Signal demodulation method and apparatus, electronic device, storage medium, and product

Abstract

Embodiments of the present application provide a signal demodulation method and device, electronic equipment, storage medium and product. The method comprises: receiving a signal demodulation request, the signal demodulation request comprising a modulated signal of noise carrier communication; inputting the modulated signal into a radiometer according to the signal demodulation request to obtain a voltage waveform signal, the voltage waveform signal being used to represent the instantaneous power of the modulated signal through average voltage; obtaining a target response characteristic parameter of an integrator of the radiometer; inputting the voltage waveform signal and the target response characteristic parameter into a pre-trained deep neural network model to obtain a demodulated signal, the deep neural network model being used to dynamically adjust the weight of a feature in the voltage waveform signal according to the target response characteristic parameter, so as to compensate for the feature affected by the response characteristic of the integrator in the voltage waveform signal. Through the pre-trained deep neural network model, the distortion caused by the response characteristic of the integrator of the radiometer is compensated for, thereby improving the accuracy of signal demodulation.

Description

Technical Field

[0001] This application relates to the field of communication technology, and in particular to a signal demodulation method, apparatus, electronic device, storage medium, and product. Background Technology

[0002] In noisy carrier communication technology, the signal transmitter uses random noise as a carrier to generate a modulated signal from a bit code sequence to transmit information. The receiver demodulates the modulated signal to obtain a demodulated signal. Noisy carrier communication technology has the advantages of low power spectral density and strong anti-interference capability.

[0003] In related technologies, the receiving end uses a radiometer to convert the modulated signal into a continuous voltage waveform signal. A decision is made based on the voltage waveform signal to demodulate the modulated signal, obtaining the demodulated signal.

[0004] However, when the modulation signal frequency is too high, the integrator, the core component of the radiometer, cannot respond quickly to signal changes, resulting in distortion of the voltage waveform signal output by the radiometer, which in turn leads to low demodulation accuracy. Summary of the Invention

[0005] This application provides signal demodulation methods, apparatus, electronic devices, storage media, and products to improve the accuracy of signal demodulation.

[0006] In a first aspect, embodiments of this application provide a signal demodulation method, comprising: receiving a signal demodulation request, the signal demodulation request including a modulated signal of noise carrier communication; inputting the modulated signal into a radiometer according to the signal demodulation request to obtain a voltage waveform signal, the voltage waveform signal being used to represent the instantaneous power of the modulated signal by means of an average voltage; obtaining target response characteristic parameters of the integrator of the radiometer; inputting the voltage waveform signal and the target response characteristic parameters into a pre-trained deep neural network model to obtain a demodulated signal, the deep neural network model being used to dynamically adjust the weights of features in the voltage waveform signal according to the target response characteristic parameters to compensate for the influence of the integrator response characteristics on the features in the voltage waveform signal.

[0007] In one possible implementation, obtaining the target response characteristic parameters of the integrator of the radiometer includes: determining the bit error rate of the historical demodulated signal; in response to the bit error rate being less than a bit error rate threshold, obtaining the target response characteristic parameters of the integrator through a first acquisition method, wherein the first acquisition method includes at least one of the following: automatic input through the configuration information storage path of the radiometer, or parameter acquisition through a manual setting interface; in response to the bit error rate being greater than or equal to the bit error rate threshold, obtaining the target response characteristic parameters of the integrator through a testing method.

[0008] In one possible implementation, the target response characteristic parameters of the integrator are obtained through testing, including: injecting a test signal into the integrator through a calibration signal source and acquiring the output response of the integrator to obtain a response waveform signal; and performing fitting processing on the response waveform signal to obtain the target response characteristic parameters.

[0009] In one possible implementation, the voltage waveform signal and the target response characteristic parameters are input into a pre-trained deep neural network model to obtain a demodulated signal, including: analyzing and processing the voltage waveform signal to obtain signal features corresponding to the voltage waveform signal, wherein the signal features are modulation rate features and / or signal-to-noise ratio; determining a target deep neural network model adapted to the signal features, wherein the target deep neural network model is obtained by model training based on sample data corresponding to the signal features; and inputting the voltage waveform signal and the target response characteristic parameters into the target deep neural network model to obtain the demodulated signal.

[0010] In one possible implementation, inputting the voltage waveform signal and the target response characteristic parameters into the target deep neural network model to obtain the demodulated signal includes: determining a target sliding window based on the data length of the sample data of the target deep neural network model; performing sliding segmentation on the voltage waveform signal according to the target sliding window to obtain multiple signal segments, each signal segment having the same length as the target sliding window; and inputting the multiple signal segments and the target response characteristic parameters into the target deep neural network model to obtain the demodulated signal.

[0011] In one possible implementation, the method further includes: determining a sample database, the sample database including multiple sample data corresponding to multiple distortion feature waveforms; obtaining multiple first sample data from the sample database; and performing model training based on the multiple first sample data to obtain the deep neural network model.

[0012] In one possible implementation, the method further includes: determining a multidimensional parameter space, wherein the dimensions of the multidimensional parameter space include at least the response characteristic parameters of the integrator, the theoretical symbol length of the noisy carrier communication, and the channel signal-to-noise ratio; sampling from the multidimensional parameter space to obtain multiple sets of parameter combinations; performing simulation processing based on the multiple sets of parameter combinations to obtain multiple second sample data; and generating the sample database based on the multiple second sample data.

[0013] In one possible implementation, for any set of parameter combinations, simulation processing is performed based on the parameter combinations to obtain second sample data, including: determining the original bit code sequence; generating an original baseband waveform based on the original bit code sequence and the theoretical symbol length; establishing a forward distortion model of the integrator based on the response characteristic parameters in the parameter combination; simulating the original baseband waveform through the forward distortion model to obtain a theoretical distortion waveform; adding noise to the theoretical distortion waveform based on the channel signal-to-noise ratio in the parameter combination to obtain sample waveform data; and constructing the second sample data using the sample waveform data as input and the original bit code sequence as a label.

[0014] Secondly, embodiments of this application provide a signal demodulation apparatus, comprising: a receiving module for receiving a signal demodulation request, the signal demodulation request including a modulated signal of noise carrier communication; a measurement module for inputting the modulated signal into a radiometer according to the signal demodulation request to obtain a voltage waveform signal, the voltage waveform signal being used to represent the instantaneous power of the modulated signal by means of an average voltage; an acquisition module for acquiring target response characteristic parameters of the integrator of the radiometer; and a demodulation module for inputting the voltage waveform signal and the target response characteristic parameters into a pre-trained deep neural network model to obtain a demodulated signal, the deep neural network model being used to dynamically adjust the weights of features in the voltage waveform signal according to the target response characteristic parameters to compensate for the influence of the integrator response characteristics on the features in the voltage waveform signal.

[0015] In one possible implementation, the acquisition module is specifically used to determine the bit error rate of the historical demodulated signal; the acquisition module is further specifically used to, in response to the bit error rate being less than a bit error rate threshold, acquire the target response characteristic parameters of the integrator through a first acquisition method, the first acquisition method including at least one of the following: automatic input through the configuration information storage path of the radiometer, or parameter acquisition through a manual setting interface; the acquisition module is further specifically used to, in response to the bit error rate being greater than or equal to the bit error rate threshold, acquire the target response characteristic parameters of the integrator through a testing method.

[0016] In one possible implementation, the acquisition module is specifically used to inject a test signal into the integrator through a calibration signal source and to acquire the output response of the integrator to obtain a response waveform signal; the acquisition module is also specifically used to perform fitting processing on the response waveform signal to obtain the target response characteristic parameters.

[0017] In one possible implementation, the demodulation module is specifically configured to analyze and process the voltage waveform signal to obtain signal features corresponding to the voltage waveform signal, wherein the signal features are modulation rate features and / or signal-to-noise ratio; the demodulation module is further configured to determine a target deep neural network model adapted to the signal features, wherein the target deep neural network model is obtained by model training based on sample data corresponding to the signal features; the demodulation module is further configured to input the voltage waveform signal and the target response characteristic parameters into the target deep neural network model to obtain the demodulated signal.

[0018] In one possible implementation, the demodulation module is specifically configured to determine a target sliding window based on the data length of the sample data of the target deep neural network model; the demodulation module is further configured to perform sliding segmentation on the voltage waveform signal according to the target sliding window to obtain multiple signal segments, the length of each signal segment being the same as the length of the target sliding window; the demodulation module is further configured to input the multiple signal segments and the target response characteristic parameters into the target deep neural network model to obtain the demodulated signal.

[0019] In one possible implementation, the apparatus further includes: a training module for determining a sample database, the sample database including multiple sample data corresponding to multiple distortion feature waveforms; the training module is further configured to obtain multiple first sample data from the sample database; the training module is further configured to perform model training based on the multiple first sample data to obtain the deep neural network model.

[0020] In one possible implementation, the apparatus further includes: a generation module, configured to determine a multidimensional parameter space, wherein the dimensions of the multidimensional parameter space include at least the response characteristic parameters of the integrator, the theoretical symbol length of the noisy carrier communication, and the channel signal-to-noise ratio; the generation module is further configured to sample from the multidimensional parameter space to obtain multiple sets of parameter combinations; the generation module is further configured to perform simulation processing based on the multiple sets of parameter combinations to obtain multiple second sample data; and the generation module is further configured to generate the sample database based on the multiple second sample data.

[0021] In one possible implementation, the generation module is specifically configured to determine the original bit code sequence and generate an original baseband waveform based on the original bit code sequence and the theoretical symbol length; the generation module is further configured to establish a forward distortion model of the integrator based on the response characteristic parameters in the parameter combination; the generation module is further configured to simulate the original baseband waveform through the forward distortion model to obtain a theoretical distortion waveform; the generation module is further configured to add noise to the theoretical distortion waveform based on the channel signal-to-noise ratio in the parameter combination to obtain sample waveform data; the generation module is further configured to construct the second sample data using the sample waveform data as input and the original bit code sequence as a label.

[0022] Thirdly, embodiments of this application provide a signal demodulation device, including: a memory and a processor;

[0023] The memory stores computer-executed instructions;

[0024] The processor executes computer execution instructions stored in the memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.

[0025] Fourthly, embodiments of this application provide a non-volatile computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible implementations of the first aspect.

[0026] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the first aspect and / or various possible implementations of the first aspect.

[0027] The signal demodulation method, apparatus, electronic device, storage medium, and product provided in this application include: receiving a signal demodulation request, the signal demodulation request including a modulated signal of noisy carrier communication; inputting the modulated signal into a radiometer according to the signal demodulation request to obtain a voltage waveform signal, the voltage waveform signal being used to represent the instantaneous power of the modulated signal through an average voltage; obtaining target response characteristic parameters of the integrator of the radiometer; inputting the voltage waveform signal and the target response characteristic parameters into a pre-trained deep neural network model to obtain a demodulated signal, the deep neural network model being used to dynamically adjust the weights of features in the voltage waveform signal according to the target response characteristic parameters to compensate for the influence of the integrator response characteristics on the voltage waveform signal. This solution, through a pre-trained deep neural network model, compensates for the distortion caused by the response characteristics of the radiometer's integrator, thereby improving the accuracy of signal demodulation. Attached Figure Description

[0028] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0029] Figure 1 This is a schematic diagram illustrating an application scenario of a signal demodulation method provided in an embodiment of this application.

[0030] Figure 2 A schematic diagram of the integrator structure provided in the embodiments of this application;

[0031] Figure 3 A schematic diagram illustrating the output pattern of the integrator provided in an embodiment of this application;

[0032] Figure 4 This is a schematic diagram of voltage waveform signal distortion provided in an embodiment of this application;

[0033] Figure 5 A schematic flowchart illustrating a signal demodulation method provided in an embodiment of this application;

[0034] Figure 6 A flowchart illustrating another signal demodulation method provided in an embodiment of this application;

[0035] Figure 7 A schematic diagram illustrating the signal demodulation results provided in an embodiment of this application;

[0036] Figure 8 A schematic diagram of the deep neural network model structure provided in the embodiments of this application;

[0037] Figure 9 A schematic diagram of a waveform segment provided in an embodiment of this application;

[0038] Figure 10 This is a schematic diagram of the structure of a signal demodulation device provided in an embodiment of this application;

[0039] Figure 11 This is a schematic diagram of another signal demodulation device provided in an embodiment of this application;

[0040] Figure 12 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0041] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0042] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0043] In this application embodiment, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0044] It should be noted that the phrase "at...time" in the embodiments of this application can refer to the instant at which a certain situation occurs, or to a period of time after the occurrence of a certain situation; the embodiments of this application do not specifically limit this. Furthermore, the display interface provided in the embodiments of this application is merely an example, and the display interface may include more or less content.

[0045] It should be noted that the signal demodulation method, apparatus, electronic device, storage medium and product of this application can be used in the field of communication technology, or in any field other than communication. The application field of the signal demodulation method, apparatus, electronic device, storage medium and product of this application is not limited.

[0046] Figure 1 This is a schematic diagram illustrating an application scenario of a signal demodulation method provided in an embodiment of this application. The following scenario is illustrated: the transmitting end sends a modulated signal to the receiving end; the radiometer at the receiving end processes the modulated signal; and the receiving end then makes a decision to obtain a demodulated signal, thereby achieving information transmission.

[0047] For example, in noisy carrier communication, the transmitting end modulates a bit code sequence onto a noisy carrier to obtain a modulated signal, and the receiving end demodulates the bit code sequence from the modulated signal through mechanisms such as correlation detection or energy detection. The noisy carrier has statistical characteristics similar to random noise (e.g., wide bandwidth or flat power spectrum).

[0048] With a scenario example, because the modulated signal diffuses its energy on a wideband noisy carrier, its power spectral density is low, making it difficult to identify in background noise, thus reducing the probability of interception. For non-target receivers, the modulated signal is indistinguishable from natural noise, further reducing the probability of identification. The wideband characteristics of the noisy carrier also make it resistant to narrowband interference, thus effectively mitigating interference.

[0049] For example, in noisy carrier communication, the receiver includes a radiometer. The radiometer measures the average power of a radio frequency signal within a specific frequency band. Specifically, the radiometer selects the specific frequency band containing the desired modulation signal from the wide spectrum received by the antenna. The radiometer squares the filtered modulation signal and outputs a DC component proportional to the modulation signal power. The radiometer performs time averaging on the DC component, filtering out high-frequency fluctuations, and outputs a smooth voltage waveform signal proportional to the modulation signal power. The voltage waveform signal is an analog voltage quantity characterizing the instantaneous average power of the received modulation signal after being measured and converted by the radiometer.

[0050] For example, the core component of a radiometer is an integrator. The integrator performs time averaging on the modulated signal, suppressing rapid fluctuations and extracting slowly changing trends to output a voltage waveform signal. The integrator's response speed is determined by its time constant. ,in, R represents the response speed, C represents the resistance, and C represents the capacitance.

[0051] Below, in conjunction with Figure 2 The structure of the integrator is explained.

[0052] Figure 2 This is a schematic diagram of the integrator structure provided in an embodiment of this application. Figure 2 As shown, the integrator is a first-order low-pass filter, which can be implemented using a resistor-capacitor circuit. Input voltage. The input represents the signal source (e.g., a modulated signal). A resistor R is located on the input path to limit current. A capacitor C is connected in series with the resistor, with the other end grounded to provide a reference potential; the capacitor is used for the output voltage and to store charge. Output voltage. The voltage waveform signal is drawn from terminal 2 of the capacitor.

[0053] Optionally, the mathematical model of the integrator theory can be expressed by the following formula:

[0054]

[0055] in, U represents the output voltage value of the integrator at time t, U represents the amplitude of the step voltage input to the integrator (i.e., the sudden change in power level of the RF signal input to the integrator when the transmitter switches from transmitting a "0" signal to transmitting a "1" signal), and t represents any time.

[0056] Below, in conjunction with Figure 3 Explain the output pattern of the integrator.

[0057] Figure 3 This is a schematic diagram illustrating the output pattern of the integrator provided in an embodiment of this application. For example... Figure 3 As shown, the horizontal axis The normalized time axis represents the integrator output voltage, which starts from 0 and approaches a stable value U. Over time, the output voltage follows an exponential curve. The signal rises in a predictable pattern, with the initial rise rate being the fastest, then slowing down until it approaches its maximum value U. In high-speed communication (where the modulation signal frequency is too high), the symbol period can approach or even be less than τ. This means that when the bit representing "1" ends and the next bit begins, the output voltage is far from reaching a stable high level U, and may then begin to slowly decline due to the arrival of the next "0" bit. This repeated occurrence of a decline before the rise is complete leads to distortion of the radiometer's output voltage waveform.

[0058] Below, in conjunction with Figure 4 Explain the distortion of voltage waveform signals.

[0059] Figure 4 This is a schematic diagram illustrating voltage waveform signal distortion provided in an embodiment of this application. Figure 4 As shown, (a) represents the original modulated signal, exhibiting high-frequency, irregular voltage variations. (b)-(f) represent the same original modulated signal after passing through different time constants. The output waveform after processing by the integrator. Refer to (b), when When 1 / Rb (where Rb represents the bit rate) is large, the integrator's response speed is slow. When the modulation signal frequency is too high, the integrator cannot complete a full charging or discharging process within a single bit cycle, and the voltage waveform signal exhibits slow rising and falling edges instead of an ideal square wave. Refer to (c)-(f), as... As the distortion decreases, the integrator's response speed increases, the voltage waveform signal gradually becomes sharper, the distortion gradually decreases, and it gradually approaches the original modulation signal.

[0060] For example, during signal demodulation, the received modulated signal is mapped into a discrete bit code sequence (0 or 1) according to certain rules or algorithms. Specifically, in noisy carrier communication, a decision threshold voltage is calculated. The voltage value of the voltage waveform signal at each moment is compared with the decision threshold; values ​​higher than the threshold are judged as "1", and values ​​lower than the threshold are judged as "0".

[0061] In related technologies, when the frequency of the modulation signal is too high, the response speed of the radiometer integrator cannot keep up with the signal change. The radiometer integrator causes waveform distortion, and the boundary between high and low levels becomes blurred, making the calculated decision threshold unreliable. This leads to inaccurate decision results and consequently, low demodulation accuracy.

[0062] The signal demodulation method provided in this application aims to solve the above-mentioned technical problems of the prior art.

[0063] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0064] Figure 5 This is a flowchart illustrating a signal demodulation method provided in an embodiment of this application. The method includes the following steps:

[0065] S501, Receive signal demodulation request, the signal demodulation request includes the modulated signal of the noise carrier communication.

[0066] The entity executing this application may be the signal receiving end.

[0067] For example, the modulated signal is the signal that the receiver needs to demodulate.

[0068] With the example of the scenario, the signal demodulation request is used to trigger the demodulation of the modulated signal of the noisy carrier communication to obtain the demodulated signal. The demodulated signal is the bit code sequence of the transmitting end, which represents the information sent by the transmitting end.

[0069] S502. According to the signal demodulation request, the modulated signal is input to the radiometer to obtain a voltage waveform signal. The voltage waveform signal is used to represent the instantaneous power of the modulated signal through the average voltage.

[0070] For example, a radiometer is used to measure the power of a modulated signal.

[0071] For example, the radiometer uses an internal square-law detector and integrator to convert the input radio frequency signal into a changing voltage signal, i.e., a voltage waveform signal, as the output. The voltage value of the voltage waveform signal directly corresponds to the instantaneous magnitude of the signal power of the modulated signal. Therefore, the voltage waveform signal is essentially a waveform signal representing the change in the power of the received signal.

[0072] In related technologies, voltage waveform signals are directly used for decision-making to obtain demodulated signals. However, when the modulation signal frequency is too high, the radiometer cannot respond quickly to signal changes, leading to distortion of the voltage waveform signal. Using distorted voltage waveform signals for decision-making results in inaccurate demodulated signals.

[0073] S503. Obtain the target response characteristic parameters of the radiometer's integrator.

[0074] For example, the integrator is a core component of the radiometer. Target response characteristic parameters are used to quantify the integrator's response characteristics. Response characteristics refer to the speed and manner in which the integrator responds to the input signal.

[0075] Optionally, the target response characteristic parameter can be the integrator's time constant. The integrator's time constant represents how fast the integrator responds. A larger time constant results in a slower integrator response, requiring a longer time for the integrator's output voltage to approach the target value, which can easily lead to waveform distortion. A smaller time constant results in a faster integrator response, allowing the integrator's output voltage to quickly keep up with changes in the input modulation signal.

[0076] With the help of scenario examples, the target response characteristics are determined by the physical structure of the integrator, and the target response characteristic parameters reflect the physical root cause of radiometer distortion.

[0077] S504. Input the voltage waveform signal and target response characteristic parameters into the pre-trained deep neural network model to obtain the demodulated signal. The deep neural network model is used to dynamically adjust the weights of features in the voltage waveform signal according to the target response characteristic parameters in order to compensate for the influence of the integrator response characteristics on the voltage waveform signal.

[0078] For example, the target response characteristic parameters are used as prior knowledge and input into a deep neural network model along with the voltage waveform signal. The deep neural network model determines the cause of distortion based on the prior knowledge and performs targeted compensation based on the cause of distortion, thereby reducing distortion. The deep neural network model then makes a decision on the compensated voltage waveform signal to obtain an accurate demodulated signal.

[0079] For example, a deep neural network model performs two actions: compensation and decision.

[0080] To illustrate with a scenario example, if the time constant is large, the deep neural network model will recognize that the current distortion pattern mainly consists of slow rising or falling edges and plateaus with decreasing flatness. Therefore, it will pay special attention to these specific morphological features in the voltage waveform signal. In this way, the deep neural network model can accurately identify the distortion pattern generated by the time constant on the waveform, perform targeted compensation in its internal calculations, and ultimately determine the bit code sequence most likely to produce the distortion pattern, thus outputting a high-precision demodulated signal.

[0081] The signal demodulation method provided in this application embodiment receives a signal demodulation request, which includes a modulated signal of noisy carrier communication; according to the signal demodulation request, the modulated signal is input to a radiometer to obtain a voltage waveform signal, which is used to represent the instantaneous power of the modulated signal through an average voltage; the target response characteristic parameters of the radiometer's integrator are obtained; the voltage waveform signal and the target response characteristic parameters are input to a pre-trained deep neural network model to obtain a demodulated signal, and the deep neural network model is used to dynamically adjust the weights of features in the voltage waveform signal according to the target response characteristic parameters to compensate for the influence of the integrator response characteristics on the voltage waveform signal. This scheme, through a pre-trained deep neural network model, compensates for the distortion caused by the response characteristics of the radiometer's integrator, thereby improving the accuracy of signal demodulation.

[0082] Based on any of the above embodiments, the following, in conjunction with Figure 6 The detailed process of signal demodulation is explained.

[0083] Figure 6 This is a schematic flowchart illustrating another signal demodulation method provided in an embodiment of this application. Figure 6 As shown, the method includes:

[0084] S601, Receive signal demodulation request, the signal demodulation request includes the modulated signal of the noise carrier communication.

[0085] It should be noted that the execution process of S601 is the same as that of S501, and will not be repeated here.

[0086] S602. According to the signal demodulation request, the modulated signal is input into the radiometer to obtain a voltage waveform signal. The voltage waveform signal is used to represent the instantaneous power of the modulated signal through the average voltage.

[0087] It should be noted that the execution process of S602 is the same as that of S502, and will not be repeated here.

[0088] S603. Determine the bit error rate of the historical demodulated signal.

[0089] The bit error rate is the ratio of the number of bits that erroneous to the total number of bits transmitted.

[0090] For example, the historical demodulated signal is the demodulated signal generated in the past. The bit error rate represents the demodulation performance of the historical demodulated signal. The historical demodulated signal can be used to determine whether the current demodulation strategy needs to be optimized.

[0091] S604. In response to the bit error rate being less than the bit error rate threshold, the target response characteristic parameters of the integrator are obtained through the first acquisition method.

[0092] The first acquisition method includes at least one of the following: automatic input through the radiometer's configuration information storage path, or parameter acquisition through a manual setting interface.

[0093] For example, the bit error rate threshold is a threshold for judging the quality of demodulation performance.

[0094] For example, if the bit error rate is less than the bit error rate threshold, it means that the demodulation performance of the integrator meets the standard, that is, the response characteristic parameters used in the past match the actual response characteristics of the integrator, and the existing response characteristic parameters can be used.

[0095] For example, the configuration information storage path stores historically used response characteristic parameters, allowing for quick retrieval of target response characteristic parameters. The manual setting interface allows maintenance personnel to quickly configure target response characteristic parameters via software or hardware DIP switches.

[0096] With scenario examples illustrating these points, both automatic input and manual setting are low-cost, high-efficiency acquisition methods. When the bit error rate meets the target, acquiring the target response characteristic parameters using the first acquisition method can effectively improve signal demodulation efficiency.

[0097] S605. Inject a test signal into the integrator through a calibration signal source and collect the output response of the integrator to obtain the response waveform signal.

[0098] With the help of scenario examples, it can be explained that the integrator may experience temperature deviation and aging during use, which will cause changes in the actual response characteristic parameters of the integrator. Online measurement is used to obtain accurate target response characteristic parameters.

[0099] For example, the test signal is a signal with known characteristics. For instance, the test signal could be a step signal or a pulse signal.

[0100] Optionally, the test signal contains rich frequency components to fully stimulate the dynamic characteristics of the integrator, and the theoretical response model of the test signal is known (e.g., the exponential response of a step), thus enabling accurate analysis of the integrator.

[0101] Optionally, the calibration signal source can be a dedicated hardware circuit or a software-programmable waveform generated by a digital-to-analog converter.

[0102] For example, the response waveform signal is the waveform signal actually output by the integrator for the test signal.

[0103] S606. Fit the response waveform signal to obtain the target response characteristic parameters.

[0104] For example, the response waveform signal is compared with the theoretical mathematical model of the integrator, and the model parameters that match the response waveform signal are found through optimization algorithms, which are the target response characteristic parameters.

[0105] Optionally, the theoretical mathematical model of the integrator is a first-order linear model. The model is fitted using optimization algorithms such as nonlinear least squares to obtain the time constant in the theoretical mathematical model, which is then determined as the target response characteristic parameter.

[0106] Based on the above implementation method, when the bit error rate of the historical demodulated signal does not meet the standard, the target response characteristic parameters that accurately reflect the current state of the integrator can be obtained by conducting actual measurements using known test signals, thereby improving the accuracy of signal demodulation.

[0107] S607. Analyze and process the voltage waveform signal to obtain the signal characteristics corresponding to the voltage waveform signal. The signal characteristics are the modulation rate characteristics and / or signal-to-noise ratio.

[0108] For example, the modulation rate characteristic is the bit rate (or symbol rate) of the current communication, which directly determines the severity of distortion. It can be obtained by analyzing the density of transition edges or the duration of plateau periods in the voltage waveform signal.

[0109] For example, the signal-to-noise ratio (SNR) reflects the noise level of the current communication, and the noise level determines how much attention a deep neural network model needs to devote to suppressing noise. It can be determined by analyzing the fluctuation amplitude of background noise in the voltage waveform signal or the stability of the signal plateau.

[0110] S608. Determine the target deep neural network model that is adapted to the signal features. The target deep neural network model is obtained by training the model based on the sample data corresponding to the signal features.

[0111] For example, a set of deep neural network models are pre-trained, each of which is optimized for a specific modulation rate range and / or signal-to-noise ratio interval.

[0112] With scenario examples, it becomes clear that a single deep neural network model cannot accurately cover all possible scenarios during training, leading to performance compromises in extreme situations. By training multiple deep neural network models based on signal characteristics, each model can focus solely on the target scenario. This results in cleaner training data, simpler tasks, and improved targeting for each model, thereby enhancing the accuracy of signal demodulation.

[0113] For example, high-rate dedicated models focus on learning how to compensate for severe distortions (such as stepped waveforms). Low signal-to-noise ratio dedicated models focus on learning how to suppress noise interference (such as filtering and feature enhancement).

[0114] S609. Input the voltage waveform signal and target response characteristic parameters into the target deep neural network model to obtain the demodulated signal.

[0115] One feasible implementation method is to obtain the demodulated signal by: determining the target sliding window based on the data length of the sample data of the target deep neural network model; performing sliding segmentation on the voltage waveform signal according to the target sliding window to obtain multiple signal segments, the length of each signal segment being the same as the length of the target sliding window; and inputting the multiple signal segments and target response characteristic parameters into the target deep neural network model to obtain the demodulated signal.

[0116] For example, each deep neural network model is trained using fixed-length sample data segments. A target sliding window is determined based on the length of the sample data to match the dimension of the target deep neural network model.

[0117] For example, each signal segment is grouped with the same target response characteristic parameter and input into a target deep neural network model. The target deep neural network model demodulates each segment independently and outputs the corresponding bit code sequence. The target deep neural network concatenates the demodulation results of all segments in chronological order to form a complete demodulated signal.

[0118] Below, in conjunction with Figure 7 Explain the signal demodulation results.

[0119] Figure 7 This is a schematic diagram illustrating the signal demodulation results provided in an embodiment of this application. Figure 7 As shown, the modulated signal exhibits a slowly changing, continuous waveform superimposed with severe high-frequency noise. This is caused by the combined effect of the integrator's response characteristics and channel noise, making it impossible to accurately determine a clear high-low level boundary from the modulated signal. The demodulated signal obtained through processing with a deep neural network model has a clean and regular waveform, allowing for accurate determination of the high-low level boundary and thus reconstruction of the bit code sequence.

[0120] Optionally, based on the integrator's response characteristic parameters and the theoretical symbol length of the noisy carrier communication, the minimum complete acquisition length is calculated, and multiple waveform segments are determined from the voltage waveform signal based on the minimum complete acquisition length. These waveform segments are then converted to a fixed length required by the target deep neural network model through interpolation or downsampling.

[0121] Optionally, the target deep neural network model can be compressed into a lightweight model and deployed on edge devices to achieve low-power demodulation. Specifically, the target deep neural network model is compressed and optimized into a lightweight model, which is then deployed on an edge computing device acting as a signal receiver. The compression optimization includes at least pruning and quantization of the model to significantly reduce its parameter count, computational complexity, and memory footprint. The edge computing device runs the lightweight model locally, performing real-time, online demodulation of the received modulated signal, thus avoiding uploading the modulated signal to a remote location and achieving low-power, low-latency demodulation.

[0122] In this feasible implementation, the target sliding window is determined based on the data length of the sample data so that the signal segment matches the dimension of the target deep neural network model, thereby improving the accuracy of the output of the target deep neural network and thus improving the accuracy of signal demodulation.

[0123] One feasible implementation method for signal demodulation may further include: determining a sample database, which includes multiple sample data corresponding to multiple distortion feature waveforms; obtaining multiple first sample data from the sample database; and training a model based on the multiple first sample data to obtain a deep neural network model.

[0124] For example, during model training, sample data for training is obtained from a sample database.

[0125] For example, each deep neural network model corresponds to a distortion feature waveform.

[0126] Below, in conjunction with Figure 8 The structure of deep neural network models is explained.

[0127] Figure 8 This is a schematic diagram of the deep neural network model structure provided in an embodiment of this application. Figure 8As shown, high-discrimination features (such as rising edge slope or plateau width) are automatically extracted from the waveform of the first sample data through convolution and pooling operations. Based on the response characteristic parameters in the first sample data, the network weights are dynamically adjusted to enhance the focus on waveform features related to the response characteristic parameters and suppress noise and irrelevant interference. An adaptive weight adjustment mechanism is used to compensate for external perturbations such as temperature drift, and the demodulated result is finally output through a fully connected layer. Training is repeated using multiple first sample data sets to improve the accuracy of the deep neural network model.

[0128] Optionally, the training process uses the PyTorch framework, Adam optimizer, and cross-entropy loss function to optimize the model, and saves the trained parameters for real-time demodulation.

[0129] In this feasible implementation, obtaining the first sample data through the constructed sample database can improve the training efficiency of the deep neural network model.

[0130] One feasible implementation method for signal demodulation may further include: determining a multidimensional parameter space, wherein the dimensions of the multidimensional parameter space include at least the response characteristic parameters of the integrator, the theoretical symbol length of the noisy carrier communication, and the channel signal-to-noise ratio; sampling from the multidimensional parameter space to obtain multiple sets of parameter combinations; performing simulation processing based on the multiple sets of parameter combinations to obtain multiple second sample data; and generating a sample database based on the multiple second sample data.

[0131] For example, response characteristic parameters determine the core hardware parameters for distortion identification. The theoretical symbol length of noisy carrier communication determines the signal change rate. The channel signal-to-noise ratio determines the degree of interference from environmental noise.

[0132] For example, multiple sets of parameter combinations are obtained by uniform and unbiased sampling, ensuring that the multiple sets of parameter combinations cover all possible extreme operating conditions and transition states.

[0133] For example, through simulation processing, the response of the integrator is simulated for each set of parameter combinations to obtain the second sample data corresponding to each set of parameter combinations.

[0134] Optionally, the voltage waveform signal in the second sample data is segmented to obtain waveform segments, which are then processed into image format and stored in the sample database. These image-formatted waveform segments allow for the use of image recognition capabilities of the convolutional kernels in deep neural networks to extract image features during model training, enriching the training content and thus improving the accuracy of the deep neural network model.

[0135] Below, in conjunction with Figure 9 Provide an explanation of the waveform segment.

[0136] Figure 9 This is a schematic diagram of a waveform segment provided in an embodiment of this application. For example... Figure 9 As shown, the voltage waveform signal is truncated according to different waveform characteristics to obtain multiple waveform segments, and the multiple waveform segments cover multiple distortion features.

[0137] In this feasible implementation, by simulating various parameter combinations separately, multiple second sample data can cover multiple distortion features, thereby training corresponding deep neural network models for each of the multiple distortion features to improve the accuracy of the deep neural network models.

[0138] One feasible implementation method involves simulation processing as follows: determining the original bit code sequence; generating the original baseband waveform based on the original bit code sequence and the theoretical symbol length; establishing a forward distortion model of the integrator based on the response characteristic parameters in the parameter combination; simulating the original baseband waveform using the forward distortion model to obtain the theoretical distortion waveform; adding noise to the theoretical distortion waveform based on the channel signal-to-noise ratio in the parameter combination to obtain sample waveform data; and constructing second sample data using the sample waveform data as input and the original bit code sequence as a label.

[0139] For example, the original baseband waveform refers to the digital signal waveform without any distortion. It is the most direct and simplest representation of the original bit code sequence when converted into an electrical signal. The duration of each symbol in the original baseband waveform corresponds to the theoretical length of the symbol.

[0140] For example, a corresponding positive distortion model is constructed based on the response characteristic parameters so that the positive distortion model covers the physical characteristics of the integrator.

[0141] For example, the original baseband waveform is calculated using a forward distortion model to obtain a pure theoretical distortion waveform caused solely by the integrator response characteristics. The theoretical distortion waveform accurately demonstrates how the response characteristic parameters distort the original baseband waveform.

[0142] For example, based on the theoretically distorted waveform, noise in the real channel environment is simulated by the channel signal-to-noise ratio, and the noise is superimposed on the theoretically distorted waveform to make the obtained sample waveform data more consistent with the actual application scenario.

[0143] For example, the original bit code sequence is used as a label so that the second sample data can be accurately used for supervised training.

[0144] In this feasible implementation, the second sample data generated by combining multiple dimensions in the parameter combination can provide pure and reliable learning material for the deep neural network model, enabling the deep neural network model to accurately learn the rules of distortion itself, thereby improving the accuracy of the deep neural network model and thus improving the accuracy of signal demodulation.

[0145] Figure 10 This is a schematic diagram of a signal demodulation device provided in an embodiment of this application. Figure 10 As shown, the signal demodulation device 100 may include: a receiving module 101, a measurement module 102, an acquisition module 103, and a demodulation module 104.

[0146] The receiving module 101 is used to receive a signal demodulation request, which includes the modulated signal of the noise carrier communication.

[0147] The measurement module 102 is used to input the modulated signal into the radiometer according to the signal demodulation request to obtain a voltage waveform signal. The voltage waveform signal is used to represent the instantaneous power of the modulated signal by means of the average voltage.

[0148] The acquisition module 103 is used to acquire the target response characteristic parameters of the radiometer's integrator.

[0149] The demodulation module 104 is used to input the voltage waveform signal and the target response characteristic parameters into a pre-trained deep neural network model to obtain a demodulated signal. The deep neural network model is used to dynamically adjust the weights of features in the voltage waveform signal according to the target response characteristic parameters in order to compensate for the influence of the integrator response characteristics on the voltage waveform signal.

[0150] Optionally, the receiving module 101 can perform... Figure 5 S501 in the embodiment.

[0151] Optionally, the measurement module 102 can perform... Figure 5 S502 in the embodiment.

[0152] Optionally, module 103 can be executed. Figure 5 S503 in the embodiment.

[0153] Optionally, demodulation module 104 can execute... Figure 5 S504 in the embodiment.

[0154] It should be noted that the signal demodulation device shown in the embodiments of this application can execute the technical solution shown in the above method embodiments, and its implementation principle and beneficial effects are similar, so they will not be described again here.

[0155] In one possible implementation, the acquisition module 103 is specifically used for:

[0156] Determine the bit error rate of the historical demodulated signal;

[0157] In response to a bit error rate less than a bit error rate threshold, the target response characteristic parameters of the integrator are obtained through a first acquisition method, which includes at least one of the following: automatic input through the configuration information storage path of the radiometer, or parameter acquisition through a manual setting interface;

[0158] In response to a bit error rate greater than or equal to the bit error rate threshold, the target response characteristic parameters of the integrator are obtained through testing.

[0159] In one possible implementation, the acquisition module 103 is specifically used for:

[0160] A test signal is injected into the integrator by a calibration signal source, and the output response of the integrator is acquired to obtain the response waveform signal.

[0161] The target response characteristic parameters are obtained by fitting the response waveform signal.

[0162] In one possible implementation, the demodulation module 104 is specifically used for:

[0163] The voltage waveform signal is analyzed and processed to obtain the signal characteristics corresponding to the voltage waveform signal. The signal characteristics are the modulation rate characteristics and / or signal-to-noise ratio.

[0164] Determine the target deep neural network model that is adapted to the signal features. The target deep neural network model is obtained by training the model based on the sample data corresponding to the signal features.

[0165] The voltage waveform signal and target response characteristic parameters are input into the target deep neural network model to obtain the demodulated signal.

[0166] In one possible implementation, the demodulation module 104 is specifically used for:

[0167] The target sliding window is determined based on the data length of the sample data of the target deep neural network model;

[0168] Based on the target sliding window, the voltage waveform signal is divided into multiple signal segments, and the length of each signal segment is the same as the length of the target sliding window.

[0169] Multiple signal segments and target response characteristic parameters are input into the target deep neural network model to obtain the demodulated signal.

[0170] Figure 11 This is a schematic diagram of another signal demodulation device provided in an embodiment of this application. Figure 10 Based on the illustrated embodiments, as Figure 11As shown, the signal demodulation device 100 also includes a training module 105 and a generation module 106.

[0171] Training module 105 is used for:

[0172] Define the sample database, which includes multiple sample data corresponding to multiple distortion feature waveforms;

[0173] Retrieve multiple first sample data from the sample database;

[0174] A deep neural network model is obtained by training the model using multiple first sample data.

[0175] Module 106 is used for:

[0176] Determine the multidimensional parameter space. The dimensions of the multidimensional parameter space should include at least the response characteristic parameters of the integrator, the theoretical symbol length of the noisy carrier communication, and the channel signal-to-noise ratio.

[0177] Multiple sets of parameter combinations are obtained by sampling from a multidimensional parameter space;

[0178] Multiple sets of parameter combinations are used for simulation processing to obtain multiple second sample data;

[0179] A sample database is generated based on multiple second sample data.

[0180] In one possible implementation, the generation module 106 is specifically used for:

[0181] Determine the original bit code sequence, and generate the original baseband waveform based on the original bit code sequence and the theoretical length of the symbol;

[0182] Establish a positive distortion model of the integrator based on the response characteristic parameters in the parameter combination;

[0183] The original baseband waveform is simulated using a positive distortion model to obtain the theoretical distortion waveform;

[0184] Based on the channel signal-to-noise ratio in the parameter combination, noise is added to the theoretically distorted waveform to obtain sample waveform data.

[0185] Using sample waveform data as input and the original bit code sequence as a label, construct the second sample data.

[0186] Figure 12 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application, such as... Figure 12 As shown, the electronic device includes:

[0187] The electronic device includes a processor 291 and a memory 292; it may also include a communication interface 293 and a bus 294. The processor 291, memory 292, and communication interface 293 can communicate with each other via the bus 294. The communication interface 293 can be used for information transmission. The processor 291 can invoke logical instructions stored in the memory 292 to execute the methods of the above embodiments.

[0188] Furthermore, the logic instructions in the aforementioned memory 292 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium.

[0189] The memory 292, as a non-volatile computer-readable storage medium, can be used to store software programs and computer-executable programs, such as program instructions / modules corresponding to the methods in the embodiments of this application. The processor 291 executes functional applications and data processing by running the software programs, instructions, and modules stored in the memory 292, that is, it implements the methods in the above-described method embodiments.

[0190] The memory 292 may include a program storage area and a data storage area. The program storage area may store the operating system and application programs required for at least one function; the data storage area may store data created based on the use of the terminal device. Furthermore, the memory 292 may include high-speed random access memory and may also include non-volatile memory.

[0191] This application provides a non-volatile computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the method as described in the foregoing embodiments.

[0192] This application provides a computer program product, including a computer program that, when executed by a processor, implements the method as described in the foregoing embodiments.

[0193] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to this application.

[0194] It should be further noted that although the steps in the flowchart are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowchart may include multiple sub-steps or multiple stages, which do not necessarily complete at the same time, but can be executed at different times. The execution order of these sub-steps or stages is also not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0195] It should be understood that the above-described device embodiments are merely illustrative, and the device of this application can also be implemented in other ways. For example, the division of units / modules in the above embodiments is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple units, modules, or components may be combined, or integrated into another system, or some features may be ignored or not executed.

[0196] Furthermore, unless otherwise specified, the functional units / modules in the various embodiments of this application can be integrated into one unit / module, or each unit / module can exist physically separately, or two or more units / modules can be integrated together. The integrated units / modules described above can be implemented in hardware or as software program modules.

[0197] When the integrated unit / module is implemented in hardware, the hardware can be digital circuits, analog circuits, etc. The physical implementation of the hardware structure includes, but is not limited to, transistors, memristors, etc. The processor can be any suitable hardware processor, such as CPU, GPU, FPGA, DSP, and ASIC. The storage unit can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc.

[0198] If the integrated unit / module is implemented as a software program module and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0199] In the above embodiments, the descriptions of each embodiment have their own emphasis. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments. The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0200] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the claims.

[0201] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A signal demodulation method, characterized in that, include: Receive a signal demodulation request, the signal demodulation request including the modulated signal of the noise carrier communication; According to the signal demodulation request, the modulated signal is input to the radiometer to obtain a voltage waveform signal, which is used to represent the instantaneous power of the modulated signal by means of average voltage; Obtain the target response characteristic parameters of the integrator of the radiometer; The voltage waveform signal and the target response characteristic parameters are input into a pre-trained deep neural network model to obtain a demodulated signal. The deep neural network model is trained based on multiple first sample data in a sample database, which includes multiple sample data corresponding to multiple distortion feature waveforms. The deep neural network model is used to extract waveform features from the voltage waveform signal through convolution and pooling operations, and dynamically adjust the weights of the features in the voltage waveform signal according to the target response characteristic parameters to compensate for the influence of the integrator response characteristics on the voltage waveform signal.

2. The method according to claim 1, characterized in that, Obtaining the target response characteristic parameters of the integrator of the radiometer includes: Determine the bit error rate of the historical demodulated signal; In response to the bit error rate being less than the bit error rate threshold, the target response characteristic parameters of the integrator are obtained through a first acquisition method, the first acquisition method including at least one of the following: automatic input through the configuration information storage path of the radiometer, or parameter acquisition through a manual setting interface; In response to the bit error rate being greater than or equal to the bit error rate threshold, the target response characteristic parameters of the integrator are obtained through testing.

3. The method according to claim 2, characterized in that, The target response characteristic parameters of the integrator are obtained through testing, including: A test signal is injected into the integrator by a calibration signal source, and the output response of the integrator is acquired to obtain the response waveform signal. The target response characteristic parameters are obtained by fitting the response waveform signal.

4. The method according to claim 1, characterized in that, The voltage waveform signal and the target response characteristic parameters are input into a pre-trained deep neural network model to obtain a demodulated signal, including: The voltage waveform signal is analyzed and processed to obtain the signal characteristics corresponding to the voltage waveform signal, wherein the signal characteristics are modulation rate characteristics and / or signal-to-noise ratio; Determine a target deep neural network model that is compatible with the signal features, wherein the target deep neural network model is obtained by training the model based on the sample data corresponding to the signal features; The voltage waveform signal and the target response characteristic parameters are input into the target deep neural network model to obtain the demodulated signal.

5. The method according to claim 4, characterized in that, The voltage waveform signal and the target response characteristic parameters are input into the target deep neural network model to obtain the demodulated signal, including: The target sliding window is determined based on the data length of the sample data of the target deep neural network model; Based on the target sliding window, the voltage waveform signal is divided into multiple signal segments, and the length of each signal segment is the same as the length of the target sliding window. The multiple signal segments and the target response characteristic parameters are input into the target deep neural network model to obtain the demodulated signal.

6. The method according to claim 5, characterized in that, The method further includes: A multidimensional parameter space is determined, wherein the dimensions of the multidimensional parameter space include at least the response characteristic parameters of the integrator, the theoretical symbol length of the noisy carrier communication, and the channel signal-to-noise ratio; Multiple sets of parameter combinations are obtained by sampling from the multidimensional parameter space; Simulation processing is performed based on the multiple sets of parameter combinations to obtain multiple second sample data; The sample database is generated based on the multiple second sample data.

7. The method according to claim 6, characterized in that, For any set of parameter combinations, simulation processing is performed based on the parameter combinations to obtain second sample data, including: Determine the original bit code sequence, and generate the original baseband waveform based on the original bit code sequence and the theoretical length of the symbol; A positive distortion model of the integrator is established based on the response characteristic parameters in the parameter combination; The original baseband waveform is simulated using the positive distortion model to obtain the theoretical distortion waveform; Based on the channel signal-to-noise ratio in the parameter combination, noise is added to the theoretical distortion waveform to obtain sample waveform data; The second sample data is constructed using the sample waveform data as input and the original bit code sequence as a label.

8. A signal demodulation device, characterized in that, include: A receiving module is configured to receive a signal demodulation request, wherein the signal demodulation request includes a modulated signal for noise carrier communication; The measurement module is used to input the modulated signal into the radiometer according to the signal demodulation request to obtain a voltage waveform signal, which is used to represent the instantaneous power of the modulated signal by means of an average voltage. The acquisition module is used to acquire the target response characteristic parameters of the integrator of the radiometer; The demodulation module is used to input the voltage waveform signal and the target response characteristic parameters into a pre-trained deep neural network model to obtain a demodulated signal. The deep neural network model is obtained by training a model based on multiple first sample data in a sample database, which includes multiple sample data corresponding to multiple distortion feature waveforms. The deep neural network model is used to dynamically adjust the weights of features in the voltage waveform signal according to the target response characteristic parameters to compensate for the influence of the integrator's response characteristics on the voltage waveform signal.

9. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the method as described in any one of claims 1-7.

10. A non-volatile computer-readable storage medium, characterized in that, The non-volatile computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-7.

11. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method of any one of claims 1-7.

Images