A method, device and electronic device for determining nonlinear model of visible light communication system based on PN sequence

Through the PN sequence-based method, Hammerstein and Wiener nonlinear models of visible light communication systems are trained and evaluated, which solves the problem of insufficient understanding of nonlinear effects in the prior art, and realizes the rapid discrimination of system nonlinear models and Parameter estimation improves the effectiveness and reliability of the system.

CN115348138BActive Publication Date: 2024-11-29SOUTHEAST UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202210984329.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-08-17
Publication Date
2024-11-29
Estimated Expiration
2042-08-17

AI Technical Summary

Technical Problem

The understanding and understanding of the nonlinear effects of existing visible light communication systems is limited, and traditional models such as Volterra, Wiener and Hammerstein models have insufficient parameters in terms of parameter estimation and nonlinear compensation, resulting in the impact of system effectiveness and reliability.

Method used

Using a PN sequence-based method, Hammerstein and Wiener nonlinear models are trained and evaluated by generating training sets, evaluation sets and verification sets, comparing the minimum evaluation error and training error of the two models, and discriminating the system nonlinear model. The type of the , determine the optimal nonlinear model parameters.

Benefits of technology

It realizes the rapid acquisition of effective estimation of nonlinear model parameters, concisely and effectively characterizes the system nonlinearity, determines the system nonlinear model type, lays the foundation for the design of visible light communication systems, and provides nonlinear distortion compensation and predistortion model assumptions .

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115348138B_ABST
    Figure CN115348138B_ABST
Patent Text Reader

Abstract

The present invention discloses a nonlinear model discrimination method, device and electronic device for a visible light communication system based on a PN sequence, including: for the Hammerstein and Wiener models, a multi-level PN sequence can be used to quickly obtain an effective estimate of the model parameters. In an actual visible light communication system, the two models are trained, evaluated and verified respectively, and the minimum system nonlinear fitting errors of the two models are compared. The effectiveness of the model classification discrimination method is judged according to the consistency between the training comparison result and the evaluation comparison result, and the nonlinear type of the model is judged according to the comparison of the evaluation results. The nonlinear model type discrimination method for a visible light communication system proposed by the present invention has far-reaching significance for in-depth exploration of the nonlinear mechanism of the visible light communication system and research on related technologies, and provides a model hypothesis basis for the research on nonlinear distortion compensation and pre-distortion technologies of visible light communication, and has high application value.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of visible light communication, and in particular to a method, device and electronic equipment for distinguishing a nonlinear model of a visible light communication system based on a PN sequence. Background Art

[0002] With the rapid development of the information industry, the demand for mobile communications has grown rapidly, and radio frequency spectrum resources are becoming increasingly scarce. Traditional wireless physical layer transmission technology faces huge challenges. Visible light communication is an emerging wireless access technology that can provide high-speed and reliable physical layer links, alleviate the current situation of tight spectrum resources, and is an important supplementary technology and potential solution for the next generation of communication systems. Visible light communication can effectively take into account both lighting needs and data transmission. It has many advantages such as rich spectrum resources, low cost, simple implementation, high spatial reuse rate, and good electromagnetic compatibility. In recent years, it has received close attention from academia and industry.

[0003] Visible light communication systems generally use light-emitting diodes and photodiodes as transmitting and receiving antennas, and use intensity modulation and direct detection technology to complete signal modulation and transmission. Intensity modulation direct detection technology requires that the transmitted signal meets the non-negative real number requirement, resulting in obvious nonlinearity of the signal. At the same time, the light-emitting diodes, power amplifiers and other devices included in the system also exhibit nonlinear characteristics with memory. Therefore, the visible light communication system has nonlinear effects that cannot be ignored. Complex nonlinear effects will introduce inevitable nonlinear distortion, seriously affecting the effectiveness and reliability of the system. Therefore, exploring the nonlinear mechanism of visible light communication systems and establishing effective nonlinear models are urgent needs for the development of current visible light communication technology.

[0004] However, the current academic and industrial knowledge and understanding of the nonlinear effects of visible light communication systems is still very limited. There are currently several potentially feasible nonlinear models, including the Volterra, Wiener and Hammerstein models. The Volterra model requires a large amount of observation data during training, and parameter estimation and nonlinear compensation require a huge amount of calculation, which is not suitable for actual systems. The Wiener and Hammerstein models represent two diametrically opposed viewpoints. The Wiener model assumes that the signal first passes through a linear time-invariant system and then passes through a memoryless nonlinear module, while the Hammerstein model is the opposite. At present, there are serious differences in the academic community on the effectiveness of these two models. In response to this problem, the present invention proposes an efficient and feasible quantitative discrimination method for nonlinear models of visible light communication systems. For visible light communication systems, two nonlinear models are trained and evaluated respectively, and the effectiveness of the two models in describing the nonlinear effects of the system can be quantitatively judged. At the same time, the optimal nonlinear model parameters can be determined, and the nonlinearity of the system can be concisely and effectively characterized, laying the foundation for the design of visible light communication systems. Summary of the invention

[0005] In view of this, the purpose of the present invention is to provide a method, device and electronic device for distinguishing nonlinear models of visible light communication systems based on PN sequences. For the Hammerstein and Wiener models, multi-level PN sequences can be used to quickly obtain effective estimates of model parameters. In an actual visible light communication system, these two models are trained, evaluated and verified respectively, and the optimal nonlinear fitting errors of the two models are compared to distinguish the type of nonlinear model of the system.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] A nonlinear model discrimination method for a visible light communication system based on a PN sequence, the method comprising:

[0008] Step S1: In a broadband visible light communication system, a data set is generated using a multi-stage PN sequence, and the data set includes: a training set, an evaluation set, and a verification set, wherein the training set, the evaluation set, and the verification set are used to train and evaluate the Hammerstein and Wiener nonlinear model;

[0009] Step S2, using the data set obtained in step S1 to train and evaluate the Hammerstein and Wiener nonlinear models respectively, and then respectively calculating the minimum evaluation error and the corresponding training error of the two models;

[0010] Step S3, for the minimum evaluation error and training error corresponding to the two models calculated in step S2, perform the following operations: compare the size relationship between the minimum evaluation error of the two models and the corresponding training error. If the size relationship is consistent, it is defined as valid and step S4 is executed; otherwise, step S1 is executed;

[0011] Step S4, perform the following judgment: if the minimum evaluation error of the Hammerstein model is smaller than the minimum evaluation error of the Wiener nonlinear model, then the Hammerstein model is more suitable for describing the system nonlinearity; if the minimum evaluation error of the Hammerstein model is larger than the minimum evaluation error of the Wiener nonlinear model, then the Wiener model is more suitable for the system nonlinearity; if the minimum evaluation errors of the two models are equal, then the two nonlinear models have equivalent effects.

[0012] Furthermore, the step S2 specifically includes:

[0013] Step S201: Set the maximum nonlinear order L max and the maximum memory depth D max , initialize the nonlinear order L = 0 and the memory depth D = 0, the Hammerstein model normalized estimation error ε H =1dB, normalized estimation error of Wiener model ε W =1dB;

[0014] Step S202: If L<L max, L=L+1, otherwise, execute step S206;

[0015] Step S203: If D < D max , D=D+1, otherwise, execute step S202;

[0016] Step S204: First, a parameter estimation algorithm of a Hammerstein nonlinear model is executed, and a Hammerstein model with a nonlinear order of L and a memory depth of D is trained using a training set to obtain an estimate of the model parameters. and Then use the parameter and The Hammerstein model and the multi-stage PN sequence in the evaluation data set estimate the corresponding received signal sequence Calculate the estimated error in,

[0017]

[0018]

[0019] if but ε H =ε h ;

[0020] Step S205: First, a parameter estimation algorithm of a Wiener nonlinear model is executed, and a Wiener model with a nonlinear order of L and a memory depth of D is trained using a training data set to obtain an estimate of the model parameters. and Then use the parameters and The Wiener model and the multi-stage PN sequence in the evaluation data estimate the corresponding received signal sequence Calculate the estimated error in,

[0021]

[0022]

[0023] if but ε W =ε w , execute step S203;

[0024] Step S206: According to the method in step S204, the parameters are and The Hammerstein model and the multi-stage PN sequence in the verification data estimate the corresponding received sequence, and calculate the model estimation error ε in the verification stageHtest ;

[0025] Step S207: According to the method in step S205, the parameters are and The Wiener model and the multi-stage PN sequence in the verification data estimate the corresponding received sequence, and calculate the model estimation error ε in the verification stage Wtest .

[0026] Furthermore, in step S2, the parameter estimation algorithm of the Hammerstein nonlinear model includes:

[0027] (1) This method uses a multi-level PN sequence and its digital received signal to complete parameter estimation: where the L-order nonlinear transfer function is φ(·), is the parameter of the nonlinear function φ(·), the linear time-invariant discrete system response with a memory depth of D is g(n), and the N-segment multi-level PN sequence is x i (n) = a i z(n),a i ∈R,i∈Z + and 1≤i≤N, z(n) is a unit power with a mean of zero and a length of N c Pseudo-noise (PN) sequence with amplitude a i ≠a j , i≠j, in the communication system, a multi-level PN sequence is sent and the digital received signal r is stored i (n);

[0028] (2) Calculate the received signal r i The cross-correlation sequence of (n) and PN sequence z(n):

[0029]

[0030] (3) Calculate the autocorrelation sequence of the PN sequence z(n):

[0031]

[0032] (4) Calculate r sz (k) = A -1 r rz (k), where

[0033]

[0034]

[0035]

[0036] (5) Solve the relationship using deconvolution Obtain an estimated sequence of linear time-invariant system responses

[0037] (6) Solve the relationship by deconvolution Get the estimated sequence

[0038] (7) Estimate the parameters of the nonlinear transfer function φ(·) using the least squares method: in,

[0039]

[0040]

[0041]

[0042] Furthermore, in step (1), N≥L, L<<(N×N c ).

[0043] Furthermore, in step S2, the parameter estimation algorithm of the Wiener nonlinear model includes:

[0044] (1) This method uses a multi-level PN sequence and its digital received signal to complete parameter estimation; the L-order nonlinear transfer function is φ(·), is the parameter of the nonlinear function, the linear time-invariant discrete system response with a memory depth of D is g(n), and the N-segment multi-level PN sequence is x i (n) = a i z(n),a i ∈R,i∈Z + and 1≤i≤N, z(n) is a unit power with a mean of zero and a length of N c PN sequence with amplitude a i ≠a j , i≠j, the communication system sends a multi-level PN sequence and stores the digital received signal r i (n);

[0045] (2) Calculate the received signal r i The cross-correlation sequence of (n) and PN sequence z(n):

[0046]

[0047] (3) Calculate the autocorrelation sequence of the PN sequence z(n):

[0048]

[0049] (4) Calculate r wz (k) = A -1 rrz (k), where

[0050]

[0051]

[0052]

[0053] (5) Solve the relationship using deconvolution Obtain an estimated sequence of linear time-invariant system responses

[0054] (6) Calculation Get intermediate signal sequence estimates;

[0055] (7) Estimate the parameters of the nonlinear transfer function φ(·) using the least squares method: in,

[0056]

[0057]

[0058]

[0059] Furthermore, in step (1), N≥L, L<<(N×(D+N c -1)).

[0060] A nonlinear model discrimination device for a visible light communication system based on a PN sequence, the device comprising:

[0061] A first device, comprising: in a broadband visible light communication system, generating a data set using a multi-stage PN sequence, and the data set comprises: a training set, an evaluation set and a verification set, wherein the training set, the evaluation set and the verification set are used to train and evaluate a Hammerstein and Wiener nonlinear model;

[0062] The second device comprises: using the data set to train and evaluate the Hammerstein and Wiener nonlinear models respectively, and then respectively calculating the minimum evaluation error and the corresponding training error of the two models;

[0063] The third device includes: for the minimum evaluation error and training error corresponding to the two models calculated, performing the following operations: comparing the size relationship between the minimum evaluation error of the two models and the corresponding training error, if the size relationship is consistent, it is defined as valid, and the operation configured in the fourth device is performed; otherwise, the operation configured in the first device is performed;

[0064] The fourth device includes: performing the following judgment: if the minimum evaluation error of the Hammerstein model is smaller than the minimum evaluation error of the Wiener nonlinear model, then the Hammerstein model is more suitable for describing the system nonlinearity; if the minimum evaluation error of the Hammerstein model is larger than the minimum evaluation error of the Wiener nonlinear model, then the Wiener model is more suitable for describing the system nonlinearity; if the minimum evaluation errors of the two models are equal, then the two nonlinear models have equivalent effects.

[0065] An electronic device, comprising:

[0066] A memory for storing executable instructions;

[0067] The processor is used to implement the above-mentioned nonlinear model determination method of the visible light communication system based on PN sequence when executing the executable instructions stored in the storage.

[0068] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the nonlinear model discrimination method of a visible light communication system based on a PN sequence is executed.

[0069] The beneficial effects of the present invention are:

[0070] The present invention is directed to a specific visible light communication system. For the Wiener nonlinear model and the Hammerstein model, the multi-level PN sequence can be used to quickly obtain parameter estimates of different models; based on the estimation error of the trained nonlinear model for the received signal, the fitting performance of different nonlinear models for the system nonlinearity can be compared, and the optimal model parameter estimates of Wiener and Hammerstein can be obtained respectively; by comparing the optimal fitting performance of Wiener and Hammerstein, it can be determined whether the system is a Wiener model type or a Hammerstein model type. In addition, the proposed nonlinear model parameter estimation algorithm based on the multi-level PN sequence is easy to implement, and synchronization and channel estimation can be achieved simultaneously. Moreover, since the PN sequence has orthogonality, the channel model parameter estimation of multiple users and the synchronization of the transceiver can also be achieved simultaneously. The nonlinear model type discrimination method of the visible light communication system proposed by the present invention has far-reaching significance for the in-depth exploration of the nonlinear mechanism of the visible light communication system, lays a solid foundation for the research of related technologies, and provides a model hypothesis basis for the research of nonlinear distortion compensation and pre-distortion technologies in visible light communication and even radio frequency communication, and has high application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 It is a flow chart of a method for qualitatively judging the type of nonlinear model of a visible light communication system;

[0072] Figure 2 It is a schematic diagram of signal transformation of Hammerstein and Weiner model;

[0073] Figure 3 is a block diagram of a visible light communication system used to specifically demonstrate the implementation of the invention;

[0074] Figure 4 This is the NMSE estimation performance graph of the Hammerstein nonlinear model. It can be seen from the figure that the third-order nonlinear Hammerstein model with a memory depth of 50 has the best estimation performance;

[0075] Figure 5 This is the NMSE estimation performance graph of the Wiener nonlinear model. It can be seen from the figure that the 4th-order nonlinear 59-memory depth Wiener model has the best estimation performance;

[0076] Figure 6 This is a comparison of the nonlinear transfer functions of the Hammerstein and Wiener models with the best performance after training on a broadband visible light communication system. It can be seen that the nonlinear curves of the Hammerstein and Wiener models are very close;

[0077] Figure 7 This is a comparison of the linear time-invariant system frequency responses of the Hammerstein and Wiener models with the best performance after training on a broadband visible light communication system. It can be seen that the linear time-invariant responses of the Hammerstein and Wiener models are also very close, and both have a low-pass effect;

[0078] Figure 8 The following is a comparison of the received signal estimation waveforms obtained by using the Hammerstein and Wiener models with the best performance after training to estimate the received signals in the evaluation data set. It can be clearly seen that the received signal estimated by the Hammerstein model is closer to the actual received signal.

[0079] Fig. 9 This is a comparison chart of the estimated waveforms of the transmitted signals obtained by using the Hammerstein and Wiener models with the best performance after training to estimate the transmitted signals in the evaluation data set using the actual received signals. It can be clearly seen that the transmitted signal reconstructed by Hammerstein is closer to the actual transmitted signal. DETAILED DESCRIPTION

[0080] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are 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 creative work are within the scope of protection of the present invention.

[0081] Example 1

[0082] See also Figure 1-Figure 9This embodiment provides a nonlinear model identification method for a visible light communication system based on a PN sequence. The specific process of the method is as follows: Figure 1 As shown, the specific steps include:

[0083] Step S1: In a broadband visible light communication system, a training set, an evaluation set and a verification set are generated using a multi-level PN sequence, wherein the training set, the evaluation set and the verification set are used to train and evaluate a Hammerstein and Wiener nonlinear model;

[0084] Specifically, in this embodiment, in step S1, the broadband visible light communication system has a structure Figure 3 As shown, it specifically includes:

[0085] At the transmitting end, the computer's MATLAB and other software generates digital signals, which are then transmitted to the programmable FPGA via Ethernet and converted to analog signals by the digital-to-analog converter DAC. The DAC sampling rate is 160Msps and the accuracy is 16bit. The analog signal is amplified by the power amplifier and then the bias-tee integrated circuit adds a DC bias to drive the LED to emit light.

[0086] At the receiving end, the light beam is first focused by a lens and filtered by a grating. The filtered light signal is detected by an avalanche diode and converted into a current signal. The current signal is converted into a voltage signal through a resistor. Then the voltage signal is amplified by a low-noise amplifier, a voltage-controlled amplifier and a buffer amplifier. The amplified signal is sampled by an analog-to-digital converter ADC and converted into a digital signal. The sampling frequency of the ADC is also 160Msps and the accuracy is 14bit. Finally, the digital signal is transmitted back to the computer via Ethernet.

[0087] Specifically, in this embodiment, in step S1, the specific structure of the Hammerstein and Wiener nonlinear model is as follows: Figure 2 As shown, it includes:

[0088] Both Hammerstein and Wiener nonlinear models consist of a memoryless nonlinear module and a linear time-invariant system;

[0089] Assuming that the input signal is x(n), the L-order nonlinear transfer function is φ(·), where The response of the linear time-invariant discrete system with a memory depth of D is g(n), q(n) is the intermediate signal, and v(n) is the background white noise. The digital received signal of the Hammerstein model is r(n)=φ(x(n))*g(n)+v(n), where * represents linear convolution. The digital received signal of the Wiener model is r i (n) = φ(q i (n))+v(n).

[0090] Specifically, in this embodiment, in step S1, a training set, an evaluation set, and a verification set are generated by using a multi-level PN sequence by the following method, which includes:

[0091] This embodiment sets the length N of the PN sequence z(n) in the training set c =2 15 -1, there are five different amplitudes, namely N = 5, a1 = 5 000, a2 = 19 000, a3 = 23 000, a4 = 26 000, a5 = 28 000, respectively for the multi-level PN sequence x i (n) = a i z(n) is upsampled by 6 times to generate the corresponding transmitted digital signal, and the receiving end downsamples the received signal by 6 times to generate the corresponding received signal r i (n), where i∈Z + And 1≤i≤N, by x i (n), r i (n) constitute the training set.

[0092] This embodiment sets the length N of the PN sequence z(n) in the evaluation set c =4000, there are 4 different amplitudes, namely N = 4, a1 = 12 000, a2 = 15 000, a3 = 20 000, a4 = 22 000, respectively for the multi-level PN sequence x i (n) = a i z(n) is upsampled by 6 times to generate the corresponding transmitted digital signal, and the receiving end downsamples the received signal by 6 times to generate the corresponding received signal r i (n), where i∈Z + And 1≤i≤N, by x i (n), r i (n) constitute the evaluation set.

[0093] This embodiment sets the length N of the PN sequence z(n) in the validation set c =4000, there are 4 different amplitudes, namely N = 4, a1 = 12 000, a2 = 15 000, a3 = 20 000, a4 = 22 000, respectively for the multi-level PN sequence x i (n) = a i z(n) is upsampled by 6 times to generate the corresponding transmitted digital signal, and the receiving end downsamples the received signal by 6 times to generate the corresponding received signal r i (n), where i∈Z + And 1≤i≤N, by x i (n), r i (n) constitutes the validation set.

[0094] Step S2, using the training set obtained in step S1 to train and evaluate the Hammerstein and Wiener nonlinear models respectively, and then respectively calculating the minimum evaluation error and the corresponding training error of the two models;

[0095] Specifically, in this embodiment, step S2 specifically includes:

[0096] Step S201: Set the maximum nonlinear order L max =5 and maximum memory depth D max = 80, initialize the nonlinear order L = 0 and the memory depth D = 0, Hammerstein model normalized estimation error ε H =1(dB), normalized estimation error of Wiener model ε W =1(dB);

[0097] Step S202: If L<L max , L=L+1, otherwise, execute step S206;

[0098] Step S203: If D < D max , D=D+1, otherwise, execute step S202;

[0099] Step S204: First, a parameter estimation algorithm of a Hammerstein nonlinear model is executed, and a Hammerstein model with a nonlinear order of L and a memory depth of D is trained using a training set to obtain an estimate of the model parameters. and Then use the parameter and The Hammerstein model and the multi-stage PN sequence in the evaluation data set estimate the corresponding received signal sequence Calculate the estimated error in,

[0100]

[0101]

[0102] if but ε H =ε h ;

[0103] Step S205: First, a parameter estimation algorithm of a Wiener nonlinear model is executed, and a Wiener model with a nonlinear order of L and a memory depth of D is trained using a training data set to obtain an estimate of the model parameters. and Then use the parameters and The Wiener model and the multi-stage PN sequence in the evaluation data estimate the corresponding received signal sequence Calculate the estimated error in,

[0104]

[0105]

[0106] if but ε W =ε w , execute step S203;

[0107] Step S206: Calculate the minimum Hammerstein estimation error ε in the evaluation phase H =-20.152 1(dB), the corresponding nonlinear order is L * =3, memory depth is D * =50, the estimated nonlinear conversion function is φ(y)=-0.0144+0.7584y+0.0764y 2 -0.3844y 3 , the estimated coefficients of the nonlinear module are The frequency domain amplitude response of Figure 6 As shown;

[0108] Step S207: Calculate the minimum Hammerstein estimation error ε in the evaluation phase W =-16.205 7 (dB), the corresponding nonlinear order is L * =4, memory depth is D * =59, the estimated nonlinear conversion function is φ(y)=-0.0608+0.9065y+0.1028y 2 -0.5124y 3 -7.4125 10 -5 y 4 , the nonlinear coefficient is The frequency domain amplitude response of Figure 6 As shown;

[0109] Step S208: According to the above-mentioned parameter estimation algorithm of the Hammerstein nonlinear model, the parameters are and The Hammerstein model and the multi-level PN sequence in the validation data are used to calculate the estimation of the corresponding received signal and verify the model estimation error ε in the phase Htest =-20.2(dB);

[0110] Step S209: According to the above-mentioned parameter estimation algorithm of the Wiener nonlinear model, the parameters are and The Wiener model and the multi-stage PN sequence in the verification data are used to calculate the estimation of the corresponding received signal and verify the model estimation error ε in the stage Wtest =-16.200 6 (dB).

[0111] More specifically, in this embodiment, the parameter estimation algorithm of the Hammerstein nonlinear model mentioned in the above step S204 and step S208 specifically includes:

[0112] (1) This method uses a multi-level pseudorandom noise (PN) sequence and its digital received signal to complete parameter estimation: where the L-order nonlinear transfer function is φ(·), is the parameter of the nonlinear function φ(·), the linear time-invariant discrete system response with a memory depth of D is g(n), and the N-segment multi-level PN sequence is x i (n) = a i z(n),a i ∈R,i∈Z + and 1≤i≤N, z(n) is a unit power with a mean of zero and a length of N c Pseudo-random noise PN sequence with amplitude a i ≠a j , i≠j, in the communication system, a multi-level PN sequence is sent and the digital received signal r is stored i (n);

[0113] (2) Calculate the received signal r i The cross-correlation sequence of (n) and PN sequence z(n):

[0114]

[0115] (3) Calculate the autocorrelation sequence of the PN sequence z(n):

[0116]

[0117] (4) Calculate r sz (k) = A -1 r rz (k), where

[0118]

[0119]

[0120]

[0121] (5) Solve the relationship using deconvolution Obtain an estimated sequence of linear time-invariant system responses

[0122] (6) Solve the relationship by deconvolution Get the estimated sequence

[0123] (7) Estimate the parameters of the nonlinear transfer function φ(·) using the least squares method: in,

[0124]

[0125]

[0126]

[0127] More specifically, in this embodiment, the parameter estimation algorithm of the Wiener nonlinear model mentioned in the above step S205 and step S209 specifically includes:

[0128] (1) This method uses a multi-level PN sequence and its digital received signal to complete parameter estimation; the L-order nonlinear transfer function is φ(·), is the parameter of the nonlinear function, the linear time-invariant discrete system response with a memory depth of D is g(n), and the N-segment multi-level PN sequence is x i (n) = a i z(n),a i ∈R,i∈Z + and 1≤i≤N, z(n) is a unit power with a mean of zero and a length of N c PN sequence with amplitude a i ≠a j , i≠j, the communication system sends a multi-level PN sequence and stores the digital received signal r i (n);

[0129] (2) Calculate the received signal r i The cross-correlation sequence of (n) and PN sequence z(n):

[0130]

[0131] (3) Calculate the autocorrelation sequence of the PN sequence z(n):

[0132]

[0133] (4) Calculate r wz (k) = A -1 r rz (k), where

[0134]

[0135]

[0136]

[0137] (5) Solve the relationship using deconvolution Obtain an estimated sequence of linear time-invariant system responses

[0138] (6) Calculation Get intermediate signal sequence estimates;

[0139] (7) Estimate the parameters of the nonlinear transfer function φ(·) using the least squares method: in

[0140]

[0141]

[0142]

[0143] Step S3: Compare ε H With ε W and ε Htest With ε Wtest If the size relationship is consistent, it is defined as valid and step S4 is executed; otherwise, step S1 is executed;

[0144] Step S4: If ε H <ε W , then the Hammerstein model is more suitable for describing the nonlinearity of the system. If ε H >ε W , then the Wiener model is more suitable for the nonlinear system. If ε H =ε W , then the two nonlinear models have equivalent effects.

[0145] Specifically, in this embodiment, there is ε H <ε W and ε Htest <ε Wtest , that is, the Hammerstein nonlinear model is more suitable for the system.

[0146] The matters not described in detail in the present invention are all known technologies to those skilled in the art.

[0147] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A nonlinear model identification method for visible light communication system based on PN sequence, It is characterized in that The method includes: Step S1: In a broadband visible light communication system, a data set is generated using a multi-stage PN sequence, and the data set includes: a training set, an evaluation set, and a verification set, wherein the training set, the evaluation set, and the verification set are used to train and evaluate the Hammerstein and Wiener nonlinear model; Step S2, using the data set obtained in step S1 to train and evaluate the Hammerstein and Wiener nonlinear models respectively, and then respectively calculating the minimum evaluation error and the corresponding training error of the two models; Step S3, for the minimum evaluation error and training error corresponding to the two models calculated in step S2, perform the following operations: compare the size relationship between the minimum evaluation error of the two models and the corresponding training error. If the size relationship is consistent, it is defined as valid and step S4 is executed; otherwise, step S1 is executed; Step S4, perform the following judgment: if the minimum evaluation error of the Hammerstein model is less than the minimum evaluation error of the Wiener nonlinear model, then the Hammerstein model is more suitable for describing the system nonlinearity; if the minimum evaluation error of the Hammerstein model is greater than the minimum evaluation error of the Wiener nonlinear model, then the Wiener model is more suitable for the system nonlinearity; if the minimum evaluation errors of the two models are equal, then the two nonlinear models have equivalent effects; The step S2 specifically includes: Step S201: Set the maximum nonlinear order L max and the maximum memory depth D max , initialize the nonlinear order L = 0 and the memory depth D = 0, the Hammerstein model normalized estimation error ε H =1dB, normalized estimation error of Wiener model ε W =1dB; Step S202: If L<L max , L=L+1, otherwise, execute step S206; Step S203: If D < D max , D=D+1, otherwise, execute step S202; Step S204: First, a parameter estimation algorithm of a Hammerstein nonlinear model is executed, and a Hammerstein model with a nonlinear order of L and a memory depth of D is trained using a training set to obtain an estimate of the model parameters. and Then use the parameter and The Hammerstein model and the multi-stage PN sequence in the evaluation data set estimate the corresponding received signal sequence Calculate the estimated error in, if but ε H =ε h ; Step S205: First, a parameter estimation algorithm of a Wiener nonlinear model is executed, and a Wiener model with a nonlinear order of L and a memory depth of D is trained using a training data set to obtain an estimate of the model parameters. and Then use the parameters and The Wiener model and the multi-stage PN sequence in the evaluation data estimate the corresponding received signal sequence Calculate the estimated error in, if but ε W =ε w , execute step S203; Step S206: According to the method in step S204, the parameters are and The Hammerstein model and the multi-stage PN sequence in the verification data estimate the corresponding received sequence, and calculate the model estimation error ε in the verification stage Htest ; Step S207: According to the method in step S205, the parameters are and The Wiener model and the multi-stage PN sequence in the verification data estimate the corresponding received sequence, and calculate the model estimation error ε in the verification stage Wtest ; In step S2, the parameter estimation algorithm of the Hammerstein nonlinear model includes: (1) Parameter estimation is performed using a multi-level PN sequence and its digital received signal: where the L-order nonlinear transfer function is φ(·), is the parameter of the nonlinear function φ(·), the linear time-invariant discrete system response with a memory depth of D is g(n), and the N-segment multi-level PN sequence is x i (n) = a i z(n),a i ∈R,i∈Z + and 1≤i≤N, z(n) is a unit power with a mean of zero and a length of N c Pseudo-noise (PN) sequence with amplitude a i ≠a j , i≠j, in the communication system, a multi-level PN sequence is sent and the digital received signal r is stored i (n); (2) Calculate the received signal r i The cross-correlation sequence of (n) and PN sequence z(n): (3) Calculate the autocorrelation sequence of the PN sequence z(n): (4) Calculate r sz (k) = A -1 r rz (k), where (5) Solve the relationship using deconvolution Obtain an estimated sequence of linear time-invariant system responses (6) Solve the relationship by deconvolution Get the estimated sequence (7) Estimate the parameters of the nonlinear transfer function φ(·) using the least squares method: in, In the step S2, the parameter estimation algorithm of the Wiener nonlinear model includes: (1) Parameter estimation is performed using a multi-level PN sequence and its digital received signal; the L-order nonlinear transfer function is φ(·), is the parameter of the nonlinear function, the linear time-invariant discrete system response with a memory depth of D is g(n), and the N-segment multi-level PN sequence is x i (n) = a i z(n),a i ∈R,i∈Z + and 1≤i≤N, z(n) is a unit power with a mean of zero and a length of N c PN sequence with amplitude a i ≠a j , i≠j, the communication system sends a multi-level PN sequence and stores the digital received signal r i (n); (2) Calculate the received signal r i The cross-correlation sequence of (n) and PN sequence z(n): (3) Calculate the autocorrelation sequence of the PN sequence z(n): (4) Calculate r wz (k) = A -1 r rz (k), where (5) Solve the relationship using deconvolution Obtain an estimated sequence of linear time-invariant system responses (6) Calculation Get intermediate signal sequence estimates; (7) Estimate the parameters of the nonlinear transfer function φ(·) using the least squares method: in, 2. According to claim 1, a nonlinear model discrimination method for a visible light communication system based on a PN sequence, It is characterized in that In step (1), N ≥ L, L < < (N × N c ).

3. According to claim 1, a nonlinear model discrimination method for a visible light communication system based on a PN sequence, It is characterized in that In step (1), N ≥ L, L < < (N × (D + N c -1)).

4. An electronic device, It is characterized in that The electronic device comprises: A memory for storing executable instructions; The processor is used to implement the nonlinear model discrimination method of a visible light communication system based on a PN sequence as described in any one of claims 1 to 3 when executing the executable instructions stored in the storage.

5. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the nonlinear model identification method of a visible light communication system based on a PN sequence as described in any one of claims 1 to 3 is executed.

Citation Information

Patent Citations

  • Leader sequence of visible light DCO-OFDM communication system, nonlinear amplitude limit estimation method and channel parameter estimation method

    CN106100813A

  • Visible light communication system nonlinear channel modeling method based on neural network

    CN111245512A