An underwater alternating magnetic field detection and magnetic induction communication integrated method

Abstract

The application discloses an underwater alternating magnetic field detection and magnetic induction communication integrated method; the method is as follows: one, an omnidirectional coil obtains time-varying magnetic field signals in the environment and converts the time-varying magnetic field signals into time-varying voltage signals. Two, based on the time-varying voltage signals, corresponding magnetic field signal I / Q data is formed; feature extraction is performed on the I / Q data to obtain a feature vector; the feature vector is input into an adaptive threshold pulse neural network model to obtain channel quality and environmental magnetic field conditions. Three, when communication equipment is used for communication, a signal modulation mode is adjusted according to the channel quality. The application trains an adaptive threshold pulse neural network model which is low in power consumption, small in calculation amount and is specially used for an underwater alternating magnetic field detection and magnetic induction communication integrated device. The model not only can realize magnetic detection requirements and monitor magnetic communication conditions in a certain range of water, but also can detect magnetic communication channel quality.

Description

Technical Field

[0001] This invention relates to the field of underwater communication technology, specifically to an integrated method for underwater alternating magnetic field detection and magnetic induction communication. Background Technology

[0002] Underwater wireless communication technology is a significant bottleneck technology restricting current underwater monitoring / observation systems. Underwater wireless communication systems have wide applications in tactical surveillance and counter-blockade, and are a current research hotspot in underwater marine technology and marine engineering, with very broad application prospects in civilian, scientific research, and military fields. Due to the complex, time-varying signal environment underwater, the effective information transmission rate of communication systems often becomes a bottleneck, contradicting the ever-increasing demand for underwater communication. Current underwater wireless communication technologies often cannot meet the mission requirements of underwater data transmission; conventional underwater acoustic communication and laser communication have their limitations and cannot simultaneously achieve various performance targets. Therefore, finding new underwater wireless data transmission technologies has become one of the core objectives in the field of underwater communication technology, and is of great significance for building sensor systems in distributed detection environments.

[0003] Compared to underwater acoustic and laser communication, magnetic induction communication offers advantages such as strong anti-interference capabilities and simplified manufacturing. Currently, most research in underwater magnetic induction communication technology employs dedicated signal modulation and demodulation chips for signal processing, fixed amplifiers to amplify the transmitted signal, and fixed filters to filter the received signal. These approaches result in a fixed modulation scheme, leading to high bit error rates in harsh environments and excessive power consumption and slow speeds in favorable conditions. Furthermore, while high-frequency signals offer higher bit rates, they also exacerbate eddy current losses, increasing magnetic signal attenuation and shortening communication distance. Conversely, low-frequency signals have lower bit rates and lower communication speeds. Balancing communication distance and speed in underwater magnetic induction communication presents another significant challenge. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention proposes an integrated underwater alternating magnetic field detection and magnetic induction communication device. To meet the need for edge computing capabilities in underwater detection equipment, this invention trains a low-power, computationally inefficient adaptive threshold pulse neural network model specifically for this integrated underwater alternating magnetic field detection and magnetic induction communication device. This model not only fulfills the magnetic detection requirement—monitoring the magnetic communication situation within a certain water area—but also detects the quality of the magnetic communication channel. The Software-Defined Radio Peripheral (USRP) dynamically adjusts the magnetic communication modulation method based on the detected magnetic communication channel quality to ensure high-quality communication. To address the issues of significant attenuation of high-frequency underwater magnetic signals and limited magnetic communication speed, this invention employs a Multiple-Input Multiple-Output (MIMO) coil system to support high-speed transmission requirements.

[0005] This invention provides an integrated method for underwater alternating magnetic field detection and magnetic induction communication. Each communication device is equipped with multiple omnidirectional coils; each omnidirectional coil comprises three coil units arranged orthogonally. Different communication devices communicate through the omnidirectional coils.

[0006] The integrated method for underwater alternating magnetic field detection and magnetic induction communication includes the following steps:

[0007] Step 1: The omnidirectional coil acquires the time-varying magnetic field signal in the environment and converts it into a time-varying voltage signal.

[0008] Step 2: Based on the time-varying voltage signal, generate corresponding magnetic field signal I / Q data; extract features from the I / Q data to obtain feature vectors; input the feature vectors into an adaptive threshold spiking neural network model. The adaptive threshold spiking neural network model outputs the channel quality (channel) and the environmental magnetic communication status (MC).

[0009] The adaptive threshold spiking neural network model comprises multiple layers of neurons connected sequentially. The threshold voltage of a neuron is adjusted based on the pulse output of the preceding neuron, as detailed in the following model:

[0010]

[0011] U(t)=(βU(t-1)+I(t))×(1-S i (t-1))

[0012]

[0013] Where I(t) is the current current of the neuron; U(t) is the current membrane voltage of the neuron; α and β are the decay coefficients of the neuronal current and membrane voltage, respectively; i is the sequence number of the current neuron; j is the sequence number of the preceding neuron of the current neuron; t is time; w ij S represents the weighting coefficient between the synapse of the current neuron i and the j-th preceding neuron; j (t-1) represents the pulse output of the j-th pre-stage neuron; S i (t-1) represents the pulse output of the current neuron at the previous time step; V th V is the threshold voltage of the current neuron; tho is the base threshold; k is the adjustment coefficient for the threshold voltage.

[0014] Step 3: When the communication equipment is communicating, adjust the signal modulation method according to the channel quality (channel). When the communication equipment is not communicating, determine whether there are other devices communicating magnetically in the environment based on the ambient magnetic communication condition (MC).

[0015] Preferably, the current magnetic communication channel quality is divided into four levels: level 2, level 1, level 0, and level -1, where the channel quality gradually deteriorates. In step three, if channel = 2, then quadrature phase shift keying is used for modulation; if channel = 1, then binary phase shift keying is used for modulation; if channel = 0 or -1, then communication is suspended.

[0016] As a preferred embodiment, in the training set of the adaptive threshold spiking neural network model, the magnetic communication channel quality channel for data with a signal-to-noise ratio (SNR) greater than 10 dB is labeled as level 2; the magnetic communication channel quality channel for data with an SNR greater than or equal to 0 and less than or equal to 10 dB is labeled as level 1; the magnetic communication channel quality channel for data with an SNR less than 0 is labeled as level 0; and the magnetic communication channel quality channel for pure noise data is labeled as level -1.

[0017] Preferably, the environmental magnetic communication status MC is divided into two levels, namely level 1 and level 0; level 1 corresponds to the presence of magnetic communication signals in the environment; level 0 corresponds to the absence of magnetic communication signals in the environment; in step three, when the communication device is not communicating, if MC = 1, the communication device sends a warning signal to other communication devices.

[0018] As a preferred option, the feature extraction process in step two is as follows:

[0019] a. Extract instantaneous features Ins based on I / Q data; instantaneous features Ins include instantaneous amplitude, instantaneous phase, and instantaneous frequency.

[0020] b. Extract the high-order cumulant vector High based on I / Q data.

[0021] c. Construct the feature vector INPUT = [Data,Ins,High,Channel,MC] for the adaptive threshold spiking neural network model used as input; Data is the segmented I / Q data.

[0022] Preferably, the coil unit is formed by winding enameled wire with a cross-sectional area of ​​0.3 square millimeters in a coil slot with a radius of 10 cm, and the number of coil turns is 30.

[0023] Preferably, a communication device contains three omnidirectional coils; the three omnidirectional coils are arranged in an equilateral triangle in the pressure chamber.

[0024] Preferably, the communication device includes a pressure-resistant chamber, and a general-purpose software-defined radio peripheral, an external amplifier circuit, a resonant matching circuit, and multiple omnidirectional coils installed within the pressure-resistant chamber. The general-purpose software-defined radio peripheral includes a processor, an automatic gain control module, a bandpass filter, and a built-in amplifier. The resonant matching circuit includes a series resonant transmitting circuit and a parallel resonant receiving circuit. During operation, a high-frequency relay is used to switch the connection between the omnidirectional coils and the series resonant transmitting circuit and the parallel resonant receiving circuit.

[0025] Preferably, the underwater pressure tank is made of titanium alloy.

[0026] Preferably, the layer mapping and precoding processes in step three of the signal transmission process are as follows:

[0027] a. Layer mapping processing divides the complex signal output by the constellation modulator into three layers of data based on the number of omnidirectional coils.

[0028] b. Precoding Processing: The Zero Forcing (ZF) precoding algorithm is used to process the three layers of data. The specific implementation is as follows: The communication system is defined with one transmitting communication device equipped with three omnidirectional coils, and one receiving communication device equipped with three omnidirectional coils. The three layers of data to be transmitted by the three omnidirectional coils of the transmitting communication device are defined as: X = [x1, x2, x3] T Where, x i The data that the transmitting omnidirectional coil i needs to send to the receiving omnidirectional coil j is defined. The channel matrix is ​​defined. Among them, H ij Let i be the channel matrix between the omnidirectional coil i of the transmitting communication device and the omnidirectional coil j of the receiving communication device. Define the encoding matrix. Among them, W i The encoding matrix for transmitting omnidirectional coil i and receiving omnidirectional coil i is given. The encoding matrix W = cH is calculated using the ZF algorithm. H (HH H ) -1 Among them, H H This is the conjugate transpose of the channel matrix; Here, P is the power normalization factor, and P is the average transmit power at the transmitter. The coding matrix obtained by the ZF algorithm has the following characteristics [H ij W k ]=0, i≠j≠k, the actual transmitted signal of the omnidirectional coil i of the transmitting communication device is send i =W i x i .

[0029] Preferably, the decoding process of the encoding matrix during signal reception in step three is as follows:

[0030] The signal received by the omnidirectional coil j of the receiving communication device is: Where n represents the noise signal. The data received by the omnidirectional coil j of the receiving communication device contains only H... jj W j x j +n represents the data originating from the omnidirectional coil j of the transmitting communication device, and is the data that the omnidirectional coil j of the receiving communication device expects to receive. Interference from omnidirectional coils of other transmitting communication devices besides the transmitting device's omnidirectional coil j is considered. The received data is decoded using the extracted and configured encoding matrix and power normalization factor to obtain... This means eliminating interference between omnidirectional coils of transmitting and receiving communication devices with different serial numbers.

[0031] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0032] 1. To address the need for edge computing capabilities in underwater detection equipment, this invention trains a low-power, computationally inefficient adaptive threshold pulse neural network model specifically designed for an integrated underwater alternating magnetic field detection and magnetic induction communication device. This model not only fulfills the magnetic detection requirement—monitoring the magnetic communication status of a certain water area—but also detects the quality of the magnetic communication channel.

[0033] 2. This invention implements communication and detection functions based on the Software Radio Peripheral (USRP) architecture. The USRP is used to perform signal modulation, filtering and other operations on the transmitted signal, which significantly reduces the circuitry involved in signal generation and reception.

[0034] 3. This invention uses a magnetic induction multiple-input multiple-output 3×3 MIMO coil system and utilizes spatial multiplexing technology to achieve three times the data rate while maintaining the same bandwidth, thus supporting high-speed transmission requirements.

[0035] 4. This invention designs a device structure that enables underwater 3×3 MIMO communication for underwater application scenarios.

[0036] 5. The system specifications of this invention are as follows: underwater communication rate: 100kbps; maximum communication distance: 10m; maximum instantaneous bandwidth during detection: 125kHz; MIMO scale: 3×3. Attached Figure Description

[0037] Figure 1 This invention provides a communication hardware framework diagram for an integrated underwater alternating magnetic field detection and magnetic induction communication device.

[0038] Figure 2 This is a schematic diagram of the structure of an integrated underwater alternating magnetic field detection and magnetic induction communication device provided by the present invention.

[0039] Figure 3 This is a schematic diagram of the omnidirectional coil in this invention.

[0040] Figure 4 This is a schematic diagram of the circuit principle of the series resonant transmitting circuit in this invention.

[0041] Figure 5 This is a schematic diagram of the parallel resonant receiving circuit in this invention.

[0042] Figure 6 This is a schematic diagram of the training process of the adaptive threshold spiking neural network model in this invention.

[0043] Figure 7 This is a schematic diagram of the communication process of the present invention. Detailed Implementation

[0044] The present invention will be further described below with reference to the accompanying drawings.

[0045] like Figure 1 , Figure 2 As shown, an integrated underwater alternating magnetic field detection and magnetic induction communication device includes a pressure-resistant chamber, and a general-purpose software-defined radio peripheral (USRP), an external amplifier circuit, a resonant matching circuit, and multiple omnidirectional coils installed in the pressure-resistant chamber. The omnidirectional coils in two integrated underwater alternating magnetic field detection and magnetic induction communication devices together form a 3×3 MIMO coil system to achieve underwater magnetic induction communication between the two terminals.

[0046] The underwater pressure tank is made of titanium alloy and meets the requirements for operation at a depth of 100 meters. It provides three watertight coil mounting interfaces and one serial watertight interface. Inside the pressure tank, the watertight coil mounting interfaces connect to a resonant matching circuit, and the serial watertight interface connects to the USRP's USB port. Outside the pressure tank, the three watertight coil mounting interfaces connect to three omnidirectional magnetic induction coils. Two integrated underwater alternating magnetic field detection and magnetic induction communication devices communicate as master and slave units. The master unit's serial watertight interface connects to the computer's USB port, while the slave unit's serial watertight interface is sealed and not used. The three omnidirectional magnetic induction coils are evenly distributed on a 1-meter diameter circular base.

[0047] In this embodiment, there are three omnidirectional coils; the three omnidirectional coils are arranged in an equilateral triangle in the withstand voltage chamber; specifically, as follows: Figure 3 As shown, each omnidirectional coil comprises three coil units; the three coil units are arranged orthogonally (i.e., perpendicular to each other in pairs). This invention uses omnidirectional coils to transmit and detect alternating magnetic fields; the coil units are wound with enameled wire with a cross-sectional area of ​​0.3 square millimeters in a coil slot with a radius of 10 cm, and the number of coil turns is set to 30 turns.

[0048] The output interfaces of the three omnidirectional coils, the resonant matching circuit, and the external amplifier circuit are connected in sequence. The Universal Software Radio Peripheral (USRP) is implemented on the GNU Radio platform and includes a processor, an automatic gain control module, a bandpass filter, and a built-in amplifier. The USRP's RX / TX interface, external amplifier circuit, resonant matching circuit, and omnidirectional coils are connected in sequence via SMA cables.

[0049] Since the transmit power of USRP is mostly in the hundreds of milliwatts, in order to improve the driving capability of the quadrature coil, this invention uses an external amplifier circuit. The external amplifier circuit uses the LM1875 chip, which can achieve a maximum power output of 30W.

[0050] The resonant matching circuit includes a series resonant transmitting circuit and a parallel resonant receiving circuit. A high-frequency relay is used to switch the connection between the coil and the series resonant transmitting circuit and the parallel resonant receiving circuit.

[0051] Series resonant transmitting circuit, such as Figure 4 As shown; Figure 4 In the diagram, R1 is the current regulating resistor, C1 is the resonant matching capacitor, L1 is the coil inductance, and Uout is the transmitted signal. The total impedance Z of the series resonant transmitting circuit is... s :

[0052]

[0053] In the resonant state, the impedance in the series resonant circuit reaches its minimum value. Using a series resonant circuit at the transmitting end allows the circuit current to reach its maximum value in the resonant state. Calculation of series resonant transmitting circuit parameters: Given the system's transmitting frequency f and the coil inductance L1, according to the formula... Where ω=2πf, the capacitor C1 can be obtained, and the current regulating resistor R1 is selected according to the power requirements.

[0054] Parallel resonant transmitting circuit, such as Figure 5 As shown: R2 is the load resistance, R L U is the DC resistance of the coil, C2 is the resonant matching capacitor, L2 is the coil inductance, and U... in This is the signal received by the coil. When the load R2 is extremely large, the parallel impedance of the equivalent circuit of R2 and C2 is approximately equal to the impedance of C2. The voltage across the load R2 is:

[0055]

[0056] Among them, I COIL The current within the loop, at resonance. Under resonant conditions, as shown by the formula, the equivalent parallel receiving circuit can be designed with appropriate parameters to make the load voltage R2 greater than U. inCalculation of resonant receiver circuit parameters: Given the system transmission frequency f and coil inductance L2, according to the formula... Where ω=2πf, the capacitance C2 can be calculated, and the load R2 should be much larger. R2 is set to 10000 times.

[0057] This underwater alternating magnetic field detection and magnetic induction communication integrated device uses a 3×3 MIMO coil system and leverages spatial multiplexing technology to achieve three times the data rate while maintaining the same bandwidth. An adaptive threshold pulse neural network model was designed to monitor the magnetic communication situation within a certain range of water and to detect the magnetic communication channel quality to ensure high-quality communication. The specific working method is as follows:

[0058] Step 1: The omnidirectional coil acquires the time-varying magnetic field signal in the environment and converts it into a time-varying voltage signal. The time-varying electrical signal is then input to the USRP through a resonant matching circuit.

[0059] Step 2: USRP samples and acquires time-varying voltage signals to form I / Q data of magnetic field signals. The I / Q data is input to the feature extraction module. The feature vector output by the feature extraction module is used as the input vector and processed by the adaptive threshold spiking neural network model to obtain the channel quality (channel) and the environmental magnetic communication status (MC).

[0060] The adaptive threshold spiking neural network model consists of multiple layers of neurons connected sequentially. The mathematical model of the neurons is optimized based on the specific characteristics of the underwater environment, and the specific process is as follows:

[0061] Due to the skin effect in water propagation, the intensity of high-frequency magnetic fields decreases rapidly with increasing transmission distance. When the underwater abnormal magnetic field source is far from this invention, the voltage amplitude detected by the detection coil will also be low. The trigger voltage of the neuron trained using a common neuron structure is a fixed value, which is insufficiently sensitive to weak voltages, failing to meet the expectations of this invention. Therefore, a neuron model with dynamically adjustable thresholds is proposed, defined as Adaptive Loihi Current Based Leaky Integrate and Fire (A-CUBA). The A-CUBA model is a modification of the Loihi Current Based Leaky Integrate and Fire (CUBA) model to meet the requirements of this system. The mathematical model of the A-CUBA model is as follows:

[0062]

[0063] U(t)=(βU(t-1)+I(t))×(1-S i(t-1)) (2)

[0064]

[0065] Where I(t) is the current current of the neuron; U(t) is the current membrane voltage of the neuron; α∈(0,1) and β∈(0,1) are the attenuation coefficients of the neuronal current and membrane voltage, respectively; i is the sequence number of the current neuron; j is the sequence number of the preceding neuron (specifically, the neuron in the layer above the current neuron); t is time; w ij S represents the weighting coefficient between the synapse of the current neuron i and the j-th preceding neuron; j (t-1)∈{0,1} represents the pulse output of the j-th pre-stage neuron; S i (t-1)∈{0,1} represents the pulse output of the current neuron at the previous time step; V th V is the threshold voltage of the current neuron; tho is the base threshold; k is the adjustment coefficient for the threshold voltage.

[0066] Compared to the CUBA model, the A-CUBA model mainly improves upon formula (3). From formula (3), the neuron threshold voltage formula shows that when the external magnetic signal is weak, the pulses of the pre-stage neurons are infrequent. The calculated result is relatively small, so the neuron threshold voltage will be appropriately reduced; when the external magnetic signal is strong, the pulses of the pre-stage neurons are frequent. If the calculated result is large, the overall threshold voltage of the neuron will be appropriately increased. Compared with other commonly used neuron models, the A-CUBA model is more suitable for this invention. The constructed spiking neural network can adapt to the complex magnetic field environment of the seabed. According to the above formula, the A-CUBA model has attenuation coefficients α∈(0,1) and β∈(0,1). When the pulse input of the A-CUBA neuron's preceding stage is 0, the neuron's membrane potential will gradually decrease until it reaches 0, and the attenuation rate is jointly controlled by the two voltage and current attenuation coefficients α∈(0,1) and β∈(0,1). It can adapt to more complex situations and is also consistent with biological characteristics. Compared with CUBA, the commonly used IF neuron does not have the function of membrane potential leakage and is more strongly affected by noise on the seabed; the membrane potential leakage of the LIF model is controlled by only one parameter β∈(0,1) to control voltage attenuation. This structure may not perform well in complex environments. After actual testing, the use of the ALIF neuron structure greatly improved the accuracy of long-distance magnetic signal detection.

[0067] The training process of the adaptive threshold spiking neural network model is as follows: Figure 6 The specific implementation method is as follows:

[0068] 1. Data Acquisition: Two integrated magnetic field detection and magnetic induction communication terminal devices are defined as the master and slave devices, respectively, with an underwater distance L∈[2,4,6] meters between them. The master device generates nine modulation signals with modulation methods of ASK, 2ASK, FSK, 4FSK, PSK, 4PSK, 4QAM, 16QAM, and 64QAM using modulation modules such as AM Modulation, FSKModulator, and Constellation Decoder on the GUNRadio platform. Each modulation signal has seven signal frequencies: F T = K×10 (kHz), K∈[1,5,10,25,20,25,30]. The host adds a noise source using the Noise Source module on the GUN Radio platform and calculates the signal-to-noise ratio (SNR) by calculating the average power of the signal and noise using the Complex to Mag^2 and Moving Average modules. The host sends signals with different SNRs, frequencies, and modulation schemes to the slave. The slave acquires the I / Q data of the time-varying electrical signal on the coil through the UHD:USRP Source module and records the modulation scheme, SNR, and frequency. Additionally, the slave also records the pure noise I / Q data collected by the underwater magnetic coil.

[0069] 2. Data tagging:

[0070] a. Marking the quality of magnetic communication channels: Based on actual testing, it was found that the bit error rate of QPSK modulation at 100kHz frequency was less than 1‰ in an environment with a signal-to-noise ratio (SNR) greater than 10dB, and the bit error rate exceeded 3% when the SNR was less than 0dB. Therefore, an SNR greater than 10dB was marked as excellent channel conditions (channel recorded as 2), an SNR between 0dB and 10dB was marked as average channel conditions (channel recorded as 1), and an SNR less than 0dB was marked as poor channel conditions (channel recorded as 0). Pure noise data was recorded as -1.

[0071] b. Mark the environmental magnetic communication status MC: Set the I / Q data MC of noise to 0, and set the I / Q data MC of the modulated signal to 1.

[0072] 3. Segment the I / Q data to obtain Data: The I / Q data can be divided into 567 categories based on 3 master-slave interval distances, 9 modulation schemes, 7 frequencies, and 3 channel qualities. For each category of I / Q data, take N groups of length and store them in the training set. Similarly, for pure noise I / Q data, take N groups of length and store them in the training set. During each training iteration, take one set of data of length N from the training set as Data.

[0073] 4. Feature Extraction

[0074] a. Extracting instantaneous features from I / Q data (Ins): Define I / Q data as... s(k) is the real part of the data; Let be the imaginary part of the data; j be the imaginary unit; k ∈ 1, 2, ..., N. According to the formula, the instantaneous feature Ins = [A, φ, F] of length N is obtained, where A, φ, and F are vectors composed of the instantaneous amplitude, instantaneous phase, and instantaneous frequency at N different times, respectively, arranged in order. The instantaneous amplitude is: The instantaneous phase is: The instantaneous frequency is:

[0075] b. Based on the I / Q data, extract the higher-order cumulant vector High; vector High includes the higher-order cumulant C. 20 C 21 C 40 C 41 C 42 C 60 C 61 C 63 C 80 Signal modulation analysis based on higher-order cumulative quantities is a traditional analysis method. This invention improves the accuracy of magnetic detection by incorporating higher-order signal quantity features into an adaptive threshold pulse neural network model.

[0076] Finally, the output of the feature extraction module is INPUT = [Data, Ins, High, Channel, MC]

[0077] 4. Iterative Training of Adaptive Threshold Spike Neural Network Model

[0078] a. The input to the coding layer of the spiking neural network is the output after feature extraction, INPUT = [Data, Ins, High, Channel, MC]. The coding layer uses population coding to convert the input vector into spiking form.

[0079] b. The network structure of the adaptive threshold spiking neural network is set to a fully connected structure. The adaptive threshold spiking neural network is trained using the STBP learning algorithm and cross-entropy loss function for iterative training.

[0080] c. The final output of the model is the current channel communication quality (channel) and the environmental magnetic communication status (MC).

[0081] Step 3: If the underwater alternating magnetic field detection and magnetic induction communication integrated device is communicating, the USRP dynamically controls the modulation mode of the communication signal according to the channel conditions output by the adaptive threshold pulse neural network to ensure communication quality. If channel = 2, QPSK (Quadrature Phase Shift Keying) modulation is used; if channel = 1, BPSK (Binary Phase Shift Keying) modulation is used; if channel = 0, communication is suspended.

[0082] If the underwater alternating magnetic field detection and magnetic induction communication integrated device is not in communication mode, but magnetic field communication exists in the current environment (MC=1), the slave device will activate the magnetic induction communication function to report the early warning status to the master device.

[0083] In this embodiment, the communication function of USRP is developed on the GUN Radio platform. USRP communication includes two processes: signal transmission and signal reception. The specific implementation method of communication using QPSK modulation is as follows (changing QPSK to BPSK modulation only requires simple modifications to the modulation and demodulation modules, which can be understood by industry professionals and therefore will not be specifically given):

[0084] Signal transmission such as Figure 7 The process marked as 1:

[0085] 1. The main code block outputs a data stream to be sent and calls the custom module: codeword processing module to perform operations such as CRC, code block segmentation, channel coding, and code block concatenation on the data to be sent to generate a data stream composed of a series of codewords (parameter settings: maximum code block length: maxsize, CRC: 24bit, channel coding: turbo).

[0086] 2. Call the Constellation Modulator module to modulate the data to be transmitted. The output is a complex modulated signal on the baseband (parameter settings: Constellation modulation mode QPSK, Differential Encoding selected as yes, Root Raised Cosine Filter Excess BW of 0.3, etc.).

[0087] 3. Call the custom: layer mapping and precoding module to perform layer mapping and precoding on the complex signal output by the constellation modulator, so that the complex codeword stream becomes multi-layer parallel data (parameter settings: preprocessing algorithm: ZF algorithm).

[0088] 4. Use the UHD:USRP Sink module to load the signal to be transmitted onto the antenna port (parameter settings: center frequency 100kHz, bandwidth 10kHz, gain value 1, sampling rate 400kHz).

[0089] Signal reception, such as Figure 7 The process marked as 2:

[0090] 1. Use the UHD:USRP Source module to acquire time-varying electrical signals (parameters set as follows: center frequency (CenterFreq) 100kHz, bandwidth (BandWide) 10kHz, sampling rate (Samp Rate) 400kHz).

[0091] 2. The acquired signal is guided to the AGC automatic gain control module to ensure that the input signal is always within a suitable amplitude range, thereby maintaining the stability and consistency of the signal (parameters are set as follows: Reference voltage is 1, initial gain parameter Gain is 1, and maximum gain value is Max Gain of 10k).

[0092] 3. Call the custom: encoding matrix decoding module, and use the encoding matrix W at the time of transmission to complete the decoding operation.

[0093] 4. Call the Polyphase Clock Sync module to implement time synchronization using multiple multiphase filters (parameter settings: Taps root raised cosine filter tap value is 5, Filter size number of filters is 32, Maximum Rate Deviation maximum phase deviation value default value is 1.5, Output SPS is 1, one sample is output per symbol).

[0094] 5. Call the Constellation Decoder module to implement hard-decision decoding of complex values ​​in the constellation diagram, and demodulate the received complex numbers. (This module requires calling the Constellation Object. The parameters of the Constellation Object are set as follows: Constellation Type is VariableConstellation, Symbol Map is manually specified as a list {0, 1, 2, 3}, ConstellationPoints are manually specified as a list of complex numbers {1+1j, 1-1j, -1+1j, -1-1j}, Rotational Symmetry is set to 4 for every 360 degrees of symmetry, Dimensionality is 1, and Soft Decisions LUT is set to auto for the vector of floating-point tuples used as a lookup table (LUT).)

[0095] 6. Call the Differential Decoder module to decode the received differential signal to obtain the original signal, resulting in char type original signal data (parameter settings: modulation order is 4, coding is set to different differential coding algorithm).

[0096] 7. Call the custom codeword decoding module to decode and verify the codeword data stream from the Differential Decoder module and generate the original data.

[0097] The implementation methods for signal transmission calling a custom layer mapping and precoding module, and signal reception calling a custom encoding matrix decoding module are as follows:

[0098] a. Layer mapping processing divides the complex signal output by the constellation modulator into three layers of data based on the number of omnidirectional coils.

[0099] b. Precoding Processing: The Zero Forcing (ZF) precoding algorithm is used to process the three layers of data. The specific implementation is as follows: The communication system is defined with one transmitting communication device equipped with three omnidirectional coils, and one receiving communication device equipped with three omnidirectional coils. The three layers of data to be transmitted by the three omnidirectional coils of the transmitting communication device are defined as: X = [x1, x2, x3] T Where, x i The data that the transmitting omnidirectional coil i needs to send to the receiving omnidirectional coil j is defined. The channel matrix is ​​defined. Among them, H ijLet i be the channel matrix between the omnidirectional coil i of the transmitting communication device and the omnidirectional coil j of the receiving communication device. Define the encoding matrix. Among them, W i The encoding matrix for transmitting omnidirectional coil i and receiving omnidirectional coil i is given. The encoding matrix W = cH is calculated using the ZF algorithm. H (HH H ) -1 Among them, H H This is the conjugate transpose of the channel matrix; Here, P is the power normalization factor, where P is the average transmit power of the transmitter. The coding matrix obtained by the ZF algorithm has the following characteristics [H ij W k ]=0, i≠j≠k, the actual transmitted signal of the omnidirectional coil i of the transmitting communication device is send i =W i x i .

[0100] c. The signal received by the omnidirectional coil j of the receiving communication device is: Where n represents the noise signal. The data received by the omnidirectional coil j of the receiving communication device contains only H... jj W j x j +n represents the data originating from the omnidirectional coil j of the transmitting communication device, and is the data that the omnidirectional coil j of the receiving communication device expects to receive. Interference from omnidirectional coils of other transmitting communication devices besides the transmitting device's omnidirectional coil j is considered. The received data is decoded using the extracted and configured encoding matrix and power normalization factor to obtain... This means eliminating interference between omnidirectional coils of transmitting and receiving communication devices with different serial numbers.

Claims

1. A method for integrating underwater alternating magnetic field detection and magnetic induction communication, characterized in that: Each communication device employs multiple omnidirectional coils; each omnidirectional coil comprises three coil units arranged orthogonally; different communication devices communicate through the omnidirectional coils. The integrated method for underwater alternating magnetic field detection and magnetic induction communication includes the following steps: Step 1: The omnidirectional coil acquires the time-varying magnetic field signal in the environment and converts it into a time-varying voltage signal; Step 2: Based on the time-varying voltage signal, generate the corresponding magnetic field signal I / Q data; extract features from the I / Q data to obtain feature vectors; input the feature vectors into the adaptive threshold spiking neural network model; the adaptive threshold spiking neural network model outputs the channel quality (channel) and the environmental magnetic communication status (MC). The adaptive threshold spiking neural network model comprises multiple layers of neurons connected sequentially; the threshold voltage of a neuron is adjusted based on the pulse output of the preceding neuron, as detailed in the following model: U(t)=(βU(t-1)+I(t))×(1-S i (t-1)) Where I(t) is the current current of the neuron; U(t) is the current membrane voltage of the neuron; α and β are the decay coefficients of the neuronal current and membrane voltage, respectively; i is the sequence number of the current neuron; j is the sequence number of the preceding neuron of the current neuron; t is time; w ij S represents the weighting coefficient between the synapse of the current neuron i and the j-th preceding neuron; j (t-1) represents the pulse output of the j-th pre-stage neuron; S i (t-1) represents the pulse output of the current neuron at the previous time step; V th V is the threshold voltage of the current neuron; tho The base threshold is denoted by k, which is the adjustment coefficient for the threshold voltage. Step 3: When the communication device is communicating, adjust the signal modulation method according to the channel quality (channel); when the communication device is not communicating, determine whether there are other devices communicating magnetically in the environment based on the ambient magnetic communication condition (MC).

2. The integrated method for underwater alternating magnetic field detection and magnetic induction communication according to claim 1, characterized in that: The current magnetic communication channel quality is divided into four levels, namely level 2, level 1, level 0, and level -1, which are gradually deteriorating in quality. In step three, if channel = 2, then quadrature phase shift keying is used for modulation; if channel = 1, then binary phase shift keying is used for modulation; if channel = 0 or -1, then communication is suspended.

3. The integrated method for underwater alternating magnetic field detection and magnetic induction communication according to claim 1, characterized in that: In the training set of the adaptive threshold spiking neural network model, the magnetic communication channel quality of data with a signal-to-noise ratio (SNR) greater than 10 dB is labeled as level 2; the magnetic communication channel quality of data with an SNR greater than or equal to 0 and less than or equal to 10 dB is labeled as level 1; the magnetic communication channel quality of data with an SNR less than 0 is labeled as level 0; and the magnetic communication channel quality of pure noise data is labeled as level -1.

4. The integrated method for underwater alternating magnetic field detection and magnetic induction communication according to claim 1, characterized in that: The environmental magnetic communication status MC is divided into two levels: level 1 and level 0. Level 1 corresponds to the presence of magnetic communication signals in the environment, while level 0 corresponds to the absence of magnetic communication signals in the environment. In step three, if MC = 1 when the communication device is not communicating, the communication device will send a warning signal to other communication devices.

5. The integrated method for underwater alternating magnetic field detection and magnetic induction communication according to claim 1, characterized in that: The feature extraction process in step two is as follows: a. Extract instantaneous features Ins based on I / Q data; instantaneous features Ins include instantaneous amplitude, instantaneous phase, and instantaneous frequency; b. Extract the high-order cumulant vector High based on I / Q data; c. Construct the feature vector INPUT = [Data,Ins,High,Channel,MC] for the adaptive threshold spiking neural network model used as input; Data is the segmented I / Q data.

6. The integrated method for underwater alternating magnetic field detection and magnetic induction communication according to claim 1, characterized in that: The coil unit is formed by winding enameled wire with a cross-sectional area of ​​0.3 square millimeters in a coil slot with a radius of 10 cm, and the number of coil turns is 30.

7. The integrated method for underwater alternating magnetic field detection and magnetic induction communication according to claim 1, characterized in that: A communication device contains three omnidirectional coils; the three omnidirectional coils are arranged in an equilateral triangle in the pressure chamber.

8. The integrated method for underwater alternating magnetic field detection and magnetic induction communication according to claim 1, characterized in that: The communication equipment includes a pressure-resistant chamber, and a general-purpose software radio peripheral, an external amplifier circuit, a resonant matching circuit, and multiple omnidirectional coils installed in the pressure-resistant chamber. The general-purpose software radio peripheral includes a processor, an automatic gain module, a bandpass filter, and a built-in amplifier. The resonant matching circuit includes a series resonant transmitting circuit and a parallel resonant receiving circuit. During operation, a high-frequency relay is used to switch the connection between the omnidirectional coils and the series resonant transmitting circuit and the parallel resonant receiving circuit.

9. The integrated method for underwater alternating magnetic field detection and magnetic induction communication according to claim 1, characterized in that: The underwater pressure tank is made of titanium alloy.

10. The integrated method for underwater alternating magnetic field detection and magnetic induction communication according to claim 7, characterized in that: In step three, the layer mapping and precoding processes during signal transmission are as follows: a. Layer mapping processing divides the communication data into three layers corresponding to the number of omnidirectional coils; b. Precoding processing: The Zero Forcing precoding algorithm is used to process the 3-layer data; the 3-layer data to be transmitted by the three omnidirectional coils of the transmitting communication device is defined as: X = [x1, x2, x3] T , where x i To determine the data that the omnidirectional coil i of the transmitting communication device needs to send to the receiving communication device; define the channel matrix. Where H ij Define the channel matrix between the omnidirectional coil i of the transmitting communication device and the omnidirectional coil j of the receiving communication device; define the encoding matrix. Among them W i The encoding matrix of the transmitting omnidirectional coil i for the receiving communication device is given; the encoding matrix W = cH is calculated using the ZF algorithm. H (HH H )- 1 H H This is the conjugate transpose of the channel matrix; Here, P is the power normalization factor, where P is the average transmit power of the transmitter; the coding matrix has the following characteristics [H ij W k ]=0, i≠j≠k, the actual transmitted signal of the omnidirectional coil i of the transmitting communication device is send i =W i x i ; In step three, the decoding process of the encoding matrix during signal reception is as follows: The signal received by the omnidirectional coil j of the receiving communication device is: Where n is the noise signal; the data received by the omnidirectional coil j of the receiving communication device contains only H. jj W j x j +n represents the data originating from the omnidirectional coil j of the transmitting communication device, and is the data that the omnidirectional coil j of the receiving communication device expects to receive. Interference from omnidirectional coils of other transmitting communication devices besides the transmitting communication device's omnidirectional coil j; the received data is decoded using the extracted and set encoding matrix and power normalization factor to obtain... This means eliminating interference between omnidirectional coils of transmitting and receiving communication devices with different serial numbers.

Images