Geometrically shaped single-carrier photonic terahertz communication system based on nonlinear-resistant SVM
By combining independent sideband modulation and geometric shaping techniques in the photon terahertz communication system, the nonlinear effects are processed using SVM algorithms, which solves the problems of low spectrum utilization and SSBI, and significantly improves the system's code error performance and spectrum efficiency.
Patent Information
- Application Number
- CN202410356965.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-03-27
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2044-03-27
AI Technical Summary
In the photon terahertz communication system, the spectrum utilization rate is low and the transmission power is large, and the SSBI generated by independent sideband modulation under nonlinear conditions leads to an increase in the bit error rate. Existing solutions such as the K-means clustering algorithm are seriously misjudged.
Combining independent sideband modulation (ISB) and geometric shaping technology, support vector machine (SVM) algorithm is used to fight against nonlinear effects, optimize constellation point distribution, and reduce bit error rate.
The data transmission rate and channel capacity of the photon terahertz communication system are significantly improved, the bit error rate is reduced, and the higher spectral efficiency and lower transmission power are achieved.
Smart Images

Figure CN118157771B_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of access communication, and in particular relates to a geometric shaping single-carrier photon terahertz communication system based on an anti-nonlinear SVM. Background Art
[0002] Since the 3G era, wireless communication technology has been committed to improving the data transmission capacity of the system. In the current 5G era, human demand for data services is also expanding. According to the law proposed by Phil Edholm, the data rate of wireless communication shows an exponential growth trend and will double in just 18 months. If this trend continues, people's demand for data transmission rate will reach tens or even hundreds of Gbps. As a result, the demand for communication bandwidth continues to increase, while spectrum resources are extremely scarce. Traditional low-frequency communication systems can no longer meet such high rate requirements.
[0003] The fifth generation (5G) mobile communication system uses the 60GHz millimeter wave frequency band as one of the key technical solutions, which can provide bandwidths ranging from hundreds of MHz to several GHz. Compared with traditional wireless communication solutions below the 5GHz frequency band, this solution can obtain more available bandwidth, and the data transmission rate can reach several Gbps, but it is still not enough to meet the growing data traffic demand in future wireless communications. In the post-5G era, the demand for service transmission rates of wireless LAN systems and virtual reality devices will exceed 10Gbps. In addition, the future transmission of video services (ultra-high-definition video, 3D video, etc.) will expand the demand for service data transmission rates to 24Gbps, and may even reach 100Gbps. Therefore, wireless communications need to develop towards higher frequencies and wider bandwidth windows. The terahertz band (0.1~10THz) is located between microwaves and infrared light waves, has rich spectrum resources, and is considered to be the next breakthrough point in the communication technology revolution.
[0004] In the design of photonic terahertz communication systems, factors such as the efficiency of spectrum resource utilization in the terahertz frequency band and the system transmission power need to be considered. In order to improve spectrum efficiency and reduce transmission power, single sideband modulation (SSB) is used for signal modulation and demodulation, which becomes an ideal design idea. This modulation has good tolerance to power fading caused by dispersion and has higher spectrum efficiency. However, SSB modulation only uses one sideband to transmit data. In order to make full use of the full sideband, independent sideband (ISB) modulation has become popular. Independent sideband (ISB) modulation doubles the data rate by transmitting two independent data in the left sideband (LSB) and the right sideband (RSB) respectively. Therefore, ISB modulation was later applied to establish bandwidth-efficient optical fiber links, which is usually called optical independent sideband (O-ISB) modulation.
[0005] In traditional communication systems, due to the randomness of transmitted data, the transmission probability of each symbol is usually considered to be the same, and the geometric position of the constellation points is generally arranged in a square. However, due to the influence of factors such as nonlinear effects, the actual optical fiber communication system is power-limited, so it cannot achieve a high signal-to-noise ratio. Therefore, the amount of information that the channel can accommodate is often less than the source entropy, resulting in a waste of transmission power. In other words, although the typical optical fiber communication system can obtain the maximum source entropy by transmitting uniformly distributed signals, the amount of information cannot be fully carried by the channel during actual transmission, so the optimal balance between transmission power and source entropy cannot be achieved. Shannon proved that in an AWGN channel, when the signal is Gaussian distributed, the optimal balance between transmission power and source entropy can be achieved, thereby increasing channel capacity and improving system performance. Constellation shaping is an effective technical means to achieve this optimal balance. Constellation shaping is a technology based on the signal constellation diagram, which is divided into two implementation methods: probability shaping and geometric shaping.
[0006] In the AWGN channel, the mutual information curve of the square QAM constellation signal shows a 1.53dB capacity gap with the Shannon limit capacity at a high signal-to-noise ratio. The fundamental reason for this capacity gap is the square layout and uniform discrete distribution of the constellation points. The purpose of geometric shaping is to design the constellation point distribution of the transmitted signal according to the optimization criteria under a given signal-to-noise ratio so that its performance is better than the traditional square constellation diagram. Generally speaking, after geometric shaping, the low-energy constellation points will be more concentrated near the origin. Geometric shaping increases the minimum Euclidean distance between constellation points by redesigning the constellation point arrangement, thereby reducing inter-symbol interference and improving the system's tolerance to noise.
[0007] In order to improve the system transmission rate, in addition to solving the coding modulation problem, the nonlinear effect problem must also be considered. Both the modulation method and the detection method are affected by nonlinear effects. In single-sideband modulation, the modulated signal is modulated with the carrier to generate a single-sideband signal. However, due to the nonlinear effects in the modulation process, such as the nonlinear characteristics of the modulator, additional spectral components are introduced, including suppressed carrier components. These additional spectral components will cause amplitude and phase distortion, namely SSBI.
[0008] Specifically, SSBI causes amplitude distortion, that is, the amplitude of the signal no longer maintains its original shape. In addition, SSBI also causes phase distortion, that is, the phase of the signal is no longer consistent with the original signal. This amplitude phase distortion may seriously affect the performance of the communication system, especially in the case of high signal power or large dynamic range. There are relatively mature solutions for SSBI with single sideband modulation, but for SSBI generated by independent sideband modulation (ISB), the solution needs to be reconsidered. Therefore, how to solve this problem is also one of the research focuses.
[0009] The SSBI problem can be solved by using machine learning related algorithms, such as the widely used K-means clustering algorithm. The traditional constellation point classification and decision method based on Euclidean distance divides the target constellation point into the category of the nearest standard constellation point according to the minimum Euclidean distance. However, this method does not take into account the signal transmission characteristics, and the obtained boundary plane is only related to the distribution of standard constellation points. Therefore, under nonlinear conditions, a large number of constellation point misjudgments will be caused, which will greatly increase the bit error rate. Another classic supervised machine learning algorithm, SVM, can solve this problem by finding the optimal classification interface. SVM learns the distribution of received constellation points based on a small amount of training data, adjusts the classification decision interface, and thus reduces misjudgments to improve bit error performance. Summary of the invention
[0010] In view of the problems of low spectrum utilization and high transmission power in photonic terahertz communication systems, the present invention proposes combining ISB with geometric shaping and SVM algorithms to construct a geometric shaping single-carrier photonic terahertz communication system based on anti-nonlinear SVM, thereby optimizing system transmission performance.
[0011] The geometric shaping single-carrier photonic terahertz communication system based on anti-nonlinear SVM has the following construction process:
[0012] Step 1: Generate two independent data as low-order orthogonal amplitude modulation (QAM) sideband signals at the transmitting end of the photonic terahertz communication system. After modulation and up-conversion, two independent intermediate frequency (IF) vector signals are obtained, and the two IF signals are added together to form the driving signal of the Mach-Zehnder modulator (MZM), namely the GS-16QAM signal.
[0013] The processing process of the two independent data is specifically as follows:
[0014] First, two independent data are modulated by regular quadrature phase shift keying (QPSK) and geometric shaped QPSK (GS-S-QPSK), respectively;
[0015] Then, the two independent modulated signals are respectively modulated by two carrier waves of complex sinusoidal form exp(-2πf s t) and exp(2πf s t) mixing to generate a radio frequency (RF) signal, and further up-converting the RF signal into an IF signal, that is, obtaining a left side band (LSB) and a right side band (RSB) signal;
[0016]
[0017]
[0018] A(t), B(t); ω l (t),ω r (t) represents the amplitude, phase and angular frequency of LSB and RSB signals, respectively.
[0019] Finally, the two IF signals are added together to obtain the MZM driving signal, namely the GS-16QAM signal.
[0020] E(t)=E LSB (t)+E RSB (t) (3)
[0021] The left band of the GS-16QAM signal is modulated by QPSK, and the right band is modulated by S-QPSK. The specific shape depends on the shape of S-QPSK. The amplitude and phase of the 16QAM constellation diagram are a combination of the amplitude and phase of the QPSK and S-QPSK constellations.
[0022] Step 2: Digitally separate the drive signal E(t) into real and imaginary parts, respectively drive the in-phase (I) and quadrature (Q) branches of the IQ modulator, and modulate the laser ECL1 to emit a continuous and stable optical carrier.
[0023] Step three, adjust the GS-16QAM signal to the terahertz frequency band, and generate a terahertz signal as a carrier through optical heterodyne beat frequency in the photonic terahertz communication system, and carry the GS-16QAM signal into the terahertz simulation channel for transmission.
[0024] The specific scheme of generating terahertz signals by optical heterodyne beat frequency is as follows:
[0025] One laser is used to carry the signal to be sent, and is cascaded with the modulator to modulate the signal to be sent; the other laser is used to generate optical signals of different frequencies, and beat the frequency with the previous optical signal carrying the signal in the photodetector, and the frequency difference between the two optical signals is used to obtain the terahertz wave, thereby realizing the terahertz wave carrying the transmission signal.
[0026] The two optical signals generated by lasers ECL1 and ECL2 are:
[0027]
[0028]
[0029] The two optical signals are represented by A1 and A2, the frequencies are represented by f1 and f2, and the initial phase information is represented by and On this basis, the baseband signal to be sent is modulated into the first optical signal using a modulator, and then:
[0030]
[0031] The baseband signal modulated by the modulator is represented by I(t)+jQ(t), and then sent to the coupler to couple with another optical signal. The result after coupling is:
[0032]
[0033] The coupling result is sent to the photodetector for beat frequency, and the result after beat frequency is:
[0034]
[0035] Among them, R is used to represent the responsivity of the photodetector, the unit is A / W, and its expression formula is:
[0036]
[0037] Among them, e represents the electron charge, η represents the quantum efficiency of the photodiode, represents Planck's constant, and f represents the center frequency of the light signal detected by the photodiode.
[0038] Step 4: During the transmission process, the received THz wave signal passes through the variable optical attenuator and is converted by the UTC-PD. Finally, the captured signal is equalized and restored by the DSP algorithm at the receiving end to obtain the output signal of the photonic terahertz communication system.
[0039] The DSP algorithm specifically includes:
[0040] Step 401, resampling, downconverter and channel equalization using a cascaded multimode algorithm.
[0041] Step 402, using a carrier frequency offset estimation algorithm based on a training sequence to compensate for a carrier frequency offset of a digital communication signal;
[0042] Step 403, using a blind phase search carrier phase recovery algorithm to compensate for the phase shift;
[0043] Step 404, deploying a support vector machine (SVM) algorithm to obtain decision constellation points.
[0044] The SVM algorithm is used to build a classification model to obtain the centroid of the convergent cluster as the standard decision constellation point, specifically:
[0045] Given a training set {(x1,y1),(x2,y2),…,(x n ,y n )}, where x i is the input feature vector, y iis the corresponding category label. The goal of the kernel function support vector machine is to find a hyperplane that separates samples of different categories. The formula is:
[0046]
[0047] Among them, K(x,x i ) is the kernel function, which is used to map the input features to a high-dimensional space. i and b are model parameters obtained through training. The steps of the SVM algorithm are as follows:
[0048] (1) Select a segment of data as a training set for building a classifier, including the I-path and Q-path features of the signal;
[0049] (2) Extract key features such as amplitude and phase to form the input components of the SVM model;
[0050] (3) Select an appropriate kernel function and use labeled data to train the model;
[0051] (4) Evaluate the performance of the trained model with the help of the test set and calculate the accuracy;
[0052] Step 405: Demap the received GS-16QAM signal using the decision constellation provided by the SVM algorithm.
[0053] The demapping refers to separating the QPSK and S-QPSK signals from the GS-16QAM signal and calculating the optimal bit error rate (BER) of the GS-16QAM, QPSK and S-QPSK signals.
[0054] The advantages and positive effects of the present invention are:
[0055] (1) The ISB modulation is used in the present invention to improve the data transmission rate of the system. Currently, there is no precedent for using ISB modulation in a photonic terahertz communication system.
[0056] (2) The present invention adds geometric shaping technology to the photonic terahertz communication system to increase the system capacity, making the system capacity closer to the Shannon limit;
[0057] (3) In order to combat SSBI generated during the modulation process, the present invention uses a support vector machine (SVM) algorithm to deal with the nonlinear effect problem. Currently, most solutions to SSBI use the k-means clustering algorithm, which is relatively simple but can cause a large number of misjudgments. SVM can better solve this problem. Experimental results show that the use of this solution can significantly improve the bit error performance. BRIEF DESCRIPTION OF THE DRAWINGS
[0058] Figure 1It is the schematic diagram of the photonic terahertz communication system;
[0059] Figure 2 This is the schematic diagram of ISB modulating two signals;
[0060] Figure 3 These are three different GS-16QAM constellations obtained after ISB modulation;
[0061] Figure 4 It is a diamond-shaped 16QAM signal symbol marking diagram;
[0062] Figure 5 It is the SVM principle diagram of the photonic terahertz communication system;
[0063] Figure 6 It is the pre-beat frequency band of the photonic terahertz communication system;
[0064] Figure 7 It is the post-beat frequency band of the photonic terahertz communication system;
[0065] Figure 8 These are the three different GS-16QAM result diagrams obtained after modulation;
[0066] Fig. 9 is the relationship between BER and SNR for different GS-16QAM constellations;
[0067] Fig.10 is the relationship between BER and ROP for different GS-16QAM constellations;
[0068] Fig.11 It is the relationship between BER and ROP with and without SVM. DETAILED DESCRIPTION
[0069] In order to facilitate those skilled in the art to understand and implement the present invention, the present invention is further described in detail below with reference to the accompanying drawings and embodiments.
[0070] The independent sideband modulation, geometric shaping and nonlinear compensation method of the photonic terahertz communication system proposed in the present invention significantly reduces the complexity, thereby saving system costs. Considering the signal-signal beat interference, the support vector machine (SVM) algorithm is used to improve the bit error rate performance. The details are as follows:
[0071] In the photonic terahertz communication system, two data channels are sent and modulated into a 16QAM signal. The two signals are used as the upper sideband and lower sideband of the 16QAM signal, which greatly saves system resources. At the same time, three different shapes of GS-16QAM signals are verified to evaluate the impact of different geometric shaping (GS) schemes on performance. The terahertz signal is generated as a carrier by beating frequency, and then the GS-16QAM signal is carried to the terahertz frequency band for transmission. Due to the existence of SSBI, the support vector machine (SVM) is used at the receiving end to obtain the actual convergence centroid, thereby replacing the standard decision constellation point for signal demodulation.
[0072] Geometric shaping single-carrier photonic terahertz communication system based on anti-nonlinear SVM, such as Figure 1 As shown in the figure, the signal processing process is as follows:
[0073] Step 1: Generate two independent data as low-order orthogonal amplitude modulation (QAM) sideband signals at the transmitting end of the photonic terahertz communication system. After modulation and up-conversion, two independent intermediate frequency (IF) vector signals are obtained, and the two IF signals are added together to form the driving signal of the Mach-Zehnder modulator (MZM), namely the GS-16QAM signal.
[0074] The process of ISB modulating two signals is as follows: Figure 2 As shown, specifically:
[0075] First, two independent data are modulated in QPSK and Staggered-QPSK (S-QPSK) formats.
[0076] Then, by comparing the two carriers in the form of complex sinusoids exp(-2πf s t) and exp(2πf s t) mixing and up-converting the generated radio frequency (RF) signal to an intermediate frequency (IF) signal. The two independent IF vector signals are LSB and RSB signals. The modulation format, symbol rate and filter roll-off factor of the two sidebands may be different.
[0077] The LSB and RSB signals are:
[0078]
[0079]
[0080] Here, A(t), B(t); ω l (t),ω r (t) represents the amplitude, phase and angular frequency of LSB and RSB signals, respectively.
[0081] Finally, the two IF signals are added to form the driving signal for the Mach-Zehnder modulator (MZM).
[0082] The formula is as follows:
[0083] E(t)=E LSB (t)+E RSB (t) (3)
[0084] Where E LSB (t) and E RSB (t) represent LSB and RSB signals respectively.
[0085] The shape of the GS-16QAM signal obtained above depends on the shape of S-QPSK. The amplitude and phase of the 16QAM constellation diagram are a combination of the amplitude and phase of the QPSK and S-QPSK constellations.
[0086] On the basis of independent sideband modulation, by adjusting S-QPSK, three different geometric S-QPSK schemes are obtained, thus obtaining three different GS-16QAM signals. The amplitude of the GS-16QAM constellation is the product of the amplitudes of the QPSK constellation and the S-QPSK constellation, and the phase of the GS-16QAM constellation is the sum of the phases of the QPSK constellation and the S-QPSK constellation. In short, the amplitude and phase of the GS-16QAM constellation diagram are the combination of the amplitude and phase of the QPSK and S-QPSK constellation diagrams. Figure 3 As shown, the dark blue constellation points represent the constellation points of QPSK, and the purple, yellow, green, and light blue constellation points represent the constellation points of S-QPSK. A rejection symbol set can be used to represent the constellation points of QPSK and S-QPSK signals, that is, the rejection symbol set {0,1,2,3} is used to represent the constellation points of QPSK and S-QPSK signals, and {0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15} is used for GS-16QAM signals.
[0087] Based on the above theory, each constellation point of the corresponding GS-16QAM signal can be specifically demodulated into a point of the QPSK signal and a point of the S-QPSK signal when the signal is decomposed, that is, a mixed color point in the GS-16QAM constellation can be demodulated into the corresponding pure color constellation points in QPSK and S-QPSK. For example, the symbol "0" of the diamond-16QAM can be specifically demodulated into "0" of QPSK and "2" of S-QPSK, such as Figure 4 As shown. Based on this theory, the constellation points of the received GS-16QAM signal can be extracted, and then demodulated for each constellation point. Each constellation point corresponds to a point of the QPSK signal and the S-QPSK signal, thereby completely demodulating the QPSK and S-QPSK signals.
[0088] Step 2: The drive signal E(t) is digitally separated into real and imaginary parts, which respectively drive the in-phase (I) and quadrature (Q) branches of the IQ modulator, and modulate the laser ECL1 to emit a continuous and stable optical carrier.
[0089] Step three: adjust the GS-16QAM signal to the terahertz frequency band, generate a terahertz signal as a carrier in the photonic terahertz communication system, and carry the GS-16QAM signal into the terahertz simulation channel for transmission.
[0090] Before sending the GS-16QAM signal to the channel, a terahertz signal needs to be generated as a carrier to carry the GS-16QAM signal. From the perspective of the generation of photonic terahertz signals, there are three methods, including direct modulation, external modulation, and optical heterodyne beat frequency.
[0091] The present invention adopts the scheme of generating terahertz signal by optical heterodyne beat frequency, specifically:
[0092] One laser is used to carry the signal to be sent, and is cascaded with the modulator to modulate the signal to be sent; the other laser is used to generate optical signals of different frequencies, and beats with the previous optical signal carrying the signal in the photodetector, and uses the frequency difference of the two optical signals to obtain terahertz waves, thereby realizing the terahertz wave carrying the transmission signal. The terahertz signal obtained in this way is relatively low in cost and easy to adjust in frequency, but it also has certain disadvantages, namely, the phase noise is relatively high. However, this problem can be solved by compensating at the receiving end, which is generally achieved by using digital signal processing algorithms. The principle of the optical heterodyne beat frequency scheme is introduced below:
[0093] ECL1 and ECL2 are used as local oscillator light sources to generate free-running optical carriers. The optical carrier generated by ECL1 is transmitted to the I / Q modulator. The optical carrier generated by ECL2 differs from the optical carrier generated by ECL1 by one terahertz frequency. A terahertz signal with a frequency band of 90 GHz is generated by optical heterodyne beat frequency as the carrier of the ISB signal. After ECL1 and ECL2 generate optical carriers, the two optical signals are coupled through PM-OC and connected to a 50 km SSMF for transmission.
[0094] The two optical signals generated by lasers ECL1 and ECL2 are:
[0095]
[0096]
[0097] The two optical signals are represented by A1 and A2, the frequencies are represented by f1 and f2, and the initial phase information is represented by and On this basis, the baseband signal to be sent is modulated into the first optical signal using a modulator, and then:
[0098]
[0099] The baseband signal modulated by the modulator is represented by I(t)+jQ(t), and then sent to the coupler to couple with another optical signal. The result after coupling is:
[0100]
[0101] The coupling result is sent to the photodetector for beat frequency, and the result after beat frequency is:
[0102]
[0103]
[0104] Among them, R is used to represent the responsivity of the photodetector, the unit is A / W, and its expression formula is:
[0105]
[0106] Among them, e represents the electron charge, η represents the quantum efficiency of the photodiode, represents Planck's constant, and f represents the center frequency of the light signal detected by the photodiode.
[0107] It can be seen from formula (8) that the required terahertz wave can be obtained by modifying the frequency difference between the two optical signals.
[0108] The band before the beat frequency of the photon terahertz system is as follows Figure 6 As shown in Figure 2, the terahertz band after the beat frequency is as follows: Figure 7 shown.
[0109] After generating the terahertz carrier signal using the above method, the GS-16QAM signal is carried to the terahertz frequency band and then sent into the terahertz simulation channel.
[0110] Step 4: During the transmission process, the received THz wave signal passes through the variable optical attenuator and is converted by the UTC-PD. Finally, the captured signal is equalized and restored at the receiving end through the digital signal processing (DSP) algorithm to obtain the output signal of the photonic terahertz communication system.
[0111] The DSP algorithm specifically includes:
[0112] Step 401, resampling, downconverter and channel equalization using a cascaded multimode algorithm.
[0113] Step 402, using a carrier frequency offset estimation algorithm based on a training sequence to compensate for a carrier frequency offset of a digital communication signal;
[0114] Step 403, using a blind phase search carrier phase recovery algorithm to compensate for the phase shift;
[0115] Step 404, deploying a support vector machine (SVM) algorithm to obtain standard decision constellation points, rather than theoretically calculated points, the distortion considered is mainly caused by inter-symbol interference (SSBI).
[0116] The SVM algorithm is used to build a classification model to obtain the centroid of the convergent cluster as the standard decision constellation point, specifically:
[0117] Given a training set {(x1,y1),(x2,y2),…,(x n ,y n )}, where x i is the input feature vector, y i is the corresponding category label. The goal of the kernel function support vector machine is to find a hyperplane that separates samples of different categories. The formula is:
[0118]
[0119] Among them, K(x,x i ) is the kernel function, which is used to map the input features to a high-dimensional space. i and b are model parameters obtained through training.
[0120] like Figure 5 As shown, the steps for training the SVM algorithm are as follows:
[0121] (1) Select a segment of data as a training set for building a classifier, including the I-path and Q-path features of the signal;
[0122] (2) Extract key features such as amplitude and phase to form the input components of the SVM model;
[0123] (3) Select an appropriate kernel function and use labeled data to train the model;
[0124] (4) Evaluate the performance of the trained model with the help of the test set and calculate the accuracy;
[0125] Step 405: Demap the received GS-16QAM signal using the decision constellation provided by the SVM algorithm.
[0126] The demapping refers to separating the QPSK and S-QPSK signals from the GS-16QAM signal and calculating the optimal bit error rate (BER) of the GS-16QAM, QPSK and S-QPSK signals.
[0127] The demodulation theory for GS-16QAM signals in the prior art is based on the standard 16QAM constellation points at the transmitting end for decision making. However, since the modulation process produces nonlinear effects, resulting in amplitude and phase distortion (SSBI), the amplitude of the signal no longer maintains its original shape, so using the standard decision constellation points for decision making will produce certain errors. The ISB modulation used in the present invention also has this problem, so it needs to be further optimized, and the support vector machine (SVM) can be used to solve the SSBI problem.
[0128] Since the SVM algorithm is a binary classification algorithm, when the SVM algorithm is used to perform constellation judgment on the GS-16QAM signal received at the receiving end, it needs to be expanded to multi-classification in order to be applicable to the 16 constellation point classification requirements of the GS-16QAM signal. For the constellation judgment scenario, the one-versus-all (OVA) strategy of SVM can be used. The OVA strategy is a solution to multi-classification problems. In this strategy, an SVM (or other binary classifier) is established for each category. When establishing the SVM model for each category, samples belonging to this category are regarded as positive (+), while samples of all other categories are regarded as negative (-). In this way, each SVM becomes a binary classification problem, which can be solved using the SVM binary classifier. In this way, the multi-classification problem is decomposed into multiple binary classification problems, thereby simplifying the complexity of the problem.
[0129] In the OVA strategy, the judgment value is calculated based on the distance from the sample point to the classification hyperplane. The farther from the hyperplane, the less likely the sample is to be misclassified. Therefore, when dealing with multi-classification problems, this strategy can decompose the problem into a more solvable form by establishing multiple binary classifiers. During the training process, each class 1 is distinguished from the other 15 classes by establishing an SVM binary classifier, thereby obtaining an SVM multi-classification model containing 16 binary classifiers. Because the constellation points are linearly separable, in order to reduce the complexity of the model, the present invention selects a linear kernel in the simulation, that is, the basic SVM form.
[0130] In the present invention, the training set ratio is selected to be 5% to balance the training effect and bandwidth usage. Then, a set of decision constellation points obtained by the SVM algorithm is used to perform constellation judgment on the GS-16QAM signal, and the judgment result is measured, compared and analyzed by the bit error rate.
[0131] The results show that:
[0132] The structure of the present invention is simulated by the simulation software VPI, and the independent sideband modulation, geometric shaping and nonlinear compensation method are implemented in the photonic terahertz communication system, and the 16QAM signal is synthesized by the beat difference of the QPSK signal in the LSB and the S-QPSK signal in the RSB, such as Figure 8 As shown, there are three different GS-16QAM result diagrams, where the S-QPSK schemes are square S-QPSK, diamond S-QPSK and circular S-QPSK, and three S-QPSK schemes with different geometric shapes are evaluated.
[0133] The study found that Fig. 9 and Fig.10 As shown in , the GS-16QAM signal obtained by square S-QPSK signal can achieve the best bit error rate (BER). Fig.11 As shown in FIG. 1 , a comparison is made between using the SVM algorithm at the receiving end to combat the phase distortion and IQ imbalance of the constellation and whether the SVM algorithm is used at the receiving end. The results show that the SVM algorithm significantly improves the bit error rate performance.
Claims
1. Geometric shaping single-carrier photonic terahertz communication system based on anti-nonlinear SVM, It is characterized in that The construction process is as follows: Step 1: Generate two independent data as QAM sideband signals at the transmitting end of the photonic terahertz communication system, obtain two independent intermediate frequency vector signals after modulation and up-conversion, and add the two IF signals to form an MZM driving signal, i.e., a GS-16QAM signal; The processing process of the two independent data is specifically as follows: First, two independent data are modulated by QPSK and geometric shaped S-QPSK respectively; Then, the two independent modulated signals are respectively modulated by two carrier waves of complex sinusoidal form exp(-2πf s t) and exp(2πf s t) mixing to generate an RF signal, and further up-converting the RF signal into an IF signal, that is, obtaining LSB and RSB signals; A(t), B(t); ω l (t),ω r (t) represents the amplitude, phase and angular frequency of LSB and RSB signals respectively; Finally, the two IF signals are added together to obtain the MZM driving signal, i.e., the GS-16QAM signal; E(t)=E LSB (t)+E RSB (t) Step 2: The driving signal E(t) is digitally separated into real and imaginary parts, which respectively drive the in-phase (I) and quadrature (Q) branches of the IQ modulator, and the modulated laser ECL1 emits a continuous and stable optical carrier. Step 3: adjust the GS-16QAM signal to the terahertz frequency band, and generate a terahertz signal as a carrier by optical heterodyne beat frequency in the photonic terahertz communication system, and carry the GS-16QAM signal into the terahertz simulation channel for transmission; The specific scheme of generating terahertz signals by optical heterodyne beat frequency is as follows: One laser is used to carry the signal to be sent, and is cascaded with the modulator to modulate the signal to be sent; the other laser is used to generate optical signals of different frequencies, and beat the optical signal carrying the signal in the photodetector, and use the frequency difference between the two optical signals to obtain the terahertz wave, thereby realizing the terahertz wave carrying the transmission signal; Step 4: During the transmission process, the received THz wave signal passes through the variable optical attenuator and is converted by the UTC-PD. Finally, the captured signal is equalized and restored by the DSP algorithm at the receiving end to obtain the output signal of the photonic terahertz communication system. The DSP algorithm specifically includes: Step 401, resampling, downconverter and channel equalization using a cascaded multimode algorithm; Step 402, using a carrier frequency offset estimation algorithm based on a training sequence to compensate for a carrier frequency offset of a digital communication signal; Step 403, using a blind phase search carrier phase recovery algorithm to compensate for the phase shift; Step 404, deploying a support vector machine (SVM) algorithm to obtain decision constellation points; Step 405: Demap the received GS-16QAM signal using the decision constellation provided by the SVM algorithm to obtain an output signal.
2. According to claim 1, the geometric shaping single-carrier photonic terahertz communication system based on anti-nonlinear SVM, It is characterized in that The left band of the GS-16QAM signal is modulated by QPSK, and the right band is modulated by S-QPSK, and the shape depends on the shape of S-QPSK; the amplitude and phase of the 16QAM constellation diagram are a combination of the amplitude and phase of the QPSK and S-QPSK constellations.
3. According to claim 1, the geometric shaping single-carrier photonic terahertz communication system based on anti-nonlinear SVM, It is characterized in that The specific process of generating terahertz signals by optical heterodyne beat frequency is as follows: The two optical signals generated by lasers ECL1 and ECL2 are: Among them, the two optical signals are respectively 1 and A 2 Indicated by f 1 and f 2 The initial phase information is expressed as and On this basis, the baseband signal to be sent is modulated into the first optical signal by a modulator, and then: The baseband signal modulated by the modulator is represented by I(t)+jQ(t), and then sent to the coupler to couple with another optical signal. The result after coupling is: The coupling result is sent to the photodetector for beat frequency, and the result after beat frequency is: Among them, R is used to represent the responsivity of the photodetector, the unit is A / W, and its expression formula is: Among them, e represents the electron charge, η represents the quantum efficiency of the photodiode, represents Planck's constant, and f represents the center frequency of the light signal detected by the photodiode.
4. The geometric shaping single-carrier photonic terahertz communication system based on anti-nonlinear SVM according to claim 1, It is characterized in that The SVM algorithm is used to build a classification model to obtain the centroid of the convergent cluster as the standard decision constellation point, specifically: Given a training set {(x 1 ,y 1 ),(x 2 ,y 2 ),…,(x n ,y n )}, where x i is the input feature vector, y i is the corresponding category label; the goal of the kernel function support vector machine is to find a hyperplane to separate samples of different categories. The formula is: Among them, K(x,x i ) is the kernel function, which is used to map the input features to a high-dimensional space; a i and b are model parameters.
5. The geometric shaping single-carrier photonic terahertz communication system based on anti-nonlinear SVM according to claim 4, It is characterized in that The specific steps of training the SVM algorithm are as follows: (1) Select a segment of data as a training set for building a classifier, including the I-path and Q-path features of the signal; (2) Extract key features to form the input components of the SVM model; (3) Select an appropriate kernel function and use labeled data to train the model; (4) Use the test set to evaluate the performance of the trained model and calculate the accuracy.
6. The geometric shaping single-carrier photonic terahertz communication system based on anti-nonlinear SVM according to claim 1, It is characterized in that The demapping refers to separating the QPSK and S-QPSK signals from the GS-16QAM signal and calculating the BER of the GS-16QAM, QPSK and S-QPSK signals.
