Method and system for reducing peak-to-average power ratio of imitated radio frequency division multiplexing based on deep learning

Abstract

The invention discloses an imitation radio frequency division multiplexing peak-to-average power ratio reduction method and system based on deep learning. The method comprises the following steps: acquiring an imitation radio frequency division multiplexing emission signal; constructing a deep neural network model based on unsupervised learning, and training the deep neural network based on the simulated radio frequency division multiplexing emission signal to obtain a peak-to-average power ratio optimization model; and acquiring an imitated radio frequency division multiplexing emission signal, inputting the imitated radio frequency division multiplexing emission signal into the peak-to-average power ratio optimization model for optimization to obtain an optimization value, and combining the optimization value with the imitated radio frequency division multiplexing emission signal to reduce the peak-to-average power ratio. The method and the system for reducing the simulated radio frequency division multiplexing peak-to-average power ratio are designed based on the deep learning technology, so that the blank of deep learning in reducing the simulated radio frequency division multiplexing peak-to-average power ratio is filled, and the application of the deep learning technology in a wireless communication system is enriched.

Description

Technical Field

[0001] The present invention belongs to the field of wireless communication technology, and specifically relates to a method and system for reducing the peak-to-average power ratio of radio frequency division multiplexing based on deep learning. Background Art

[0002] With the development of science and technology, people's demand for wireless communications under high-speed mobility is gradually increasing, such as high-speed rail and vehicle networking. Therefore, the next generation of wireless communications has put forward higher requirements for wireless communications under high-speed mobility. In 4G and 5G wireless communication technologies, orthogonal frequency division multiplexing has been widely used due to its good resistance to multipath fading, high spectrum utilization, and excellent performance under linear time-invariant channels. However, in linear time-varying channels, the orthogonality of orthogonal frequency division multiplexing will drop sharply, resulting in interference between subcarriers, which greatly reduces the performance of orthogonal frequency division multiplexing. In order to achieve good performance in linear time-varying channels and meet the needs of wireless communications under high-speed mobility, many new signal modulation technologies have been proposed, and Affine Frequency Division Multiplexing (AFDM) was born in this context.

[0003] AFDM is based on the discrete affine Fourier transform (DAF), a generalization of the discrete Fourier transform (DFT). It can achieve full diversity because its linear frequency modulation (LFM) pulse parameters can adapt to the channel characteristics, thereby achieving a complete delay-Doppler representation of the channel in the DFT domain. Experiments have shown that AFDM has excellent performance in linear time-varying channels and is capable of meeting the needs of wireless communications under high-speed mobility.

[0004] As a generalization of Orthogonal Frequency Division Multiplexing (OFDM), AFDM faces the same problem as OFDM: high Peak to Average Power Ratio (PAPR). High PAPR makes it easy for signals to enter the nonlinear region of the power amplifier, resulting in nonlinear distortion, which in turn leads to increased bit error rate (BER) and spectrum leakage. To effectively reduce PAPR, numerous PAPR suppression algorithms have been proposed, such as clipping, selective mapping (SLM), and partial transmission sequence (PTS). In recent years, with the rapid development of computer computing power, deep learning technology has also developed rapidly. Deep learning technology is widely used in many fields, such as computer vision and natural language processing. Furthermore, deep learning technology has also found widespread application in wireless communications, and its potential in this field is constantly being explored. Inspired by this, the present invention applies deep learning technology to a pseudo-RFDM system to reduce its PAPR. Simulations show that the present invention effectively reduces PAPR with only a minimal loss in the BER performance of the pseudo-RFDM communication system. Summary of the Invention

[0005] The present invention aims to solve the deficiencies of the prior art and provides the following solutions:

[0006] A method for reducing peak-to-average power ratio of radio frequency division multiplexing based on deep learning, comprising the following steps:

[0007] Acquire a simulated RFDM transmission signal;

[0008] Constructing a deep neural network model based on unsupervised learning, and training the deep neural network based on the simulated radio frequency division multiplexing transmission signal to obtain a peak-to-average power ratio optimization model;

[0009] Acquire a simulated radio frequency division multiplexing transmission signal and input it into the peak-to-average power ratio optimization model for optimization to obtain an optimized value, and combine the optimized value with the simulated radio frequency division multiplexing transmission signal to reduce the peak-to-average power ratio.

[0010] Preferably, the simulated radio frequency division multiplexing transmission signal includes: a pilot and data;

[0011] The pilot signal is data known to both the sender and the receiver, and is used to estimate the channel when receiving the signal;

[0012] The data carries information that needs to be sent.

[0013] Preferably, the deep neural network model includes: an input layer, several hidden layers and an output layer;

[0014] The number of hidden layers is 8, the activation function of the hidden layer is ReLU, and the activation function of the output layer is Sigmoid;

[0015] Adam is used as the optimizer for the deep neural network model.

[0016] Preferably, the method for training the deep neural network includes:

[0017] Get the peak-to-average power ratio expression for RFDM:

[0018]

[0019] Among them, E{|s[n]| 2} represents the average power of the simulated RFDM transmitted signal, represents the peak power of the simulated RFDM transmit signal, s[n] represents the time domain signal of the simulated RFDM transmit signal after discrete affine Fourier transform, N represents the number of linear frequency modulation subcarriers, x[m] represents the simulated RFDM transmit signal, c1 and c2 represent two parameters of the simulated RFDM;

[0020] The peak-to-average power ratio expression is used as a loss function to train the deep neural network model to obtain the peak-to-average power ratio optimization model.

[0021] Preferably, the method for reducing the peak-to-average power ratio includes:

[0022] Acquire a simulated radio frequency division multiplexing transmission signal and input it into the peak-to-average power ratio optimization model for optimization to obtain an optimized value, wherein the optimized value includes: a guard interval insertion point and an active constellation extension point;

[0023] The optimized value is combined with the simulated radio frequency division multiplexing transmission signal to reduce the peak-to-average power ratio: the guard interval insertion point is embedded between the pilot and the data to prevent interference between the pilot and the data and reduce the peak-to-average power ratio; the active constellation extension point reduces the peak-to-average power ratio by finding the optimal value within the extended range.

[0024] The present invention also provides a deep learning-based RFDM peak-to-average power ratio reduction system, wherein the system applies any of the above-mentioned methods and comprises: a signal acquisition module, a model construction module, and a power ratio reduction module;

[0025] The signal acquisition module is used to acquire the simulated radio frequency division multiplexing transmission signal;

[0026] The model construction module constructs a deep neural network model based on unsupervised learning, and trains the deep neural network based on the simulated radio frequency division multiplexing transmission signal to obtain a peak-to-average power ratio optimization model;

[0027] The power ratio reduction module is used to obtain the simulated radio frequency division multiplexing transmission signal and input it into the peak-to-average power ratio optimization model for optimization to obtain an optimized value, and combine the optimized value with the simulated radio frequency division multiplexing transmission signal to reduce the peak-to-average power ratio.

[0028] Preferably, the simulated radio frequency division multiplexing transmission signal includes: a pilot and data;

[0029] The pilot signal is data known to both the sender and the receiver, and is used to estimate the channel when receiving the signal;

[0030] The data carries information that needs to be sent.

[0031] Preferably, the deep neural network model includes: an input layer, several hidden layers and an output layer;

[0032] The number of hidden layers is 8, the activation function of the hidden layer is ReLU, and the activation function of the output layer is Sigmoid;

[0033] Adam is used as the optimizer for the deep neural network model.

[0034] Preferably, the process of training the deep neural network includes:

[0035] Get the peak-to-average power ratio expression for RFDM:

[0036]

[0037] Among them, E{|s[n]| 2} represents the average power of the simulated RFDM transmitted signal, represents the peak power of the simulated RFDM transmit signal, s[n] represents the time domain signal of the simulated RFDM transmit signal after discrete affine Fourier transform, N represents the number of linear frequency modulation subcarriers, x[m] represents the simulated RFDM transmit signal, c1 and c2 represent two parameters of the simulated RFDM;

[0038] The peak-to-average power ratio expression is used as a loss function to train the deep neural network model to obtain the peak-to-average power ratio optimization model.

[0039] Preferably, the working process of the power ratio reduction module includes:

[0040] Acquire a simulated radio frequency division multiplexing transmission signal and input it into the peak-to-average power ratio optimization model for optimization to obtain an optimized value, wherein the optimized value includes: a guard interval insertion point and an active constellation extension point;

[0041] The optimized value is combined with the simulated radio frequency division multiplexing transmission signal to reduce the peak-to-average power ratio: the guard interval insertion point is embedded between the pilot and the data to prevent interference between the pilot and the data and reduce the peak-to-average power ratio; the active constellation extension point reduces the peak-to-average power ratio by finding the optimal value within the extended range.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] (1) The present invention designs a method and system for reducing the peak-to-average power ratio of simulated radio frequency division multiplexing based on deep learning technology, filling the gap in deep learning in reducing the peak-to-average power ratio of simulated radio frequency division multiplexing and enriching the application of deep learning technology in wireless communication systems;

[0044] (2) The present invention innovatively uses a deep neural network to simultaneously optimize the guard interval insertion point and the active constellation expansion point, which greatly reduces the peak-to-average power ratio while maintaining almost the same system complexity;

[0045] (3) While fully reducing the peak-to-average power ratio (PAPR), the present invention has minimal impact on the bit error rate (BER) performance of the wireless communication system. Furthermore, because the guard interval insertion points and active constellation extension points used to optimize the PAPR do not carry any side information, the transmitter does not need to transmit any additional data to the receiver, resulting in no additional spectrum loss in the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0047] Figure 1 Schematic diagram of a method flow in an embodiment of the present invention;

[0048] Figure 2 A schematic diagram of a symbol architecture according to an embodiment of the present invention;

[0049] Figure 3 This is a constellation distribution diagram of optimized points according to an embodiment of the present invention;

[0050] Figure 4This is a comparison diagram of peak-to-average power ratio performance of a simulated radio frequency division multiplexing signal after optimizing the peak-to-average power ratio and an original signal according to an embodiment of the present invention;

[0051] Figure 5 This is a comparison diagram of bit error rate performance simulation of the simulated RFDM signal according to the embodiment of the present invention and the original simulated RFDM signal under the existing channel estimation and symbol detection method. DETAILED DESCRIPTION

[0052] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0053] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0054] Example 1

[0055] In this embodiment, if Figure 1 As shown, a method for reducing peak-to-average power ratio of radio frequency division multiplexing based on deep learning includes the following steps:

[0056] S1. Obtain a simulated RFDM transmission signal.

[0057] The simulated RFDM transmission signal includes: pilot and data; the pilot is data known to both the sender and the receiver, and is used to estimate the channel when receiving the signal; the data carries the information to be sent.

[0058] In this embodiment, the transmission signal structure of the RFDM is as follows: Figure 2 As shown. In the transmission architecture before optimization, the pilot is data known to both the sender and the receiver, and is used to estimate the channel when receiving the signal; the data carries the information to be sent; the guard interval is a string of empty symbols of length Q located at both ends of the pilot, which is used to prevent interference between the pilot and the data, where Q = (2α max +1)(l max +1)-1,α max With l max are the normalized maximum Doppler shift and delay, respectively.

[0059] S2. Build a deep neural network model based on unsupervised learning, and train the deep neural network based on simulated RFDM transmission signals to obtain a peak-to-average power ratio optimization model.

[0060] The deep neural network model includes: an input layer, several hidden layers and an output layer; the number of hidden layers is 8, the activation function of the hidden layer is ReLU, and the activation function of the output layer is Sigmoid; Adam is used as the optimizer of the deep neural network model.

[0061] The method for training a deep neural network includes obtaining an expression for the peak-to-average power ratio of RF division multiplexing:

[0062]

[0063] Among them, E{|s[n]| 2} represents the average power of the simulated RFDM transmitted signal, represents the peak power of the simulated RFDM transmit signal, s[n] represents the time domain signal of the simulated RFDM transmit signal after discrete affine Fourier transform, N represents the number of linear frequency modulation subcarriers, x[m] represents the simulated RFDM transmit signal, c1 and c2 represent two parameters of the simulated RFDM; in this embodiment, c1=(2α max +1) / 2N, where c2 is any irrational number or a rational number much smaller than 1 / 2N. The peak-to-average power ratio expression is used as the loss function to train the deep neural network model and obtain the peak-to-average power ratio optimization model.

[0064] S3. Obtain the simulated RFDM transmission signal and input it into the peak-to-average power ratio optimization model for optimization to obtain an optimized value, and combine the optimized value with the simulated RFDM transmission signal to reduce the peak-to-average power ratio.

[0065] The method for reducing the peak-to-average power ratio includes: obtaining a simulated radio frequency division multiplexing transmission signal and inputting it into a peak-to-average power ratio optimization model for optimization to obtain an optimized value, which includes: a guard interval insertion point and an active constellation extension point; and combining the optimized value with the simulated radio frequency division multiplexing transmission signal to reduce the peak-to-average power ratio: the guard interval insertion point will be embedded between the pilot and the data to prevent interference between the pilot and the data and reduce the peak-to-average power ratio; the active constellation extension point reduces the peak-to-average power ratio by finding the optimal value within the extended range.

[0066] In this embodiment, the pilot and data are used as inputs of the neural network to optimize the peak-to-average power ratio, and the optimized value for reducing the peak-to-average power ratio can be output, including: a guard interval insertion point and an active constellation expansion point. In the optimized transmission signal architecture, the guard interval insertion point is located at the position of the guard interval in the original architecture, and is used to optimize the peak-to-average power ratio while preventing interference between the pilot and the data; the data after active constellation expansion will replace the original data, and the peak-to-average power ratio will be optimized while the transmission bit information remains unchanged and the minimum Euclidean distance does not decrease. In this embodiment, taking Quadrature Phase Shift Keying (QPSK) as an example, the constellation distribution of the optimization points for reducing the peak-to-average power ratio is as follows: Figure 3 As shown. For the guard interval optimization point, its constellation range is a circle with a radius of R. Increasing R can improve the peak-to-average power ratio reduction performance. However, due to the increase in the energy of the guard interval insertion point, its interference with the pilot and data will increase, thereby reducing the bit error rate performance in the symbol detection process. Therefore, it is crucial to choose a suitable R to balance the bit error rate and peak-to-average power ratio reduction performance. For the active constellation extension point, its constellation range extends outward from the original constellation point of the data. In order to prevent the energy of the active constellation extension point from being too large, we set the upper limit of the normalized amplitude expansion of its real and imaginary parts to 1. Since this extension range ensures that the minimum Euclidean distance between constellation points will not decrease, it can ensure to the greatest extent that the bit error rate performance will not be affected.

[0067] This embodiment optimizes the peak-to-average power ratio (PAPR) of RFDM-like transmission signals by deploying a pre-trained deep neural network at the signal transmitter. This embodiment comprises two phases: an offline training phase and an online deployment phase. During the offline training phase, a corresponding deep neural network is constructed based on the architecture of the pilot and data signals and the architecture of the PAPR optimization point. A loss function for reducing the PAPR is applied to the network, and the optimal parameter combination is found by continuously adjusting the various neural network parameters. Next, a large number of RFDM-like transmission signals (i.e., pilot signals and data) form a training dataset for the neural network and are input into the neural network for training until optimal performance is achieved. During the online deployment phase, the trained deep neural network model is deployed at the transmitter. The user inputs the RFDM-like signal to be optimized into the pre-trained neural network, which then outputs the optimization point for reducing the PAPR. The transmitted pilot and data are combined with the optimization point to construct a RFDM-like transmission signal with low PAPR characteristics.

[0068] Example 2

[0069] In this embodiment, a deep learning-based RFDM peak-to-average power ratio reduction system includes: a signal acquisition module, a model construction module and a power ratio reduction module.

[0070] The signal acquisition module is used to acquire the simulated RFDM transmission signal.

[0071] The simulated RFDM transmission signal includes: pilot and data; the pilot is data known to both the sender and the receiver, and is used to estimate the channel when receiving the signal; the data carries the information to be sent.

[0072] The model construction module constructs a deep neural network model based on unsupervised learning, and trains the deep neural network based on the simulated RF division multiplexing transmission signal to obtain a peak-to-average power ratio optimization model.

[0073] The deep neural network model includes: an input layer, several hidden layers and an output layer; the number of hidden layers is 8, the activation function of the hidden layer is ReLU, and the activation function of the output layer is Sigmoid; Adam is used as the optimizer of the deep neural network model.

[0074] The process of training a deep neural network includes: obtaining the peak-to-average power ratio expression of RFDM:

[0075]

[0076] Among them, E{|s[n]| 2} represents the average power of the simulated RFDM transmitted signal, represents the peak power of the simulated RFDM transmit signal, s[n] represents the time domain signal of the simulated RFDM transmit signal after discrete affine Fourier transform, N represents the number of linear frequency modulation subcarriers, x[m] represents the simulated RFDM transmit signal, c1 and c2 represent two parameters of the simulated RFDM; the peak-to-average power ratio expression is used as the loss function to train the deep neural network model and obtain the peak-to-average power ratio optimization model.

[0077] The power ratio reduction module is used to obtain the simulated RFDM transmission signal and input it into the peak-to-average power ratio optimization model for optimization, obtain the optimized value, and combine the optimized value with the simulated RFDM transmission signal to reduce the peak-to-average power ratio.

[0078] The workflow of the power ratio reduction module includes: obtaining the simulated RF division multiplexing transmission signal and inputting it into the peak-to-average power ratio optimization model for optimization to obtain the optimized value, which includes: the guard interval insertion point and the active constellation extension point; and combining the optimized value with the simulated RF division multiplexing transmission signal to reduce the peak-to-average power ratio: the guard interval insertion point will be embedded between the pilot and the data to prevent interference between the pilot and the data and reduce the peak-to-average power ratio; the active constellation extension point reduces the peak-to-average power ratio by finding the optimal value within the extended range.

[0079] Example 3

[0080] In order to illustrate the technical advancement of the present invention, the peak-to-average power ratio and bit error rate performance of the deep learning-based RFDM peak-to-average power ratio reduction method proposed in this embodiment are simulated on a Matlab platform.

[0081] The simulation parameters are set as follows: the carrier frequency and bandwidth are 4 GHz and 32 kHz respectively; the number of linear frequency modulation subcarriers N = 32; the maximum delay of the channel is 62.5 μs, which is 2 after normalization; the maximum Doppler shift of the channel is 1 kHz, which is 1 after normalization; the length of the guard interval is 2Q = 2((2α max +1)(l max +1)-1)=16; the number of channels P=2; the complex gain of the channel is generated by an independent complex random variable with zero mean and variance 1 / P; the normalized delay of the channel l=[0,2]; the normalized Doppler shift of the channel in the ith channel is α i =α max cosθ i , where α max represents the maximum Doppler shift, θ i The data is evenly distributed on [-π, π]. The ratio of the pilot signal-to-noise ratio to the data signal-to-noise ratio is SNRp / SNRd = 20dB. Regarding the neural network parameter settings, the input layer, hidden layer, and output layer are included. The neural network has a total of 8 hidden layers. The activation function of the hidden layer is ReLU, and the activation function of the output layer is Sigmoid. Adam is used as the neural network optimizer. Regarding the peak-to-average power ratio optimization parameter, the distribution radius of the guard interval insertion point is R = 0.2. Specifically, Figure 4 The peak-to-average power ratio performance of this method is demonstrated by comparing the complementary cumulative distribution function (CCDF), where CCDF(PAPR0) = prob(PAPR>PAPR0), which is the probability that the peak-to-average power ratio of the transmitted signal sample is greater than PAPR0. Figure 5 The bit error rate performance of this method is demonstrated. Figure 4 Compare with the CCDF curve of the original RFDM signal; Figure 5 The existing embedded channel estimation method and minimum mean square error (MMSE) symbol detection method are used, and the bit error rate performance is compared with the original RFDM signal.

[0082] Through simulation, we can see that compared with the original RFDM signal, the signal optimized by deep neural network can effectively reduce the peak-to-average power ratio. During the simulation process, we also tried to optimize only the guard interval insertion point and optimize the guard interval insertion point and active constellation extension point at the same time. Figure 4 As shown in the figure, the scheme of optimizing the guard interval insertion point and the active constellation extension point at the same time can achieve a lower peak-to-average power ratio than the scheme of optimizing only the guard interval insertion point. It is worth noting that since there is no significant difference in the network architecture of the two optimization methods, the complexity of the two optimization methods is almost equal, which means that optimizing the guard interval insertion point and the active constellation extension point at the same time can bring higher peak-to-average power ratio reduction efficiency. At the same time, the signal optimized by deep neural network still performs very well in bit error rate. Figure 5 As shown, the solution that only optimizes the guard interval insertion point can achieve performance almost identical to the original signal. While the bit error rate performance of optimizing both the guard interval insertion point and the active constellation extension point is somewhat lower than that of optimizing only the guard interval insertion point, it still maintains relatively good performance. In summary, if a system has a high demand for reducing the peak-to-average power ratio, then the solution of optimizing both the guard interval insertion point and the active constellation extension point is suitable. If a system needs to reduce the peak-to-average power ratio to a certain extent but has more stringent requirements on the system's bit error rate performance, then the solution of optimizing only the guard interval insertion point is more suitable.

[0083] The embodiments described above are merely descriptions of preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by persons skilled in the art should fall within the scope of protection defined by the claims of the present invention.

Claims

1. A method for reducing peak-to-average power ratio of radio frequency division multiplexing based on deep learning, characterized in that: The following steps are involved: Acquire a simulated RFDM transmission signal; Constructing a deep neural network model based on unsupervised learning, and training the deep neural network based on the simulated radio frequency division multiplexing transmission signal to obtain a peak-to-average power ratio optimization model; Acquire a simulated radio frequency division multiplexing transmission signal and input it into the peak-to-average power ratio optimization model for optimization to obtain an optimized value, and combine the optimized value with the simulated radio frequency division multiplexing transmission signal to reduce the peak-to-average power ratio.

2. The method for reducing peak-to-average power ratio of radio frequency division multiplexing based on deep learning according to claim 1, characterized in that: The simulated radio frequency division multiplexing transmission signal includes: a pilot and data; The pilot signal is data known to both the sender and the receiver, and is used to estimate the channel when receiving the signal; The data carries information that needs to be sent.

3. The method for reducing peak-to-average power ratio of radio frequency division multiplexing based on deep learning according to claim 1, characterized in that: The deep neural network model includes: an input layer, several hidden layers and an output layer; The number of hidden layers is 8, the activation function of the hidden layer is ReLU, and the activation function of the output layer is Sigmoid; Adam is used as the optimizer for the deep neural network model.

4. The method for reducing peak-to-average power ratio of radio frequency division multiplexing based on deep learning according to claim 1, characterized in that: The method for training the deep neural network includes: Get the peak-to-average power ratio expression for RFDM: Among them, E{|s[n]| 2 } represents the average power of the simulated RFDM transmitted signal, represents the peak power of the simulated RFDM transmit signal, s[n] represents the time domain signal of the simulated RFDM transmit signal after discrete affine Fourier transform, N represents the number of linear frequency modulation subcarriers, x[m] represents the simulated RFDM transmit signal, c1 and c2 represent two parameters of the simulated RFDM; The peak-to-average power ratio expression is used as a loss function to train the deep neural network model to obtain the peak-to-average power ratio optimization model.

5. The method for reducing peak-to-average power ratio of radio frequency division multiplexing based on deep learning according to claim 2, characterized in that: The method for reducing the peak-to-average power ratio includes: Acquire a simulated radio frequency division multiplexing transmission signal and input it into the peak-to-average power ratio optimization model for optimization to obtain an optimized value, wherein the optimized value includes: a guard interval insertion point and an active constellation extension point; The optimized value is combined with the simulated radio frequency division multiplexing transmission signal to reduce the peak-to-average power ratio: the guard interval insertion point is embedded between the pilot and the data to prevent interference between the pilot and the data and reduce the peak-to-average power ratio; the active constellation extension point reduces the peak-to-average power ratio by finding the optimal value within the extended range.

6. A deep learning-based RFDM peak-to-average power ratio reduction system, wherein the system applies the method according to any one of claims 1 to 5, characterized in that: include: signal acquisition module, model building module and power ratio reduction module; The signal acquisition module is used to acquire the simulated radio frequency division multiplexing transmission signal; The model construction module constructs a deep neural network model based on unsupervised learning, and trains the deep neural network based on the simulated radio frequency division multiplexing transmission signal to obtain a peak-to-average power ratio optimization model; The power ratio reduction module is used to obtain the simulated radio frequency division multiplexing transmission signal and input it into the peak-to-average power ratio optimization model for optimization to obtain an optimized value, and combine the optimized value with the simulated radio frequency division multiplexing transmission signal to reduce the peak-to-average power ratio.

7. The deep learning-based RFDM peak-to-average power ratio reduction system according to claim 6, characterized in that: The simulated radio frequency division multiplexing transmission signal includes: a pilot and data; The pilot signal is data known to both the sender and the receiver, and is used to estimate the channel when receiving the signal; The data carries information that needs to be sent.

8. The deep learning-based RFDM peak-to-average power ratio reduction system according to claim 6, characterized in that: The deep neural network model includes: an input layer, several hidden layers and an output layer; The number of hidden layers is 8, the activation function of the hidden layer is ReLU, and the activation function of the output layer is Sigmoid; Adam is used as the optimizer for the deep neural network model.

9. The deep learning-based RFDM peak-to-average power ratio reduction system according to claim 6, characterized in that: The process of training the deep neural network includes: Get the peak-to-average power ratio expression for RFDM: Among them, E{|s[n]| 2 } represents the average power of the simulated RFDM transmitted signal, represents the peak power of the simulated RFDM transmit signal, s[n] represents the time domain signal of the simulated RFDM transmit signal after discrete affine Fourier transform, N represents the number of linear frequency modulation subcarriers, x[m] represents the simulated RFDM transmit signal, c1 and c2 represent two parameters of the simulated RFDM; The peak-to-average power ratio expression is used as a loss function to train the deep neural network model to obtain the peak-to-average power ratio optimization model.

10. The deep learning-based RFDM peak-to-average power ratio reduction system according to claim 7, characterized in that: The working process of the power ratio reduction module includes: Acquire a simulated radio frequency division multiplexing transmission signal and input it into the peak-to-average power ratio optimization model for optimization to obtain an optimized value, wherein the optimized value includes: a guard interval insertion point and an active constellation extension point; The optimized value is combined with the simulated radio frequency division multiplexing transmission signal to reduce the peak-to-average power ratio: the guard interval insertion point is embedded between the pilot and the data to prevent interference between the pilot and the data and reduce the peak-to-average power ratio; the active constellation extension point reduces the peak-to-average power ratio by finding the optimal value within the extended range.

Images