A nonlinear blind equalization method for VLC system based on variational autoencoder

Abstract

This invention belongs to the field of visible light communication technology and relates to a nonlinear blind equalization method for a VLC system based on a variational autoencoder. The method includes: inputting the received signal into a trained blind equalization network model to obtain an equalized signal; the training process of the blind equalization network model includes: preprocessing the transmitted and received signals of the VLC to obtain a preprocessed transmitted signal Tx and its counterpart to the received signal Rx; and inputting Rx into an equalizer network to obtain the latent variable signal Tx. * The prior probability distribution and the posterior probability distribution of a given received signal Rx are obtained; the prior probability distribution and the posterior probability distribution are input into the channel estimation network to obtain the likelihood probability distribution and the equalization signal; the loss function value is calculated based on the prior probability distribution, the posterior probability distribution and the likelihood probability distribution, and the parameters of the model are updated based on the loss function value until the stopping condition is reached; the present invention integrates multiple loss functions to avoid the local optimum problem caused by a single loss function.

Description

Technical Field

[0001] This invention belongs to the field of visible light communication technology and relates to a nonlinear blind equalization method for VLC systems based on variational autoencoders. Background Technology

[0002] Visible Light Communication (VLC) is an emerging optical wireless communication technology that utilizes the visible light spectrum to transmit information. This technology primarily uses light-emitting diodes (LEDs) to modulate optical signals for data transmission. Due to its advantages such as high cost-effectiveness, strong security, no electromagnetic pollution, and no need for spectrum licensing, VLC technology has become a hot topic in scientific research and a technology with great development potential. Research on VLC is continuously deepening, mainly focusing on optical materials, advanced modulation techniques, signal equalization processing, and multiplexing techniques, especially in the field of equalization. However, signals in VLC communication systems are subject to severe linear and nonlinear effects during transmission in the channel. Nonlinear effects in visible light dominate and significantly impact the system's transmission performance.

[0003] The rapid development of machine learning has led to diverse applications of equalization techniques. Traditional neural networks have demonstrated high equalization performance in handling nonlinear impairments, but most of them are based on non-blind equalizer implementations. Currently, blind equalizers (BE) can perform channel equalization without training sequences. This approach provides an effective method to reduce signal distortion without consuming additional bandwidth. This means that during signal transmission, blind equalization technology can avoid using training sequences, thus saving valuable communication bandwidth and effectively increasing the communication capacity of fiber optic communication systems while conserving communication resources.

[0004] In recent years, some methods have been proposed, such as blind equalization algorithms based on constant modulus (CMA) and least mean square (LMS). These methods estimate the channel through statistical signal characteristics to achieve blind equalization. However, these methods have high computational complexity, poor equalization effect, and poor robustness. Summary of the Invention

[0005] To address the aforementioned problems in the prior art, this invention employs a nonlinear blind equalization method for VLC systems based on variational autoencoders, comprising: acquiring the received signal at the receiver of the VLC system; inputting the received signal into a trained blind equalization network model to obtain an equalized signal; the blind equalization network model includes: an equalizer network and a channel estimation network; the training process of the blind equalization network model includes:

[0006] S1. Obtain the transmitted signal from the transmitter of the VLC system and the corresponding received signal from the receiver; preprocess the transmitted signal and the corresponding received signal to obtain training samples; the training samples include: the preprocessed transmitted signal Tx and the corresponding received signal Rx from the receiver.

[0007] S2. Input the received signal Rx into the equalizer network to obtain the latent variable signal Tx. * The prior probability distribution and the posterior probability distribution for a given received signal Rx;

[0008] S3, transfer the latent variable signal Tx * The prior probability distribution and the posterior probability distribution for a given received signal Rx are input into the channel estimation network to obtain the given latent variable signal Tx. * Likelihood probability distribution of received signal Rx and equalization signal Rx * ;

[0009] S4. Based on the given latent variable signal Tx * The likelihood probability of the received signal Rx and the latent variable signal Tx * The loss function value is calculated based on the prior probability distribution and the posterior probability distribution for a given received signal Rx. The parameters of the blind equalization network model are updated based on the loss function value. When the pre-set training stopping condition is met, the training of the blind equalization network model is completed.

[0010] Beneficial effects:

[0011] The mean square loss function of this invention ensures that the prior probability distribution of the latent variable signal is close to the preset ideal distribution. It measures the difference between the prior distribution and the channel data distribution using KL divergence, preventing excessively discrete or unreasonable signal values. The singleton loss function measures the error between the posterior probability distribution predicted by the equalizer network and the actual data, ensuring that the equalizer network can extract effective features from the received signal to compensate for signal distortion. During training, it can dynamically adjust the output signal to match the expected latent variable distribution, maintaining good performance even under unseen channel conditions. The reconstruction loss function measures the accuracy of reconstructing the received signal from the latent variables, evaluating the channel estimation network's ability to model the channel response model. Unlike the singleton loss function, it verifies the rationality of the equalized signal from another perspective. By fusing these three loss functions, the local optimum problem that may be caused by a single loss function is avoided, improving the equalization effect. Simultaneously, the "adversarial" mechanism between the equalizer network and the channel estimator network improves the model's robustness and generalization ability; that is, the equalizer network attempts to recover the original signal, while the channel estimator verifies the rationality of the recovered signal. Attached Figure Description

[0012] Figure 1This is a diagram of the PAM-VLC system architecture provided in an embodiment of the present invention;

[0013] Figure 2 The flowchart illustrates a nonlinear blind equalization method for a VLC system based on a variational autoencoder, as provided in an embodiment of the present invention. Detailed Implementation

[0014] 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.

[0015] like Figure 1 The PAM-VLC system architecture includes a transmitter and a receiver. The transmitter includes an arbitrary waveform generator, two electronic amplifiers, a bias unit, an LED, and an optical filter assembly. The receiver includes a PIN photodetector, a digital storage oscilloscope, and an offline signal processing module. The workflow is as follows: The raw binary data to be transmitted, i.e., the transmission signal, is mapped (e.g., using PAM-8 mapping) into PAM symbols. After upsampling, upconversion, and pulse shaping, it is sent to an arbitrary waveform generator (AWG). The first electronic amplifier further enhances the PAM signal generated by the AWG. The bias unit applies a DC bias voltage to the signal enhanced by the first electronic amplifier to ensure that the signal output by the bias unit can activate the LED and make it reach the operating voltage threshold. The LED then converts the signal output by the bias unit into an optical signal, and transmits it to the receiving end after the optical signal is processed by the optical filtering components (i.e., grating and filter system). The processed optical signal is received by the PIN photodetector after the transmission process is completed. The second electronic amplifier amplifies the signal received by the PIN photodetector. The digital storage oscilloscope samples and records the amplified signal. The offline signal processing module analyzes and processes the signal recorded by the digital storage oscilloscope to obtain the equalized signal. The equalized signal is converted back into raw binary bits through PAM mapping to complete the signal compensation.

[0016] Based on the above PAM-VLC system, such as Figure 2 As shown, this invention employs a nonlinear blind equalization method for a VLC system based on a variational autoencoder, comprising the following steps: acquiring the received signal from the receiver of the VLC system, inputting the received signal into a trained blind equalization network model to obtain the equalization signal; the blind equalization network model includes: an equalizer network and a channel estimation network; the training process of the blind equalization network model includes:

[0017] S1. Obtain the transmitted signal from the transmitter of the VLC system and the corresponding received signal from the receiver; preprocess the transmitted signal and the corresponding received signal to obtain training samples; the training samples include: the preprocessed transmitted signal Tx and the corresponding received signal Rx from the receiver.

[0018] The preprocessing of the transmitted signal and its corresponding received signal includes: setting the sliding window size tap = n, dividing the transmitted signal and its corresponding received signal according to the set sliding window size n, and dividing the divided transmitted signal Tx and its corresponding received signal Rx into training samples and validation samples in a 7:3 ratio.

[0019] S2. Input the received signal Rx of the training samples into the equalizer network to obtain the latent variable signal Tx. * The prior probability distribution and the posterior probability distribution for a given received signal Rx;

[0020] The equalizer network is a deep learning network model; in this embodiment, the entity extraction neural network (EXNN) model is selected. The EXNN model includes an input layer, hidden layers, and an output layer. The input layer receives the received signal Rx, and the hidden layers extract features from the received signal Rx. Each hidden layer includes a cascaded convolutional layer and an activation function layer. The convolutional layers extract linear and nonlinear impairments of the received signal Rx by sampling the convolutional kernel. The activation function layer uses the ReLU function as the activation function to capture and model nonlinear features. The output layer includes two softmax layers, which output latent variable signals Tx based on the features extracted by the hidden layers. * The prior probability distribution P θ (Tx * ) and the posterior probability distribution for a given received signal Rx.

[0021] The prior probability distribution describes the statistical description of the possible states of the latent variable signal when the probability distribution of the received signal at the receiver is not given. This prior probability should be as close as possible to the set prior probability. Therefore, the mean square loss function is calculated between the prior probability distribution and the probability distribution of the received signal Rx. Among them, D KL () indicates the calculation of the Kullback-Leibler divergence, where λ is the adjustment parameter and θ is the parameter of the equalizer network. To estimate the parameters of the network for the channel, Regarding the probability distribution of the received signal Rx, in this embodiment, the distribution of Tx in the PAM8 VLC system is observed to be closer to a uniform distribution; therefore, the probability distribution of the received signal Rx is set to a uniform distribution. This is achieved using the mean square loss function L... prior It can effectively adapt to and predict latent variable signal data, especially under complex or incompletely observable signal conditions, thereby constraining the generation of equalization signals.

[0022] The posterior probability distribution reflects how the equalizer network adjusts the output signal to compensate for signal distortion using specific information from the received signal Rx. The posterior probability distribution predicted by the equalizer network should be as close as possible to the predicted latent variable signal Tx. * Given the prior probability distribution, the latent variable signal Tx is calculated using the Kullback-Leibler divergence. * The singleton loss function of the posterior probability distribution given the received signal Rx. Alternatively, cross-entropy calculation can be used to calculate the error in multi-class label classification data.

[0023] S3, transfer the latent variable signal Tx * The prior probability distribution and the posterior probability distribution for a given received signal Rx are input into the channel estimation network to obtain the given latent variable signal Tx. * Likelihood probability distribution of received signal Rx and equalization signal Rx * ;

[0024] The channel estimation network is an EXNN, the same as the equalizer network, consisting of an input layer, hidden layers, and an output layer; the input layer is used to receive the latent variable signal Tx. * The prior probability distribution and the posterior probability distribution given the received signal Rx are used by the hidden layer to extract the latent variable signal Tx. * The features of the prior probability distribution and the posterior probability distribution for a given received signal Rx are used to define the output layer, which includes two softmax layers. These two softmax layers output the given latent variable signal Tx based on the features extracted from the hidden layer. * The likelihood probability distribution P of the received signal Rx θ (Rx|Tx * ) and equalization signal Rx * ;

[0025] The primary task of the channel estimation network is to recover the received signal Rx from the prior and posterior probability distributions output by the equalization network. * At the same time, it measures the various distortions caused by the channel response when the signal passes through the communication channel, such as noise, attenuation and interference, from another perspective.

[0026] Specifically, the reconstruction loss function can be expressed as follows: The reconstruction loss function is represented in the posterior probability distribution. Next, calculate the expected value E of the logarithm of the likelihood probability, such that the latent variable signal Tx... * Recover the original signal Rx as accurately as possible * .

[0027] S4. Based on the given latent variable signal Tx * The likelihood probability of the received signal Rx and the latent variable signal Tx * The loss function value is calculated based on the prior probability distribution and the posterior probability distribution for a given received signal Rx. The parameters of the blind equalization network model are updated based on the loss function value. When the pre-set training stopping condition is met, the training of the blind equalization network model is completed.

[0028] The loss function is:

[0029] Loss=αL prior +βL sig +γL Rec +δL mse

[0030] Here, α, β, γ, and δ are weights used to balance the impact of different losses on the overall training process.

[0031] L mse To measure the loss between the calculated equalized signal and the actual signal data transmitted by the corresponding transmitter, MSE is typically used as the loss metric.

[0032]

[0033] Where N represents the length of the signal, and this loss function is typically used to quantify the error between the model and the target signal. The latent variable signal Tx * It can be obtained by randomly sampling its prior probability distribution and its posterior probability distribution for a given received signal Rx.

[0034] The process of determining whether the preset training stopping condition has been met includes: determining whether the preset maximum number of training iterations M has been reached; if so, the preset training stopping condition has been met; otherwise, the verification sample is input into the blind equilibrium network model for training. If the accuracy of the blind equilibrium network model does not improve after m consecutive training iterations, the preset training stopping condition has been met. Here, M is generally set to 600 and m is generally set to 20.

[0035] The optimizer used for backpropagation updates is typically the Adam optimizer. In this embodiment, the Adam optimizer performs well in handling coefficient gradients and non-stationary objectives, and also performs well in handling the parameters θ of the equalizer network and the channel estimation network. Update them separately. During backpropagation and parameter update, the loss is calculated using the chain rule to update θ and θ respectively. The parameters of the two network models gradually approach the optimal solution after each iteration. The optimizer updates the model weights and then trains the model for the next iteration.

[0036] The above-described embodiments further illustrate the purpose, technical solution, and advantages of the present invention. It should be understood that the above-described embodiments are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made to the present invention within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A nonlinear blind equalization method for a VLC system based on a variational autoencoder, characterized in that, include: The received signal from the receiver of the VLC system is acquired, and then input into the trained blind equalization network model to obtain the equalization signal Rx. * ; The blind equalization network model includes: an equalizer network and a channel estimation network; the training process of the blind equalization network model includes: S1. Obtain the transmitted signal from the transmitter of the VLC system and the corresponding received signal from the receiver; preprocess the transmitted signal and the corresponding received signal to obtain training samples; the training samples include: the preprocessed transmitted signal Tx and the corresponding received signal Rx from the receiver. S2. Input the received signal Rx into the equalizer network to obtain the latent variable signal Tx. * The prior probability distribution and the posterior probability distribution for a given received signal Rx; S3, transfer the latent variable signal Tx * The prior probability distribution and the posterior probability distribution for a given received signal Rx are input into the channel estimation network to obtain the given latent variable signal Tx. * Likelihood probability distribution of received signal Rx and equalization signal Rx * ; S4. Based on the given latent variable signal Tx * The likelihood probability of the received signal Rx and the latent variable signal Tx * The loss function value is calculated based on the prior probability distribution and the posterior probability distribution for a given received signal Rx. The parameters of the blind equalization network model are updated based on the loss function value. When the pre-set training stopping condition is met, the training of the blind equalization network model is completed.

2. The nonlinear blind equalization method for a VLC system based on a variational autoencoder according to claim 1, characterized in that, The equalizer network is an EXNN; the EXNN includes an input layer, hidden layers, and an output layer; the input layer receives the received signal Rx, the hidden layers extract features from the received signal Rx, and the output layer includes two softmax layers, which output latent variable signals Tx based on the features extracted by the hidden layers. * The prior probability distribution and the posterior probability distribution for a given received signal Rx; where EXNN is an entity extraction neural network.

3. The nonlinear blind equalization method for a VLC system based on a variational autoencoder according to claim 1, characterized in that, The channel estimation network is the same EXNN as the equalizer network; the EXNN includes an input layer, a hidden layer, and an output layer; wherein, the input layer is used to receive the latent variable signal Tx. * The prior probability distribution and the posterior probability distribution given the received signal Rx are used by the hidden layer to extract the latent variable signal Tx. * The features of the prior probability distribution and the posterior probability distribution for a given received signal Rx are used to define the output layer, which includes two softmax layers. These two softmax layers output the given latent variable signal Tx based on the features extracted from the hidden layer. * Likelihood probability distribution of received signal Rx and equalization signal Rx * ; where EXNN is an entity extraction neural network.

4. The nonlinear blind equalization method for a VLC system based on a variational autoencoder according to claim 1, characterized in that, The loss function is: Loss=αL prior +βL sig +γL Rec +δL mse Among them, α, β, γ, δ are weights, L prior For the latent variable signal Tx * The mean square loss function of the prior probability distribution and the probability distribution of the received signal Rx, L sig For the latent variable signal Tx * The singleton loss function, L, is given by the posterior probability distribution of the received signal Rx. Rec L is the reconstruction loss function of the received signal. mse This is the error loss function between the latent variable signal and the actual signal transmitted by the transmitter.

5. A nonlinear blind equalization method for a VLC system based on a variational autoencoder according to claim 4, characterized in that, Loss function L prior The calculation formula is: Among them, D KL () indicates the calculation of the Kullback-Leibler divergence, where λ is the adjustment parameter and P is the weight of the Kullback-Leibler divergence. θ (Tx * ) represents the latent variable signal Tx * The prior probability distribution, Let Rx be the probability distribution of the received signal.

6. The nonlinear blind equalization method for a VLC system based on a variational autoencoder according to claim 4, characterized in that, Loss function L sig The calculation formula is: Among them, D KL () indicates the calculation of the Kullback-Leibler divergence. Let P be the posterior probability distribution given the received signal Rx. θ (Tx * ) represents the latent variable signal Tx * The prior probability distribution.

7. A nonlinear blind equalization method for a VLC system based on a variational autoencoder according to claim 4, characterized in that, The formula for calculating the reconstruction loss function is: in, Let P be the posterior probability distribution given the received signal Rx. θ (Rx|Tx * Given the latent variable signal Tx * The likelihood probability distribution of the received signal Rx.

8. A nonlinear blind equalization method for a VLC system based on a variational autoencoder according to claim 4, characterized in that, loss function Among them, the latent variable signal Tx * It is obtained by randomly sampling its prior probability distribution and its posterior probability distribution for a given received signal Rx, where N represents the length of the signal and i is the index of the data point of the signal.

9. A nonlinear blind equalization method for a VLC system based on a variational autoencoder according to claim 1, characterized in that, Preprocessing the transmitted signal and its corresponding received signal includes: setting the sliding window size, segmenting the transmitted signal and its corresponding received signal according to the set sliding window size, and dividing the segmented transmitted signal Tx and its corresponding received signal Rx into training samples and validation samples.

10. A nonlinear blind equalization method for a VLC system based on a variational autoencoder according to claim 9, characterized in that, The process of determining whether the preset training stopping condition has been met includes: determining whether the preset maximum number of training iterations has been reached; if so, the preset training stopping condition has been met; otherwise, the verification samples are input into the blind equilibrium network model for training. If the accuracy of the blind equilibrium network model does not improve after multiple consecutive training iterations, the preset training stopping condition has been met.

Images