A signal processing method, storage medium and system based on a DBR laser

Abstract

The present application provides a signal processing method, storage medium and system based on a DBR laser, relating to the technical field of signal processing. The method includes: first, obtaining a signal to be processed, then performing sampling of a preset length to obtain observed values, then establishing a mathematical model to obtain theoretical values, and calculating the residuals between the observed values and the theoretical values. The model parameters are iteratively updated by the nonlinear least squares method to minimize the residuals. Finally, the central wavelength of the pulse signal is output, and it is determined whether it is within a preset range to determine whether the signal is normal. The present application can quickly and accurately determine the central wavelength of the pulse signal and improve the accuracy of the signal processing result.

Description

Technical Field

[0001] This application relates to the technical field of signal processing, and in particular, to a signal processing method, storage medium, and system based on a DBR laser. Background Art

[0002] In the field of signal processing based on DBR (Distributed Bragg Reflector) lasers, the weak light reflectivity is a key technical parameter. Due to its unique structural characteristics, the DBR laser can achieve high reflectivity within a specific wavelength range, thus playing an important role in applications such as optical communication, spectral analysis, and laser sensing.

[0003] However, in practical applications, when the incident light passes through the weak grating, only a small part of the laser signal will be reflected, and most of the laser signal will be transmitted and continue to propagate backward. However, due to the influence of the spectral shadow effect and multiple reflections, the intensity of the reflected light will become weaker and weaker, and the reflected laser signal will undergo a series of complex changes during the propagation process, resulting in a certain drift of the central wavelength of the reflected light, reducing the accuracy of the signal processing result. Summary of the Invention

[0004] In order to quickly and accurately determine the central wavelength of the pulse signal in the signal to be processed and improve the accuracy of the signal processing result, this application provides a signal processing method, storage medium, and system based on a DBR laser.

[0005] In the first aspect, this application provides a signal processing method based on a DBR laser, adopting the following technical solution:

[0006] A signal processing method based on a DBR laser includes the following steps:

[0007] First acquisition: Obtain the signal to be processed;

[0008] Sampling: Sample the signal to be processed with a preset length to obtain the observed value of the signal to be processed;

[0009] Fitting: Establish a mathematical model of the signal to be processed, and obtain the theoretical value of the signal to be processed based on the mathematical model;

[0010] Calculate the residual: Calculate the residual value based on the observed value and the theoretical value;

[0011] Update: Use the non-linear least squares method to iteratively update the parameters in the mathematical model until the residual value is the smallest, and then output the central wavelength of the pulse signal in the signal to be processed;

[0012] Wavelength judgment: Judge whether the central wavelength is within the preset wavelength range. If so, output that the signal is normal; otherwise, output that the signal is abnormal.

[0013] In this application, the signal to be processed is first obtained, and then the signal to be processed is sampled at a preset length to obtain the observed values of the signal to be processed. Then, a mathematical model of the signal to be processed is established, and the theoretical values of the signal to be processed are obtained based on the mathematical model. The residual value is calculated based on the observed values and the theoretical values. Then, the parameters in the mathematical model are iteratively updated using the nonlinear least squares method until the residual value is minimized, and then the central wavelength of the pulse signal in the signal to be processed is output. By iteratively updating the model parameters, the fitting effect of the model can be continuously optimized until the minimum residual value is reached, so as to more accurately extract the central wavelength of the pulse signal in the signal to be processed. This application uses a mathematical model to fit the signal to be processed, and iteratively updates the parameters in the mathematical model using the nonlinear least squares method, realizing the accurate fitting of the signal to be processed, thereby accurately extracting the central wavelength of the signal to be processed and improving the processing accuracy.

[0014] Optionally, after performing the step of the first acquisition and before performing the step of sampling, it further includes:

[0015] Second acquisition: Obtain training data, where the training data includes historical processed signals and denoised historical processed signals;

[0016] Modeling: Based on the deep learning algorithm, establish a CNN-LSTM model;

[0017] Model training: Use the training data to train the CNN-LSTM model to obtain the trained CNN-LSTM model;

[0018] Prediction: Input the signal to be processed into the trained CNN-LSTM model to obtain the denoised signal to be processed, and update the denoised signal to be processed as the signal to be processed.

[0019] In this application, the training data is first obtained, where the training data includes historical processed signals and denoised historical processed signals. Then, based on the deep learning algorithm, a CNN-LSTM model is established. The CNN-LSTM model is suitable for processing complex signal data with time series dependence. Then, the training data is used to train the CNN-LSTM model to obtain the trained CNN-LSTM model. Through training, the CNN-LSTM model can learn the noise patterns in the historical processed signals and the characteristics of the clean signals, so as to have the ability to denoise new signals to be processed. Then, the signal to be processed is input into the trained CNN-LSTM model to obtain the denoised signal to be processed, and the denoised signal to be processed is updated as the signal to be processed to improve the accuracy and reliability of subsequent sampling, fitting, and analysis steps.

[0020] Optionally, in the fitting step, the calculation model of the mathematical model is as follows:

[0021] ;

[0022] Wherein, is the theoretical value of the signal to be processed; A is the peak value of the pulse signal in the signal to be processed; B is the baseline slope; C is the baseline intercept; is the central wavelength of the pulse signal in the signal to be processed; is the width of the pulse signal in the signal to be processed; is the wavelength of the signal to be processed.

[0023] Based on the approximate Gaussian distribution characteristic of the grating reflection spectrum, this application realizes high-precision parameter estimation by combining the least squares optimization method.

[0024] Optionally, the fitting step further includes: setting the initial value of the peak value A of the pulse signal in the signal to be processed as the maximum value of the pulse signal, setting the initial value of the baseline slope B as 0, setting the initial value of the baseline intercept C as the minimum value of the pulse signal, setting the initial value of the central wavelength of the pulse signal in the signal to be processed as the central wavelength of the grating, and setting the width of the pulse signal in the signal to be processed as the full width at half maximum of the grating.

[0025] By setting the initial values of the various parameters of the mathematical model, this application speeds up the convergence rate and improves the calculation efficiency.

[0026] Optionally, after the step of performing prediction and before the step of performing sampling, it further includes:

[0027] Third acquisition: Obtain the acquisition frequency of the photodetector and the refractive index of the grating;

[0028] Calculation of quantity: Determine the sampling quantity of the signal to be processed, and the calculation model is as follows:

[0029] ;

[0030] Wherein, m is the sampling quantity; c is the speed of light; is the period of the laser signal generated by the DBR laser; is the refractive index of the grating;

[0031] Calculation of length: Calculate the value of the sampling length, and the calculation model is as follows:

[0032] ;

[0033] Wherein, d is the sampling length;

[0034] Length update: Update the sampling length to the preset length.

[0035] The present application first obtains the period of the laser signal generated by the DBR laser and the refractive index of the grating in the optical path, and then obtains the number of samples of the signal to be processed. The number of samples in the sampling step is determined according to the period of the laser signal generated by the DBR laser, realizing the dimensional alignment between signal acquisition and the laser signal generated by the DBR laser. Subsequently, the sampling length is calculated and updated to a preset length. In the sampling step, the sampling strategy can be customized according to the physical characteristics and processing requirements of the signal, thereby improving the accuracy and effectiveness of sampling, which helps the subsequent signal processing and analysis to be based on high-quality and representative sampling data, thus enhancing the performance and reliability of the entire signal processing process.

[0036] Optionally, after performing the sampling step and before performing the fitting step, it further includes:

[0037] Fourth acquisition: Acquire background noise data and calculate the average value of the background noise data using a statistical analysis algorithm and standard deviation ;

[0038] Set threshold: Set a coarse screening threshold, and the calculation model of the coarse screening threshold is as follows:

[0039] ;

[0040] where D is the coarse screening threshold; k is a coefficient, and the value range is 3 - 5;

[0041] Traversal: Traverse the signal to be processed, and obtain the region where the signal value of the pulse signal in the signal to be processed is greater than the coarse screening threshold, denoted as the target region;

[0042] Signal clipping: Obtain the sampling point corresponding to the maximum signal value in the target region, denoted as the first sampling point. Take the first sampling point as the clipping center and clip the signal to be processed in the target region with a preset window size to obtain the clipped signal to be processed, and update the clipped signal to be processed as the signal to be processed.

[0043] This application first collects background noise data and uses a statistical analysis algorithm to calculate the mean and standard deviation of the background noise data. By calculating the mean and standard deviation, the distribution of the noise can be understood, providing a basis for setting a reasonable threshold. Subsequently, this application sets a coarse screening threshold. By setting the coarse screening threshold, the areas in the signal that may contain useful information can be initially screened out, excluding most of the noise. The setting of the threshold takes into account the statistical characteristics of the background noise and is adjusted by a coefficient k to adapt to different noise levels and signal processing requirements. Subsequently, the signal to be processed is traversed to obtain the areas where the signal values of the pulse signals in the signal to be processed are greater than the coarse screening threshold, which are recorded as target areas. By traversing the signal to be processed, the areas where the signal values exceed the coarse screening threshold are found, and these areas contain useful pulse signals. Subsequently, this application obtains the sampling point corresponding to the maximum signal value within the target area, which is recorded as the first sampling point. Taking the first sampling point as the cropping center, the signal to be processed within the target area is cropped with a preset window size to obtain the cropped signal to be processed, and the cropped signal to be processed is updated as the signal to be processed. The cropping operation can further focus on the part of the signal that is most likely to contain useful information. Cropping with the sampling point with the maximum signal value as the center ensures that the cropped signal segment contains the main part of the pulse signal, and thus can more accurately extract and analyze the useful information in the signal.

[0044] In a second aspect, this application provides a computer-readable storage medium, on which a computer program is stored. When the processor processes the computer program, it is used to implement the method described above.

[0045] In a third aspect, this application provides a signal processing system based on a DBR laser, adopting the following technical solutions:

[0046] A signal processing system based on a DBR laser includes a memory and a processor.

[0047] The memory stores the computer-readable storage medium described above.

[0048] When the processor processes the computer program stored on the computer-readable storage medium, it is used to implement the method described above.

[0049] Optionally, the system further includes:

[0050] A DBR laser, which is used to generate a laser signal according to a control current; and is also used to modulate the laser signal according to modulation parameters and emit a modulated signal.

[0051] A first current control source, which is communicatively connected to the DBR laser and is used to send a control current to the DBR laser according to a current control signal.

[0052] A voltage control source, communicatively connected to the DBR laser, for sending modulation parameters to the DBR laser according to a modulation control signal;

[0053] An optical path, including a grating, for reflecting the modulation signal;

[0054] A photodetector, for capturing the reflected modulation signal and converting the captured reflected modulation signal into an analog signal;

[0055] An analog-to-digital converter, communicatively connected to the photodetector and the control module respectively, for converting the analog signal into a signal to be processed;

[0056] A control module, communicatively connected to the first current control source for sending a current control signal; communicatively connected to the voltage control source for sending a modulation control signal; communicatively connected to the analog-to-digital converter for receiving and processing the signal to be processed, and outputting a signal normal or signal abnormal.

[0057] The DBR (Distributed Bragg Reflector) laser can generate a laser signal according to a control current. In addition, it can modulate the laser signal according to the modulation parameters and emit a modulation signal. The first current control source, communicatively connected to the DBR laser, sends a control current to the DBR laser according to the current control signal, achieving precise control of the generation of the laser signal, and can adjust characteristics such as the power and frequency of the laser as needed. The voltage control source is communicatively connected to the DBR laser and sends modulation parameters to the DBR laser according to the modulation control signal, realizing the modulation of the laser signal, making the laser signal narrower. The grating in the optical path reflects the modulation signal so that the subsequent photodetector can capture the signal more effectively. The photodetector can capture the reflected modulation signal and convert it into an analog signal. The analog-to-digital converter can convert the analog signal into a signal to be processed. The control module is communicatively connected to the first current control source, the voltage control source, and the analog-to-digital converter, responsible for sending the current control signal and the modulation control signal, receiving and processing the signal to be processed, and outputting a signal normal or signal abnormal. This application realizes a complete process from the generation, modulation, capture, conversion to processing of the optical signal, has a high degree of integration and flexibility, and is applicable to various application scenarios that require precise control and processing of optical signals.

[0058] Optionally, the system further includes a second current control source, which is communicatively connected to the control module and the DBR laser respectively, for sending an amplification signal to the DBR laser according to the amplification control signal of the control module, so that the DBR laser amplifies the modulation signal according to the amplification signal.

[0059] The second current control source is communicatively connected to the control module and the DBR laser respectively. It sends an amplification signal to the DBR laser according to the amplification control signal sent by the control module. After receiving the amplification signal, the DBR laser amplifies the modulation signal according to the signal. By introducing the second current control source, the present application realizes the function of amplifying the modulation signal, thereby enhancing the signal strength, improving the signal transmission distance and anti-interference ability.

[0060] In summary, the present application includes at least one of the following beneficial technical effects:

[0061] 1. The present application first obtains the signal to be processed, then samples the signal to be processed with a preset length to obtain the observed value of the signal to be processed, then establishes a mathematical model of the signal to be processed, obtains the theoretical value of the signal to be processed based on the mathematical model, calculates the residual value based on the observed value and the theoretical value, and then uses the nonlinear least squares method to iteratively update the parameters in the mathematical model until the residual value is minimized and then outputs the central wavelength of the pulse signal in the signal to be processed. By iteratively updating the model parameters, the fitting effect of the model can be continuously optimized until the minimum residual value is reached, so as to more accurately extract the central wavelength of the pulse signal in the signal to be processed. The present application uses a mathematical model to fit the signal to be processed and iteratively updates the parameters in the mathematical model by the nonlinear least squares method, realizing the precise fitting of the signal to be processed, thereby accurately extracting the central wavelength of the signal to be processed and improving the processing accuracy.

[0062] 2. The present application first obtains the training data, where the training data includes the historical processed signal and the denoised historical processed signal, and then based on the deep learning algorithm, a CNN-LSTM model is established. The CNN-LSTM model is suitable for processing complex signal data with time series dependence. Then the training data is used to train the CNN-LSTM model to obtain the trained CNN-LSTM model. Through training, the CNN-LSTM model can learn the noise pattern in the historical processed signal and the characteristics of the clean signal, so as to have the ability to denoise the new signal to be processed. Then the signal to be processed is input into the trained CNN-LSTM model to obtain the denoised signal to be processed, and the denoised signal to be processed is updated as the signal to be processed to improve the accuracy and reliability of subsequent sampling, fitting and analysis steps. Description of the Drawings

[0063] Figure 1 is the flowchart of the method of Embodiment 1 of the present application;

[0064] Figure 2 is the flowchart of the method of Embodiment 2 of the present application;

[0065] Figure 3 is the flowchart of the method of Embodiment 3 of the present application;

[0066] Figure 4 It is a schematic structural diagram of Embodiment 5 of this application. Detailed implementation manners

[0067] The following Figures 1 to 4 is used to further elaborate on this application in detail.

[0068] Embodiment 1: This embodiment discloses a signal processing method based on a DBR laser. Referring to Figure 1 , this embodiment first acquires the signal to be processed, then samples it with a preset length to obtain observation values, then establishes a mathematical model to obtain theoretical values, and calculates the residuals between the observation values and the theoretical values. The model parameters are iteratively updated through the nonlinear least squares method to minimize the residuals. Finally, the central wavelength of the pulse signal is output, and it is determined whether it is within the preset range to determine whether the signal is normal. The process of this embodiment is as follows:

[0069] S11 First acquisition, acquire the signal to be processed at the output end of the analog-to-digital converter.

[0070] S12 Sampling, sample the signal to be processed with a preset length to obtain the observation values of the signal to be processed at each sampling point.

[0071] S13 Fitting, establish a mathematical model of the signal to be processed, and obtain the theoretical value of the signal to be processed based on the mathematical model.

[0072] Based on the approximate Gaussian distribution characteristic of the grating reflection spectrum in this embodiment, combined with the least squares optimization method, high-precision parameter estimation is realized. The calculation model of the mathematical model is as follows:

[0073] ;

[0074] where is the theoretical value of the signal to be processed; A is the peak value of the pulse signal in the signal to be processed; B is the baseline slope; C is the baseline intercept; is the central wavelength of the pulse signal in the signal to be processed; is the width of the pulse signal in the signal to be processed; is the wavelength of the signal to be processed. In this embodiment, by introducing a linear baseline term to characterize background drift (such as Rayleigh scattering).

[0075] After that, set the initial value of the peak value A of the pulse signal in the signal to be processed as the maximum value of the pulse signal, set the initial value of the baseline slope B as 0, set the initial value of the baseline intercept C as the minimum value of the pulse signal, set the initial value of the central wavelength of the pulse signal in the signal to be processed as the central wavelength of the grating, and set the initial value of the width is the full width at half maximum of the grating.

[0076] In other embodiments, the calculation model of the mathematical model can also be:

[0077] 。

[0078] S14 Calculate the residuals, which are the differences between the observed values and the corresponding theoretical values at each sampling point. The magnitude of the residuals reflects the matching degree between the mathematical model and the actual signal. If the residuals are small, it indicates that the mathematical model can well describe the actual signal; if the residuals are large, it means that the mathematical model needs to be adjusted or improved.

[0079] S15 Update, using the nonlinear least squares method (such as the Levenberg-Marquardt (LM) algorithm) to iteratively update the parameters in the mathematical model. The nonlinear least squares method is an optimization algorithm used to minimize the sum of the squares of the residuals.

[0080] In this embodiment, the nonlinear least squares method is used to iteratively update the parameters of the mathematical model to minimize the residuals. The calculation model during iterative update is as follows:

[0081] ;

[0082] where , is the Jacobian matrix composed of the partial derivatives of the residuals with respect to each parameter; is the damping factor at the b-th iteration; is the residual at the b-th iteration, and its value is equal to the observed value minus the theoretical value at the b-th iteration 。

[0083] The elements of the Jacobian matrix need to be explicitly calculated. For example, the partial derivative with respect to the parameter is:

[0084] ;

[0085] For the partial derivative is:

[0086] ;

[0087] The partial derivatives of the remaining parameters can be deduced by analogy.

[0088] The iterative process will continue until the residuals reach a predetermined minimum value or converge to a stable value. At this time, it is considered that the mathematical model has accurately described the actual signal.

[0089] Output the central wavelength of the signal to be processed after the residuals reach the minimum value.

[0090] S16 Wavelength judgment: Determine whether the central wavelength is within a preset wavelength range. If so, output a normal signal; otherwise, output an abnormal signal.

[0091] In this embodiment, a mathematical model is established to fit the signal to be processed, and the theoretical value is obtained based on the mathematical model. Then, the residual between the observed value and the theoretical value is calculated, and the model parameters are iteratively updated using the nonlinear least squares method to minimize the residual. The central wavelength of the pulse signal is output. Through the above solution, this embodiment can obtain a more accurate central wavelength, and thus obtain a more accurate judgment result.

[0092] Embodiment 2: Refer to Figure 2 , the difference between this embodiment and Embodiment 1 is that after performing S11 First Acquisition and before performing S12 Sampling, it further includes:

[0093] S21 Second Acquisition: Obtain training data, where the training data includes historical processed signals and denoised historical processed signals.

[0094] Historical processed signals are signals that have been processed before, and they contain noise and other unwanted components. The denoised historical processed signals are signals after denoising processing, and they are closer to the real, noise-free signals.

[0095] The purpose of obtaining these training data is to enable the CNN-LSTM model to learn the characteristics of the signal and the pattern of the noise, so as to better denoise the signal to be processed in subsequent processing.

[0096] S22 Modeling: Based on deep learning algorithms, establish a CNN-LSTM model. CNN (Convolutional Neural Network) is good at extracting the spatial features of signals, while LSTM (Long Short-Term Memory Network) is good at processing time series data and capturing the time dependence of signals.

[0097] In the noise reduction process of weak grating demodulation, the CNN-LSTM model constructs a hierarchical noise reduction architecture by combining the local feature extraction ability of the convolutional neural network (CNN) with the temporal modeling advantage of the long short-term memory network (LSTM). In the signal noise reduction task, noise usually shows non-linear superposition (such as the mixture of impulse interference and Gaussian noise), and is highly coupled with the effective signal in the time-frequency domain. In this embodiment, the CNN-LSTM model uses the ReLU activation function to introduce non-linear transformation, enhancing the expression ability for complex noise patterns. In the LSTM stage, the gating mechanism dynamically adjusts the information flow through the combination of the Sigmoid and Tanh functions, distinguishing the short-term fluctuations caused by noise from the long-term trend of the real signal. In addition, the introduction of the Dropout layer further improves the generalization ability of the model. By randomly masking some neurons during the training process, Dropout forces the network to avoid overfitting to specific noise patterns as much as possible, thus enhancing the adaptability to unknown noise types. This regularization strategy has been proven to effectively reduce the validation set error in experiments, especially outstanding in the scenario of unbalanced noise distribution.

[0098] Through the CNN-LSTM model, this embodiment can not only extract the local features of the signal, but also obtain the temporal changes through LSTM modeling. Moreover, the parallelization of CNN will reduce the temporal calculation burden, and the number of layers and parameters of CNN and LSTM can be adjusted to adapt to different noise scenarios.

[0099] S23 Model training: Use the training data to train the CNN-LSTM model to obtain the trained CNN-LSTM model. During the training process, the CNN-LSTM model will learn the mapping relationship between the historical processed signal and the denoised historical processed signal, that is, how to convert the noisy signal into a noise-free signal.

[0100] S24 Prediction: Input the signal to be processed into the trained CNN-LSTM model. The CNN-LSTM model will perform noise reduction processing on the signal to be processed according to the previously learned mapping relationship and output the denoised signal to be processed.

[0101] First, the signal to be processed enters the CNN layer composed of one-dimensional convolution and pooling from the input layer. The local noise patterns in the signal (such as high-frequency jitter or pulse interference) are extracted layer by layer through the convolution kernel, and the redundant information is compressed. Then, the output of the convolution is downsampled by the pooling layer to retain its main features. Subsequently, the high-dimensional feature sequence output by the CNN is input into the bidirectional LSTM layer, and its gating mechanism (forget gate, input gate, output gate) is used to capture the long-range dependence relationship of the signal on the time axis, and distinguish the temporal evolution law of the real grating reflection peak and random noise. During the training process, 30% of the neurons are randomly discarded through the Dropout layer to reduce the overfitting phenomenon of the CNN-LSTM model to the noise patterns in the training data. Finally, the high-dimensional features of the LSTM are mapped back to the original signal space through the fully connected layer, and the signal to be processed after noise reduction is output. Then, the signal to be processed after noise reduction is updated as the signal to be processed.

[0102] S25 Third acquisition, obtain the acquisition frequency of the photodetector and the refractive index of the grating.

[0103] S26 Calculate the quantity, determine the sampling quantity of the signal to be processed, and the calculation model is as follows:

[0104] ;

[0105] Where m is the sampling quantity; c is the speed of light; is the acquisition frequency of the photodetector; is the refractive index of the grating, which takes the value of 1.5 in this embodiment, and its value can also be set according to requirements in other embodiments.

[0106] The sampling quantity determines how many data points need to be considered when processing the signal, that is, how many gratings are set in the optical path.

[0107] S27 Calculate the length, calculate the value of the sampling length, and the calculation model is as follows:

[0108] ;

[0109] Where d is the sampling length. The sampling length determines the interval length between each grating.

[0110] S28 Length update, update the sampling length to the preset length.

[0111] By updating the sampling length, the sampling process can meet the actual requirements, so as to obtain more accurate sampling results and subsequent processing results.

[0112] On the premise that the Bragg wavelength of the grating is the same as the wavelength of the laser signal output by the DBR laser, a wavelength dynamic change range of at least 5 nm including the Bragg wavelength is reserved to cope with the change of the central wavelength of the grating array caused by environmental changes such as temperature and strain.

[0113] For example, when the acquisition frequency of the photodetector is set to 100 MHz, if the numerical value of the acquired data volume m is 1000, then there are approximately 500,000 data within a wavelength scanning range of 5 nm. According to the formula for calculating the length in S27, it can be known that when the pulse width of the optical pulse is less than the speed of light transmitted between two adjacent gratings, the two gratings can be distinguished. If the pulse width of the optical pulse is 50 ns, the minimum interval between two adjacent gratings cannot be less than 5 meters, otherwise crosstalk will occur in the reflected light (i.e., the reflected laser signal), resulting in spectral overlap and inability to distinguish the reflected light.

[0114] In this embodiment, training data is first obtained, a model is established, the model is trained, and the denoised signal is predicted. Then, the sampling quantity and sampling length are calculated, and the sampling length is updated to a preset length, so as to effectively process the signal and improve the quality and accuracy of the signal.

[0115] Embodiment 3: According to the Neyman-Pearson criterion, the signal region can be preliminarily screened through threshold detection, and then the region of potential grating signals can be quickly located, narrowing the processing range. Refer to Figure 3 This embodiment is different from Embodiment 2 in that after performing S12 sampling and before performing S13 fitting, it further includes:

[0116] S31 Fourth acquisition, acquiring background noise data, and calculating the average value of the background noise data by using a statistical analysis algorithm and the standard deviation .

[0117] S32 Setting the threshold, setting the coarse screening threshold, and the calculation model of the coarse screening threshold is as follows:

[0118] ;

[0119] where D is the coarse screening threshold; k is a coefficient, and the value range is 3 - 5.

[0120] This step sets a coarse screening threshold for preliminarily screening the pulse signals in the signal to be processed.

[0121] By adjusting the value of k, the strictness of the coarse screening threshold can be controlled. The larger the value of k, the stricter the threshold, and the fewer the pulse signals screened out; the smaller the value of k, the looser the threshold, and the more the pulse signals screened out.

[0122] In S33 traversal, for each sampling point, if its signal value is greater than the rough screening threshold D, it is regarded as part of the potential pulse signal. Then, the regions with consecutive signal values greater than D are recorded and denoted as the target regions.

[0123] By setting an intensity threshold (such as 3 times the standard deviation higher than the background noise), mark the time-domain window that may contain the grating signal. If , then retain this region.

[0124] In S34 signal clipping, obtain the sampling point corresponding to the maximum signal value within the target region, denoted as the first sampling point. Take the first sampling point as the center or representative point of the target region, which will be used as the clipping center.

[0125] With the first sampling point as the center point, set a preset window size. Use the preset window size (in this embodiment, 10 sampling points before and after the center point are selected) to clip the signal to be processed, and obtain the clipped signal to be processed, so as to retain the pulse signal that may be contained within the target region and remove the surrounding noise or interference at the same time. In this embodiment, the rough screening threshold can significantly reduce the data dimension through screening, providing a grating focusing region for subsequent processing. Then, update the clipped signal to be processed as the signal to be processed.

[0126] Perform block processing on the signal to be processed obtained after S34 signal clipping. For example, classify the data according to the arrival time sequence of the signal to be processed (such as a total of 30 gratings, divided into groups of five gratings each), or classify according to the wavelength (such as the center wavelength interval of the spectrum is 2nm, then it can be split into intervals of 0.5nm to obtain 4 data blocks). Then, use a multi-core CPU / GPU for parallel processing and calculation. Use multi-threading (OpenMP) to allocate sub-blocks to different computing units for independent processing, and then store the independently processed data blocks into an SQL database.

[0127] In this embodiment, first collect background noise data, set the rough screening threshold, traverse the signal to be processed to obtain the target region, and then clip the signal within the target region and update the signal to be processed, so as to effectively extract the pulse signal in the signal to be processed and provide an accurate data basis for subsequent signal analysis and processing.

[0128] Embodiment 4: This embodiment provides a computer-readable storage medium, on which a computer program is stored. When the processor processes the computer program, it is used to implement the described method.

[0129] Embodiment 5: Referring to Figure 4 , this embodiment provides a signal processing system based on a DBR laser. The system includes a memory and a processor,

[0130] The computer-readable storage medium described above is stored in the memory;

[0131] When the processor processes the computer program stored on the computer-readable storage medium, the method described above is implemented.

[0132] Before signal processing in this embodiment, the DBR laser is controlled to reach the expected state. Specifically: the control module controls the DBR laser to reach the preset temperature and controls the gain current of the DBR laser to reach the preset current value. Then, the current parameters pre-stored in the parameter table are sent to the DBR laser to control the DBR laser to emit a laser signal. The system further includes:

[0133] A DBR laser, which can generate a laser signal according to the input control current by changing the carrier concentration inside the laser, thereby changing the refractive index of the gain medium. The DBR laser can also modulate the laser signal according to the input modulation parameters to emit a modulated signal to obtain a narrower laser signal.

[0134] The DBR (Distributed Bragg Reflector) laser is a special type of semiconductor laser, which is characterized by integrating a Bragg reflector in the waveguide structure of the laser.

[0135] A first current control source, communicatively connected to the DBR laser, for sending a control current to the DBR laser according to a current control signal. By adjusting the magnitude of the control current, the output characteristics of the DBR laser can be precisely controlled to meet different application requirements.

[0136] In this embodiment, multiple first current control sources are provided. One of the first current control sources sends a control current Gain to the DBR laser according to the current control signal. The control current Gain is also called the gain region control current; another first current control source sends a control current Phase to the DBR laser according to the current control signal. The control current Phase is also called the phase region control current; the remaining first current control sources send a control current Grating to the DBR laser according to the current control signal. The control current Grating is also called the grating region control current.

[0137] In other embodiments, the remaining first current control sources can also be multiple, and the sum of the control currents sent by these remaining first current control sources to the DBR laser is equal to the control current Grating.

[0138] A voltage control source, communicatively connected to the DBR laser, for sending a modulation parameter EA to the DBR laser according to a modulation control signal to obtain a laser signal with a target width.

[0139] Optical path, the optical path includes optical elements such as gratings, which are used to reflect the modulation signal. A grating is an optical element with a periodic structure that can diffract and reflect the incident light (i.e., the laser signal), thereby changing the propagation direction and characteristics of the incident light.

[0140] Photodetector, which is used to capture the reflected modulation signal and convert it into an analog signal. A photodetector is a device that can convert optical signals into electrical signals and works based on the principle of the photoelectric effect. Through the photodetector, in this embodiment, the modulation signal in the optical path can be converted into an electrical signal.

[0141] Analog-to-Digital Converter, the Analog-to-Digital Converter (ADC) is communicatively connected to the photodetector and the control module respectively, and its main function is to convert the analog signal output by the photodetector into a signal to be processed.

[0142] Control module, which is communicatively connected to the first current control source and is used to send a current control signal; communicatively connected to the voltage control source and is used to send a modulation control signal; communicatively connected to the analog-to-digital converter and is used to receive and process the signal to be processed, and output whether the signal is normal or abnormal.

[0143] Second current control source, the second current control source is communicatively connected to the control module and the DBR laser respectively, and is used to send an amplification signal SOA to the DBR laser according to the amplification control signal of the control module, so that the DBR laser amplifies the modulation signal according to the amplification signal SOA to achieve precise amplification and control of the modulation signal.

[0144] In this embodiment, through the collaborative work of components such as the DBR laser, the first current control source, the voltage control source, the optical path, the photodetector, the analog-to-digital converter, and the control module, the whole process of generating, modulating, reflecting, capturing, converting, processing, and judging the laser signal is realized. At the same time, by introducing the second current control source, the amplification and control functions of the modulation signal are also realized. This system has high flexibility and practicality and can be widely applied in fields such as optical communication, spectral analysis, and laser ranging.

[0145] The above are all preferred embodiments of this application. The protection scope of this application is not limited thereby. Therefore, all equivalent changes made according to the structure, shape, and principle of this application should be covered within the protection scope of this application.

Claims

1. A signal processing method based on a DBR laser, characterized in that, Including: First acquisition: Obtain the signal to be processed; Sampling: Sample the signal to be processed with a preset length to obtain the observed values of the signal to be processed; Fitting: Establish a mathematical model of the signal to be processed, and obtain the theoretical values of the signal to be processed based on the mathematical model; Calculate the residual: Calculate the residual value based on the observed value and the theoretical value; Update: Iteratively update the parameters in the mathematical model using the nonlinear least squares method until the residual value is minimized, and then output the central wavelength of the pulse signal in the signal to be processed; Wavelength judgment: Judge whether the central wavelength is within the preset wavelength range. If so, output that the signal is normal; Otherwise, output that the signal is abnormal; Before performing the step of sampling, the calculation method of the preset length further includes: Third acquisition: Obtain the acquisition frequency of the photodetector and the refractive index of the grating; Calculate the quantity: Determine the sampling quantity of the signal to be processed, and the calculation model is as follows: ； where m is the number of samplings; c is the speed of light; is the period for the DBR laser to generate a laser signal; is the refractive index of the grating; Calculate the length: Calculate the value of the sampling length, and the calculation model is as follows: ； where d is the sampling length; Length update: Update the sampling length to the preset length.

2. The signal processing method based on a DBR laser according to claim 1, wherein After performing the step of the first acquisition and before performing the step of sampling, it further includes: Second acquisition: Obtain training data, where the training data includes historical processed signals and denoised historical processed signals; Modeling: Based on the deep learning algorithm, establish a CNN-LSTM model; Model training: Use the training data to train the CNN-LSTM model to obtain the trained CNN-LSTM model; Prediction: Input the signal to be processed into the trained CNN-LSTM model to obtain the denoised signal to be processed, and update the denoised signal to be processed as the signal to be processed.

3. The signal processing method based on a DBR laser according to claim 2, wherein In the step of fitting, the calculation model of the mathematical model is as follows: ； Among them, is the theoretical value of the signal to be processed; A is the peak value of the pulse signal in the signal to be processed; B is the baseline slope; C is the baseline intercept; is the central wavelength of the pulse signal in the signal to be processed; is the width of the pulse signal in the signal to be processed; is the wavelength of the signal to be processed.

4. The signal processing method based on a DBR laser according to claim 3, wherein The steps of the fitting further include: setting the initial value of the peak A of the pulse signal in the signal to be processed as the maximum value of the pulse signal, setting the initial value of the baseline slope B as 0, setting the initial value of the baseline intercept C as the minimum value of the pulse signal, and setting the center wavelength of the pulse signal in the signal to be processed as the center wavelength of the grating, and setting the width of the pulse signal in the signal to be processed as the full width at half maximum of the grating.

5. The signal processing method based on a DBR laser according to claim 1, wherein After performing the step of sampling and before performing the step of fitting, it further includes: Fourth collection: Collect background noise data and calculate the average value of the background noise data using a statistical analysis algorithm and standard deviation ; Set the threshold: Set the coarse screening threshold, and the calculation model of the coarse screening threshold is as follows: ； where D is the coarse screening threshold; k is a coefficient, and the value range is 3 - 5; Traverse: Traverse the signal to be processed, and obtain the area where the signal value of the pulse signal in the signal to be processed is greater than the coarse screening threshold, and record it as the target area; Signal clipping: Obtain the sampling point corresponding to the maximum signal value in the target area, record it as the first sampling point, use the first sampling point as the clipping center, and clip the signal to be processed in the target area with a preset window size to obtain the clipped signal to be processed, and update the clipped signal to be processed as the signal to be processed.

6. A computer-readable storage medium, on which a computer program is stored, characterized in that, When the processor processes the computer program, it is used to implement the method according to any one of claims 1 - 5.

7. A signal processing system based on a DBR laser, characterized in that, The system includes a memory and a processor, The memory stores a computer-readable storage medium as described in claim 6; When the processor processes the computer program stored on the computer-readable storage medium, it is used to implement the method according to any one of claims 1 - 5.

8. The signal processing system based on a DBR laser according to claim 7, wherein The system further includes: A DBR laser, which is used to generate a laser signal according to a control current; and is also used to modulate the laser signal according to modulation parameters and emit a modulated signal; A first current control source, which is communicatively connected to the DBR laser and is used to send a control current to the DBR laser according to a current control signal; A voltage control source, communicatively connected to the DBR laser, for sending modulation parameters to the DBR laser according to a modulation control signal; An optical path, including a grating, for reflecting the modulation signal; A photodetector, for capturing the reflected modulation signal and converting the captured reflected modulation signal into an analog signal; An analog-to-digital converter, communicatively connected to the photodetector and the control module respectively, for converting the analog signal into a signal to be processed; A control module, communicatively connected to a first current control source for sending a current control signal; communicatively connected to the voltage control source for sending a modulation control signal; communicatively connected to the analog-to-digital converter for receiving and processing the signal to be processed and outputting that the signal is normal or abnormal.

9. The signal processing system based on a DBR laser according to claim 8, wherein The system further includes a second current control source, communicatively connected to the control module and the DBR laser respectively, for sending an amplification signal to the DBR laser according to the amplification control signal of the control module, so that the DBR laser amplifies the modulation signal according to the amplification signal.

Images