Front-haul optical fiber sensing noise reduction method, electronic device, and storage medium

Abstract

The application provides a front-haul optical fiber sensing noise reduction method, an electronic device and a storage medium. Phase extraction and phase unwrapping are performed on a low-frequency vibration signal. Phase unwrapping solves the phase jump caused by excessive vibration, guarantees the continuity of the time-domain waveform, ensures the integrity of the vibration sensing data, realizes high spatial resolution and high sensitivity vibration sensing, and accurately determines the sensing result. The time-frequency amplitude spectrum of the high-frequency communication signal is extracted, and a deep convolutional neural network is used to generate an amplitude spectrum mask to accurately distinguish the energy distribution of noise and effective signals. The original phase information of the reference laser is retained when the original amplitude spectrum with noise is weighted and the timing signal is reconstructed. In the demodulation and recovery stage, the reference laser is used to eliminate phase ambiguity, ensuring the accuracy of binary phase decision, effectively removing vibration crosstalk, improving the signal-to-noise ratio, and improving the communication reliability of the system in complex environments.

Description

Technical Field

[0001] This application relates to the field of optical fiber communication technology, and in particular to a fronthaul optical fiber inductive noise reduction method, electronic device and storage medium. Background Technology

[0002] Fiber optic sensing technology has attracted significant attention due to its ability to simultaneously achieve high-speed data transmission and real-time environmental awareness. By integrating intelligent sensing functions into existing networks, it is possible not only to monitor the operational status of fiber optic links in real time but also to accurately sense changes in the external environment. Fiber optic sensing technology is divided into fronthaul and scattering types. Both fronthaul and scattering fiber optic sensing systems suffer from low signal-to-noise ratios. Summary of the Invention

[0003] In view of this, the purpose of this application is to propose a fronthaul fiber optic sensing noise reduction method, electronic device and storage medium, which separates the modulated optical signal into a high-frequency communication signal and a low-frequency vibration signal, performs separate processing on the dual links, reduces crosstalk of the vibration signal to the communication signal, and improves the signal-to-noise ratio of the communication while ensuring accurate sensing.

[0004] To achieve the above objectives, this application provides a fronthaul fiber optic induction noise reduction method, characterized in that it includes:

[0005] A reference laser signal with the same frequency and phase as the received modulated optical signal is generated. Based on the reference laser signal, the received modulated optical signal is separated into a dual-link signal to obtain a high-frequency communication signal and a low-frequency vibration signal. The low-frequency vibration signal is subjected to phase extraction and phase unwrapping to obtain a continuous phase change signal. The continuous phase change signal is then subjected to vibration sensing to obtain a sensing result. The high-frequency communication signal is subjected to time-frequency feature extraction to obtain the time-frequency amplitude spectrum, and the time-frequency amplitude spectrum is subjected to forward propagation noise reduction processing according to the deep convolutional neural network to obtain the amplitude spectrum mask. The signal is reconstructed based on the original amplitude spectrum of the reference laser signal and the amplitude spectrum mask to obtain the noise-reduced reconstructed timing signal; The reconstructed timing signal is demodulated and recovered based on the reference laser to obtain the recovered phase corresponding to each sampling point. Binary phase decision is performed on each recovered phase to obtain noise-reduced communication data.

[0006] In one embodiment, the step of performing dual-link signal separation on the received modulated optical signal based on the reference laser to obtain a high-frequency communication signal and a low-frequency vibration signal includes: The reference laser signal and the modulated optical signal are subjected to in-phase interference to obtain in-phase components in the optical domain; Phase difference interference is performed on the reference laser signal and the modulated optical signal to obtain orthogonal components in the optical domain; The optical in-phase component and the optical quadrature component are photoelectrically converted to obtain a digital in-phase component and a digital quadrature component, and the signal frequencies of the digital in-phase component and the digital quadrature component are determined. In response to the signal frequency being greater than or equal to the lowest frequency of the communication signal, the digital in-phase component and the digital quadrature component are determined as the high-frequency communication signal; In response to the signal frequency being less than or equal to the highest frequency of the vibration signal, the digital in-phase component and the digital quadrature component are determined as the low-frequency vibration signal.

[0007] In one embodiment, the step of performing phase extraction and phase unwrapping on the low-frequency vibration signal to obtain a continuously phase-changing signal includes: Phase extraction is performed by determining the instantaneous phase of each of the low-frequency vibration signals to obtain an instantaneous phase sequence; The continuous phase change signal is obtained by performing phase unwrapping by correcting the phase jump of the instantaneous phase sequence.

[0008] In one embodiment, the sensing result includes vibration amplitude and vibration event; the vibration sensing of the continuous phase change signal to obtain the sensing result includes: The vibration amplitude is obtained by quantizing the vibration amplitude based on the light wavelength and the continuous phase change signal; The time-domain and frequency-domain information of the continuous phase change signal is extracted, and event matching is performed based on the time-domain and frequency-domain information to obtain the vibration event.

[0009] In one embodiment, the step of extracting time-frequency features from the high-frequency communication signal to obtain a time-frequency amplitude spectrum includes: The high-frequency communication signal is continuously sampled according to a preset sampling frequency to obtain a discrete time series; The discrete time series is subjected to anti-aliasing filtering based on the preset ideal low-pass filter impulse response to obtain the filtered time series. The filtered time series is subjected to overlapping frame segmentation and windowing processing based on the continuously sampled frame length and the preset overlap rate to obtain the windowed frame signal; Perform a short-time Fourier transform on the windowed frame signal to obtain a complex matrix in the time-frequency domain, and extract the amplitude spectrum of the complex matrix in the time-frequency domain to obtain the time-frequency amplitude spectrum.

[0010] In one embodiment, the step of performing forward propagation denoising on the time-frequency amplitude spectrum using a deep convolutional neural network to obtain an amplitude spectrum mask includes: By moving the convolution kernel vertically or horizontally along the time-frequency plot of the time-frequency amplitude spectrum, local time-frequency features of the time-frequency amplitude spectrum are extracted to obtain a feature image; Add a preset offset to the feature image to obtain an offset feature image; The offset feature image is normalized and subjected to nonlinear transformation to obtain enhanced features; The enhanced features are mapped to the amplitude spectrum mask by linear activation convolution.

[0011] In one embodiment, the step of reconstructing the signal based on the original amplitude spectrum of the reference laser signal and the amplitude spectrum mask to obtain the denoised reconstructed time-series signal includes: The original amplitude spectrum and the amplitude spectrum mask are multiplied element-wise to obtain the reconstructed amplitude spectrum; The reconstructed amplitude spectrum is subjected to inverse short-time Fourier transform to obtain the reconstructed time-domain signal; The reconstructed time-domain signal is obtained by weighting each frame of the reconstructed time-domain signal using a Kaiser window function. The time-domain sampled signal is then superimposed and spliced ​​according to the frame shift and overlap addition method of the original amplitude spectrum to obtain the reconstructed time-series signal.

[0012] In one embodiment, performing binary phase decision on each recovered phase to obtain noise-reduced communication data includes: In response to the absolute value of the phase value of the recovered phase being less than a preset phase threshold, the recovered phase is determined to be a first binary identifier; In response to the absolute value of the phase value of the recovered phase being greater than or equal to a preset phase threshold, the recovered phase is determined to be a second binary identifier; Differential decoding is performed on the binary data stream composed of the first binary identifier and the second binary identifier to obtain the noise-reduced communication data.

[0013] Based on the same inventive concept, this application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable by the processor, wherein the processor implements the method described above when executing the computer program.

[0014] Based on the same inventive concept, this application also provides a non-transitory computer-readable storage medium that stores computer instructions for causing a computer to perform the method described above.

[0015] As described above, the fronthaul fiber optic sensing noise reduction method, electronic device, and storage medium provided in this application utilize the interference between a reference laser and the modulated optical signal to separate the modulated optical signal into a high-frequency communication signal and a low-frequency vibration signal, performing separate processing on two links. Using the same modulated optical signal, through parallel processing of dual-link separation (high-frequency communication / low-frequency sensing), the high-frequency communication link outputs normal communication data, while the low-frequency sensing link outputs vibration event identification results, reducing system cost and complexity. No separate fiber optic cable or sensor deployment is required for sensing, achieving true sensing integration and hardware resource reuse, perfectly meeting the agile control requirements of all-optical heterogeneous relays and computing networks. Phase extraction and phase unwrapping are performed on the low-frequency vibration signal. Phase unwrapping solves the phase jump caused by excessive vibration, ensuring the continuity of the time-domain waveform and the integrity of the vibration sensing data, achieving high spatial resolution and high sensitivity vibration sensing for accurate determination of the sensing results. By extracting the time-frequency amplitude spectrum from the high-frequency communication signal and using a deep convolutional neural network to generate an amplitude spectrum mask, the energy distribution of noise and effective signals is accurately distinguished. An amplitude spectrum mask is used to weight the noisy original amplitude spectrum, and the original phase information of the reference laser is preserved when reconstructing the time-series signal. In the demodulation and recovery stage, the reference laser is used to eliminate phase ambiguity, ensuring the accuracy of binary phase decision and thus ensuring the integrity of the bitstream of the denoised communication data. Vibration crosstalk is effectively removed, improving the signal-to-noise ratio and enhancing the system's communication reliability in complex environments. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in this application or related technologies, the drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of a front-pass integrated sensing device according to an embodiment of this application; Figure 2 This is a flowchart of the front-haul fiber optic induction noise reduction method according to an embodiment of this application; Figure 3 This is a schematic diagram of a deep convolutional neural network according to an embodiment of this application; Figure 4 This is a schematic diagram of the front-haul fiber optic sensing noise reduction device according to an embodiment of this application; Figure 5 This is a schematic diagram of an electronic device according to an embodiment of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with specific embodiments and the accompanying drawings.

[0019] It should be noted that, unless otherwise defined, the technical or scientific terms used in the embodiments of this application should have the ordinary meaning understood by one of ordinary skill in the art to which this application pertains. The terms "first," "second," and similar terms used in the embodiments of this application do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed after the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are only used to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0020] In this article, it is important to understand that any number of elements in the accompanying figures is for illustrative purposes and not for limitation, and any naming is for distinction only and has no limiting meaning.

[0021] Based on the above background description, the following situations also exist in the related technologies: Most fiber optic induction systems in related technologies do not use a noise reduction module to reduce the impact of external noise. This means that external noise will cause crosstalk to the communication signal, thereby reducing the quality of communication.

[0022] The fronthaul fiber optic sensing noise reduction method, electronic device, and storage medium provided in this application utilize the interference between a reference laser and a modulated optical signal to separate the modulated optical signal into a high-frequency communication signal and a low-frequency vibration signal, performing separate processing on two links. Using the same modulated optical signal, through parallel processing of the two links (high-frequency communication / low-frequency sensing), the high-frequency communication link outputs normal communication data, while the low-frequency sensing link outputs vibration event identification results, reducing system cost and complexity. No separate fiber optic cable or sensor deployment is required for sensing, achieving true sensing integration and hardware resource reuse, perfectly meeting the agile control requirements of all-optical heterogeneous relays and computing networks. Phase extraction and phase unwrapping are performed on the low-frequency vibration signal. Phase unwrapping solves the phase jump caused by excessive vibration, ensuring the continuity of the time-domain waveform and the integrity of the vibration sensing data, achieving high spatial resolution and high sensitivity vibration sensing for accurate determination of the sensing results. By extracting the time-frequency amplitude spectrum from the high-frequency communication signal and using a deep convolutional neural network to generate an amplitude spectrum mask, the energy distribution of noise and effective signals is accurately distinguished. An amplitude spectrum mask is used to weight the noisy original amplitude spectrum, and the original phase information of the reference laser is preserved when reconstructing the time-series signal. In the demodulation and recovery stage, the reference laser is used to eliminate phase ambiguity, ensuring the accuracy of binary phase decision and thus ensuring the integrity of the bitstream of the denoised communication data. Vibration crosstalk is effectively removed, improving the signal-to-noise ratio and enhancing the system's communication reliability in complex environments.

[0023] Low-frequency signals (sensing results) and high-frequency signals (noise-reduced communication data) are output in parallel, independently serving different functions. The core of this application's embodiment is to reuse the same fiber optic link to simultaneously achieve communication and sensing. The goal of the high-frequency link is to recover the original binary noise-reduced communication data (such as user services and control information), while the goal of the low-frequency link is to extract vibration features and identify sensing events (such as traffic vibration and bridge deformation). After separation, the two types of signals are processed independently throughout the process. The noise-reduced communication data in binary data stream form carries communication services (such as user data transmission and system control signaling), and the sensing results include vibration amplitude and event type (such as "vehicle passage" and "bridge vibration"). This serves sensing applications (such as intelligent traffic monitoring and infrastructure health monitoring). The two are output independently; the communication link outputs normal communication data, while the sensing link outputs vibration event identification results, achieving integrated sensing.

[0024] The forward-looking fiber optic induction noise reduction method provided in this application embodiment is applied to, for example, Figure 1The receiver of the fronthaul-type integrated sensing device is shown. The fronthaul-type integrated sensing device achieves communication and sensing integration through five core steps: ① fronthaul-end sensing carrier generation, ② receiver-end signal separation and photoelectric conversion, ③ low-frequency link vibration sensing processing, ④ high-frequency link communication signal noise reduction and demodulation, and ⑤ scenario-based application implementation. The entire process reuses fiber optic links, eliminating the need for additional sensing equipment. It uses Convolutional Neural Networks (CNNs) to suppress crosstalk from external noise and utilizes low-frequency vibration signals for environmental sensing.

[0025] For the sending end, including: The Communication Data Source (CDS) is used to generate the binary data stream to be transmitted (such as 010110...), and is the source of information for the entire system.

[0026] Tunable laser diodes (TLDs) are used to output highly stable, narrow-linewidth optical carriers (e.g., wavelength 1550nm, linewidth ≤100Hz) to provide a reference light source for binary phase shift keying (BPSK) modulation.

[0027] An Arbitrary Waveform Generator (AWG) is used to map the binary data stream output by the CDS into a bipolar non-return-to-zero (NRZ) analog drive signal, completing the electrical domain mapping of "digital bits → phase change" (e.g., ±2.8V voltage corresponds to 0 / 1 phase difference).

[0028] An electro-optic modulator (EOM) is used to convert the electrical phase changes generated by an AWG into phase changes of an optical carrier, thereby enabling the encoding of communication data in the optical domain (the bias point is set to the minimum transmission point to optimize modulation efficiency).

[0029] An isolator (ISO) allows optical signals to be transmitted in only one direction, preventing reflected light from the fiber from backtracking to the laser and avoiding interference with the stability of the TLD (preventing wavelength drift or power fluctuations).

[0030] The receiving end includes: The local oscillator (LO) is used to generate a reference laser (the reference for coherent demodulation) that is in the same frequency and phase as the received optical signal. This reference laser is then used for subsequent interferometric extraction of the in-phase component (I component) and the quadrature component (Q component).

[0031] A 90° optical hybrid is used to interfere the received modulated optical signal with the reference laser of the receiver (LO), outputting two orthogonal optical signals: In-phase component (I): The interference result when the phase difference between the modulated optical signal and the reference laser is 0°; Quadrature component (Q): The interference result when the phase difference between the modulated optical signal and the reference laser is 90°.

[0032] The in-phase component and the quadrature component satisfy the phase extraction formula φ=arctan(Q / I).

[0033] A balanced photodetector (BPD) is used to convert the optical domain I / Q components of the 90° Hybrid output into electrical domain I / Q signals (analog electrical signals), and to suppress common-mode noise (such as laser relative intensity noise) through differential detection, thereby improving the signal-to-noise ratio.

[0034] The core logic of the transmitter is to load communication data onto the optical carrier through the modulation of "electrical signal → optical phase". At the same time, the modulation effect of fiber strain on the optical phase naturally realizes the sensing function (sensory multiplexing).

[0035] The core logic of the receiver is to extract I / Q components through coherent demodulation (LO+90°Hybrid) to provide a foundation for subsequent dual-link separation (high-frequency communication / low-frequency sensing).

[0036] These modules together form the hardware foundation of "integrated sensing," and together with digital signal processing (such as CNN noise reduction and phase unwrapping), they improve communication quality and enhance sensing capabilities.

[0037] Because sensing information resides in the low-frequency band while communication information resides in the high-frequency band, this frequency domain separation of communication and sensing information is a key foundation for achieving deep integration of communication and sensing, fully utilizing the inherent low-frequency characteristics of external vibration signals and the high-frequency characteristics of communication signals. The input light undergoes phase modulation, encoding the communication data into optical carriers propagating along the optical fiber. When the light is disturbed, the phase of the charge carriers is further modulated. In the detection phase, the high-frequency communication signal and the low-frequency vibration signal are first separated and processed separately, and then the communication data and vibration information are recovered through high-frequency signal processing and low-frequency signal processing, respectively.

[0038] The following describes in detail, with reference to the accompanying drawings, the front-haul fiber optic induction noise reduction method provided by the embodiments of this application.

[0039] In some embodiments, such as Figure 2 As shown, the front-haul fiber optic induction noise reduction method includes steps 201-205.

[0040] Step 201: Generate a reference laser signal with the same frequency and phase as the received modulated optical signal. Based on the reference laser signal, perform dual-link signal separation on the received modulated optical signal to obtain a high-frequency communication signal and a low-frequency vibration signal.

[0041] In some embodiments, the received modulated optical signal is subjected to dual-link signal separation based on a reference laser to obtain a high-frequency communication signal and a low-frequency vibration signal, including: By performing in-phase interference on the reference laser signal and the modulated optical signal, in-phase components in the optical domain are obtained; Phase difference interference is performed on the reference laser signal and the modulated optical signal to obtain orthogonal components in the optical domain; The optical in-phase component and the optical quadrature component are photoelectrically converted to obtain the digital in-phase component and the digital quadrature component, and the signal frequencies of the digital in-phase component and the digital quadrature component are determined. In response to a signal frequency greater than or equal to the lowest frequency of the communication signal, the digital in-phase component and the digital quadrature component are identified as high-frequency communication signals; In response to a signal frequency less than or equal to the highest frequency of the vibration signal, the digital in-phase component and the digital quadrature component are identified as low-frequency vibration signals.

[0042] In practice, the local oscillator at the receiving end generates a reference laser signal with the same frequency and stable phase as the received modulated optical signal, which serves as the reference light source for coherent demodulation. The reference laser signal and the modulated optical signal are then input into a 90° optical mixer for interference: the in-phase interference of the modulated optical signal and the reference laser signal outputs an in-phase component (optical domain I component), while the interference of the modulated optical signal and the reference laser signal with a 90° phase difference outputs an orthogonal component (optical domain Q component). These two optical components are converted into digital in-phase components (I[n]) and digital orthogonal components (Q[n]) in voltage form by a balanced photodetector. During the conversion process, differential detection suppresses common-mode noise (such as laser relative intensity noise), improving the signal-to-noise ratio.

[0043] Based on the frequency domain characteristics of communication signals and vibration signals (communication signals are high-frequency, with energy concentrated in the MHz~GHz range; vibration signals are low-frequency, with urban traffic vibration frequencies ≤1kHz), the digital I / Q components are frequency-determined: if the signal frequency is ≥ the lowest frequency of the communication signal (e.g., 10MHz), it is determined to be a high-frequency communication signal; if the signal frequency is ≤ the highest frequency of the vibration signal (e.g., 1kHz), it is determined to be a low-frequency vibration signal. Dual-link separation is achieved using a digital FIR filter: a high-pass filter (cutoff frequency 10MHz) extracts the high-frequency communication signal, and a low-pass filter (cutoff frequency 1kHz) extracts the low-frequency vibration signal, ensuring no crosstalk between the two signals. For example, an FIR filter using a Kaiser window design has a transition band width ≤1MHz and a stopband attenuation ≥60dB, ensuring effective separation.

[0044] Step 202: Perform phase extraction and phase unwrapping on the low-frequency vibration signal to obtain a continuous phase change signal, and perform vibration sensing on the continuous phase change signal to obtain the sensing result.

[0045] In practical implementation, the low-frequency vibration signal sensing module first calculates the instantaneous phase, extracting phase information from the low-frequency I and Q components. Then, phase unwrapping is used to process the 2π jump, ensuring phase continuity. Next, the quantitative relationship between phase difference and optical path difference is used to calculate the fiber length change, accurately quantifying the vibration amplitude. To identify different types of vibration events, short-time Fourier transform or wavelet transform is used to obtain the time and frequency domain information of the vibration signal, ultimately extracting the characteristics of the vibration signal to achieve the identification of different vibration events.

[0046] In some embodiments, phase extraction and phase unwrapping are performed on the low-frequency vibration signal to obtain a continuously phase-changing signal, including: Phase extraction is performed by determining the instantaneous phase of each low-frequency vibration signal to obtain an instantaneous phase sequence; Phase unwrapping is performed by correcting the phase jumps in the instantaneous phase sequence to obtain a continuously phase-changing signal.

[0047] In practical implementation, for low-frequency vibration signals, the instantaneous phase is first calculated: φ(t)=arctan(I(t) / Q(t))(1) Wherein, φ(t) is the instantaneous phase of the low-frequency vibration signal; I(t) is the in-phase component of the low-frequency vibration signal; and Q(t) is the quadrature component of the low-frequency vibration signal. The phase change caused by vibration is extracted by formula (1) to obtain the instantaneous phase sequence containing 2π jumps. Then, phase unwrapping processing is performed to correct the 2π phase jumps in the instantaneous phase sequence (such as judging the jump direction by the phase difference threshold between adjacent sampling points) to ensure the continuity of phase over time, and the continuous phase change signal Δφ(t) is obtained.

[0048] In some embodiments, the sensing results include vibration amplitude and vibration events; vibration sensing of a continuously phase-changing signal yields sensing results, including: The vibration amplitude is obtained by quantizing the vibration amplitude based on the optical wavelength and the continuous phase change signal. The time-domain and frequency-domain information of the continuous phase-change signal is extracted, and event matching is performed based on the time-domain and frequency-domain information to obtain the vibration event.

[0049] In practical implementation, the linear relationship between vibration phase change and fiber strain is used: Δφ=2π ΔL / λ(2) Here, ΔL represents the change in fiber length, i.e., the vibration amplitude, and λ represents the optical carrier wavelength (e.g., 1550nm). The continuous phase-change signal is converted into vibration amplitude ΔL, thus quantifying the vibration intensity. To further identify vibration event types, time-frequency analysis is performed on the continuous phase-change signal: time-domain (e.g., vibration duration) and frequency-domain (e.g., vibration dominant frequency) features are extracted using Short-Time Fourier Transform (STFT) or Wavelet Transform. These features are then combined with a pre-defined event feature library (e.g., the periodicity of approximately 20Hz for vehicle traffic and the low-frequency, slowly varying characteristics of bridge vibration) for pattern matching to obtain the sensing results, including vibration amplitude (e.g., 0.05mm) and vibration event type (e.g., "vehicle traffic" or "bridge vibration").

[0050] In urban intelligent transportation scenarios, the low-pass filter cutoff frequency is adjusted to 500Hz based on traffic vibration characteristics, and the phase unwrapping algorithm is optimized to adapt to the strain changes of short-distance optical fibers, achieving an event recognition accuracy of over 95%.

[0051] Step 203: Extract time-frequency features from the high-frequency communication signal to obtain the time-frequency amplitude spectrum, and perform forward propagation noise reduction on the time-frequency amplitude spectrum using a deep convolutional neural network to obtain the amplitude spectrum mask.

[0052] In practical implementation, the high-frequency communication signal processing module first performs ADC sampling on the high-frequency I and Q components to convert the analog signal into a digital signal. Then, the signal undergoes preprocessing. Specifically, the sampled I and Q components are first low-pass filtered to prevent aliasing. Subsequently, the I and Q components are framed and windowed to obtain the continuous changes in the signal. Next, a short-time Fourier transform is performed on each preprocessed frame, and the amplitude spectrum of each frame is extracted as input to the convolutional neural network.

[0053] The deep convolutional neural network then comprises N 2D convolutional layers, partially batch normalized layers, and activation layers. The convolutional layers move vertically and horizontally along the input, calculate the dot product of the weights and the input, and then add a bias term to introduce a learnable offset. Next, the local time-frequency features extracted by the convolutional layers are passed through batch normalized layers to stabilize the training of the convolutional neural network and accelerate its convergence. Finally, a ReLU activation layer is introduced for non-linear transformation to enhance the model's expressive power. Through repeated convolutions, a linearly activated convolutional layer is used in the final layer to generate an amplitude spectrum mask.

[0054] In some embodiments, time-frequency feature extraction is performed on the high-frequency communication signal to obtain the time-frequency amplitude spectrum, including: The high-frequency communication signal is continuously sampled according to the preset sampling frequency to obtain a discrete time series; The discrete time series is subjected to anti-aliasing filtering based on the preset ideal low-pass filter impulse response to obtain the filtered time series. The filtered time series is subjected to overlapping frame segmentation and windowing processing based on the continuously sampled frame length and the preset overlap rate to obtain the windowed frame signal; Perform a short-time Fourier transform on the windowed frame signal to obtain a complex matrix in the time-frequency domain. Extract the amplitude spectrum of the complex matrix in the time-frequency domain to obtain the time-frequency amplitude spectrum.

[0055] In practice, in order to reduce the crosstalk of external noise to communication signals, a forward-propagating fiber optic induction noise reduction algorithm is introduced into high-frequency communication signal processing. This noise reduction algorithm is constructed based on a deep convolutional neural network.

[0056] The high-frequency I / Q signal is sampled by an ADC at a sampling rate of 2.5 GS / s to obtain the discrete-time sequence I / Q[n]: (3) Where Ts is the sampling period, representing the time interval between the continuous-time signal being discretized into sampling points, in seconds (s). The relationship between the sampling period and the sampling frequency is: Ts = 1 / fs, where fs is the sampling frequency (in Hz), i.e., the number of samples collected per second, and n represents the nth sampling point.

[0057] The discrete-time series is convolved and filtered using the impulse response of an ideal low-pass filter to eliminate spectral aliasing, resulting in the filtered time series: (4) (5) Where h[k] is the impulse response of the ideal low-pass filter, M is the half-length of the filter, and is the upper limit (absolute value of the lower limit) of the summation, representing the half-length of the filter. It is used to control the total length of the filter (2M+1 taps) and frequency response characteristics: then the total length of the filter = 2M+1 (e.g., if M=10, there are 21 taps). The larger M is, the steeper the transition band of the filter (closer to the ideal low-pass characteristics), but the higher the computational complexity. M corresponds to the truncation length of the impulse response of the ideal low-pass filter, because the impulse response of the ideal low-pass filter is infinitely long, and in actual implementation it needs to be truncated to a finite length. k is the summation index variable (filter tap position), which is the index variable for traversing the impulse response h[k] of the ideal low-pass filter, indicating the starting point from the end of the summation. The integer range from M to M. In the convolution operation, k corresponds to each tap of the filter (i.e., the position of the filter coefficients), I / Q[n [k] represents the value of the input signal I / Q at a delay of k sampling points. For example: if M=2, then k= 2, 1,0,1,2 correspond to the 5 tap positions of the filter.

[0058] To analyze the changes in I / Q signals over time, based on the short-term stationarity of speech, the filtered time series needs to be subjected to overlapping frame segmentation and windowing processing according to the frame length of continuous sampling and a preset overlap rate to obtain the windowed frame signal. The overlapping frame segmentation process is as follows: (6) Where L is the frame length, H is the frame shift, satisfying H=|L(1-α)|, and α is the overlap ratio. m is the frame number (representing the m-th frame); l is the intra-frame sampling point index (from 0 to L). 1) mH+l is the sampling point index of the filtered discrete sequence (must be an integer, conforming to the characteristics of a discrete sequence), indicating that the frame length L and frame shift H are index rules for the discrete sequence I / Qfiltered[n], and are core parameters in the processing.

[0059] To avoid loss of speech signal after window function processing, the overlap between windows is set so that the sliding step size of the window is smaller than the window length. To obtain the periodic changes of the speech signal and prevent spectral leakage caused by sudden truncation, Kaiser windowing is also used for windowing processing. Therefore, the expression for the m-th frame signal after Kaiser windowing is: (7) (8) Where w[l] is a function of the Kaiser window, I0 is a zero-order Bezier function, and β is a shape parameter.

[0060] Subsequently, a short-time Fourier transform (STFT) is performed on each preprocessed frame signal to obtain a complex matrix in the time-frequency domain. The expression for the STFT of the m-th frame signal in the time-frequency domain complex matrix is ​​as follows: (9) Where X[m,k] represents the sampling of the short-time Fourier transform of the m-th frame signal at the k-th frequency point.

[0061] Extract the amplitude spectrum of X[m,k] and use it as the input to the convolutional neural network. The expression for the time-frequency amplitude spectrum A[m,k] corresponding to X[m,k] is: (10) Where Re denotes taking the real part, and Im denotes taking the imaginary part. In some embodiments, forward propagation denoising is performed on the time-frequency amplitude spectrum using a deep convolutional neural network to obtain an amplitude spectrum mask, including: By moving the convolution kernel vertically or horizontally along the time-frequency plot of the time-frequency amplitude spectrum, local time-frequency features of the time-frequency amplitude spectrum are extracted to obtain a feature image; Add a preset offset to the feature image to obtain an offset feature image; Enhanced features are obtained by normalizing and nonlinearly transforming the offset feature image; The enhanced features are mapped to an amplitude spectrum mask through linear activation convolution.

[0062] In practice, the deep convolutional neural network takes the time-frequency amplitude spectrum A[m,k] of the high-frequency communication signal as input and performs forward propagation noise reduction processing to generate an amplitude spectrum mask M[m,k]. The deep convolutional neural network is as follows: Figure 3 As shown, a deep convolutional neural network (CNN) consists of N 2D convolutional layers, a partially batch normalized layer, and activation layers. The convolutional layers move along the vertical and horizontal directions of the input, calculate the dot product of the weights and the input, and then add a bias term to introduce a learnable offset. Next, the local time-frequency features extracted by the convolutional layers are passed through the batch normalized layer to stabilize the training of the CNN and accelerate its convergence. Finally, a ReLU activation layer is introduced for non-linear transformation to enhance the model's expressive power. Through repeated convolutions, a linearly activated convolutional layer is used in the final layer to generate an amplitude spectrum mask. The operation process of the deep convolutional neural network is as follows: Convolutional layers, as the fundamental components of convolutional neural networks, are responsible for extracting local time-frequency features. By moving the convolutional kernel vertically or horizontally along the time-frequency plot of the time-frequency amplitude spectrum, local time-frequency features of the time-frequency amplitude spectrum are extracted, resulting in a feature image. (11) Where Y[i,j] represents the feature image generated by the convolution kernel through the amplitude spectrum, i represents the vertical index (corresponding to the time axis / frame number direction of the time-frequency amplitude spectrum), j represents the horizontal index of the feature image (corresponding to the frequency axis / frequency point direction of the time-frequency amplitude spectrum), W[m,n] represents the convolution kernel feature map, m represents the vertical index of the convolution kernel feature map, n represents the horizontal index of the convolution kernel feature map, Kh represents the height of the convolution kernel feature map, and Kw represents the width of the convolution kernel feature map.

[0063] Then, a preset offset is added to the feature image to obtain the offset feature image Z[i,j]: (12) Where b[i,j] is the offset.

[0064] To stabilize the training of convolutional neural networks and accelerate their convergence, the local time-frequency features extracted by the convolutional layers need to be processed through a batch normalization layer. Z[i,j] is input into the batch normalization layer to normalize the offset feature image, resulting in normalized features. ; (13) Among them, u B σ is the batch mean. B γ is the batch standard deviation, representing the standard deviation of all samples in the current training batch along the feature dimension, reflecting the dispersion of the batch data. γ and β are learnable parameters. It is an extremely small constant (such as 1e) 5) Add to variance to avoid the denominator being 0 (to prevent numerical instability).

[0065] Finally, a ReLU activation layer is introduced to perform a nonlinear transformation on the normalized features, enhancing the model's ability to express complex time-frequency patterns (such as the randomness of noise and the periodicity of signals), resulting in enhanced features. ; (14) The enhanced features are mapped to an amplitude spectrum mask through linear activation convolution. Amplitude spectral mask It is a scaling factor in the time-frequency domain, representing the proportion of energy occupied by pure speech in each time-frequency unit of the amplitude spectrum. After multiple layers of convolutional feature extraction, the final layer uses a linearly activated convolutional layer to map the enhanced features into an amplitude spectrum mask of the same size as the input time-frequency amplitude spectrum. (Value range [0,1]), representing the percentage of pure communication signal energy in each time-frequency unit.

[0066] Step 204: Reconstruct the signal based on the original amplitude spectrum and amplitude spectrum mask of the reference laser signal to obtain the reconstructed time sequence signal after noise reduction.

[0067] In practice, to reconstruct the denoised signal, an amplitude spectrum mask is first applied to the amplitude spectrum of the noisy reference laser signal to obtain an estimated amplitude spectrum value for each frame of the reconstructed signal. Then, through inverse short-time Fourier transform, the time-domain expression of each frame is obtained. Finally, the Kaiser window function and overlapping addition method are applied to reconstruct the time-domain expressions of the denoised I and Q signals, resulting in the reconstructed time-series signal after denoising. The entire convolutional neural network uses root mean square error as the loss function for training and the Adam algorithm as the optimization algorithm for the neural network parameters.

[0068] In some embodiments, signal reconstruction is performed based on the original amplitude spectrum and amplitude spectrum mask of the reference laser signal to obtain a denoised reconstructed time-series signal, including: The original amplitude spectrum and the amplitude spectrum mask are multiplied element by element to obtain the reconstructed amplitude spectrum; The reconstructed amplitude spectrum is subjected to inverse short-time Fourier transform to obtain the reconstructed time-domain signal; The reconstructed time-domain signal is obtained by weighting each frame of the reconstructed time-domain signal using the Kaiser window function. The time-domain sampled signal is then superimposed and spliced ​​according to the frame shift and overlap addition method of the original amplitude spectrum to obtain the reconstructed time-series signal.

[0069] In practice, the original amplitude spectrum A[m,k] containing noise is first multiplied element-wise using the amplitude spectrum mask M[m,k] to obtain the reconstructed amplitude spectrum after noise reduction. ; (15) in, This represents the estimate of the amplitude spectrum of the signal in each frame after reconstruction. This represents the original amplitude spectrum of the noisy reference laser signal for each frame.

[0070] Then, by performing an inverse short-time Fourier transform on the amplitude spectrum estimate of each frame of the reconstructed amplitude spectrum, the time-domain sampled value of the m-th frame signal can be obtained, resulting in the reconstructed time-domain signal. Finally, the reconstructed time-domain signal is weighted by a Kaiser window function for each frame signal, yielding the time-domain sampled signal. (16) in, Let w[n] represent the reconstructed time-domain signal, and w[n] represent the value of the Kaiser window function at the nth sampling point.

[0071] Finally, the time-domain sampled signals are superimposed and stitched together according to the frame shift and overlap addition method of the original amplitude spectrum to reconstruct the complete reconstructed time-series signal (also known as the reconstructed I / Q signal): (17) The root mean squared error (RMSE), also known as the standard error, is the square root of the ratio of the square of the deviation between the predicted and actual values ​​to the number of observations, n. It is used as the loss function for convolutional neural networks. (18) Where y is the original time-series signal of the reference laser signal (also known as the original I / Q signal). MSE, as the loss function, reflects the deviation between the reconstructed I / Q signal after denoising and the original I / Q signal, thereby measuring the denoising effect of the model. Root mean square error is used as the basis for measuring the model training effect. The model is trained to the state with the minimum root mean square error, which is also considered to be the optimal model parameter, achieving the best denoising effect.

[0072] Step 205: Demodulate and recover the reconstructed time-series signal based on the reference laser to obtain the recovered phase corresponding to each sampling point. Perform binary phase decision on each recovered phase to obtain noise-reduced communication data.

[0073] In practice, the reconstructed timing signal after noise reduction is first decomposed and orthogonalized using I / Q methods to ensure precise orthogonality of the I / Q signals and compensate for amplitude imbalance. Then, timing error detection technology is used to precisely lock the optimal sampling time before resampling. Next, frequency domain equalization compensation is used to compensate for various impairments introduced during transmission, including dispersion and polarization mode dispersion. Then, a Costas loop is used to estimate and compensate for frequency offset and phase noise. Finally, using the I(n) and Q(n) signals, the phase value of the high-frequency communication signal at each sampling point is extracted to obtain the recovered phase φ(n). Based on the recovered phase, binary phase decision is performed: when the absolute phase value is less than 90°, it is judged as "0"; when it is greater than or equal to 90°, it is judged as "1", and a binary data stream is output. Then, differential decoding is performed to solve the 0-π phase ambiguity problem of the binary data stream, resulting in the noise-reduced communication data.

[0074] In some embodiments, binary phase decision is performed on each recovered phase to obtain noise-reduced communication data, including: If the absolute value of the recovered phase value is less than a preset phase threshold, the recovered phase is determined to be the first binary identifier; In response to the absolute value of the recovered phase being greater than or equal to a preset phase threshold, the recovered phase is determined to be a second binary identifier; Differential decoding is performed on the binary data stream consisting of the first binary identifier and the second binary identifier to obtain noise-reduced communication data.

[0075] In practical implementation, phase extraction is performed as follows: The recovered phase φ(n) = arctan(Q(n)I(n)) is calculated for each sampling point. Then, a binary decision is made on the recovered phase. When the absolute value of the phase is less than 90°, it is determined as the first binary identifier "0", and when it is greater than or equal to 90°, it is determined as the second binary identifier "1". The binary decision is then expressed as: (19) Then, differential decoding is performed on the binary data stream after binary decision to eliminate the 0-π phase ambiguity of BPSK modulation (e.g., XORing the current bit with the previous bit, setting the initial reference bit to 0), and the original binary data stream is restored. After demodulation, the bit error rate is reduced from 1e-3 to 1e-5, meeting the communication quality requirements.

[0076] For example, in an urban intelligent traffic monitoring scenario, the transmitter uses a highly stable tunable laser diode with a center wavelength of 1550nm, a linewidth ≤100Hz, and an output power of 12mW. After splitting by a 95:5 fiber coupler, 95% of the light is used for the BPSK modulation module, and 5% is used as the local oscillator. In the modulation module, the electrical signal generated by the arbitrary waveform generator includes a communication signal based on BPSK modulation with a symbol rate of 200MBaud. The communication signal uses bipolar NRZ encoding with a voltage swing of ±2.8V. To meet the optical power requirements of the urban environment, the erbium-doped fiber amplifier at the modulator output amplifies the optical power to 18dBm, and an optical filter is used to further optimize signal quality and suppress urban electromagnetic interference.

[0077] The optical signal travels tens of kilometers through a single-mode fiber optic network before reaching the receiving end, covering major urban traffic arteries. The communication signal receiving module acquires the signal using a balanced photodetector and a high-speed ADC (sampling rate 2.5 GS / s), followed by noise reduction processing using a 17-layer deep convolutional neural network. The network input is the time-frequency amplitude spectrum, and the output is an amplitude spectrum mask. By applying the amplitude spectrum mask and the time-frequency amplitude spectrum of the noisy signal, the time-frequency amplitude spectrum of the denoised signal can be generated. Then, a series of processes are performed, including I / Q orthogonalization, clock recovery and resampling, frequency domain equalization compensation, carrier phase recovery, and phase extraction and recovery, ultimately obtaining the communication data. In addition to the basic modules used in the basic implementation scheme, the sensor signal receiving module also employs a high-sensitivity circulator, an erbium-doped fiber amplifier, and a narrowband optical filter to ensure effective extraction of traffic vibration signals. During data processing, the sensor signal receiving module also adjusts the filtering range according to the characteristics of urban traffic vibration signals and obtains strain distribution information along the fiber optic network through a precise phase unwrapping algorithm. By combining time-frequency analysis, it can accurately identify events such as vehicle traffic and bridge vibration, and convert phase changes into vibration amplitude and frequency information to achieve intelligent traffic monitoring.

[0078] It should be noted that the method in this embodiment can be executed by a single device, such as a computer or server. The method can also be applied in a distributed scenario, where multiple devices cooperate to complete the task. In such a distributed scenario, one of these devices may execute only one or more steps of the method in this embodiment, and the multiple devices will interact with each other to complete the method described.

[0079] It should be noted that the above description describes some embodiments of this application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in a different order than that shown in the above embodiments and still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0080] Based on the same inventive concept, and corresponding to any of the above embodiments, this application also provides a forward-transmission fiber optic sensing noise reduction device.

[0081] refer to Figure 4 The forward-transmission fiber optic sensing noise reduction device includes: The dual-link signal separation module 10 is configured to: generate a reference laser signal with the same frequency and phase as the received modulated optical signal, and perform dual-link signal separation on the received modulated optical signal based on the reference laser signal to obtain a high-frequency communication signal and a low-frequency vibration signal; The low-frequency signal processing module 20 is configured to: extract and unwrap the phase of the low-frequency vibration signal to obtain a continuous phase change signal, perform vibration sensing on the continuous phase change signal, and obtain a sensing result; The high-frequency signal noise reduction module 30 is configured to: extract time-frequency features from the high-frequency communication signal to obtain the time-frequency amplitude spectrum, and perform forward-propagation noise reduction processing on the time-frequency amplitude spectrum according to the deep convolutional neural network to obtain the amplitude spectrum mask; The timing signal reconstruction module 40 is configured to: reconstruct the signal based on the original amplitude spectrum and amplitude spectrum mask of the reference laser signal to obtain the denoised reconstructed timing signal; The communication data recovery module 50 is configured to: demodulate and recover the reconstructed time sequence signal based on the reference laser to obtain the recovered phase corresponding to each sampling point, and perform binary phase decision on each recovered phase to obtain noise-reduced communication data.

[0082] For ease of description, the above devices are described in terms of function, divided into various modules. Of course, in implementing this application, the functions of each module can be implemented in one or more software and / or hardware.

[0083] The apparatus of the above embodiments is used to implement the corresponding forward-transmission fiber optic induction noise reduction method in any of the foregoing embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0084] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, this application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the forward transmission fiber optic induction noise reduction method described in any of the above embodiments.

[0085] Figure 5 This embodiment illustrates a more specific hardware structure of an electronic device. The device may include a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, memory 1020, input / output interface 1030, and communication interface 1040 are interconnected internally via the bus 1050.

[0086] The processor 1010 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0087] The memory 1020 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1020 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program code is stored in the memory 1020 and is called and executed by the processor 1010.

[0088] The input / output interface 1030 is used to connect input / output modules to realize information input and output. Input / output modules can be configured as components within the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touchscreens, microphones, various sensors, etc., while output devices may include displays, speakers, vibrators, indicator lights, etc.

[0089] The communication interface 1040 is used to connect a communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0090] Bus 1050 includes a pathway for transmitting information between various components of the device, such as processor 1010, memory 1020, input / output interface 1030, and communication interface 1040.

[0091] It should be noted that although the above-described device only shows the processor 1010, memory 1020, input / output interface 1030, communication interface 1040, and bus 1050, in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the embodiments of this specification, and not necessarily all the components shown in the figures.

[0092] The electronic devices described above are used to implement the corresponding forward-transmission fiber optic induction noise reduction method in any of the foregoing embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0093] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, this application also provides a non-transitory computer-readable storage medium storing computer instructions for causing the computer to execute the fronthaul fiber optic induction noise reduction method as described in any of the above embodiments.

[0094] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0095] The computer instructions stored in the storage medium of the above embodiments are used to cause the computer to execute the forward fiber optic induction noise reduction method as described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0096] It is understood that before using the technical solutions of the various embodiments in this application, users will be informed of the type, scope of use, and usage scenarios of the personal information involved in an appropriate manner, and user authorization will be obtained.

[0097] For example, upon receiving a user's active request, a prompt message is sent to the user to explicitly inform them that the requested operation will require the acquisition and use of the user's personal information. This allows the user to independently choose, based on the prompt message, whether to provide personal information to the software or hardware such as electronic devices, applications, servers, or storage media performing the operations described in this application.

[0098] As an optional but not limited implementation, in response to a user's active request, sending a prompt message to the user can be done via a pop-up window, where the prompt message can be presented in text format. Furthermore, the pop-up window can also include a selection control allowing the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0099] It is understood that the above notification and user authorization process are merely illustrative and do not constitute a limitation on the implementation of this application. Other methods that comply with relevant laws and regulations may also be applied to the implementation of this application.

[0100] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of this application is limited to these examples; under the concept of this application, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of the embodiments of this application as described above, which are not provided in detail for the sake of brevity.

[0101] Additionally, to simplify the description and discussion, and to avoid obscuring the embodiments of this application, the well-known power / ground connections to integrated circuit (IC) chips and other components may or may not be shown in the provided drawings. Furthermore, the apparatus may be shown in block diagram form to avoid obscuring the embodiments of this application, and this also takes into account the fact that the details of the implementation of these block diagram apparatuses are highly dependent on the platform on which the embodiments of this application will be implemented (i.e., these details should be fully understood by those skilled in the art). While specific details (e.g., circuits) have been set forth to describe exemplary embodiments of this application, it will be apparent to those skilled in the art that the embodiments of this application can be implemented without these specific details or with variations thereof. Therefore, these descriptions should be considered illustrative rather than restrictive.

[0102] Although this application has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed.

[0103] The embodiments of this application are intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the claims of this application. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the embodiments of this application should be included within the protection scope of this application.

Claims

1. A method for noise reduction in fronthaul fiber optic induction, characterized in that, include: A reference laser signal with the same frequency and phase as the received modulated optical signal is generated. Based on the reference laser signal, the received modulated optical signal is separated into a dual-link signal to obtain a high-frequency communication signal and a low-frequency vibration signal. The low-frequency vibration signal is subjected to phase extraction and phase unwrapping to obtain a continuous phase change signal. The continuous phase change signal is then subjected to vibration sensing to obtain a sensing result. The high-frequency communication signal is subjected to time-frequency feature extraction to obtain the time-frequency amplitude spectrum, and the time-frequency amplitude spectrum is subjected to forward propagation noise reduction processing according to the deep convolutional neural network to obtain the amplitude spectrum mask. The signal is reconstructed based on the original amplitude spectrum of the reference laser signal and the amplitude spectrum mask to obtain the noise-reduced reconstructed timing signal; The reconstructed timing signal is demodulated and recovered based on the reference laser to obtain the recovered phase corresponding to each sampling point. Binary phase decision is performed on each recovered phase to obtain noise-reduced communication data.

2. The method according to claim 1, characterized in that, The step of performing dual-link signal separation on the received modulated optical signal based on the reference laser to obtain a high-frequency communication signal and a low-frequency vibration signal includes: The reference laser signal and the modulated optical signal are subjected to in-phase interference to obtain in-phase components in the optical domain; Phase difference interference is performed on the reference laser signal and the modulated optical signal to obtain orthogonal components in the optical domain; The optical in-phase component and the optical quadrature component are photoelectrically converted to obtain a digital in-phase component and a digital quadrature component, and the signal frequencies of the digital in-phase component and the digital quadrature component are determined. In response to the signal frequency being greater than or equal to the lowest frequency of the communication signal, the digital in-phase component and the digital quadrature component are determined as the high-frequency communication signal; In response to the signal frequency being less than or equal to the highest frequency of the vibration signal, the digital in-phase component and the digital quadrature component are determined as the low-frequency vibration signal.

3. The method according to claim 1, characterized in that, The step of performing phase extraction and phase unwrapping on the low-frequency vibration signal to obtain a continuously phase-changing signal includes: Phase extraction is performed by determining the instantaneous phase of each of the low-frequency vibration signals to obtain an instantaneous phase sequence; The continuous phase change signal is obtained by performing phase unwrapping by correcting the phase jump of the instantaneous phase sequence.

4. The method according to claim 1, characterized in that, The sensing results include vibration amplitude and vibration events; the vibration sensing of the continuous phase change signal to obtain the sensing results includes: The vibration amplitude is obtained by quantizing the vibration amplitude based on the light wavelength and the continuous phase change signal; The time-domain and frequency-domain information of the continuous phase change signal is extracted, and event matching is performed based on the time-domain and frequency-domain information to obtain the vibration event.

5. The method according to claim 1, characterized in that, The step of extracting time-frequency features from the high-frequency communication signal to obtain the time-frequency amplitude spectrum includes: The high-frequency communication signal is continuously sampled according to a preset sampling frequency to obtain a discrete time series; The discrete time series is subjected to anti-aliasing filtering based on the preset ideal low-pass filter impulse response to obtain the filtered time series. The filtered time series is subjected to overlapping frame segmentation and windowing processing based on the continuously sampled frame length and the preset overlap rate to obtain the windowed frame signal; Perform a short-time Fourier transform on the windowed frame signal to obtain a complex matrix in the time-frequency domain, and extract the amplitude spectrum of the complex matrix in the time-frequency domain to obtain the time-frequency amplitude spectrum.

6. The method according to claim 1, characterized in that, The step of performing forward propagation noise reduction on the time-frequency amplitude spectrum using a deep convolutional neural network to obtain an amplitude spectrum mask includes: By moving the convolution kernel vertically or horizontally along the time-frequency plot of the time-frequency amplitude spectrum, local time-frequency features of the time-frequency amplitude spectrum are extracted to obtain a feature image; Add a preset offset to the feature image to obtain an offset feature image; The offset feature image is normalized and subjected to nonlinear transformation to obtain enhanced features; The enhanced features are mapped to the amplitude spectrum mask by linear activation convolution.

7. The method according to claim 1, characterized in that, The step of reconstructing the signal based on the original amplitude spectrum of the reference laser signal and the amplitude spectrum mask to obtain the denoised reconstructed time-series signal includes: The original amplitude spectrum and the amplitude spectrum mask are multiplied element-wise to obtain the reconstructed amplitude spectrum; The reconstructed amplitude spectrum is subjected to inverse short-time Fourier transform to obtain the reconstructed time-domain signal; The reconstructed time-domain signal is obtained by weighting each frame of the reconstructed time-domain signal using a Kaiser window function. The time-domain sampled signal is then superimposed and spliced ​​according to the frame shift and overlap addition method of the original amplitude spectrum to obtain the reconstructed time-series signal.

8. The method according to claim 1, characterized in that, The step of performing binary phase decision on each recovered phase to obtain noise-reduced communication data includes: In response to the absolute value of the phase value of the recovered phase being less than a preset phase threshold, the recovered phase is determined to be a first binary identifier; In response to the absolute value of the phase value of the recovered phase being greater than or equal to a preset phase threshold, the recovered phase is determined to be a second binary identifier; Differential decoding is performed on the binary data stream composed of the first binary identifier and the second binary identifier to obtain the noise-reduced communication data.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1 to 8.

10. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to perform the method according to any one of claims 1 to 8.

Images