle of wavelet transform threshold denoising is to set a threshold based on the wavelet coefficients obtained by wavelet transforming the data to remove the wavelet transform coefficients corresponding to the noise signal, while retaining the wavelet transform coefficients corresponding to the effective reflected wave signal, thereby performing an inverse wavelet transform and reconstructing the data.

[0091] Signal features are extracted through Fourier transform features. Fourier transform is a mathematical tool that converts a signal from the time domain (time) to the frequency domain (frequency). It achieves this conversion by decomposing the signal into a series of sine waves and cosine waves. The specific application steps are as follows:

[0092] 1. Signal preprocessing: Preprocess the amplified and filtered signal, such as removing DC components and normalizing, to ensure signal stability and consistency.

[0093] 2. Perform Fourier transform: Perform Fourier transform on the preprocessed signal to convert it from time domain to frequency domain.

[0094] 3. Extract frequency domain features: In the frequency domain, various frequency domain features can be extracted, such as spectrum, power spectrum, spectrum density, etc. These features can reflect information such as the frequency component and power distribution of the signal.

[0095] 4. Subsequent processing: Depending on the specific application scenario, the extracted frequency domain features can be further processed, such as feature selection and dimensionality reduction, to improve the performance of subsequent machine learning models.

[0096] There are two forms of Fourier transform: the continuous Fourier transform (CFT) and the discrete Fourier transform (DFT). The CFT is suitable for processing continuous signals, while the DFT is suitable for processing discrete signals. In the field of deep learning, the DFT is more commonly used. It can be implemented through computer algorithms and digitized signals.

[0097] In the field of time series prediction, the Fourier transform can be used to obtain rich frequency domain features in sequence data. Since the original time series data is sampled to obtain discrete values in the time domain, the Fourier transform also obtains discrete values. DFT converts the signal from its original domain (usually time or space) to the frequency domain. Assume that after sampling the time series data, N values are obtained: x(0), x(1), x(2), ..., x(N-1). These N numbers are used as the input of the discrete Fourier transform. The formula is as follows:

[0098]

[0099] Among them, x(n) represents the time domain signal, and f(k) represents the k+1th frequency domain signal. From the formula, we can see that for each output point, the discrete Fourier transform needs to be calculated N times. For a total of N input observations, the time complexity of the discrete Fourier algorithm is O(n 2 ).

[0100] In practical applications, wavelet and Fourier transforms can be used in combination to leverage their respective strengths. The Fourier transform is first used to extract the signal's global frequency domain features, and then the wavelet transform is used to further extract the signal's local time-frequency features. This not only improves the accuracy and comprehensiveness of feature extraction but also provides better support for subsequent machine learning model training and prediction. By extracting key characteristic parameters such as the signal's frequency, amplitude, and phase, a feature vector is constructed, providing the foundational data for subsequent intelligent recognition.

[0101] S104 In the intelligent recognition stage, the present invention uses support vector machine (SVM) and deep learning machine learning algorithms to build a training model to learn and identify the extracted feature parameters. Through a large amount of training data and optimization algorithms, accurate classification and recognition of weak electromagnetic signals in the ocean are achieved, improving the accuracy and efficiency of signal recognition. The specific implementation steps are as follows ( Figure 6 ):

[0102] S501 extracts characteristic parameters from the signal and their corresponding labels or category information, prepares sufficient and accurately labeled training data sets, and preprocesses the data, including data cleaning and data normalization or standardization, to ensure stability and efficiency during model training.

[0103] S502 selects a suitable kernel function, as well as hyperparameters such as regularization parameter C and kernel parameter, to construct a support vector machine (SVM). The constructed SVM model will be used to classify or identify the extracted feature parameters.

[0104] S503: When using a recurrent neural network (RNN) for sequence data recognition tasks, key factors such as the model's depth, width, activation function, and loss function must be considered. The constructed deep learning model will be used to perform deep learning and recognition on the extracted feature parameters.

[0105] After the model is built, it is necessary to train the model using the training dataset. During the training process, the model parameters need to be continuously adjusted to minimize the loss function and improve the model's accuracy and generalization ability. For SVM models, optimization methods such as gradient descent and the SMO algorithm can be used. For deep learning models, training techniques such as backpropagation algorithms and optimizers (such as Adam and SGD) can be used. During the training process, model validation is also required to monitor model performance and prevent overfitting.

[0106] S505 After the model training is completed, the model needs to be evaluated to verify its performance and accuracy in actual applications. Evaluation indicators may include accuracy, recall rate, F1 score, etc. If the model performance does not meet the requirements, you can return to the previous steps to perform feature selection, model optimization and other adjustments. If the model performance meets the requirements, the model can be deployed to the actual application scenario to perform intelligent recognition of new signal data. Factors such as the real-time, stability, and scalability of the model need to be considered during deployment to ensure its effectiveness and reliability in actual applications. Through a large amount of training data and optimization algorithms, accurate classification and recognition of weak electromagnetic signals in the ocean can be achieved, and the accuracy and efficiency of signal recognition can be improved. The specific implementation steps are as follows ( Figure 7 )

[0107] S105 In the data acquisition and transmission stage, the present invention adopts a low-power data acquisition and transmission system. The system includes a low-power microcontroller, a data compression module, an encryption module and a low-power wireless communication module. The low-power microcontroller is responsible for controlling the operation of the entire system, including signal acquisition, processing and transmission. The data compression module uses advanced compression algorithms to compress the collected data to reduce the amount of data and improve transmission efficiency. The encryption module uses encryption technology to ensure the security of data transmission and prevent data from being stolen or tampered with during transmission. The low-power wireless communication module is responsible for transmitting the compressed and encrypted data to the remote receiving end, realizing real-time data transmission and remote monitoring. By optimizing the data acquisition frequency and transmission protocol, power consumption is further reduced to ensure long-term stable operation of the system. The specific implementation steps are as follows:

[0108] During the data acquisition phase, the S601's microcontroller precisely controls the sensor to collect signals at a preset frequency, ensuring data validity and integrity. It also performs preliminary processing of this data, preparing it for subsequent compression and transmission. This step not only demonstrates the microcontroller's efficient control capabilities but also ensures accurate and timely data acquisition.

[0109] Next, at step S602, the data compression module comes into play. This module uses a compression algorithm to deeply compress the data processed by the microcontroller. This step is crucial because it significantly reduces the amount of data required for transmission, thereby alleviating the burden on the wireless communication module and improving overall transmission efficiency. Furthermore, compressed data saves valuable resources during storage and transmission, which is particularly important in low-power systems.

[0110] Before data transmission, the S603's encryption module encrypts the data to ensure data security. This step utilizes advanced encryption technology, using complex encryption algorithms and key management strategies to effectively prevent data theft or tampering during transmission. The application of encryption technology not only enhances system security but also strengthens user trust in data transmission.

[0111] The S604 low-power wireless communication module receives the encrypted data and transmits it to the remote receiver. This module utilizes a low-power design to minimize energy consumption while ensuring transmission quality. By optimizing the transmission protocol and selecting the appropriate communication frequency band, the wireless communication module achieves real-time and reliable data transmission, providing strong support for remote monitoring.

[0112] To further optimize the S605's power consumption, we've incorporated a power management strategy into the system design. By precisely controlling the data acquisition frequency and transmission protocol, the system minimizes power consumption while ensuring data quality and transmission efficiency. This strategy not only extends the system's operating time but also improves its reliability and stability in practical applications.

[0113] The method proposed in the present invention has significant advantages. First, through innovative signal preprocessing technology, the present invention can effectively remove noise and enhance signal quality, laying the foundation for subsequent high-precision identification. Secondly, the use of low-power amplification and filtering technology enables the present invention to achieve a significant reduction in energy consumption while maintaining high-precision identification, and is suitable for application scenarios of long-term, continuous monitoring. In addition, the optimization of feature extraction and intelligent recognition technology improves the accuracy and efficiency of recognition and can cope with complex and changeable marine environments. Finally, the implementation of low-power data acquisition and transmission technology further reduces the energy consumption of the entire system and improves the practicality and reliability of the system. In summary, the method proposed in the present invention has significant advantages and broad application prospects in the identification and acquisition of weak electromagnetic signals in the ocean.

[0114] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for identifying and collecting low-power weak electromagnetic signals in the ocean, characterized in that: The steps include: Step 1: Adaptive filtering is used to remove noise and interference from weak ocean electromagnetic signals, and gain control and phase adjustment techniques are used to enhance signal strength and stability. Step 2: Amplify the pre-processed signal using a low-power amplifier and perform secondary filtering on the amplified signal using a digital filter to eliminate residual noise; Step 3: Use wavelet transform and Fourier transform to extract the frequency, amplitude and phase characteristics of the signal to construct feature vectors, and classify and identify the feature vectors based on machine learning algorithms; Step 4: Control the signal acquisition frequency through a low-power microcontroller and compress and encrypt the identified data; The method for controlling the signal acquisition frequency includes: constructing an electromagnetic field gradient threshold model based on Kalman prediction, when the real-time signal S(t) meets When η is the dynamic threshold coefficient, σ noise is the standard deviation of background bioelectric noise, To monitor the signal gradient, a high-precision sampling mode is started; when the signal is stable, it switches to a low-power monitoring mode and avoids sampling conflicts with neighboring nodes through a time slot allocation algorithm; Step 5: The signal processed in step 4 is transmitted to the remote terminal using a low-power hierarchical transmission method based on an energy cost function, wherein the data acquisition frequency control and the hierarchical transmission method are used to reduce the overall power consumption of the system.

2. The method for identifying and collecting low-power marine weak electromagnetic signals according to claim 1, characterized in that: The specific steps of step 1 are as follows: In step 1, the adaptive filtering adopts a variable step size LMS algorithm to dynamically adjust the filter parameters, specifically including: The error between the filtered output signal and the expected signal is calculated, the filter coefficients are iteratively adjusted until the error reaches a preset threshold, and the quality of the filtered signal is evaluated based on the signal-to-noise ratio.

3. The method for identifying and collecting low-power marine weak electromagnetic signals according to claim 1, characterized in that: The gain control in step 1 adopts an automatic gain control system, which keeps the signal amplitude stable by dynamically adjusting the amplifier gain value, calculates the gain coefficient in real time according to the input signal amplitude, and controls the output signal amplitude through negative feedback.

4. The method for identifying and collecting low-power marine weak electromagnetic signals according to claim 1, characterized in that: The digital filter in step 2 is a finite impulse response (FIR) filter or an infinite impulse response (IIR) filter, whose order and cutoff frequency are designed according to the frequency domain characteristics of the ocean electromagnetic signal, and the computational complexity is reduced by optimizing the filter algorithm.

5. The method for identifying and collecting low-power marine weak electromagnetic signals according to claim 1, characterized in that: The machine learning algorithm in step three includes a support vector machine and a deep learning model, wherein the support vector machine uses a Gaussian kernel function to classify feature vectors, and the deep learning model is a recurrent neural network for processing time-related electromagnetic signal features.

6. The method for identifying and collecting low-power marine weak electromagnetic signals according to claim 1, characterized in that: The method for compressing the identified data in step 4 includes: (1) Wavelet transform: Multi-scale wavelet transform is performed on the collected weak ocean electromagnetic signals to decompose the signals into wavelet coefficients at different frequencies and time scales. The wavelet transform formula is: Among them, s(n) is the original signal, ψ j,k (n) is the wavelet basis function, j and k represent the scale and translation parameters respectively, W j,k is the wavelet coefficient; (2) Adaptive quantization: Adaptive quantization strategy is used to quantize the wavelet coefficients according to their amplitudes. The quantization step length Δ is calculated as follows: Among them, σ is the standard deviation of the signal, β is the quantization coefficient, which is used to control the quantization accuracy; the quantization step size Δ is dynamically adjusted with the amplitude of the wavelet coefficient. For coefficients with larger amplitudes, a smaller quantization step size is used to retain more details, while for coefficients with smaller amplitudes, a larger quantization step size is used to reduce the amount of data; (3) Data encoding: Encode the quantized data using Huffman coding or other efficient lossless coding algorithms to further compress the data volume; the encoded data is transmitted to the remote receiving end through a low-power wireless communication module.

7. The method for identifying and collecting low-power marine weak electromagnetic signals according to any one of claims 1 to 6, characterized in that: The layered transmission protocol in step 5 includes: defining a transmission energy cost function Etx: Among them C CSI is the adaptive modulation channel state information, D is the compressed data packet size, P TX is the transmission power; when the node is located in a strong current disturbance area, short burst forward error correction (FEC) coding transmission is preferred; in the silent period, it switches to the underwater acoustic beacon synchronization mode with a low duty cycle. α and β are the data transmission weight coefficient and the power consumption weight coefficient, respectively, to balance the data transmission efficiency and priority: when the sea condition is stable, the data transmission weight coefficient α increases, the power consumption weight coefficient β decreases, and the focus is on high-speed transmission; when energy is limited, the data transmission weight coefficient α decreases, the power consumption weight coefficient β increases, and the focus is on energy saving. T active is the node activation time.

8. A marine low-power weak electromagnetic signal recognition and acquisition system, characterized in that: include: The pre-processing module removes noise and interference from weak ocean electromagnetic signals through adaptive filtering, and uses gain control and phase adjustment technology to enhance signal strength and stability; The low-power amplification and filtering module uses a low-power amplifier to amplify the pre-processed signal and performs secondary filtering on the amplified signal through a digital filter to eliminate residual noise; The intelligent recognition module uses wavelet transform and Fourier transform to extract the frequency, amplitude and phase characteristics of the signal to construct feature vectors, and classifies and identifies the feature vectors based on machine learning algorithms; The data acquisition and transmission module controls the signal acquisition frequency through a low-power microcontroller and compresses and encrypts the identified data.

9. A computer storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the frequency conversion synchronization method according to any one of claims 1 to 7 is implemented.

Images