A method and system for radio signal enhancement based on multipath fading channel modeling

Abstract

The present application relates to the technical field of signal processing and data enhancement, in particular to a radio signal enhancement method and system based on multipath fading channel modeling. First, according to the multipath fading channel characteristic parameters, the real channel environment is simulated by combining FFT and IFFT, and the channel simulation signal is generated; then the channel simulation signal and the original modulation signal are generated into a training set and a test set according to a set proportion, the training set is input into the ResNet model for training, and the signal enhancement model is obtained; the test set is input into the signal enhancement model for signal enhancement until the enhanced signal meets the set condition, and the signal enhancement model is output, so that the subsequent radio signal enhancement is carried out by using the signal enhancement model. According to the multipath fading channel characteristic parameters, various conditions of the real physical channel are simulated, the channel simulation signal and the original modulation signal generated by simulation are used as samples for model construction, and the performance of the modulation classification system in the non-ideal channel environment can be improved.

Description

Technical Field

[0001] This invention relates to the field of signal processing and data enhancement technology, and in particular to a radio signal enhancement method and system based on multipath fading channel modeling. Background Technology

[0002] Automatic Modulation Classification (AMC) is a key technology in wireless communication receivers, widely used in military communications, cognitive radio, spectrum monitoring, and interference identification. However, the performance of deep learning models is highly dependent on the richness and representativeness of the training data. Traditional datasets are generally constructed based on ideal or simplified channel conditions, lacking simulation of complex real-world wireless channel environments (such as multipath fading, time-varying distortion, and shadowing effects). This leads to a significant drop in model performance and insufficient generalization ability in actual deployments, and data augmentation is the core means to compensate for this deficiency. Some works have attempted to use simple methods such as Gaussian white noise or Rayleigh fading for simulated signal enhancement, but most methods lack systematic channel modeling capabilities and cannot accurately express key channel parameters such as multipath quantity, delay distribution, path gain, and noise level.

[0003] Multipath fading is a common physical phenomenon in wireless channels. When a signal reaches the receiver through multiple paths, the differences in delay, amplitude, and phase between the different paths can cause severe signal distortion. Therefore, designing a radio signal enhancement method that can simulate various conditions in real physical channels has become a pressing technical problem in this field. Summary of the Invention

[0004] The purpose of this invention is to provide a radio signal enhancement method and system based on multipath fading channel modeling, so as to simulate various situations of real physical channels and thus improve the performance of modulation classification system in non-ideal channel environments.

[0005] To achieve the above objectives, the present invention provides a radio signal enhancement method based on multipath fading channel modeling, the method comprising:

[0006] Step S1: Obtain multipath fading channel characteristic parameters; the multipath fading channel characteristic parameters include: the total number of signal propagation paths L, the delay of each signal propagation path, the gain of each signal propagation path, and the phase corresponding to each signal propagation path.

[0007] Step S2: Combining FFT and IFFT, simulate the real channel environment based on the multipath fading channel characteristic parameters to generate a channel simulation signal;

[0008] Step S3: Generate a training set and a test set by combining the channel analog signal and the original modulated signal according to a set ratio. Input the training set into the ResNet model for training to obtain a signal enhancement model. Input the test set into the signal enhancement model for signal enhancement to obtain an enhanced signal.

[0009] Step S4: Determine whether the enhanced signal is greater than the set condition; if it is greater than the set condition, regenerate the multipath fading channel characteristic parameters and return to "Step S2"; if it is less than or equal to the set condition, output the signal enhancement model so that the signal enhancement model can be used for subsequent radio signal enhancement.

[0010] Optionally, the step of combining FFT and IFFT to simulate the real channel environment based on the multipath fading channel characteristic parameters and generate a channel simulation signal specifically includes:

[0011] Acquire multiple raw modulated signals;

[0012] The original modulation signals are subjected to anti-offset processing to obtain the anti-offset time domain signal corresponding to each original modulation signal;

[0013] Based on the multipath fading channel characteristic parameters, a modeling and discretization process is performed to obtain the discrete-time impulse response time-domain signal.

[0014] The anti-offset time-domain signal and the discrete-time impulse response time-domain signal corresponding to each of the original modulation signals are sequentially processed by FFT and IFFT, and complex Gaussian noise is added to obtain the channel simulation signal.

[0015] Optionally, the step of performing anti-offset processing on each of the original modulated signals to obtain an anti-offset time-domain signal corresponding to each of the original modulated signals specifically includes:

[0016] Remove the DC bias term from each of the original modulation signals to obtain the DC-debiased signal corresponding to each of the original modulation signals;

[0017] The average power is calculated based on the DC-de-DC signal corresponding to each of the original modulation signals.

[0018] The DC-de-DC signal corresponding to each of the original modulation signals is normalized based on the average power to obtain the anti-offset time-domain signal corresponding to each of the original modulation signals.

[0019] Optionally, the step of modeling and discretizing based on the multipath fading channel characteristic parameters to obtain the discrete-time impulse response time-domain signal specifically includes:

[0020] Based on the multipath fading channel characteristic parameters, multiple signal propagation paths are modeled to obtain the continuous-time impulse response of the multipath fading channel;

[0021] The continuous-time impulse response of the multipath fading channel is discretized according to the sampling rate to obtain the discrete-time impulse response time-domain signal.

[0022] Optionally, the step of sequentially performing FFT and IFFT processing on the anti-offset time-domain signals and the discrete-time impulse response time-domain signals corresponding to each of the original modulation signals, and adding complex Gaussian noise to obtain the channel analog signal, specifically includes:

[0023] The anti-offset time-domain signal and the discrete-time impulse response time-domain signal corresponding to each of the original modulation signals are zeroed up to the set number of points required for the FFT.

[0024] The zero-padded anti-offset time-domain signal and the discrete-time impulse response time-domain signal are convolved using the Fast Fourier Transform (FFT) to obtain the anti-offset frequency-domain signal and the discrete-time impulse response frequency-domain signal, respectively.

[0025] The anti-offset frequency domain signal and the discrete-time impulse response frequency domain signal are multiplied in the frequency domain and then IFFT is performed to obtain the IFFT signal.

[0026] A predetermined number of points are extracted from the IFFT signal as the effective signal, and complex Gaussian noise is added to the effective signal to obtain the channel analog signal.

[0027] The present invention also provides a radio signal enhancement system based on multipath fading channel modeling, the system comprising:

[0028] The acquisition module is used to acquire multipath fading channel characteristic parameters; the multipath fading channel characteristic parameters include: the total number of signal propagation paths L, the delay of each signal propagation path, the gain of each signal propagation path, and the phase corresponding to each signal propagation path.

[0029] The real channel environment simulation module is used to combine FFT and IFFT to simulate the real channel environment based on the multipath fading channel characteristic parameters and generate channel simulation signals.

[0030] The signal enhancement model training module is used to generate training and test sets from the channel analog signal and the original modulated signal according to a set ratio, and input the training set into the ResNet model for training to obtain the signal enhancement model; the test set is input into the signal enhancement model for signal enhancement to obtain the enhanced signal.

[0031] The judgment module is used to determine whether the enhanced signal is greater than a set condition; if it is greater than the set condition, the multipath fading channel characteristic parameters are regenerated and returned to the "real channel environment simulation module"; if it is less than or equal to the set condition, the signal enhancement model is output so that the signal enhancement model can be used for subsequent radio signal enhancement.

[0032] Optionally, the real channel environment simulation module specifically includes:

[0033] The acquisition unit is used to acquire multiple raw modulation signals;

[0034] An anti-offset time-domain signal generation unit is used to perform anti-offset processing on each of the original modulated signals to obtain an anti-offset time-domain signal corresponding to each of the original modulated signals.

[0035] The modeling and discretization unit is used to model and discretize based on the multipath fading channel characteristic parameters to obtain the discrete-time impulse response time domain signal.

[0036] The channel analog signal generation unit is used to sequentially perform FFT and IFFT processing on the anti-offset time domain signal and the discrete-time impulse response time domain signal corresponding to each of the original modulation signals, and add complex Gaussian noise to obtain the channel analog signal.

[0037] Optionally, the anti-offset time-domain signal generation unit specifically includes:

[0038] The DC bias term removal subunit is used to remove the DC bias term from each of the original modulation signals to obtain the de-DC signal corresponding to each of the original modulation signals;

[0039] An average power calculation subunit is used to calculate the average power based on the de-DC signal corresponding to each of the original modulation signals.

[0040] The normalization processing subunit is used to normalize the DC-de-DC signal corresponding to each of the original modulation signals according to the average power, so as to obtain the anti-offset time domain signal corresponding to each of the original modulation signals.

[0041] Optionally, the modeling and discretization unit specifically includes:

[0042] The modeling subunit is used to model multiple signal propagation paths based on the multipath fading channel characteristic parameters to obtain the continuous-time impulse response of the multipath fading channel.

[0043] The discretization processing subunit is used to discretize the continuous-time impulse response of the multipath fading channel according to the sampling rate to obtain the discrete-time impulse response time-domain signal.

[0044] Optionally, the channel analog signal generation unit specifically includes:

[0045] The zero-padding subunit is used to zero-padding the anti-offset time-domain signal and the discrete-time impulse response time-domain signal corresponding to each of the original modulation signals to the set number of points required by the FFT.

[0046] The FFT processing subunit is used to perform convolution calculations on the zero-padded anti-offset time-domain signal and the discrete-time impulse response time-domain signal using the Fast Fourier Transform (FFT) to obtain the anti-offset frequency-domain signal and the discrete-time impulse response frequency-domain signal.

[0047] The IFFT processing subunit is used to multiply the anti-offset frequency domain signal and the discrete-time impulse response frequency domain signal in the frequency domain and then perform IFFT to obtain the IFFT signal.

[0048] The channel analog signal generation subunit is used to extract a predetermined number of points from the IFFT signal as the effective signal, and add complex Gaussian noise to the effective signal to obtain the channel analog signal.

[0049] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0050] This invention discloses a radio signal enhancement method and system based on multipath fading channel modeling. First, by combining FFT and IFFT, a real channel environment is simulated based on multipath fading channel characteristic parameters to generate a simulated channel signal. Then, the simulated channel signal and the original modulated signal are used to generate training and testing sets according to a set ratio. The training set is input into a ResNet model for training to obtain a signal enhancement model. The testing set is input into the signal enhancement model for signal enhancement until the enhanced signal meets set conditions, at which point the signal enhancement model is output, enabling subsequent radio signal enhancement using the model. This invention simulates various real physical channel conditions based on multipath fading channel characteristic parameters, using the simulated channel signal and the original modulated signal as samples for model construction, thereby improving the performance of the modulation classification system in non-ideal channel environments. Attached Figure Description

[0051] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0052] Figure 1 This is a flowchart of a radio signal enhancement method based on multipath fading channel modeling according to an embodiment of the present invention;

[0053] Figure 2 This is a schematic diagram illustrating the multipath fading channel modeling principle of an embodiment of the present invention.

[0054] Figure 3 This is a diagram of the ResNet model structure used in an embodiment of the present invention. Detailed Implementation

[0055] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] The purpose of this invention is to provide a radio signal enhancement method and system based on multipath fading channel modeling, so as to simulate various situations of real physical channels and thus improve the performance of modulation classification system in non-ideal channel environments.

[0057] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0058] like Figure 1 As shown, this invention discloses a radio signal enhancement method based on multipath fading channel modeling, the method comprising:

[0059] Step S1: Obtain multipath fading channel characteristic parameters; the multipath fading channel characteristic parameters include: the total number of signal propagation paths L, the delay of each signal propagation path, the gain of each signal propagation path, and the phase of each signal propagation path;

[0060] Step S2: Combining FFT and IFFT, simulate the real channel environment based on the multipath fading channel characteristic parameters to generate a channel simulation signal;

[0061] Step S3: Generate a training set and a test set by combining the channel analog signal and the original modulated signal according to a set ratio. Input the training set into the ResNet model for training to obtain a signal enhancement model. Input the test set into the signal enhancement model for signal enhancement to obtain an enhanced signal.

[0062] Step S4: Determine whether the enhanced signal is greater than the set condition; if it is greater than the set condition, regenerate the multipath fading channel characteristic parameters and return to "Step S2"; if it is less than or equal to the set condition, output the signal enhancement model so that the signal enhancement model can be used for subsequent radio signal enhancement.

[0063] The following is a detailed discussion of each step:

[0064] Step S2: Combining FFT and IFFT, simulate the real channel environment based on the multipath fading channel characteristic parameters to generate a channel simulation signal, specifically including:

[0065] Step S21: Acquire multiple raw modulation signals; in this embodiment, a uniform sampling rate is used to acquire multiple raw modulation signals. Taking the RML2016.10a dataset as an example, the sampling rate is 200kHz. Each raw modulation signal is in complex baseband I / Q form, and the specific formula is as follows:

[0066]

[0067] in, , These represent the nth original modulation signal. The same-direction components and orthogonal components, where n is a positive integer.

[0068] Step S22: Perform anti-offset processing on each original modulation signal to obtain the anti-offset time-domain signal corresponding to each original modulation signal, specifically including:

[0069] Step S221: Remove the DC bias term from each original modulation signal to obtain the de-DC signal corresponding to each original modulation signal. The specific calculation formula is as follows:

[0070]

[0071]

[0072]

[0073] in, , These represent the nth original modulation signal. The same-direction components and orthogonal components, and These represent the nth original modulation signal. Corresponding DC signal The sequence consists of co-directional and orthogonal components, where N represents the total length of the sequence, and here N = 1024. Represents the k-th original modulation signal The I channel (in the same direction) component, Represents the k-th original modulation signal The Q-channel (orthogonal) components. Mentioned here... , Both contain both DC and AC components.

[0074] In this embodiment, for better subsequent description, the signal obtained after removing the DC bias term will be referred to as the de-DC signal.

[0075] Step S222: Calculate the average power based on the DC-DC de-signal corresponding to each original modulation signal. The specific calculation formula is as follows:

[0076]

[0077] Where P represents average power. Represents the nth original modulation signal The corresponding DC signal, where N represents the total length of the sequence. and These represent DC signals respectively. The same-direction components and orthogonal components.

[0078] Step S223: Normalize the DC-de-DC signal corresponding to each original modulation signal based on the average power to obtain the anti-offset time-domain signal corresponding to each original modulation signal. The specific formula is as follows:

[0079]

[0080] in, Represents the nth original modulation signal The corresponding anti-offset time domain signal, Represents the nth original modulation signal The corresponding DC signal, P represents the target power, and P represents the average power. This represents a small constant to prevent division by zero.

[0081] In this embodiment, the original modulation signal is acquired at a uniform sampling rate to ensure that data from different sources have a consistent time scale. The DC bias of each original modulation signal is removed, which can effectively avoid energy shift in subsequent power normalization and convolution calculation, and keep the modulation characteristics unaffected by low-frequency drift. After removing the DC component of each original modulation signal, the amplitude of the de-DC signal is normalized to a uniform power level. This makes the calculation of signal-to-noise ratio more accurate and controllable when adding noise or passing through the channel.

[0082] This invention establishes multiple independent signal propagation paths, each with independent multipath fading channel characteristics such as amplitude attenuation, propagation delay, and phase. These multipath fading channel characteristics are used to reflect the multipath effects of signals in scenarios such as building reflection and ground reflection. The multipath fading channel modeling principle is as follows: Figure 2 As shown, from Figure 2It is clear that the core of multipath fading channels is that there are multiple propagation paths for the signal from the transmitter to the receiver, and the differences in signal characteristics of each path collectively lead to the fading effect of the received signal. Signals mainly reach the receiver through three types of paths: Direct Path (LOS): The signal propagates directly from the transmitter to the receiver, which is usually the path with the shortest propagation delay and least attenuation; Reflected Path: The signal reaches the receiver after being reflected by obstacles (such as buildings or the ground). Figure 2 It includes path 1, which is reflected by "obstacles", and path 2, which is reflected by "ground"; Scattering path: The signal reaches the receiver after being scattered by scattering objects (such as particles in the air or small obstacles), and will undergo multiple scatterings during the propagation process.

[0083] The signal along each path undergoes three key changes independently: Amplitude attenuation: Different paths have different propagation distances and number of reflections / scatterings, resulting in varying degrees of signal amplitude attenuation (e.g., a direct path attenuates by a certain amount). The reflection path 1 attenuates to The reflection path 2 attenuates to The scattering path attenuates to Phase shift: Signals undergo phase changes during propagation, reflection / scattering (e.g., the phase of a direct path is...). The phase of reflection path 1 is The phase of reflection path 2 is The scattering path phase is Time delay: Due to the different path lengths, the time it takes for the signal to reach the receiver varies (e.g., the delay of a direct path is...). The delay of reflection path 1 is The delay of reflection path 2 is The scattering path delay is ).

[0084] Step S23: Based on the characteristic parameters of the multipath fading channel, model and discretize the signal to obtain the discrete-time impulse response in the time domain. Specific steps include:

[0085] Step S231: Model multiple signal propagation paths based on the multipath fading channel characteristic parameters to obtain the continuous-time impulse response (CIR) of the multipath fading channel. The specific formula is as follows:

[0086]

[0087] in, This represents the continuous-time impulse response of a multipath fading channel. The total number of signal propagation paths. Let i be the gain of the i-th signal propagation path. The delay of the i-th signal propagation path, The phase corresponding to the i-th signal propagation path. Let t be the Dirac function, and t represent the time dimension parameter.

[0088] In this embodiment, the total number of signal propagation paths Supports presets based on scenarios (e.g., L is larger in densely populated urban areas and smaller in suburban areas). exist Internal random generation; gain of signal propagation path Follows a Rayleigh distribution; phase obey .

[0089] Step S232: Discretize the continuous-time impulse response of the multipath fading channel according to the sampling rate to obtain the discrete-time impulse response time-domain signal. The specific formula is as follows:

[0090]

[0091]

[0092]

[0093] in, It is a discrete-time impulse response time-domain signal. The total number of signal propagation paths. Let i be the gain of the i-th signal propagation path. The phase corresponding to the i-th signal propagation path. For the Dirac function, Represents discrete time series parameters. Let be the number of delay sampling points for the i-th signal propagation path. Sampling rate, The delay of the i-th signal propagation path, This represents the length of the discrete channel impulse response.

[0094] In this embodiment, the length of the discrete-time impulse response is limited to To avoid excessive computational complexity, the final output format is a complex array.

[0095] Step S24: Perform FFT and IFFT processing on the anti-offset time-domain signals and discrete-time impulse response time-domain signals corresponding to each original modulation signal in sequence, and add complex Gaussian noise to obtain the channel simulation signal. The specific steps include:

[0096] Step S241: Pad the anti-offset time-domain signal and discrete-time impulse response time-domain signal corresponding to each original modulation signal with zeros to the set number of points required for the FFT. In this invention, the number of points M required for the FFT is set to... Where N represents the total length of the sequence. This represents the length of the discrete channel impulse response.

[0097] Step S242: Use Fast Fourier Transform (FFT) to convert the zero-padded anti-offset time-domain signal. and discrete-time impulse response time-domain signal Convolution calculations are performed separately to obtain the offset-prevented frequency domain signal and the discrete-time impulse response frequency domain signal. The specific formulas are as follows:

[0098]

[0099]

[0100] in, To prevent the signal from being offset in the frequency domain, The signal is a discrete-time impulse response in the frequency domain, where M is the number of points required for the FFT. is the discrete time index, and k is the frequency domain index parameter.

[0101] Step S243: Convert the anti-offset frequency domain signal and discrete-time impulse response frequency domain signal After multiplying in the frequency domain, an IFFT is performed to obtain the IFFT signal. The specific formula is as follows:

[0102]

[0103] Wherein, IFFT stands for Inverse Fast Fourier Transform. is the IFFT signal, and m is the time-domain index parameter.

[0104] In this embodiment, the IFFT signal is the signal after fading through a multipath channel. For ease of subsequent discussion, the signal after IFFT calculation is referred to as the IFFT signal.

[0105] Step S244: Extract a predetermined number of points from the IFFT signal as the effective signal, and add complex Gaussian noise to the effective signal to obtain the channel analog signal. The specific formula is as follows:

[0106]

[0107] in, For the z-th channel analog signal, For the z-th valid signal, It is the z-th noise element with a complex Gaussian distribution. , SNR stands for Signal-to-Noise Ratio.

[0108] In this embodiment, to ensure the effectiveness of the signal, the first 1024 points are truncated as the enhanced signal, and the redundant part at the end of the convolution is removed to obtain the effective signal. In addition, the signal power is calculated according to the set signal-to-noise ratio range (SNR ∈ [0, 20] dB), and noise that satisfies a complex Gaussian distribution is generated and stored in vector form.

[0109] In this embodiment, after the signal passes through the multipath fading channel, the first 1024 points of the IFFT signal are extracted, the tail redundant points are removed to obtain the effective signal, and complex Gaussian distributed noise is generated according to the set signal-to-noise ratio range. Then, complex Gaussian noise is added to the effective signal to simulate the unavoidable thermal noise interference in the wireless channel and obtain the channel simulation signal.

[0110] Step S3: Generate a training set and a test set by combining the channel analog signal and the original modulated signal according to a set ratio. Input the training set into the ResNet model for training to obtain a signal enhancement model. Input the test set into the signal enhancement model for signal enhancement to obtain an enhanced signal.

[0111] like Figure 3As shown, the signal enhancement model in this embodiment is a convolutional neural network based on a residual network (ResNet). The signal enhancement model consists of three stages: an input signal layer, a feature extraction layer, and a classification output layer. First, the input signal layer is used for data input, and the modulated signal to be enhanced and recognized can be input later. The feature extraction layer consists of a one-dimensional convolutional layer, batch normalization (BN) + ReLU activation, max pooling, and a high-level feature extraction layer. The feature extraction layer first extracts low-order time-frequency features through a one-dimensional convolutional layer, then performs normalization and downsampling through batch normalization + ReLU activation and max pooling, and finally extracts features through the high-level feature extraction layer. Specifically, the high-level feature extraction layer consists of four residual layers: Residual Layer 1, Residual Layer 2, Residual Layer 3, and Residual Layer 4. Each residual layer contains two 3×3 convolutional feature extraction units (i.e., Conv3x3), two batch normalization (BN) operations, one ReLU activation (ReLU), and one identity mapping path (shortcut). Each residual layer is finally output through a shortcut identity mapping path that does not undergo convolution processing. The number of channels in the four residual layers increases progressively (64→64→128→256 in the example), while the temporal / sequence dimension of the four residual layers decreases progressively (32→32→16→8 in the example) to achieve multi-scale extraction from local short-term features to global high-order features. This invention avoids the gradient vanishing problem by setting a feature extraction layer to maintain high expressive power in deep networks. Following this, a classification output layer is used for output processing. Specifically, a global average pooling layer averages all sequence positions for each channel. Compared to local pooling, global pooling can perform statistical aggregation across the entire sequence. After global average pooling, a feature flattening layer expands the data into a continuous one-dimensional vector, for example, flattening (Batch, 512, 1) into (Batch, 512). The feature flattening layer does not participate in feature transformation; it only flattens the two-dimensional features into one-dimensional data for subsequent input to the fully connected layer for discrimination or enhanced prediction. The feature vector after the feature flattening layer is input to the fully connected layer to produce a fixed-dimensional output. The fully connected layer maps deep features to the final enhanced discriminant vector, for example, the output dimension is (Batch, 11), where 11 corresponds to the 11 modulation types in the RML2016.10a dataset.

[0112] This structure improves the stability of deep training through residual connections and enhances robustness to multipath fading and noise through multi-level convolution and pooling, thereby providing a more robust feature representation for subsequent signal enhancement and automatic modulation recognition.

[0113] Step S4: Determine whether the enhanced signal is greater than a set condition; if it is greater than the set condition, regenerate the multipath fading channel characteristic parameters and return to "Step S2"; if it is less than or equal to the set condition, output the signal enhancement model so that it can be used for subsequent radio signal enhancement. In this embodiment, each batch of samples is a randomly generated set of multipath fading channel characteristic parameters, and the training set and test set are composed of channel analog signals and original modulated signals in a 2:1 ratio. Furthermore, the set condition can be a threshold or a range, depending on the actual needs.

[0114] This invention receives the original modulated signal and performs DC bias removal and amplitude normalization on the input original modulated signal (i.e., complex baseband signal) to eliminate the effects of amplitude drift and DC offset, thereby providing high-quality input samples for subsequent data enhancement.

[0115] This invention constructs a parameterizable continuous-time impulse response for a multipath fading channel, supporting the setting of parameters such as the number of multipaths, delay distribution, path gain, and noise level, and can be dynamically adjusted according to preset rules or random methods to simulate the characteristics of wireless channels under different physical propagation environments.

[0116] The present invention inputs the preprocessed modulated signal into the multipath channel and performs fading processing to obtain an effective signal. Then, additive white Gaussian noise (AWGN) is superimposed on this signal to generate samples under different signal-to-noise ratio conditions, thereby expanding the coverage of the dataset.

[0117] In the automatic modulation model training process, this invention inputs a channel simulation signal into the ResNet model to start training, thereby achieving data augmentation and improving sample diversity and model robustness.

[0118] This invention can also introduce non-ideal distortion factors such as frequency offset, IQ imbalance, and nonlinear power amplifier distortion to further increase the diversity of training data, enabling the subsequently obtained signal enhancement model to have stronger generalization ability when facing hardware defects or channel mismatches in real communication systems. The distortion factors include, but are not limited to, frequency offset, IQ imbalance, and nonlinear power amplifier distortion.

[0119] In this invention, the modulation label of the original modulation signal remains unchanged to ensure the accuracy of supervised learning. When the error vector amplitude of the enhanced signal exceeds a set threshold, the system will regenerate the multipath fading channel feature parameters to avoid classification difficulties due to excessive distortion.

[0120] This invention combines a parameterizable multipath fading channel modeling method with an automatic modulation classification model training process, enabling the dynamic introduction of physical layer channel characteristics during the data augmentation stage and improving the model's recognition performance in complex and non-ideal wireless environments.

[0121] The present invention also provides a radio signal enhancement system based on multipath fading channel modeling, the system comprising:

[0122] The acquisition module is used to acquire multipath fading channel characteristic parameters; the multipath fading channel characteristic parameters include: the total number of signal propagation paths L, the delay of each signal propagation path, the gain of each signal propagation path, and the phase corresponding to each signal propagation path.

[0123] The real channel environment simulation module is used to combine FFT and IFFT to simulate the real channel environment based on the multipath fading channel characteristic parameters and generate channel simulation signals.

[0124] The signal enhancement model training module is used to generate training and test sets by taking the channel simulation signal and the original modulated signal according to a set ratio, and inputting the training set into the ResNet model for training to obtain the signal enhancement model; and inputting the test set into the signal enhancement model for signal enhancement to obtain the enhanced signal.

[0125] The judgment module is used to determine whether the enhanced signal is greater than a set condition; if it is greater than the set condition, the multipath fading channel characteristic parameters are regenerated and returned to the "real channel environment simulation module"; if it is less than or equal to the set condition, the signal enhancement model is output so that the signal enhancement model can be used for subsequent radio signal enhancement.

[0126] As an optional implementation, the real channel environment simulation module of the present invention specifically includes:

[0127] The acquisition unit is used to acquire multiple raw modulated signals.

[0128] An anti-offset time-domain signal generation unit is used to perform anti-offset processing on each of the original modulation signals to obtain an anti-offset time-domain signal corresponding to each of the original modulation signals.

[0129] The modeling and discretization unit is used to model and discretize based on the multipath fading channel characteristic parameters to obtain the discrete-time impulse response time-domain signal.

[0130] The channel analog signal generation unit is used to sequentially perform FFT and IFFT processing on the anti-offset time domain signal and the discrete-time impulse response time domain signal corresponding to each of the original modulation signals, and add complex Gaussian noise to obtain the channel analog signal.

[0131] As an optional implementation, the anti-offset time-domain signal generation unit of the present invention specifically includes:

[0132] The DC bias term removal subunit is used to remove the DC bias term from each of the original modulation signals to obtain the de-DC signal corresponding to each of the original modulation signals.

[0133] The average power calculation subunit is used to calculate the average power based on the de-DC signal corresponding to each of the original modulation signals.

[0134] The normalization processing subunit is used to normalize the DC-de-DC signal corresponding to each of the original modulation signals according to the average power, so as to obtain the anti-offset time domain signal corresponding to each of the original modulation signals.

[0135] As an optional implementation, the modeling and discretization processing unit of the present invention specifically includes:

[0136] The modeling subunit is used to model multiple signal propagation paths based on the multipath fading channel characteristic parameters to obtain the continuous-time impulse response of the multipath fading channel.

[0137] The discretization processing subunit is used to discretize the continuous-time impulse response of the multipath fading channel according to the sampling rate to obtain the discrete-time impulse response time-domain signal.

[0138] As an optional implementation, the channel analog signal generation unit of the present invention specifically includes:

[0139] The zero-padding subunit is used to zero-padding the anti-offset time-domain signal and the discrete-time impulse response time-domain signal corresponding to each of the original modulation signals to the set number of points required by the FFT.

[0140] The FFT processing subunit is used to perform convolution calculations on the zero-padded anti-offset time-domain signal and the discrete-time impulse response time-domain signal using the Fast Fourier Transform (FFT) to obtain the anti-offset frequency-domain signal and the discrete-time impulse response frequency-domain signal.

[0141] The IFFT processing subunit is used to multiply the anti-offset frequency domain signal and the discrete-time impulse response frequency domain signal in the frequency domain and then perform IFFT to obtain the IFFT signal.

[0142] The channel analog signal generation subunit is used to extract a predetermined number of points from the IFFT signal as the effective signal, and add complex Gaussian noise to the effective signal to obtain the channel analog signal.

[0143] The specific steps and formulas are the same as those in the method, and will not be repeated here.

[0144] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0145] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A radio signal enhancement method based on multipath fading channel modeling, characterized in that, The method includes: Step S1: Obtain multipath fading channel characteristic parameters; the multipath fading channel characteristic parameters include: the total number of signal propagation paths L, the delay of each signal propagation path, the gain of each signal propagation path, and the phase corresponding to each signal propagation path. Step S2: Combining FFT and IFFT, simulate the real channel environment based on the multipath fading channel characteristic parameters to generate a channel simulation signal; Step S3: Generate a training set and a test set by combining the channel analog signal and the original modulated signal according to a set ratio. Input the training set into the ResNet model for training to obtain a signal enhancement model. Input the test set into the signal enhancement model for signal enhancement to obtain an enhanced signal. Step S4: Determine whether the enhanced signal is greater than the set condition; if it is greater than the set condition, regenerate the multipath fading channel characteristic parameters and return to "Step S2"; if it is less than or equal to the set condition, output the signal enhancement model so that the signal enhancement model can be used for subsequent radio signal enhancement.

2. The radio signal enhancement method based on multipath fading channel modeling according to claim 1, characterized in that, The process of combining FFT and IFFT to simulate the real channel environment based on the multipath fading channel characteristic parameters and generate a channel simulation signal specifically includes: Acquire multiple raw modulated signals; The original modulation signals are subjected to anti-offset processing to obtain the anti-offset time domain signal corresponding to each original modulation signal; Based on the multipath fading channel characteristic parameters, a modeling and discretization process is performed to obtain the discrete-time impulse response time-domain signal. The anti-offset time-domain signal and the discrete-time impulse response time-domain signal corresponding to each of the original modulation signals are sequentially processed by FFT and IFFT, and complex Gaussian noise is added to obtain the channel simulation signal.

3. The radio signal enhancement method based on multipath fading channel modeling according to claim 2, characterized in that, The step of performing anti-offset processing on each of the original modulated signals to obtain the anti-offset time-domain signal corresponding to each of the original modulated signals specifically includes: Remove the DC bias term from each of the original modulation signals to obtain the DC-debiased signal corresponding to each of the original modulation signals; The average power is calculated based on the DC-de-DC signal corresponding to each of the original modulation signals. The DC-de-DC signal corresponding to each of the original modulation signals is normalized based on the average power to obtain the anti-offset time-domain signal corresponding to each of the original modulation signals.

4. The radio signal enhancement method based on multipath fading channel modeling according to claim 2, characterized in that, The process of modeling and discretizing based on the multipath fading channel characteristic parameters to obtain the discrete-time impulse response time-domain signal specifically includes: Based on the multipath fading channel characteristic parameters, multiple signal propagation paths are modeled to obtain the continuous-time impulse response of the multipath fading channel; The continuous-time impulse response of the multipath fading channel is discretized according to the sampling rate to obtain the discrete-time impulse response time-domain signal.

5. A radio signal enhancement method based on multipath fading channel modeling according to claim 2, characterized in that, The step of sequentially performing FFT and IFFT processing on the anti-offset time-domain signals and the discrete-time impulse response time-domain signals corresponding to each of the original modulation signals, and adding complex Gaussian noise to obtain the channel analog signal, specifically includes: The anti-offset time-domain signal and the discrete-time impulse response time-domain signal corresponding to each of the original modulation signals are zeroed up to the set number of points required for the FFT. The zero-padded anti-offset time-domain signal and the discrete-time impulse response time-domain signal are convolved using the Fast Fourier Transform (FFT) to obtain the anti-offset frequency-domain signal and the discrete-time impulse response frequency-domain signal, respectively. The anti-offset frequency domain signal and the discrete-time impulse response frequency domain signal are multiplied in the frequency domain and then IFFT is performed to obtain the IFFT signal. A predetermined number of points are extracted from the IFFT signal as the effective signal, and complex Gaussian noise is added to the effective signal to obtain the channel analog signal.

6. A radio signal enhancement system based on multipath fading channel modeling, characterized in that, The system includes: The acquisition module is used to acquire multipath fading channel characteristic parameters; the multipath fading channel characteristic parameters include: the total number of signal propagation paths L, the delay of each signal propagation path, the gain of each signal propagation path, and the phase corresponding to each signal propagation path. The real channel environment simulation module is used to combine FFT and IFFT to simulate the real channel environment based on the multipath fading channel characteristic parameters and generate channel simulation signals. The signal enhancement model training module is used to generate training and test sets from the channel analog signal and the original modulated signal according to a set ratio, and input the training set into the ResNet model for training to obtain the signal enhancement model; the test set is input into the signal enhancement model for signal enhancement to obtain the enhanced signal. The judgment module is used to determine whether the enhanced signal is greater than a set condition; if it is greater than the set condition, the multipath fading channel characteristic parameters are regenerated and returned to the "real channel environment simulation module"; if it is less than or equal to the set condition, the signal enhancement model is output so that the signal enhancement model can be used for subsequent radio signal enhancement.

7. A radio signal enhancement system based on multipath fading channel modeling according to claim 6, characterized in that, The real channel environment simulation module specifically includes: The acquisition unit is used to acquire multiple raw modulation signals; An anti-offset time-domain signal generation unit is used to perform anti-offset processing on each of the original modulated signals to obtain an anti-offset time-domain signal corresponding to each of the original modulated signals. The modeling and discretization unit is used to model and discretize based on the multipath fading channel characteristic parameters to obtain the discrete-time impulse response time domain signal. The channel analog signal generation unit is used to sequentially perform FFT and IFFT processing on the anti-offset time domain signal and the discrete-time impulse response time domain signal corresponding to each of the original modulation signals, and add complex Gaussian noise to obtain the channel analog signal.

8. A radio signal enhancement system based on multipath fading channel modeling according to claim 7, characterized in that, The anti-offset time-domain signal generation unit specifically includes: The DC bias term removal subunit is used to remove the DC bias term from each of the original modulation signals to obtain the de-DC signal corresponding to each of the original modulation signals; An average power calculation subunit is used to calculate the average power based on the de-DC signal corresponding to each of the original modulation signals. The normalization processing subunit is used to normalize the DC-de-DC signal corresponding to each of the original modulation signals according to the average power, so as to obtain the anti-offset time domain signal corresponding to each of the original modulation signals.

9. A radio signal enhancement system based on multipath fading channel modeling according to claim 7, characterized in that, The modeling and discretization unit specifically includes: The modeling subunit is used to model multiple signal propagation paths based on the multipath fading channel characteristic parameters to obtain the continuous-time impulse response of the multipath fading channel. The discretization processing subunit is used to discretize the continuous-time impulse response of the multipath fading channel according to the sampling rate to obtain the discrete-time impulse response time-domain signal.

10. A radio signal enhancement system based on multipath fading channel modeling according to claim 7, characterized in that, The channel analog signal generation unit specifically includes: The zero-padding subunit is used to zero-padding the anti-offset time-domain signal and the discrete-time impulse response time-domain signal corresponding to each of the original modulation signals to the set number of points required by the FFT. The FFT processing subunit is used to perform convolution calculations on the zero-padded anti-offset time-domain signal and the discrete-time impulse response time-domain signal using the Fast Fourier Transform (FFT) to obtain the anti-offset frequency-domain signal and the discrete-time impulse response frequency-domain signal. The IFFT processing subunit is used to multiply the anti-offset frequency domain signal and the discrete-time impulse response frequency domain signal in the frequency domain and then perform IFFT to obtain the IFFT signal. The channel analog signal generation subunit is used to extract a predetermined number of points from the IFFT signal as the effective signal, and add complex Gaussian noise to the effective signal to obtain the channel analog signal.

Images