Adaptive O-OFDM symbol decomposition noise suppression method based on rough set information decision system

Abstract

The invention discloses an adaptive O-OFDM symbol decomposition noise suppression method based on a rough set information decision system, which suppresses additive Gaussian white noise at a system receiving end, and comprises the following steps: establishing a rough set theory noise suppression model by using an information decision system of a rough set theory and an indistinguishable relationship; constructing an information decision-making system according to the time domain sampling value of the decomposition symbol of the receiving end, making a decision-making rule according to the information decision-making system, and performing classification decision-making on the time domain sampling value which is originally equal to an amplitude limiting threshold and a zero value; the method comprises the following steps: exporting an indistinguishable relationship according to a formulated decision rule, carrying out attribute reduction on an information decision system by using the indistinguishable relationship, carrying out attribute reduction to obtain time domain sampling values between amplitude limiting thresholds, reconstructing the time domain sampling values, and finally carrying out bitwise merging and adding to recover an O-OFDM symbol. According to the method, the noise of self-adaptive O-OFDM symbol decomposition is suppressed by using a rough set theory, and a suppression algorithm is designed, so that the bit error rate performance of the system can be improved, and the complexity of the algorithm can be effectively reduced.

Description

Technical Field

[0001] The invention relates to the technical field of visible light communication, and in particular to an adaptive O-OFDM symbol decomposition noise suppression method for a rough set information decision system. Background Art

[0002] The spectrum resources of traditional radio frequency (RF) communication are in short supply and can no longer meet the growing demand for communication capacity. As an innovative solution, visible light communication (VLC) can achieve high-speed data transmission by transmitting data through light-emitting diodes (LEDs) without interfering with the lighting function. VLC technology, with its advantages such as immunity to electromagnetic interference, abundant spectrum resources and unlicensed bandwidth, can not only effectively alleviate the spectrum resource pressure faced by traditional RF communication, but also has the ability to interconnect and integrate with existing RF communication systems. Therefore, it has become a focus of research in the field of wireless communications. In particular, VLC's ability to operate at terahertz frequencies makes it considered as one of the strong candidates for 6G communication technology.

[0003] Although the technical potential of VLC has been initially verified, it still encounters many technical difficulties in the process of transition to practical application. For example, due to the multipath propagation of optical signals, it will cause interference between symbols, artificial light sources may bring narrowband interference, and LED nonlinear distortion and other challenges. In response to these problems, optical orthogonal frequency division multiplexing (O-OFDM) technology has shown the ability to cope with multipath transmission of optical signals and improve spectrum efficiency. Nevertheless, the high peak-to-average power ratio (PAPR) of O-OFDM signals is still a problem to be solved. In addition, VLC also faces the problem of nonlinear distortion caused by the nonlinear characteristics of LEDs. Although the nonlinear mapping of LED electro-optical conversion can usually be linearized with the help of equalization technology and other methods, the linear working range of LEDs is still limited. At the same time, since VLC uses intensity modulation / direct detection (IM / DD), this method only uses the intensity of the optical signal to transmit information, while ignoring the frequency and phase information, which makes O-OFDM symbols more susceptible to LED nonlinear distortion. Therefore, in order to improve the performance of O-OFDM systems and promote the practical application of VLC technology, it is crucial to reduce the nonlinear distortion of LEDs.

[0004] The nonlinear distortion of LED can be suppressed by making O-OFDM symbols in the linear region of LED, such as reducing the peak-to-average ratio or optimizing DC bias. The literature "Mesleh R. LED clipping distortion compensation in optical wireless communication via multiple transmit LEDs [J]. Photonic Network Communications, 2013, 26: 25-31." proposed iterative signal limiting (ISC), which can split O-OFDM with high PAPR into multiple low-amplitude symbols and restore O-OFDM symbols at the receiving end using a single photodetector (PD). The principle of ISC is simple and does not sacrifice the information transmission rate, but there is a certain complexity in synchronization and hardware implementation. The literature "Jia Kejun, Yang Boran, Lu Hao, et al. Research on symbol decomposition technology for suppressing LED nonlinear distortion in visible light communication optical orthogonal frequency division multiplexing system [J]. Chinese Laser, 2020.47(4):234-242." proposed O-OFDM symbol decomposition serial transmission (SDST), which serially frames the O-OFDM decomposed symbols and then inputs them into a single LED. At the receiving end, the decomposed symbols are added according to the corresponding bits to restore the O-OFDM symbol. Although SDST is easier to implement than ISC, both ISC and SDST systems may have improper symbol decomposition times.

[0005] The literature "Jia Kejun, Yang Boran, Cao Minghua, et al. Design and research of adaptive O-OFDM symbol decomposition serial transmission system for visible light communication [J]. Journal of Communications, 2020.41(9):180-189." proposed an adaptive SDST (ASDST), which automatically adjusts the number of symbol decompositions according to the amplitude of the O-OFDM symbol. It optimizes the luminous efficiency of the LED, enhances the system's spectrum utilization and information transmission speed, and improves the performance of the VLC system.

[0006] But whether it is SDST or ASDST system, at the receiving end, multiple received decomposed symbols need to be combined and added to restore the O-OFDM symbol before decomposition. However, since each decomposed symbol passes through the random channel independently and serially, if these received decomposed symbols are directly added, the additive white Gaussian noise (AWGN) superimposed on the decomposed symbols will also be superimposed, resulting in a deterioration of the system bit error rate (BER) performance, which in turn affects the performance of the VLC system. Summary of the invention

[0007] The purpose of the present invention is to provide an adaptive O-OFDM symbol decomposition noise suppression method based on a rough set information decision system; to solve the problem that if the SDST and ASDST systems directly add these received decomposed symbols at the receiving end, the AWGN superimposed on the decomposed symbols will also be superimposed, resulting in poor system BER performance; to further improve the performance of the O-OFDM system and promote the practical application of the VLC system.

[0008] The present invention proposes an adaptive O-OFDM symbol decomposition noise suppression method based on a rough set information decision system, the steps of which are:

[0009] Sending side:

[0010] Step 1: The information sequence is subjected to M-order quadrature amplitude modulation (QAM);

[0011] Step 2: Map the modulated signal so that the output signal satisfies Hermitian symmetry;

[0012] Step 3: Perform inverse fast Fourier transform (IFFT) to obtain a bipolar real O-OFDM signal;

[0013] Step 4: After the O-OFDM signal is converted from parallel to serial (P / S), perform adaptive O-OFDM symbol decomposition; set the LED linear working range and the maximum number of symbol decompositions, and perform decomposition according to the dynamic range allowed by the O-OFDM symbol. After each decomposition, subtract the input symbol from the output symbol to check whether the difference is zero. If the difference is zero, stop symbol decomposition; otherwise, continue decomposition until the difference between the input symbol and the output symbol after a certain decomposition is zero, or the maximum number of symbol decompositions is reached;

[0014] Step 5: Decompose the symbol, add a cyclic prefix (CP), and then serially frame to obtain the frame symbol;

[0015] Step 6: Perform digital-to-analog conversion (D / A) on the obtained frame symbol, and then add a direct current bias (DC) to obtain a non-negative real signal to drive the LED lamp to send a light signal;

[0016] Receiver:

[0017] Step 1: The photodetector (PD) detects the received optical signal and converts it into an electrical signal. Each decomposed symbol passes through the AWGN channel independently and serially. Therefore, in addition to the limiting noise, each decomposed symbol at the receiving end is also superimposed with the AWGN noise. At this time, the limiting noise can be ignored, while the AWGN noise is the main noise affecting the system performance at this stage.

[0018] Step 2: Each decomposed symbol is converted to digital (A / D), framed, CP and DC are deleted, and the AWGN range of the decomposed symbol is determined; in order to suppress the AWGN on each decomposed symbol, the AWGN range of each decomposed symbol needs to be determined, that is, the maximum AWGN modulus value is determined. The AWGN can also be suppressed by only defining the threshold value with the modulus value combined with the limit threshold; however, since the maximum modulus value of the AWGN of each decomposed symbol varies randomly, the time domain sampling value may be misjudged during preprocessing using this method, thereby affecting the recovery of the O-OFDM symbol. On the other hand, when the number of IFFT points and the number of symbol decompositions are large, the amount of computation required for preprocessing and reconstruction processing of this method is large, and the algorithm complexity is very high;

[0019] Step 3: Perform rough set classification preprocessing on each decomposed symbol; use the information decision system of rough set theory to construct the information decision system with the time domain sampling values ​​of the decomposed symbols at the receiving end, formulate decision rules based on the characteristics of the constructed information decision system and the classification conditions, and classify and decide on the time domain sampling values ​​that are originally equal to the limiting threshold and zero value and assign them.

[0020] Step 4: The decomposed symbols that have been classified and preprocessed are reconstructed by attribute simplification; the indiscernible relations of the rough set theory are used to derive the indiscernible relations according to the established decision rules to simplify the attributes of the information decision system, and the attributes are simplified to obtain the time domain sampling values ​​between the limiting thresholds and reconstruct them; then the decomposed symbols are merged and added to restore the O-OFDM symbols;

[0021] Step 5: After the O-OFDM symbol is serial-to-parallel converted (S / P), it is input into the Fourier transform (FFT);

[0022] Step 6: After frequency domain equalization and information extraction of the O-OFDM symbol, it is finally input into the QAM demodulator and demodulated to restore the original binary information sequence.

[0023] Compared with the prior art, the beneficial effects of the present invention include:

[0024] (1) New artificial intelligence methods, soft computing, and uncertainty mathematical theory knowledge rough set theory (RST) are introduced to suppress adaptive O-OFDM symbol decomposition noise.

[0025] Both the SDST system and the ASDST system can effectively suppress the nonlinearity of visible light communication LED and improve the performance of the VLC system. However, when the O-OFDM symbols are restored by merging and adding at the receiving end, the AWGN superimposed on each decomposed symbol will also be added, thereby increasing the AWGN and causing the system BER performance to deteriorate. The adaptive O-OFDM symbol decomposition noise suppression scheme based on the rough set information decision system involved in the present invention utilizes the good performance of soft computing rough set theory in classification decision and attribute reduction and its powerful data mining ability, combines the visible light nonlinear noise suppression problem with the rough set theory, finds a new idea for visible light communication LED nonlinear suppression, and establishes an adaptive O-OFDM symbol decomposition noise suppression model based on the rough set theory.

[0026] (2) The time domain sampling values ​​of the decomposed symbols are used as preprocessed data to construct a rough set theory information decision system. According to the constructed information decision system, the corresponding optimal decision rules are formulated to classify and decide on the information decision system. The time domain sampling values ​​that are originally equal to the limiting threshold and the zero value are classified as much as possible, and the classification decision of the time domain sampling values ​​is avoided when the maximum noise modulus method is used to remove noise, which will affect the system performance.

[0027] The rough set classification preprocessing algorithm designed by the present invention utilizes the rough set theory information decision system, takes the time domain sampling value of the decomposed symbol at the receiving end as the domain U, the pre-classification category at the sending end as the pre-processed data signal feature at the receiving end and converted into a symbolic attribute as the conditional attribute C, and the limit threshold as the decision attribute D, to construct an information decision system. It is only necessary to formulate decision rules according to the information decision system and make classification decisions on the information decision system, so that the time domain sampling values ​​that are originally equal to the limit threshold and zero value can be classified and decided. This method can effectively classify and preprocess the time domain sampling values ​​that are equal to the limit threshold and zero value and assign values. Although the maximum noise modulus method determines the AWGN noise range, the classification decision can also be made according to the threshold defined by the limit threshold and the modulus value. However, if the time domain sampling value at the receiving end is originally between the limit threshold or equal to the limit threshold and zero value, it may be greater than or less than the set threshold at the receiving end after passing through the AWGN channel, and the classification decision error will occur, affecting the recovery of subsequent O-OFDM symbols, thereby causing BER performance deterioration. Therefore, the present invention utilizes the rough set theory information decision system to avoid the classification decision errors and assignment errors of the time domain sampling values, and avoids the misjudgment and assignment errors when utilizing the maximum noise modulus value method to classify and decide the time domain sampling values, resulting in the deterioration of the system BER performance.

[0028] (3) Based on the indiscernibility relation of rough set theory, the information decision system is reconstructed by attribute simplification, which can effectively reconstruct the decomposed symbols between the limiting thresholds and avoid the errors and large amount of algorithm calculation when using the maximum noise modulus method for reconstruction, which leads to increased algorithm complexity.

[0029] The attribute simplification and reconstruction algorithm designed by the present invention uses the decision rules formulated in the preprocessing stage to derive indistinguishable relations to simplify the attributes of the information decision system, and the attributes are simplified to obtain time domain sampling values ​​between the limiting thresholds, and then reconstruct them. When the maximum noise modulus value is used for reconstruction, the reconstruction is performed according to the structural law of the decomposed symbols, but misjudgment may occur during preprocessing, and then the reconstruction may be wrong, resulting in errors in O-OFDM symbol recovery. And when the maximum noise modulus value method is used for reconstruction, all decomposed symbols need to be processed, but if the number of symbol decomposition times and the number of FFT / IFFT points increase, the amount of calculation will increase, and then the complexity of the reconstruction algorithm will increase. Therefore, in addition to being able to effectively reconstruct the decomposed symbols between the limiting thresholds, the rough set theory attribute simplification and reconstruction algorithm can also avoid the problems of misjudgment in the preprocessing stage of the maximum noise modulus value method, and then errors in the reconstruction stage, as well as increased algorithm complexity. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is the principle diagram of the RST-ASDST system of the present invention;

[0031] Figure 2 It is the ACO-OFDM-ASDST decomposition symbol diagram of the present invention;

[0032] Figure 3 It is a DCO-OFDM-ASDST decomposition symbol diagram of the present invention;

[0033] Figure 4 This is a comparison chart of the Accuracy of different decision rules of ACO-OFDM-ASDST of the present invention;

[0034] Figure 5 It is a comparison diagram of recall of different decision rules of ACO-OFDM-ASDST of the present invention;

[0035] Figure 6 This is a comparison chart of the Accuracy of different decision rules of DCO-OFDM-ASDST of the present invention;

[0036] Figure 7 It is a comparison diagram of recalls of different decision rules of DCO-OFDM-ASDST of the present invention;

[0037] Figure 8This is a system bit error rate analysis diagram under different modulation modes of ACO-OFDM-ASDST of the present invention;

[0038] Fig. 9 This is a diagram of system bit error rate analysis under different modulation modes of DCO-OFDM-ASDST of the present invention;

[0039] Fig.10 The present invention is different from 4QAM DCO-OFDM-ASDST DC System bit error rate analysis diagram under mode;

[0040] Fig.11 The present invention is different from 16QAM DCO-OFDM-ASDST DC System bit error rate analysis diagram under mode;

[0041] Fig.12 The present invention is different from 64QAM DCO-OFDM-ASDST DC System bit error rate analysis diagram under mode;

[0042] Fig.13 It is a complexity analysis diagram of the reconstruction algorithm under different modulation modes of ACO-OFDM-ASDST of the present invention;

[0043] Fig.14 It is a complexity analysis diagram of the reconstruction algorithm under different modulation modes of DCO-OFDM-ASDST of the present invention;

[0044] Fig.15 The present invention is different from 4QAM DCO-OFDM-ASDST DC Reconstruct the algorithm complexity analysis diagram in this way;

[0045] Fig.16 The present invention is different from 16QAM DCO-OFDM-ASDST DC Reconstruct the algorithm complexity analysis diagram in this way;

[0046] Fig.17 The present invention is different from 64QAM DCO-OFDM-ASDST DC Reconstruct the algorithm complexity analysis diagram in this way; DETAILED DESCRIPTION

[0047] The invention discloses an adaptive O-OFDM symbol decomposition noise suppression method based on a rough set information decision system.

[0048] The present invention takes two most typical O-OFDM systems: asymmetric limiting optical OFDM (ACO-OFDM) and direct current offset optical OFDM (DCO-OFDM) systems as examples to illustrate specific implementation methods, and other O-OFDM systems can be implemented based on this principle.

[0049] See attached Figure 1 , an adaptive O-OFDM symbol decomposition noise suppression method based on a rough set information decision system, the specific implementation method of the transmitting end is:

[0050] (1) The transmitting source generates a binary information sequence mapped to the M-order quadrature amplitude modulation (M-QAM) constellation point, where X(k) represents the power-normalized modulation symbol, satisfying E[|X(k)| 2 ]=1.

[0051] (2) Map the modulated signal X(k) so that the output signal X satisfies the Hermitian symmetry, that is, X mapping , that is, the second half of its mapping vector is equal to the conjugate mirror image of the first half. The mapping signals of ACO-OFDM and DCO-OFDM are:

[0052]

[0053]

[0054] Where (·) * represents the conjugate operation, (·) Τ represents the matrix transpose and N is the length of the inverse fast Fourier transform.

[0055] (3) Perform IFFT on the mapped signal and output a bipolar real signal x IFFT , T represents the O-OFDM symbol period.

[0056] (4) For x IFFT After parallel-to-serial (P / S) conversion, adaptive symbol decomposition is performed.

[0057] Assume that the linear operating region of the LED is V min ~V max , set the maximum number of symbol decompositions to L, and the upper and lower limit thresholds of the limit are determined by the working range of the LED and the DC bias. The upper and lower limit thresholds of ACO-OFDM are expressed as ε top =V max -B DC , ε bottom =max(V min -B DC , 0), the limiting thresholds of DCO-OFDM are ε top =Vmax -B DC , ε bottom =V min -B DC , the adaptive symbol decomposition process is.

[0058] First, x IFFT Perform the first limiting to obtain the limiting signal Then use x IFFT minus Get the decision symbol of the second symbol decomposition, that is like Then stop symbol decomposition; if Will As the input of the second symbolic decomposition, Perform the second limit. minus Get the decision symbol of the third symbol decomposition, that is like In other words. If this condition is met, the terminal symbol is further decomposed; if Then It is used as the input data for the third symbol decomposition. And so on, until the decision symbol of a symbol decomposition is zero or the maximum number of symbol decompositions is reached.

[0059] The input of the lth symbolic decomposition is Delay the limited output by (l-1)T to obtain the decomposition symbol The limit is expressed as:

[0060]

[0061] Wherein, k=1,2,…,K represents the time domain sampling signal index of the decomposed symbol.

[0062] Limiting the O-OFDM symbol with Gaussian distribution, zero mean and large PAPR is equivalent to attenuating the amplitude of the O-OFDM symbol. Since there is no channel at the transmitting end, the main factor affecting the system performance at the transmitting end is the limiting noise. Therefore, the decomposed symbols after symbol decomposition plus the Gaussian distributed limiting noise can be expressed as:

[0063]

[0064] Where δ represents the amplitude attenuation factor, n clip It is nonlinear limiting noise and is the main noise affecting system performance at this stage.

[0065] (5) Add a cyclic prefix (CP) to the decomposed symbols and serially frame them to obtain frame symbols.

[0066] (6) Perform digital-to-analog conversion (D / A) on the frame symbol and add a DC bias (DC) to obtain a non-negative real signal to drive the LED lamp to send a light signal:

[0067] x LED (t) = x frame (t)+B DC

[0068] In the formula, x frame is the frame symbol obtained by serial framing, B DC is the DC bias.

[0069] Specific implementation method of the receiving end:

[0070] (1) After the intensity modulation / direct detection (IM / DD) optical signal is transmitted through the indoor optical wireless channel, the PD detects the optical signal and converts it into an electrical signal, namely:

[0071] y(t)=γx LED (t)+n(t)

[0072] Where γ is the photoelectric conversion factor, and n(t) represents the sum of background light noise and circuit thermal noise, which is additive white Gaussian noise (AWGN) independent of the signal.

[0073] (2) After the decomposition symbol is converted by A / D and the CP is removed, the decomposition symbol is y (1) ,y (2) ,…,y (l) , where the lth decomposition symbol is expressed as:

[0074]

[0075] Where n clip is the limiting noise, at this time n clip The impact on system performance is small and can be ignored; n l is the AWGN superimposed on the lth decomposition symbol, which is the main noise affecting the system performance at this stage; then the decomposition symbol y( 1 ),y( 2 ),…,y( l ) are added according to the corresponding bits to obtain the combined O-OFDM symbol, that is:

[0076]

[0077] In the formula, B=[B DC B DC …B DC ] T; l is the number of symbol decompositions; it is obvious that AWGN is superimposed on each decomposed symbol. If the received decomposed symbols are directly added and merged, the AWGN will increase. Therefore, the present invention mainly suppresses the AWGN and improves the performance of the VLC system;

[0078] See attached Figure 2 With attached Figure 3 In order to suppress the AWGN on each decomposition symbol, it is necessary to determine the AWGN range on each decomposition symbol, that is, to determine the maximum modulus of AWGN. First, delete the decomposition symbol y (1) ,y (2) ,…,y (l) The DC in Since each decomposed symbol at the transmitter is equal to ε top , ε bottom The time domain sampling value of 0 has a significant indicative effect, so at the receiving end, it is less than ε bottom or greater than ε top The time domain sampling value of will inevitably have random noise superimposed on it, and thus the range of the random noise can be roughly determined, and the maximum noise amplitude modulus F can be determined accordingly as:

[0079]

[0080] Where, k = 0, 1, 2, L, K-1, min(·) means taking the minimum value; the AWGN can also be suppressed by using only the modulus value combined with the limit threshold to define the threshold;

[0081] (3) In order to suppress the AWGN, the decomposed symbols are preprocessed by rough set classification. The specific steps of the algorithm are as follows:

[0082] The RST classification preprocessing algorithm for asymmetric clipped OFDM symbol decomposition (ACO-OFDM-ASDST) is:

[0083] Data preparation: The upper and lower limit thresholds of ACO-OFDM-ASDST are ε top , ε bottom ;

[0084] S1: Time domain sampling value of the sending end Pre-classification, ε top is type 1, ε bottom For 2 types, in ε top and ε bottom The time domain sampling values ​​between are 3 categories;

[0085] S2: Time domain sampling value of the receiving end Add the pre-classification category number bit by bit to create the signal characteristics of the time domain sample value. Construct the ACO-OFDM-ASDST information decision system, where the domain U is composed of the time domain sample value with DC deleted The corresponding signal features are converted into symbolic attributes as conditional attributes C, ε top , ε bottom and is the decision attribute D, Y and N represent yes or no, T represents the decision as this category, and F represents the decision not to this category;

[0086] S3: Time domain sampling value of the receiving end According to the corresponding signal characteristics, the time domain sampling values ​​with signal characteristics of 1 and 2 in the information decision system are preprocessed and classified using RST: φ1: (1 = Y) ∧ (2 = N), φ2: (2 = Y) ∧ (1 = N), then the indistinguishable relationship can be described as a disjunctive normal form (DNF)

[0087] E(A)=φ(1)=φ1∨φ2

[0088] Among them, φ(1) is the decision rule; φ1, φ2 are classification conditions; ∧ represents conjunction operation; ∨ represents disjunction operation;

[0089] See attached Figure 4 With attached Figure 5 In order to formulate the best decision rule, two evaluation indicators, accuracy and recall, are selected to evaluate the formulated decision rule. According to the properties of attribute dependence and importance in rough set theory, the decision rule is formulated, and the best decision rule is formulated. Among them, the formulated decision rule φ(1) is better than the other rules.

[0090] According to this indiscernibility relation E(A), the classification decision and attribute reduction of the domain U can be described as

[0091]

[0092] Where E(A) is an indiscernibility relation, is the equivalence class determined by the classification decision; according to the decision attribute D, the judgment decision is ε top and ε bottom ;

[0093] S4: After simplifying the information decision system attributes according to the indiscernibility relation E(A), only the time domain sampling values ​​between the limiting threshold values ​​remain.

[0094] The RST classification preprocessing algorithm for DC-biased optical OFDM symbol decomposition (DCO-OFDM-ASDST) is:

[0095] Data preparation: The upper and lower limit thresholds of DCO-OFDM-ASDST are ε top , ε bottom , because it is a bipolar signal, there is also 0V;

[0096] S1: Time domain sampling value of the sending end Pre-classification, ε top is type 1, ε bottom is Class 2, 0V is Class 3, for top With 0V and ε bottom The time domain sampling values ​​between 0V and 0V are pre-classified into 4 categories;

[0097] S2: Time domain sampling value of the receiving end Add the pre-classification category number bit by bit to create the signal characteristics of the time domain sample value. Construct a DCO-OFDM-ASDST information decision system, where the domain U is composed of the time domain sample value with DC deleted The corresponding signal features are converted into symbolic attributes as conditional attributes C, ε top , ε bottom , 0V, and is the decision attribute D;

[0098] S3: Time domain sampling value of the receiving end According to the corresponding signal characteristics, the time domain sampling values ​​with signal characteristics of 1, 2 and 3 in the information decision system are preprocessed and classified using RST: φ1: (1 = Y) ∧ (2 = N), φ2: (2 = Y) ∧ (1 = N), φ3: (3 = Y) ∧ (2 = N), then the indistinguishable relationship can be described as a DNF

[0099] φ(1)=φ1∨φ2

[0100] φ(2)=φ2∨φ3

[0101] E(A)=ψ=φ(1)∨φ(2)

[0102] Among them, ψ and φ(2) are decision rules, and φ3 is the classification condition;

[0103] See attached Figure 6 With attached Figure 7 , and two indicators are also selected to evaluate the formulated decision rules, among which the formulated decision rule ψ(1) is better than the other decision rules.

[0104] According to this indiscernibility relation E(A), the classification decision and attribute reduction of the domain U can be described as

[0105]

[0106] S4: After simplifying the attributes of the information decision system according to the indiscernibility relation E(A), only the time domain sampling values ​​between the clipping threshold value and the zero value remain.

[0107] (4) The attributes of the decomposed symbols after preprocessing are simplified and reconstructed, and then the decomposed symbols are merged and added to restore the O-OFDM symbols. The attribute simplification and reconstruction algorithm is:

[0108] The ACO-OFDM-ASDST attribute reduction reconstruction algorithm is:

[0109] See attached Figure 2 Since the information decision system is attribute-simplified in the preprocessing stage, it is only necessary to reconstruct top , ε bottom The time domain sampling value between is defined according to the maximum noise modulus and the limit threshold to reconstruct the upper threshold β top and the lower threshold β bottom ;

[0110] β top =ε top -F

[0111] β bottom =ε bottom +F

[0112] Calculate all The mean of the time domain sampling values ​​between ;

[0113]

[0114] Where l is the number of symbol decompositions, k = 1, 2, ..., K represents the time domain sampling signal index of the decomposed symbol;

[0115] Preprocessing decomposition symbols When the time domain sample value of the decomposed symbol When the time domain sampling value is subtracted from the mean value m, it is restored to the value in ε top , ε bottom The decomposed symbols between, when the time domain sampling value of the decomposed symbols When the time domain sampling value is added with the mean value m, it is restored to the state in ε top , ε bottom The decomposition symbol between .

[0116] The DCO-OFDM-ASDST attribute reduction reconstruction algorithm is:

[0117] See attached Figure 3 Since the information decision system is attribute-simplified in the preprocessing stage, it is only necessary to reconstruct top , ε bottomand 0V respectively; DCO-OFDM-ASDST system needs to define reconstruction β top and β bottom , and the upper threshold of zero value needs to be defined according to the maximum noise modulus value and zero lower threshold

[0118]

[0119] Preprocessing decomposition symbols when Decompose the time domain sample values ​​of the symbol When the time domain sampling value is subtracted from the mean value m, it is restored to the value in ε top , ε bottom The decomposition symbol between 0V and Decompose the time domain sample values ​​of the symbol When the time domain sampling value is added with the mean value m, it is restored to the state in ε top , ε bottom The decomposition symbols between the limiting threshold values ​​​​between 0V and 0V respectively.

[0120] (4) Perform serial-to-parallel conversion (S / P) and Fourier transform (FFT) on the O-OFDM symbols

[0121] (5) Perform frequency domain equalization and information extraction on the O-OFDM symbols, and finally input them into the demodulator to restore the original binary information sequence.

[0122] Simulation experiment

[0123] The Monte Carlo simulation method is used to simulate and analyze the method of the present invention, verifying the rationality and feasibility of an adaptive O-OFDM symbol decomposition noise suppression method based on a rough set information decision system.

[0124] Simulation parameters: Set the Monte Carlo simulation times to 1000, the FFT / IFFT length to 256, and the LED linear working area V min ~V max is 0.1~1V, the maximum number of symbol decompositions is L=8, E b is the signal energy per bit, and N0 is the power spectral density of the noise.

[0125] Simulation Results

[0126] Figure 8 and Fig. 9The following are bit error rate analysis diagrams of different noise suppression algorithms for ACO-OFDM-ASDST and DCO-OFDM-ASDST systems under different modulation modes. The horizontal axis represents the signal-to-noise ratio, in decibels (dB), and the vertical axis represents the BER. The dotted line with the symbol "Δ" in the figure represents the traditional ASDST of 4QAM modulated ACO-OFDM-ASDST and DCO-OFDM-ASDST systems; the dotted line with the symbol "○" represents the maximum noise modulus suppression algorithm of 4QAM modulated ACO-OFDM-ASDST and DCO-OFDM-ASDST systems; the solid line with the symbol "■" represents the RST noise suppression algorithm of 4QAM modulated ACO-OFDM-ASDST and DCO-OFDM-ASDST systems; the solid line with the symbol "X" represents the RST noise suppression algorithm of 4QAM modulated ACO-OFDM-ASDST and DCO-OFDM-ASDST systems; The dotted line represents the traditional ASDST of 16QAM modulated ACO-OFDM-ASDST and DCO-OFDM-ASDST systems; the dotted line with the symbol “◇” represents the maximum noise modulus suppression algorithm of 16QAM modulated ACO-OFDM-ASDST and DCO-OFDM-ASDST systems; the solid line with the symbol “★” represents the RST noise suppression algorithm of 16QAM modulated ACO-OFDM-ASDST and DCO-OFDM-ASDST systems; The dashed lines represent the traditional ASDST of 64QAM modulated ACO-OFDM-ASDST and DCO-OFDM-ASDST systems; The dotted line represents the maximum noise modulus suppression algorithm of the 64QAM modulation ACO-OFDM-ASDST and DCO-OFDM-ASDST systems; the symbol The solid lines represent the RST suppression of 64QAM modulated ACO-OFDM-ASDST and DCO-OFDM-ASDST systems.

[0127] from Figure 8 and Fig. 9 It can be seen that the increase and decrease trend of the signal-to-noise ratio gain is consistent for both the ACO-OFDM-ASDST system and the DCO-OFDM-ASDST system, and the RST noise suppression algorithm has a significant improvement in BER performance compared to the maximum noise modulus suppression algorithm. For example, in 4QAM, 16QAM, and 64QAM modulation, the BER is 10 -5 When compared with the maximum noise modulus suppression algorithm, the RST noise suppression algorithm can achieve SNR gains of about 1dB, 1.2dB, and 1.4dB for the ACO-OFDM-ASDST system. Similarly, in 4QAM and 16QAM modulation, the RST noise suppression algorithm can achieve SNR gains of about 1dB and 2dB for the DCO-OFDM-ASDST system, respectively.

[0128] Fig.10 , Fig.11 and Fig.12 B DC =0.3V, B DC =0.4V and B DC = =0.6V DCO-OFDM-ASDST system under 4QAM, 16QAM and 64QAM modulation, the bit error rate analysis chart of different noise suppression algorithms, the horizontal axis represents the signal-to-noise ratio, the unit is decibel (dB), the vertical axis represents the BER, Fig.10 , 11 The dashed line with the symbol "Δ" in 12 represents B DC = 0.3V, DCO-OFDM-ASDST system conventional ASDST under 4QAM, 16QAM and 64QAM; the dotted line with symbol “○” represents B DC =0.3V, DCO-OFDM-ASDST system maximum noise modulus suppression algorithm under 4QAM, 16QAM and 64QAM; the solid line with the symbol “■” represents B DC =0.3V, RST noise suppression algorithm for DCO-OFDM-ASDST system under 4QAM, 16QAM and 64QAM; with sign The dotted line represents B DC =0.4V, DCO-OFDM-ASDST system conventional ASDST under 4QAM, 16QAM and 64QAM; the dotted line with symbol “◇” represents B DC =0.4V, DCO-OFDM-ASDST system maximum noise modulus suppression algorithm under 4QAM, 16QAM and 64QAM; the solid line with the symbol “★” represents B DC =0.4V, RST noise suppression algorithm for DCO-OFDM-ASDST system under 4QAM, 16QAM and 64QAM; with sign The dotted line represents B DC = 0.6V, DCO-OFDM-ASDST system conventional ASDST at 4QAM, 16QAM and 64QAM; with sign The dotted line represents B DC =0.6V, maximum noise modulus suppression algorithm for DCO-OFDM-ASDST system under 4QAM, 16QAM and 64QAM; with sign The solid line represents B DC =0.6V, RST noise suppression algorithm for DCO-OFDM-ASDST system under 4QAM, 16QAM and 64QAM.

[0129] from Fig.10 , Fig.11 and Fig.12 It can be seen that with B DC The SNR gain of the RST noise suppression algorithm under the same modulation mode shows a gradually decreasing trend. For example, when the BER is 10 -5 , Fig.10 B is 4QAM modulation DC They are 0.3V, 0.4V and 0.6V respectively. The RST noise suppression algorithm obtains SNR gains of approximately 1.6dB, 1dB and 0.4dB respectively compared with the maximum noise modulus suppression algorithm. Fig.11 B is 16QAM modulation DC They are 0.3V, 0.4V and 0.6V respectively. Compared with the maximum noise modulus suppression algorithm, the RST noise suppression algorithm obtains SNR gains of approximately 2.2dB, 2dB and 0.8dB respectively. Fig.12 B is the 64QAM modulation mode DC When the voltages are 0.3V, 0.4V, and 0.6V respectively, the RST noise suppression algorithm achieves SNR gains of approximately 3.2dB, 3dB, and 1dB respectively compared to the maximum noise modulus suppression algorithm.

[0130] Fig.13 and Fig.14 These are the complexity analysis diagrams of different noise suppression reconstruction algorithms for ACO-OFDM-ASDST system and DCO-OFDM-ASDST system under different modulation modes. The horizontal axis represents the number of symbol decompositions, and the vertical axis represents the complexity. The solid line with the symbol "■" represents the maximum noise modulus suppression reconstruction algorithm for 4QAM, 16QAM and 64QAM modulation; the solid line with the symbol "○" represents the 4QAM modulation attribute simplification reconstruction algorithm; the solid line with the symbol "◇" represents the 4QAM modulation attribute simplification reconstruction algorithm; the solid line with the symbol " The solid line represents the 4QAM modulation attribute reduction reconstruction algorithm.

[0131] l is the number of symbol decompositions, N S is the number of time domain sampling values ​​of the decomposed symbol. Since the maximum noise modulus reconstruction algorithm needs to reconstruct all time domain sampling values, when the number of symbol decomposition is l times, the number of time domain sampling values ​​of each decomposition symbol is N. S , so the maximum noise modulus reconstruction algorithm complexity is l*N S The attribute reduction and reconstruction algorithm has only time domain sample values ​​between the clipping threshold values ​​due to attribute reduction. Therefore, when the number of symbol decompositions is l, it is only necessary to add the number of time domain sample values ​​within the clipping threshold. Therefore, the complexity of the attribute reduction and reconstruction algorithm is

[0132] from Fig.13 , 14It can be seen that with the increase of the number of symbol decompositions, the complexity of the maximum noise modulus suppression algorithm shows a growing trend, while the attribute simplification reconstruction algorithm only needs to process the time domain sampling values ​​between the limit thresholds, so there is a very large improvement in complexity. For example, when the number of symbol decompositions is 4, the time domain sampling values ​​of the decomposed symbols processed by the maximum noise modulus reconstruction algorithm in the ACO-OFDM-ASDST systems modulated by 4QAM, 16QAM and 64QAM are all 1024, while the improved reconstruction algorithm is 102, 150 and 184 respectively, that is, the attribute simplification reconstruction algorithm is 1 / 10 of the complexity of the maximum noise modulus reconstruction algorithm, and the same result is obtained in the DCO-OFDM-ASDST system. Although the amount of data processed increases with the increase of the number of symbol decompositions, the complexity of the reconstruction algorithm can still be greatly reduced due to the characteristics of RST attribute simplification.

[0133] Fig.15 , Fig.16 and 17 B DC =0.3V, B DC =0.4V and B DC = =0.6V DCO-OFDM-AS DST system under 4QAM, 16QAM and 64QAM modulation, different noise suppression reconstruction algorithm complexity analysis chart. The horizontal axis represents the number of symbol decompositions, and the vertical axis represents the complexity. The solid line with the symbol “■” represents B DC =0.3V, B DC =0.4V and B DC =0.6V, DCO-OFDM-ASDST system maximum noise modulus suppression reconstruction algorithm under 4QAM, 16QAM and 64QAM; the dotted line with the symbol “○” represents B DC =0.3V, DCO-OFDM-ASDST system attribute reduction reconstruction algorithm under 4QAM, 16QAM and 64QAM; the dotted line with the symbol “◇” represents B DC =0.4V, DCO-OFDM-ASDST system attribute reduction and reconstruction algorithm under 4QAM, 16QAM and 64QAM; with sign The dotted line represents B DC =0.6V, simplified reconstruction of DCO-OFDM-ASDST system properties under 4QAM, 16QAM and 64QAM.

[0134] from Fig.15 , 16 and 17 can be seen in different B DC The complexity of the following two reconstruction algorithms increases with B DCAs the number of symbol decompositions increases, the number of time domain sampling values ​​of the decomposed symbols that need to be processed increases, and the complexity of the maximum noise modulus suppression reconstruction algorithm continues to increase, but the attribute reduction reconstruction algorithm has greatly improved in terms of complexity. For example, when the number of symbol decompositions is 5, under the 4QAM modulation mode, B DC The time domain sampling values ​​that the maximum noise modulus suppression reconstruction algorithm needs to process are 1280, and the time domain sampling values ​​that the attribute simplification reconstruction algorithm needs to process are 135, 156 and 190. That is, the attribute simplification reconstruction algorithm is 1 / 10 of the complexity of the maximum noise modulus suppression reconstruction algorithm. In the 16QAM and 64QAM modulation DCO-OFDM-ASDST system, different B DC The attribute reduction reconstruction algorithm is also 1 / 10 of the complexity of the maximum noise modulus suppression reconstruction algorithm.

[0135] The above are specific implementation methods and simulation verifications of the present invention. It should be pointed out that those skilled in the art can clearly understand that the above embodiments and simulations cited in the system design scheme of the present invention are only used to illustrate and verify the rationality and feasibility of the method, and are not used to limit the method of the present invention. Although the present invention can be effectively illustrated and described through the embodiments, there are many changes in the present invention without departing from the spirit of the present invention. Without departing from the spirit and essence of the method of the present invention, those skilled in the art can make various corresponding changes or deformations according to the method of the present invention, but these corresponding changes or deformations all belong to the protection scope required by the method of the present invention.

Claims

1. An adaptive O-OFDM symbol decomposition noise suppression method based on a rough set information decision system, characterized in that: include: Sending side: Step 1.1: The original binary information sequence is subjected to M-order orthogonal amplitude modulation; Step 1.2: Map the modulated signal to ensure that its output satisfies Hermitian symmetry; Step 1.3: Perform inverse Fourier transform and output bipolar real O-OFDM symbols; Step 1.4: Perform adaptive O-OFDM symbol decomposition after parallel-to-serial conversion of the signal; Step 1.5: Add a cyclic prefix to each decomposed symbol, and then serially frame it to obtain a frame symbol; Step 1.6: Perform digital-to-analog conversion on the frame symbol, and then add a DC bias to obtain a non-negative real signal to drive the LED lamp to send a light signal; Receiver: Step 2.1: The photodetector at the receiving end detects the optical signal transmitted through the optical wireless channel and converts it into an electrical signal; Step 2.2: Perform analog-to-digital conversion, frame division, remove cyclic prefixes and DC offsets on each decomposed symbol, and determine the range of additive white Gaussian noise on the decomposed symbol; Step 2.3: Perform rough set classification preprocessing on the decomposed symbols; Step 2.4: reconstruct the decomposed symbols by simplifying their attributes, and then merge and add the decomposed symbols to restore the O-OFDM symbols; Step 2.5: Perform serial-to-parallel conversion and Fourier transform on the O-OFDM symbols; Step 2.6: Perform frequency domain equalization and information extraction on the O-OFDM symbol, and finally input it into the demodulator to restore the original binary information sequence.

2. The adaptive O-OFDM symbol decomposition noise suppression method based on rough set information decision system according to claim 1, characterized in that: Step 2.4 of the transmitting end, for the parallel-to-serial converted O-OFDM symbol, T represents the O-OFDM symbol period, performs adaptive O-OFDM symbol decomposition, and decomposes the O-OFDM symbol into multiple symbols with smaller PAPR; Set the LED linear working range and the maximum number of symbol decompositions, and perform decomposition according to the dynamic range allowed by the O-OFDM symbol; after each decomposition, subtract the input symbol from the output symbol to check whether the difference is zero; if the difference is zero, stop the symbol decomposition; otherwise, continue to decompose until the difference between the input symbol and the output symbol after a certain decomposition is zero, or the maximum number of symbol decompositions is reached; then the input of the lth symbol decomposition is Delay the limited output by (l-1)T to obtain the decomposition symbol The limit is expressed as: In the formula, is the lth limiting output symbol The variables in , k = 1, 2, L, K represent the time domain sample value index of the decomposed symbol; At the transmitting end, since there is no channel, the main factor affecting the system performance at the transmitting end is the limiting noise. Therefore, each decomposed symbol after symbol decomposition plus the Gaussian distributed limiting noise is expressed as: In the formula, δ represents the amplitude attenuation factor, n clip It is nonlinear limiting noise, which is the main noise affecting system performance at this stage.

3. The adaptive O-OFDM symbol decomposition noise suppression method based on rough set information decision system according to claim 2, characterized in that: In step 2.2 of the receiving end, the photoelectric detector at the receiving end detects the optical signal transmitted through the optical wireless channel and converts it into an electrical signal, performs analog-to-digital conversion on the electrical signal, receives all the decomposed symbols in the frame in sequence, decomposes the frame and deletes the cyclic prefix, and obtains the decomposed symbol y (1) ,y (2) ,…,y (l) , where the lth decomposition symbol is expressed as: In the formula, γ is the photoelectric conversion factor, n clip is the limiting noise; n l It is the noise superimposed on the lth decomposed symbol. It is the sum of the noise caused by the background light and the thermal noise generated by the receiver circuit. It is an additive Gaussian white noise independent of the signal and is the main noise affecting the system performance at this stage. Then decompose the symbol y (1) ,y (2) ,…,y (l) Add the corresponding bits to get the combined O-OFDM symbol, that is, In the formula, B=[B DC B DC …B DC ] T ; l is the number of symbol decompositions.

4. The adaptive O-OFDM symbol decomposition noise suppression method based on rough set information decision system according to claim 3, characterized in that: Step 2.3 of the receiving end uses the information decision system and indistinguishable relationship of the rough set theory to classify and pre-process the time domain sampling values ​​of the decomposed symbols; including two O-OFDM technologies, asymmetric limiting O-OFDM and DC bias optical O-OFDM: The rough set theory classification preprocessing algorithm for asymmetric clipped OFDM symbol decomposition is: Data preparation: The upper and lower limit thresholds of asymmetric clipping OFDM symbol decomposition are ε top , ε bottom ; S1.1: Time domain sampling value of the transmitter Conduct pre-classification; S1.2: Time domain sampling value of the receiving end Add pre-classification category numbers bit by bit to create time domain sample value signal characteristics; build an asymmetric limited OFDM symbol decomposition information decision system; S1.3: Time domain sampling value of the receiving end Preprocessing is performed according to the corresponding signal characteristics, decision rules are formulated according to the information decision system, and indistinguishable relations are derived, which can be described as a disjunctive paradigm: E(A)=φ(1)=φ1∨φ2 In the formula, φ(1) is the decision rule; φ1, φ2 are classification conditions; ∨ represents the disjunction operation; S1.4: After simplifying the information decision system attributes according to the indiscernibility relation E(A), only the time domain sampling values ​​between the clipping threshold values ​​remain; The rough set theory classification preprocessing algorithm for DC bias optical OFDM symbol decomposition is: Data preparation: The upper and lower limit thresholds of DC bias optical OFDM symbol decomposition are ε top , ε bottom , because it is a bipolar signal, there is also 0V; S2.1: Time domain sampling value of the transmitter Conduct pre-classification; S2.2: Time domain sampling value of the receiving end Add pre-classification category numbers bit by bit to create time domain sampling value signal features; build a DC bias optical OFDM symbol decomposition information decision system; S2.3: Time domain sampling value of the receiving end Preprocessing is performed according to the corresponding signal characteristics, decision rules are formulated according to the information decision system, and indistinguishable relations are derived, which can be described as a disjunctive paradigm. φ(1)=φ1∨φ2 φ(2)=φ2∨φ3 E(A)=ψ=φ(1)∨φ(2) In the formula, ψ, φ(2) are decision rules, and φ3 is the classification condition; S2.4: After simplifying the attributes of the information decision system according to the indiscernibility relation E(A), only the time domain sampling values ​​between the limiting threshold and the zero value remain.

5. The adaptive O-OFDM symbol decomposition noise suppression method based on rough set information decision system according to claim 4, characterized in that: Step 2.4 of the receiving end uses the attribute simplification characteristics of the rough set theory to simplify and reconstruct the attributes of the information decision system, reduce the algorithm calculation amount, and thus reduce the algorithm complexity. The attribute simplification and reconstruction algorithms of the asymmetric limited OFDM symbol decomposition and DC biased optical OFDM symbol decomposition systems are: The asymmetric clipped OFDM symbol decomposition attribute reduction and reconstruction algorithm is: Since the information decision system is attribute-simplified in the preprocessing stage, it is only necessary to reconstruct the time domain sampling values ​​between the clipping thresholds, and define the reconstruction upper threshold β top and the lower threshold β bottom ; b top =e top -F b bottom =e bottom +F Calculate all The mean of the time domain sampling values ​​between ; Where l is the number of symbol decompositions, k = 1, 2, ..., K represents the time domain sampling signal index of the decomposed symbol; Preprocessing decomposition symbols When the time domain sample value of the decomposed symbol When the time domain sampling value is subtracted from the mean value m, it is restored to the decomposed symbol between the clipping thresholds. When , the mean value m is added to the time domain sample value to restore it to the decomposed symbol between the clipping thresholds; The DC bias optical OFDM symbol decomposition attribute reduction and reconstruction algorithm is: Since the information decision system is attribute-simplified in the preprocessing stage, it is only necessary to reconstruct top , ε bottom and 0V respectively; define the reconstructed β top and β bottom , zero value upper threshold and zero lower threshold Preprocessing decomposition symbols when Decompose the time domain sample values ​​of the symbol When the time domain sampling value is subtracted from the mean value m, it is restored to the value in ε top , ε bottom Respectively with the decomposition symbol between 0V; when Decompose the time domain sample values ​​of the symbol When the time domain sampling value is added with the mean value m, it is restored to the state in ε top , ε bottom The decomposition symbols between the limiting threshold values ​​​​between 0V and 0V respectively.

Images