Visible light communication pre-distortion method based on twin support vector regression

Abstract

The application provides a visible light communication pre-distortion method based on twin support vector regression, which comprises the following steps: step S1, a visible light communication offline digital pre-distortion platform is built; step S2, based on the signal information obtained by the visible light communication offline digital pre-distortion platform in step S1, a training set of an amplitude enhancement structure is constructed; step S3, based on the training set of the amplitude enhancement structure constructed in step S2, a visible light communication pre-distortion nonlinear suppression method based on amplitude TSVR is established, and a test set is formed; step S4, the training set of the training set of the amplitude enhancement structure and the test set are used for model training; step S5, after model training, pre-distortion device parameters are obtained, and a visible light communication pre-distortion system is further designed and constructed. The amplitude enhancement-based TSVR time delay structure is obviously improved in modeling accuracy, compared with the existing memory polynomial modeling, the NSME is improved by more than 22 dB; compared with the generalized memory polynomial, the accuracy is improved by about 5.5 dB on average.

Description

Technical Field

[0001] This invention belongs to the field of visible light communication technology, specifically relating to a visible light communication predistortion method based on twin support vector regression. Background Technology

[0002] With traditional radio frequency (RF) spectrum resources becoming increasingly limited, and with advancements in light-emitting diode (LED) chip design technology and the widespread use of LED devices, research into visible light communication (VLC) technology has gradually attracted widespread attention. However, visible light communication (VLC) systems contain many nonlinear components, which can cause signal distortion and limit the system performance of visible light communication (VLC).

[0003] Furthermore, as transmitters in visible light communication (VLC) systems, LED devices exhibit a non-linear relationship between LED current and optical power during photoelectric conversion, and are subject to saturation power limitations. Simultaneously, for high-speed, high-bandwidth signal transmission, LEDs also display severe nonlinear memory effects. These multiple factors make LEDs a major source of nonlinearity. Therefore, mitigating nonlinearity is a crucial aspect of VLC system design.

[0004] Furthermore, with the development of visible light communication, some higher-order signal modulation methods have been introduced into visible light communication (VLC) systems, most notably OFDM signals. OFDM has a high peak-to-average power ratio, and these OFDM signals are highly sensitive to nonlinearity. Coupled with the limited bandwidth of LEDs, the frequency response in visible light communication (VLC) systems exhibits a low-pass characteristic, resulting in more severe distortion. Furthermore, it causes in-band distortion and out-of-band interference during transmission, ultimately degrading the performance of the entire visible light communication system and limiting the bit error rate.

[0005] Generally, there are two ways to mitigate nonlinear distortion in visible light communication (VLC) systems. The most straightforward approach is to design the waveform to be insensitive to nonlinearity, such as reducing PAPR or using bipolar modulation. Another approach is to design distortion compensation methods, typically including post-distortion and pre-distortion. By adding equalizers or pre-distortors, the distorted signal is compensated and corrected, thereby mitigating nonlinearity. Post-distortion is designed at the receiver, consuming significant computational resources and increasing energy consumption during transmission. Conversely, pre-distortion is designed at the transmitter, offering simplicity and effectiveness, and thus attracting increasing attention. Furthermore, almost all pre-distortion techniques are based on behavioral models; therefore, employing appropriate models to model the behavior of LEDs is crucial.

[0006] In existing technologies, Voterra series are often used to increase the nonlinear order to improve accuracy, but this also increases the computational complexity of modeling. Furthermore, the memory effect exhibited by LEDs means that Voterra series cannot adequately characterize the system, and its performance improvement for optical communication systems is limited. Nonlinear compensation methods based on neural network predistorters suffer from overfitting during training and modeling, leading to modeling failures. Additionally, neural networks are sensitive to noise and easily affected by system channel conditions. Therefore, existing technologies suffer from insufficient modeling accuracy for nonlinear transmission characteristics and memory effects in visible light communication, and have limited ability to suppress nonlinearity. Summary of the Invention

[0007] Based on the technical problems existing in the prior art, the present invention provides a predistortion method for visible light communication based on twin support vector regression. By adopting a time delay structure and taking into account the intensity modulation / direct detection characteristics of visible light communication itself, the method further incorporates amplitude as a training feature, thereby solving the technical problems existing in the prior art.

[0008] To achieve the above objectives, this invention provides a visible light communication predistortion method based on twin support vector regression, which includes the following steps:

[0009] Step S1: Build an offline digital predistortion platform for visible light communication;

[0010] Step S2: Based on the signal information obtained from the visible light communication offline digital predistortion platform in Step S1, construct a training set for the amplitude enhancement structure;

[0011] Step S3: Based on the training set of the amplitude enhancement structure constructed in step S2, establish a predistortion nonlinear suppression method for visible light communication based on amplitude TSVR, and form a test set;

[0012] Step S4: Use the training set and test set used to construct the amplitude enhancement structure to train the model;

[0013] Step S5: After model training, the predistorter parameters are obtained, and a visible light communication predistortion system is further designed and constructed.

[0014] In step S1, the Wiener model is used to describe the nonlinear transmission characteristics and memory effect of the visible light communication system, and the measured input signal x(n) and output signal y(n) of the LED are obtained.

[0015] Further, in step S2, after normalizing the input signal x(n) and the output signal y(n), a training set for the amplitude enhancement structure is constructed, and the training set is operated on based on the TSVR algorithm.

[0016] Preferably, the Wiener model is decomposed into a cascade of a linear time-invariant system and a nonlinear system, and a time-delay tap structure is used to introduce memory nonlinearity for the memory property of LEDs.

[0017] More preferably, the linear time-invariant module describing the memory effect of a linear time-invariant system is represented as:

[0018]

[0019] Where L is the maximum delay tap, b l It is the memory effect factor, x(nl) represents the delay term of the input variable, x(n) represents the input signal, and l represents the delay depth.

[0020] More preferably, the Rapps model describing the nonlinear modules of a nonlinear system is as follows:

[0021]

[0022] Among them, I max V is the current maximum output, k is the inflection point coefficient, which controls the smoothness from the linear region to the saturation region. TOV It is the turn-on voltage value of the light-emitting diode.

[0023] Furthermore, the TSVR algorithm has the following representation relationship:

[0024]

[0025] CC structure, such as Figure 1 As shown in the diagram, f1(x) is the upper bound function, f2(x) is the lower bound function, abs(x(n)) represents the signal amplitude, x(n) represents the input signal, x(n-1) represents the first-order delay of the signal, x(n-2) represents the second-order delay of the signal, and TSVR Machine represents the constructed twin support vector regression machine. The training set is trained by the TSVR Machine to obtain the predicted output signal.

[0026] Preferably, in step S3, a method for suppressing nonlinear predistortion in visible light communication based on amplitude TSVR is established. This method involves cascading the TSVR predistorter with the transmitter of the visible light communication system to form an approximately linear system, thereby compensating for nonlinearity and improving the overall performance of the system.

[0027] More preferably, in the TSVR predistorter, the bit stream signal to be transmitted is first mapped by 16QAM. Visible light communication uses intensity modulation / direct detection. After Hermite mapping, it is converted into a real signal by IFFT (Inverse Fast Fourier Transform). Then, a prefix and guard interval are added. After parallel-to-serial conversion, the signal passes through the TSVR predistorter. After digital-to-analog conversion, the DC signal is coupled to the signal to be transmitted and loaded onto the LED through a Bias-Tee.

[0028] Furthermore, at the TSVR predistorter receiver, an APD detector is used to photoelectrically convert the received signal for offline processing.

[0029] Compared with the prior art, the beneficial effects of the present invention are:

[0030] 1. This invention presents a visible light communication predistortion method based on twin support vector regression, which significantly improves modeling accuracy based on amplitude-enhanced TSVR delay structure. Compared with existing memory polynomial (MP) modeling, NSME is improved by more than 22 dB; compared with generalized memory polynomial (GMP), the accuracy is improved by an average of about 5.5 dB.

[0031] 2. This invention presents a visible light communication predistortion method based on Siamese Support Vector Regression (TSVR). The TSVR-based modeling method significantly improves modeling speed compared to traditional SVR. When M=2, the NMSE index of TSVR is 16.5403dB higher than that of SVR. The CPU computation efficiency is at least four times that of SVR.

[0032] 3. This invention has significant advantages in mitigating the low-pass effect of LEDs and also has a strong suppression effect on the spectral interference of out-of-band signals. The low-pass portion of in-band signals is compensated, while the regeneration of out-of-band spectra is also suppressed.

[0033] 4. The TSVR-based modeling method of this invention achieves excellent power spectral density, constellation diagram, and bit error rate in compensating for nonlinearity. For example, Figure 7 For the receiver constellation diagram without pre-distortion, from Figure 7 As can be seen, the constellation chart has been rotated, the amplitude has been distorted, and the entire constellation chart appears relatively blurry. Figures 8-10 This is achieved by using a pre-distorted constellation chart, and... Figure 7 Compared to the previous version, significant performance improvements have been achieved. To further verify the effectiveness of this solution, Figure 11 The graph shows a comparison of bit error rates. As can be seen from the graph, TSVR exhibits superior performance at high signal-to-noise ratios, while it is similar to GMP at low signal-to-noise ratios. Attached Figure Description

[0034] Figure 1 This is a schematic diagram of the amplitude-enhanced TSVR used for LED modeling according to the present invention;

[0035] Figure 2 This is a schematic diagram of a visible light communication system with a predistorter;

[0036] Figure 3 It is the input / output AM / AM curve;

[0037] Figure 4 This is the transmission characteristic curve of the Weiner model;

[0038] Figure 5 It is the power spectral density map of the output signal based on TSVR modeling;

[0039] Figure 6 It is a time-domain matching diagram of the output signal and the actual signal based on TSVR modeling;

[0040] Figure 7 It is the constellation diagram of the output signal before pre-distortion;

[0041] Figure 8 It is the constellation diagram of the output signal after predistortion using memory polynomials;

[0042] Figure 9 It is the constellation diagram of the output signal after predistortion by generalized memory polynomial;

[0043] Figure 10 This is the constellation diagram of the output signal after TSVR predistortion;

[0044] Figure 11 This is a comparison chart of bit error rate curves;

[0045] Figure 12 This is a comparison table of modeling accuracy for visible light communication systems modeled according to the present invention. Detailed Implementation

[0046] 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.

[0047] This invention proposes a predistortion method for visible light communication based on Siamese Support Vector Regression (TSVR). It is an adaptive predistortion method for visible light communication systems based on amplitude-enhanced TSVR. The method uses an amplitude-enhanced time delay structure to model the LED. After establishing the LED model based on amplitude-enhanced TSVR, the model is inverted to obtain the predistorter, which is then cascaded with the LED at the transmitting end of the visible light communication system to obtain the corresponding visible light communication system with the predistorter. Compared with existing memory polynomial and Chebyshev polynomial systems, the visible light communication system obtained by this invention shows significant improvements in modeling accuracy and bit error rate performance.

[0048] The following is in conjunction with the appendix Figure 1 -Appendix Figure 12 This invention will be explained in detail as a visible light communication predistortion method based on twin support vector regression.

[0049] This invention provides a predistortion method for visible light communication based on twin support vector regression, which includes the following steps:

[0050] Step S1: Build an offline digital predistortion platform for visible light communication; at the same time, use the Wiener model to describe the nonlinear transmission characteristics and memory effect of the visible light communication system, and obtain the measured input signal x(n) and output signal y(n) of the LED.

[0051] Step S2: Based on the signal information obtained from the visible light communication offline digital predistortion platform in Step S1, a training set for the amplitude enhancement structure is constructed; specifically, after normalizing the input signal x(n) and the output signal y(n), a training set for the amplitude enhancement structure is constructed, and the training set is calculated based on the TSVR algorithm.

[0052] Step S3: Based on the training set of the amplitude enhancement structure constructed in step S2, establish a predistortion nonlinear suppression method for visible light communication based on amplitude TSVR, and form a test set;

[0053] Step S4: Use the training set and test set used to construct the amplitude enhancement structure to train the model;

[0054] Step S5: After model training, the predistorter parameters are obtained, and a visible light communication predistortion system is further designed and constructed.

[0055] In a specific embodiment, step S1 involves building an offline digital predistortion platform for visible light communication. Simultaneously, the Wiener model is used to describe the nonlinear transmission characteristics and memory effect of the visible light communication system, and the measured input signal x(n) and output signal y(n) of the LED are obtained. The Wiener model can be decomposed into a cascade of a linear time-invariant system and a nonlinear system. For the memory property of the LED, a time-delay tap structure is used to introduce memory nonlinearity. The linear time-invariant module of the linear time-invariant system describing the memory effect is represented as follows:

[0056]

[0057] Among them, I max V is the current maximum output, k is the inflection point coefficient, which controls the smoothness from the linear region to the saturation region. TOV It is the turn-on voltage value of the light-emitting diode.

[0058] The nonlinear modules describing nonlinear systems can be described using the Rapps model. To model the VI curve of an LED, the Rapps model, after some simple modifications, yields the following results:

[0059]

[0060] Among them, I max V is the current maximum output, k is the inflection point coefficient, which controls the smoothness from the linear region to the saturation region. TOV It is the turn-on voltage value of the light-emitting diode.

[0061] Step S2: After normalizing the input signal x(n) and the output signal y(n), a training set for the amplitude enhancement structure is constructed. The TSVR algorithm is then used for computation, resulting in the following relationship:

[0062]

[0063] in Indicates amplitude, Indicates a delay tap, structured as follows: Figure 1 As shown in the diagram, f1(x) is the upper bound function, f2(x) is the lower bound function, abs(x(n)) represents the signal amplitude, x(n) represents the input signal, x(n-1) represents the first-order delay of the signal, x(n-2) represents the second-order delay of the signal, and TSVR Machine represents the constructed twin support vector regression machine. The training set is trained by the TSVR Machine to obtain the predicted output signal.

[0064] Step S3: Establish a predistortion nonlinearity suppression method for visible light communication based on amplitude-dependent TSVR (Transmitter-Switch Transformer). This method cascades the TSVR predistorter with the transmitter of the visible light communication system to form an approximately linear system, thereby compensating for nonlinearity and improving the overall system performance. The location of the TSVR predistorter at the transmitter and the visible light communication system with the predistorter are shown in Figure S3. Figure 2 As shown. First, the bitstream signal to be transmitted undergoes 16QAM mapping. Since visible light communication uses intensity modulation / direct detection, it needs to undergo Hermite mapping, then IFFT (Inverse Fast Fourier Transform) to become a real number signal. Then, a prefix and guard interval are added, the signal is converted from parallel to serial, then passes through a TSVR predistorter, and after digital-to-analog conversion, the DC signal is coupled to the signal to be transmitted via a Bias-Tee and applied to the LED. At the receiving end, an APD detector is used to photoelectrically convert the received signal for offline processing. Furthermore, in... Figure 2 The predistorter is set using an indirect learning structure. The parameters of the predistorter are estimated at Estimation(A) and then copied to the Predistorter.

[0065] The predistortion nonlinearity suppression method for visible light communication based on amplitude-enhanced TSVR compensates for nonlinear effects by adding an inverse model response at the transmitter, so that the final signal response exhibits a linear state.

[0066] The nonlinear amplitude distortion correction effect of the visible light communication predistortion nonlinear suppression method based on amplitude enhancement TSVR is as follows: Figure 3 As shown. By Figure 3 It can be observed that the signal without predistortion processing exhibits strong nonlinearity and memory effect, while after predistortion, the amplitude distortion curve is corrected to be approximately linear.

[0067] Step S4: Train the model using the training and test sets. (See below) Figure 1 As shown, it specifically includes the following steps:

[0068] ① A behavioral model of LEDs is constructed using TSVR. By introducing a kernel function, the original TSVR expression is extended to the nonlinear case. The regression function of the entire model is as follows:

[0069]

[0070] in Represents the kernel function, x i ,x j Let b1, ω2, b2, b1 ∈ R be the sample points. nEach row of matrix A represents a set of training data. Using kernel function theory, we can further derive the following equivalent objective function:

[0071]

[0072]

[0073] Where ε1 and ε2 represent insensitive parameters, ξ * ξ represents slack variables, e is the identity matrix, Y is the response output signal, and C1 and C2 are penalty factors;

[0074] ② By using the Lagrange multiplier method, we obtain

[0075]

[0076] ③ Further, we can obtain the dual forms of (5) and (6) as follows:

[0077]

[0078]

[0079] Where H=[κ(A,A T [e], where α, μ, β are Lagrange multipliers, I is the identity matrix, and C3 and C4 represent the penalty factors. The regression function that ultimately yields the expected value can be expressed as the average of the upper and lower bound functions, and its functional expression is shown below:

[0080]

[0081] Where f1(x) is the upper bound function, f2(x) is the lower bound function, abs(x(n)) represents the signal amplitude, x(n) represents the input signal, x(n-1) represents the first-order delay of the signal, x(n-2) represents the second-order delay of the signal, and TSVR Machine represents the constructed twin support vector regression machine.

[0082] Step S5 involves training the model to obtain the predistorter parameters, and then designing and constructing a visible light communication predistortion system. This process includes the following sub-steps:

[0083] The input-output relationship of a predistorter cascaded with a visible light communication transmitter is approximately linear. That is, before the transmitted signal is converted into an analog signal, it undergoes predistortion processing to produce a distortion opposite to that of the LED, compensating for the LED's nonlinear memory characteristics. This invention uses an indirect learning structure for predistortion. First, the LED's input and output information are used as the model's input and output information to perform inverse modeling of the LED. Then, the calculated LED post-inverse model parameters are copied into the predistorter as pre-inverse parameters. The predistorter and power amplifier are cascaded to construct a complete predistortion system. For example... Figure 2 First, the inverse model parameters are estimated at Estimation(A), i.e., the parameters are obtained through training using the method described in step S4. After training, the obtained parameter results are set to... Figure 2 At TSVR(A), a visible light communication predistortion system based on TSVR with an indirect learning structure is thus constructed.

[0084] In the above steps, Figure 3 It is the input / output AM / AM curve; Figure 3 The diagram compares the input and output amplitude distortion before and after TSVR predistortion. The relationship between the input and output amplitudes after predistortion is approximately linear, while the curve with scattered points is the same as that without predistortion. This illustrates the nonlinearity and memory effect inherent in LEDs. The comparison demonstrates the effectiveness of the proposed method.

[0085] Figure 4 It is the transmission characteristic curve of the Weiner model; Figure 4 The Wiener memory model is used for data simulation, which describes the nonlinear transformation relationship of LEDs.

[0086] Figure 5 It is the power spectral density map of the output signal based on TSVR modeling; Figure 5 The figure shows a comparison of the power spectral density of the LED model built using TSVR. As can be seen from the figure, the established model has high accuracy, the actual output of the LED is highly matched with the output of the model, and the error signal is very small. Figure 6 It is a time-domain matching diagram of the output signal and the actual signal based on TSVR modeling; Figure 6 The time-domain waveforms of the model output and the actual output are in high agreement, and the two figures together illustrate the accuracy and effectiveness of the proposed model.

[0087] Figures 7 to 10 This represents a comparison of constellation diagrams obtained using different pre-distortion schemes. Constellation diagrams without pre-distortion are blurry and have severe amplitude distortion, while those with pre-distortion show performance improvements to some extent. Figure 7 It is the constellation diagram of the output signal without pre-distortion. Figure 8 It is the constellation diagram of the output signal after memory polynomial predistortion. Figure 9 It is the constellation diagram of the output signal after generalized memory polynomial predistortion. Figure 10 This is the constellation diagram of the output signal after TSVR predistortion. For example, Figure 7 For the receiver constellation diagram without pre-distortion, from Figure 7 As can be seen, the constellation chart has been rotated, the amplitude has been distorted, and the entire constellation chart appears relatively blurry. Figures 8-10This is achieved by using a pre-distorted constellation chart, and... Figure 7 Compared to the previous version, there has been a significant performance optimization.

[0088] To further verify the effectiveness of this scheme, Figure 11 The graph shows a comparison of bit error rates. As can be seen, TSVR exhibits superior performance at high signal-to-noise ratios (SNR), while remaining comparable to GMP at low SNRs. In other words, Figure 11 The comparison of bit error rates using different predistortion schemes shows that TSVR and GMP perform similarly at low signal-to-noise ratios, while TSVR outperforms other comparison schemes at high signal-to-noise ratios. Figure 12 This is a comparison table of modeling accuracy of visible light communication systems modeled according to the present invention. When M=2, the NMSE index of TSVR is improved by 16.5403dB compared with SVR; the CPU computing efficiency is at least four times that of SVR.

[0089] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A predistortion method for visible light communication based on Siamese support vector regression, characterized in that, It includes the following steps: Step S1: Build an offline digital predistortion platform for visible light communication; Step S2: Based on the signal information obtained from the visible light communication offline digital predistortion platform in Step S1, construct a training set for the amplitude enhancement structure; Step S3: Based on the training set of the amplitude enhancement structure constructed in step S2, establish a predistortion nonlinear suppression method for visible light communication based on amplitude TSVR, and form a test set; Step S4: Use the training and test sets of the constructed amplitude enhancement structure to train the model; Step S5: After model training, the predistorter parameters are obtained, and a visible light communication predistortion system is further designed and constructed. Step S4 includes the following steps: A behavioral model of LEDs is constructed using TSVR. By introducing a kernel function, the original TSVR expression is extended to the nonlinear case. The regression function of the entire model is as follows: in , representing the kernel function, x i x j Let ω1, ω2, b1, b2 ∈ R be the sample points. n Each row of matrix A represents a set of training data. Using kernel function theory, we can further obtain the following equivalent objective function. Where ε1 and ε2 represent insensitive parameters, and ζ * ζ and ζ represent slack variables, e is the identity matrix, Y is the response output signal, and C1 and C2 are penalty factors; Using the Lagrange multiplier method, we obtain Furthermore, the dual forms of (5) and (6) can be obtained as follows: Where H = [k(A, A)] T [e], α, μ, β are Lagrange multipliers, I is the identity matrix, C3, C4 represent the penalty factor. The regression function of the final expected value can be expressed as the average of the upper and lower bound functions, and its functional expression is shown below: Where f1(x) is the upper bound function and f2(x) is the lower bound function.

2. The visible light communication predistortion method based on Siamese support vector regression according to claim 1, characterized in that, Step S1 uses the Wiener model to describe the nonlinear transmission characteristics and memory effect of the visible light communication system, and obtains the measured input signal x(n) and output signal y(n) of the LED.

3. The visible light communication predistortion method based on Siamese support vector regression according to claim 2, characterized in that, Step S2 involves normalizing the input signal x(n) and the output signal y(n) to construct a training set for the amplitude enhancement structure. The training set is then processed based on the TSVR algorithm.

4. The visible light communication predistortion method based on Siamese support vector regression according to claim 2, characterized in that, The Wiener model is decomposed into a cascade of a linear time-invariant system and a nonlinear system. For the memory properties of LEDs, a time-delay tap structure is used to introduce memory nonlinearity.

5. A visible light communication predistortion method based on Siamese support vector regression according to claim 4, characterized in that, The linear time-invariant module describing the memory effect of a linear time-invariant system is represented as: Where L is the largest delay tap, b1 is the memory effect factor, x(nl) represents the delay term of the input signal, x(n) represents the input signal, and l represents the delay depth.

6. A visible light communication predistortion method based on Siamese support vector regression according to claim 5, characterized in that, The Rapps model describing the nonlinear module of a nonlinear system is as follows: Among them, I max V is the current maximum output, k is the inflection point coefficient, which controls the smoothness from the linear region to the saturation region. TOV It is the turn-on voltage value of the light-emitting diode.

7. A visible light communication predistortion method based on Siamese support vector regression according to claim 6, characterized in that, The TSVR algorithm has the following representation: in Indicates amplitude, Indicates a delay tap. Let x(n) represent the signal amplitude, x(n-1) represent the first-order delay of the signal, x(n-2) represent the second-order delay of the signal, and TSVR Machine represent the constructed Siamese Support Vector Regression Machine. The training set is trained by the TSVR Machine to obtain the predicted output signal. .

8. A visible light communication predistortion method based on Siamese support vector regression according to claim 5, characterized in that, Step S3: Establish a nonlinear suppression method for predistortion in visible light communication based on amplitude TSVR. This method cascades the TSVR predistorter with the transmitter of the visible light communication system to form an approximately linear system, thereby compensating for nonlinearity and improving the overall performance of the system.

9. A visible light communication predistortion method based on Siamese support vector regression according to claim 8, characterized in that, In the TSVR predistorter, the bitstream signal first undergoes 16QAM mapping. Since intensity modulation / direct detection is used in visible light communication, it needs to be mapped by Hermite and then converted into a real number signal by IFFT (Inverse Fast Fourier Transform). Then, a prefix and guard interval are added. After parallel-to-serial conversion, the signal passes through the TSVR predistorter. After digital-to-analog conversion, the DC signal is coupled to the signal to be transmitted and loaded onto the LED through a Bias-Tee.

10. A visible light communication predistortion method based on twin support vector regression according to claim 9, characterized in that, At the receiving end, an APD detector is used to convert the received signal into photoelectric value for offline processing.

Images