Optical fiber communication noise reduction method and system based on singular value decomposition, and computer program

Through singular value decomposition, the problem of increasing system cost by increasing nonlinear damage compensation algorithm in optical fiber communication systems is solved, efficient noise reduction effect is achieved, and hardware requirements are simplified.

CN115441956BActive Publication Date: 2024-11-22XIAN UNIV OF POSTS & TELECOMM
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202211046616.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-08-30
Publication Date
2024-11-22
Estimated Expiration
2042-08-30

AI Technical Summary

Technical Problem

The nonlinear damage compensation algorithm in existing fiber optic communication systems needs to process the encoding method of the original system, which increases the system cost.

Method used

Using a method based on singular value decomposition, the coherent received signal is decomposed singular value, the signal subspace and noise subspace are divided, the singular value in the noise subspace is zeroed, the signal is denoised, and the signal is reduced through the unitary matrix and orthogonal The matrix restores the noise-reduced signal.

Benefits of technology

Effective noise reduction of fiber optic communication systems is achieved, avoiding the processing of the original system encoding method, reducing system costs, simplifying hardware requirements, and improving noise reduction effect.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115441956B_ABST
    Figure CN115441956B_ABST
Patent Text Reader

Abstract

The present invention belongs to a noise reduction method. In order to solve the technical problem that the current algorithm for compensating nonlinear damage needs to process the encoding method of the original system, which increases the system cost, a fiber optic communication noise reduction method and system based on singular value decomposition and a computer program are provided. By performing singular value decomposition on coherent received signals, a noisy space is obtained, and the noise subspace in the noisy space is denoised. Then, the denoised signal is restored according to the singular value decomposition result, and the coherent received signal can be accurately denoised to obtain a good noise reduction effect. While ensuring the noise reduction effect, there is no need to process the encoding method of the original system, and the system cost will not be increased.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to a noise reduction method, and in particular to an optical fiber communication noise reduction method and system based on singular value decomposition, and a computer program. Background Art

[0002] In recent years, with the continuous emergence of mobile Internet and emerging technology applications, more requirements have been put forward in terms of transmission speed, transmission time, transmission distance, transmission bandwidth, etc. It has become particularly urgent to solve the problems of large-scale data transmission and damage compensation in the operation of high-speed wireless optical communication systems. How to achieve higher transmission rates, longer transmission distances, and larger transmission bandwidths has become the main focus of researchers.

[0003] The nonlinear effects and dispersion in optical fiber network communication technology have seriously restricted the rapid development of modern network optical fiber communication. Therefore, the research on the suppression of nonlinear effects has become increasingly important. Nonlinear damage in optical fiber transmission can be divided into two categories: one is deterministic nonlinear damage, such as self-phase modulation (SPM), cross-phase modulation (XPM) and four-wave mixing (FWM), often referred to as nonlinear interference noise (NLIN). This type of damage is not strictly speaking noise, but distortion. The other is random nonlinear damage. The noise generated by the erbium-doped fiber amplifier will randomly change the amplitude of the signal pulse. Due to the existence of the Kerr effect, this amplitude change will be converted into a random phase change, which is called nonlinear phase noise (NLPN).

[0004] In response to nonlinear damage, scholars have proposed a variety of nonlinear compensation algorithms to compensate for nonlinear damage. Among them, the most common is the digital reverse propagation algorithm (DBP), which can, in principle, simultaneously compensate for deterministic linear damage and non-deterministic nonlinear damage, and is a key technology for the DSP of the receiving module. However, due to the high complexity of the nonlinear compensation of the standard DBP algorithm, researchers have also proposed a large number of improved algorithms for DBP. For example, the probability shaping algorithm adjusts the constellation mapping in coding so that the constellation symbols can be presented with different probabilities, so that the signal power of the outer circle can be less than the signal power of the inner circle, thereby reducing the bit rate and increasing the gap between the channel capacity and the Shannon limit. However, this algorithm requires processing the coding method of the original system, which increases the cost of the system. Summary of the invention

[0005] The present invention aims to solve the technical problem that the current algorithm for compensating nonlinear damage needs to process the encoding method of the original system, which increases the system cost. The present invention provides a fiber optic communication noise reduction method and system based on singular value decomposition, and a computer program.

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

[0007] A method for reducing noise in optical fiber communication based on singular value decomposition is characterized in that it comprises the following steps:

[0008] S1, perform singular value decomposition calculation on the coherent received signal to obtain the singular value sequence

[0009] S1.1, the autocorrelation matrix R of the signal ss and the autocorrelation matrix R of the noise ww Sum them up to get the autocorrelation matrix of the coherent received signal;

[0010] S1.2, performing singular value decomposition on the autocorrelation matrix of the coherent received signal to obtain a singular value matrix of the coherent received signal containing m singular values ​​and a corresponding orthogonal matrix V, and recording the space spanned by the m singular values ​​in the singular value matrix of the coherent received signal as a noisy space; wherein m is equal to the dimension of the autocorrelation matrix of the coherent received signal;

[0011] S2, get the signal subspace and noise subspace

[0012] Set the spatial demarcation point g, and use the space spanned by the first g singular values ​​in the noisy space as the signal subspace, and the space spanned by the last mg singular values ​​as the noise subspace;

[0013] S3, setting the singular values ​​of the noise subspace in the noisy space to zero to obtain the noise reduction space;

[0014] S4, signal reconstruction

[0015] S4.1, obtaining a unitary matrix U according to the coherent received signal, the singular values ​​in the singular value matrix of the coherent received signal, and the orthogonal matrix V;

[0016] S4.2, obtaining the denoised signal motion trajectory according to the unitary matrix U, the orthogonal matrix V and the denoised space;

[0017] S4.3, restoring the coherent received signal after noise reduction according to the motion trajectory of the noise reduction signal.

[0018] Furthermore, step S2 is specifically as follows:

[0019] S2.1, arranging the main diagonal elements of the singular value matrix into a sequence D;

[0020] S2.2, performing differential calculations on two adjacent items in sequence D in sequence, and the differential calculation results obtained constitute a gradient sequence C;

[0021] S2.3, compare each item in the gradient sequence C with the sum of the next five adjacent items in turn, and the comparison results form a ratio sequence B. Take the estimated value of the sequence number corresponding to the maximum value in the ratio sequence B as the dividing point g of the noisy space, and record the space spanned by the first g singular values ​​in the noisy space as the signal subspace, and record the space spanned by the last mg singular values ​​as the noise subspace.

[0022] Furthermore, step S1.2 is specifically as follows:

[0023] S1.2.1, the autocorrelation matrix of the coherent received signal is subjected to singular value decomposition by the following formula to obtain the orthogonal matrix corresponding to the singular value matrix of the coherent received signal:

[0024] R yy =V(Λ s +Λ w )V H

[0025] Among them, R yy represents the autocorrelation matrix of the coherent received signal, V represents the corresponding orthogonal matrix, H represents the conjugate transpose, Λ s represents the singular value matrix of the signal, Λ w The singular value matrix representing the noise;

[0026] S1.2.2, the singular value matrix Λ of the coherent received signal is obtained by the following formula y

[0027] Λ y =Λ s +Λ w

[0028] S1.2.3, the singular value matrix Λ of the coherent received signal y The m singular values ​​in span the space as the noisy space.

[0029] Furthermore, step S4.1 is specifically as follows:

[0030] The unitary matrix U is obtained by the following formula:

[0031]

[0032] Where y represents the received signal, λ r Represents the rth entry of the singular value matrix of the coherent received signal.

[0033] Furthermore, step S4.2 is specifically as follows:

[0034] The motion trajectory h of the noise reduction signal is obtained by the following formula:

[0035] h=UΛ1V

[0036] Wherein, Λ1 represents the noise reduction space.

[0037] Furthermore, step S4.3 is specifically as follows:

[0038] The coherent received signal y1 after noise reduction is restored by the following formula:

[0039]

[0040] Where N represents the sequence length of the coherent received signal, M represents the variable, N ≥ M ≥ 1, k represents the dimension of the signal subspace, and h i,M-i+1 It represents the motion trajectory of the noise reduction signal corresponding to the i-th row and the M-i+1-th column in the coherent received signal sequence.

[0041] Furthermore, in step S1.2, the value of m is 50-100.

[0042] The present invention also provides an optical fiber communication noise reduction system based on singular value decomposition, which is special in that it includes a singular value decomposition module, a space demarcation point determination module, a noise reduction module and a signal reconstruction module connected in sequence;

[0043] The singular value decomposition module is used to calculate the autocorrelation matrix R of the signal ss and the autocorrelation matrix R of the noise ww Sum the autocorrelation matrix of the coherent received signal, and perform singular value decomposition on the autocorrelation matrix of the coherent received signal to obtain a singular value matrix of the coherent received signal, a corresponding orthogonal matrix V and a noisy space;

[0044] The space demarcation point determination module is used to determine the space demarcation point and divide the noisy space into a signal subspace and a noise subspace;

[0045] The noise reduction module is used to set the singular values ​​of the noise subspace in the noisy space to zero to obtain the noise reduction space;

[0046] The signal reconstruction module is used to restore the coherent received signal after noise reduction according to the coherent received signal, the singular values ​​in the singular value matrix of the coherent received signal, the orthogonal matrix V, and the noise reduction space.

[0047] At the same time, the present invention also provides a computer program product, including a computer program, the special feature of which is that when the program is executed by a processor, the steps of the above-mentioned optical fiber communication noise reduction method based on singular value decomposition are implemented.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] 1. The present invention provides a fiber optic communication noise reduction method based on singular value decomposition, which performs singular value decomposition on coherent received signals to obtain a noisy space, denoises the noise subspace in the noisy space, and then restores the denoised signal according to the singular value decomposition result, which can accurately denoise the coherent received signal and obtain a good noise reduction effect. While ensuring the noise reduction effect, there is no need to process the encoding method of the original system, and the system cost will not be increased.

[0050] 2. The noise reduction method provided by the present invention only needs to perform singular value decomposition calculation after coherent reception. There is no need to adjust the constellation mapping during encoding, and there is no need to learn a huge data set required by an artificial neural network. The method can complete the noise reduction. The method is simple and easy to operate, which is conducive to promotion and use.

[0051] 3. The noise reduction method provided by the present invention has fewer hardware requirements beyond the standard optical fiber communication system.

[0052] 4. The present invention also provides a noise reduction system that can implement the above-mentioned noise reduction method, and can achieve good noise reduction effects in a modular form.

[0053] 5. The present invention also provides a computer program product capable of executing the above method steps, so as to facilitate the promotion and application of the method of the present invention and realize integration on corresponding hardware devices. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 A schematic diagram of a process of an embodiment of a method for optical fiber communication noise reduction based on singular value decomposition according to the present invention;

[0055] Figure 2 A schematic diagram of the receiving principle of a coherent received signal in an optical fiber communication system applied in an embodiment of the present invention;

[0056] Figure 3 : The constellation diagrams of the coherent received signal and the corresponding noise reduction signal obtained in the embodiment of the present invention; wherein, (a) is the constellation diagram of the 100km coherent received signal, (b) is the constellation diagram of the 100km noise reduction signal, (c) is the constellation diagram of the 300km coherent received signal, (d) is the constellation diagram of the 300km noise reduction signal, (e) is the constellation diagram of the 500km coherent received signal, (f) is the constellation diagram of the 500km noise reduction signal, (g) is the constellation diagram of the 1000km coherent received signal, and (h) is the constellation diagram of the 1000km noise reduction signal.

[0057] Among them, 1-laser, 2-arbitrary signal generator, 3-I / Q modulator, 4-optical fiber, 5-amplifier, 6-coherent receiver. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0059] The present invention provides a fiber optic communication noise reduction method based on singular value decomposition, and proposes to perform singular value decomposition on the autocorrelation matrix of a coherent received signal, divide the singular value sequence in the singular value matrix through a series of difference and ratio operations, and obtain a signal subspace and a noise subspace. Finally, the noise subspace is subjected to noise reduction processing, and the coherent received signal after noise reduction is restored, thereby realizing the noise reduction function of the fiber optic communication system.

[0060] The specific steps of the noise reduction method of the present invention are as follows:

[0061] 1) Determine the dimension m of each correlation function matrix, where m is generally 50-100;

[0062] 2) Construct the autocorrelation matrix of the coherent received signal:

[0063] The modulated signal s(n) passes through the optical fiber channel and coherent reception, and the coherent reception signal y(n) is sampled without distortion and can be expressed as:

[0064] y(n)=s(n)+w(n)=s(n)+w1(n)+w2(n)

[0065] n=1,2,…,N

[0066] Among them, w1(n) represents nonlinear noise, w2(n) represents ASE noise, and w(n) represents the sum of nonlinear noise and ASE noise, with a mean of 0 and a variance of σ 2 , N represents the sequence length of the coherent received signal. The signal and noise distributions are independent, and the autocorrelation matrix of the coherent received signal is:

[0067] R yy =E[y(n)y H (n)]=E[(s(n)+w(n))(s(n)+w(n)) H ]=E[s(n)s H (n)]+E[w(n)w H (n)]=R ss +R ww

[0068] Among them, H represents the conjugate transpose and E represents the expectation. It can be seen that the autocorrelation matrix R of the coherent received signal is yyEqual to the autocorrelation matrix R of the signal s(n) ss and the autocorrelation matrix R of the noise w(n) ww sum.

[0069] 3) Singular Value Decomposition:

[0070] The autocorrelation matrix R of the coherent received signal yy is a symmetric matrix, so it can be decomposed into singular values ​​(SVD):

[0071] R yy =R ss +R ww =VΛ y V H =V(Λ s +A W )V H

[0072] Where V represents an orthogonal matrix; Λ s represents the singular value matrix of the signal,

[0073] Λ s =diag(γ1,γ2,...,γ p , 0, .., 0), where diag(·) represents a diagonal matrix, 1<p<m, γ p Represents Λ s For the pth element in , γ1>γ2>…>γ p , Λ w represents the singular value matrix of the noise, Λ w =diag(σ 2 , σ 2 , ..., σ 2 ), σ 2 represents the elements in the singular value matrix of the noise, Λ y =diag(λ1,λ2,...,λ m )=Λ s +Λ w , Λ y Represents the singular value matrix of the coherent received signal.

[0074] The space spanned by the m singular values ​​in the singular value matrix of the coherent received signal is called the noisy space, where the space spanned by the first g singular values ​​is called the signal subspace, and the space spanned by the last mg singular values ​​is called the noise subspace.

[0075] 4) Determine the dividing point g between the signal subspace and the noise subspace

[0076] The sequence composed of the main diagonal elements of the singular value matrix of the coherent received signal is defined as sequence D. Starting from the head, the gradient sequence C is obtained by differentiating the two adjacent terms in sequence D, as shown in the following formula:

[0077] C(e)=D(d)-D(d-1), d=1, 2,..., m-1

[0078] Wherein, D(d) represents the d-th element of the singular value sequence D, C(e) represents the e-th element of the gradient sequence, and e=1, 2, ..., m-1.

[0079] Compare each item of the gradient sequence C with the sum of the next five items to obtain the ratio sequence B:

[0080]

[0081] The estimated value of the sequence number corresponding to the maximum value in the ratio sequence B is taken as the dividing point g of the noisy space, and the space spanned by the first g singular values ​​in the noisy space is recorded as the signal subspace, and the space spanned by the last mg singular values ​​is recorded as the noise subspace.

[0082] Among them, the dimension g of the signal subspace is as follows:

[0083]

[0084] Through analysis, we can see that the value of B(f) is always less than 1, and the sequence D and gradient sequence C have the following properties:

[0085] D(1)>…>D(g)>D(g+1)≈…≈D(m)≈σ 2

[0086] C(1)>…>C(g)>C(g+1)≈…≈C(m-1)≈Δ

[0087] Here, Δ represents a symbol approaching 0.

[0088] It can be seen from this that Therefore, B(f) reaches its maximum at g.

[0089] 5) Singular value processing of noise subspace

[0090] The singular values ​​with dimension greater than g in the noisy space are changed to 0, that is, the singular values ​​in the noise subspace are set to zero, that is, the singular values ​​of the signal subspace and the noise subspace are: λ1, λ2, …, λ g ,0,0,…,0, and get the denoising space.

[0091] 6) Signal reconstruction

[0092] There are two parts, one is to obtain the motion trajectory of the noise reduction signal, and the other is to restore the noise reduction signal:

[0093] ①Movement trajectory of noise reduction signal

[0094] A. The unitary matrix U is obtained by using the coherent received signal, the singular values ​​in the singular value matrix of the coherent received signal, and the orthogonal matrix V. The calculation formula is as follows:

[0095]

[0096] Where V(i) represents the i-th column of the orthogonal matrix V, λ i represents the i-th singular value in the singular value matrix of the coherent received signal, and y represents the coherent received signal.

[0097] B. According to the unitary matrix U, the orthogonal matrix V and the noise reduction space, the noise reduction signal motion trajectory matrix h is obtained, and the formula is as follows:

[0098] h=U H Λ1V

[0099] ②Diagonal averaging

[0100] The noise reduction signal is restored through the motion trajectory of the noise reduction signal. The formula is as follows:

[0101]

[0102] Where m represents the dimension of the signal subspace, N represents the sequence length of the coherently received signal, k is the number of columns in the trajectory matrix, and h is i,M-i+1 represents the motion trajectory of the noise reduction signal corresponding to the i-th row and the M+i+1-th column in the coherent received signal sequence, where M is a variable.

[0103] Based on the above-mentioned noise reduction method, the present invention also provides an optical fiber communication noise reduction system based on singular value decomposition, including a singular value decomposition module, a spatial demarcation point determination module, a noise reduction module and a signal reconstruction module. Among them, the singular value decomposition module is used to perform autocorrelation matrix operations on the acquired coherent received data, and then calculate the singular values ​​of its autocorrelation matrix to obtain the singular value matrix of the coherent received signal, the corresponding orthogonal matrix V and the noisy space. The spatial demarcation point determination module is used to perform difference and ratio operations on the singular value sequence, and determine the estimated value of the dimension corresponding to the maximum value of the ratio result as the demarcation point between the signal subspace and the noise subspace. The noise reduction module is used to set the singular values ​​of the noise subspace in the noisy space to zero to obtain the noise reduction space. The signal reconstruction module is used to restore the noise-reduced signal according to the coherent received signal, the singular values ​​in the singular value matrix of the coherent received signal, the orthogonal matrix V, and the noise reduction space.

[0104] In addition, an optical fiber link transmission data acquisition module can also be provided to collect signals after transmission over a long distance optical fiber link, and to obtain coherent reception data transmitted under different optical fiber distances through a coherent receiver.

[0105] like Figure 1 , the following is a specific example of the noise reduction method of the present invention:

[0106] Step 1: Obtaining optical fiber communication transmission data

[0107] like Figure 2 , is a schematic diagram of the receiving principle of the coherent receiving signal in the optical fiber communication system. The laser 1 emits a laser, the arbitrary signal generator 2 generates an electrical signal, and the I / Q modulator 3 receives the electrical signal and modulates the laser signal emitted by the laser 1. After passing through N groups of optical fibers 4 and amplifiers 5, the signal is received to obtain the coherent receiving signal. By changing the input power of the laser 1 and the transmission distance of the optical fiber 4, the coherent receiver 6 is used to obtain the coherent receiving data after transmission under different optical signal-to-noise ratios. Specifically, the experimental center wavelength of the laser 1 as the signal light source is 1550nm. After the coupler, one beam of light is converted into two beams of light, and then the arbitrary waveform generator drives the I / Q modulator 3 for modulation. After the coupler combines the light, a single polarization multiplexing quadrature phase shift keying (QPSK) signal with a symbol rate of 28GBaud is obtained. After passing through the optical fiber 4 and the amplifier 5, the optical signal enters the optical fiber communication system data acquisition module.

[0108] Generally, erbium-doped fiber amplifiers and nonlinear noise of optical fibers are used as noise sources, and the coupling power ratio of the noise source to the signal light is controlled by changing the input power of the laser 1. The sampling rate of the coherent receiver 6 used is 4 times or more of the signal symbol rate, so that the accuracy of the algorithm estimation can be achieved. The potential value sampling multiple is 16. The higher the sampling rate, the more complex information of the signal is collected, and the more accurate the singular value decomposition algorithm value calculated is.

[0109] Step 2: Use the singular value decomposition algorithm to calculate the coherent received data and obtain the corresponding singular value sequence value

[0110] In this step, the dimension of the autocorrelation matrix is ​​set to 50-100 to achieve accurate calculation of the singular value sequence. If the dimension of the autocorrelation matrix is ​​too large, unnecessary calculations will be increased, and if it is too small, the constructed autocorrelation matrix will be distorted.

[0111] Step 3: Process the singular value sequence to obtain the dividing point between the signal subspace and the noise subspace.

[0112] Step 4: Process the noise subspace and set the singular value sequence of the noise subspace to 0.

[0113] Step 5: Use the coherent received signal and the denoised singular value sequence to calculate the signal trajectory matrix, and then diagonalize the trajectory matrix to obtain the denoised signal.

[0114] The coherent received signal and the signal after noise reduction in this embodiment are subjected to dispersion compensation and adaptive filtering respectively, and the following is obtained: Figure 3 As shown in the figure, the corresponding constellation diagrams before and after noise reduction. The right side represents the constellation diagram after noise reduction using the singular value algorithm (the optical signal-to-noise ratio range is 5-30dB), (a), (c), (e), and (g) are the constellation diagrams of the coherent received signal at different distances, and (b), (d), (f), and (h) are the constellation diagrams of the signal after noise reduction at different distances.

[0115] The noise reduction method of the present invention may also be formed into a computer program product, which includes a computer program. When the program is executed by a processor, the steps of the noise reduction method are implemented.

[0116] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for optical fiber communication noise reduction based on singular value decomposition, characterized in that: The following steps are involved: S1, perform singular value decomposition calculation on the coherent received signal to obtain the singular value sequence S1.1, the autocorrelation matrix R of the signal ss and the autocorrelation matrix R of the noise ww Sum them up to get the autocorrelation matrix of the coherent received signal; S1.2, performing singular value decomposition on the autocorrelation matrix of the coherent received signal to obtain a singular value matrix of the coherent received signal containing m singular values ​​and a corresponding orthogonal matrix V, and recording the space spanned by the m singular values ​​in the singular value matrix of the coherent received signal as a noisy space; wherein m is equal to the dimension of the autocorrelation matrix of the coherent received signal; S2, get the signal subspace and noise subspace Set the spatial demarcation point g, and use the space spanned by the first g singular values ​​in the noisy space as the signal subspace, and the space spanned by the last mg singular values ​​as the noise subspace; S3, setting the singular values ​​of the noise subspace in the noisy space to zero to obtain the noise reduction space; S4, signal reconstruction S4.1, the unitary matrix U is obtained by the following formula: Where y represents the received signal, λ r represents the rth entry of the singular value matrix of the coherent received signal; S4.2, the noise reduction signal motion trajectory h is obtained by the following formula: h=UΛ1V Among them, Λ1 represents the noise reduction space; S4.3, restore the noise-reduced coherent received signal y1 through the following formula: Where N represents the sequence length of the coherent received signal, M represents the variable, N ≥ M ≥ 1, k represents the dimension of the signal subspace, and h i,M-i+1 It represents the motion trajectory of the noise reduction signal corresponding to the i-th row and the M-i+1-th column in the coherent received signal sequence.

2. The optical fiber communication noise reduction method based on singular value decomposition according to claim 1, characterized in that: Step S2 is specifically as follows: S2.1, arranging the main diagonal elements of the singular value matrix into a sequence D; S2.2, performing differential calculations on two adjacent items in sequence D in sequence, and the differential calculation results obtained constitute a gradient sequence C; S2.3, compare each item in the gradient sequence C with the sum of the next five adjacent items in turn, and the comparison results form a ratio sequence B. Take the estimated value of the sequence number corresponding to the maximum value in the ratio sequence B as the dividing point g of the noisy space, and record the space spanned by the first g singular values ​​in the noisy space as the signal subspace, and record the space spanned by the last mg singular values ​​as the noise subspace.

3. The optical fiber communication noise reduction method based on singular value decomposition according to claim 2, characterized in that: Step S1.2 is specifically as follows: S1.2.1, the autocorrelation matrix of the coherent received signal is subjected to singular value decomposition by the following formula to obtain the orthogonal matrix corresponding to the singular value matrix of the coherent received signal: R yy =V(Λ s +L w )V H Among them, R yy represents the autocorrelation matrix of the coherent received signal, V represents the corresponding orthogonal matrix, H represents the conjugate transpose, Λ s represents the singular value matrix of the signal, Λ w The singular value matrix representing the noise; S1.2.2, the singular value matrix Λ of the coherent received signal is obtained by the following formula y L y =L s +L w S1.2.3, the singular value matrix Λ of the coherent received signal y The m singular values ​​in span the space as the noisy space.

4. The optical fiber communication noise reduction method based on singular value decomposition according to any one of claims 1 to 3, characterized in that: In step S1.2, the value of m is 50-100.

5. An optical fiber communication noise reduction system based on singular value decomposition, used to implement the optical fiber communication noise reduction method based on singular value decomposition according to any one of claims 1 to 4, characterized in that: It includes a singular value decomposition module, a space demarcation point determination module, a noise reduction module and a signal reconstruction module which are connected in sequence; The singular value decomposition module is used to calculate the autocorrelation matrix R of the signal ss and the autocorrelation matrix R of the noise ww Sum the autocorrelation matrix of the coherent received signal, and perform singular value decomposition on the autocorrelation matrix of the coherent received signal to obtain a singular value matrix of the coherent received signal, a corresponding orthogonal matrix V and a noisy space; The space demarcation point determination module is used to determine the space demarcation point and divide the noisy space into a signal subspace and a noise subspace; The noise reduction module is used to set the singular values ​​of the noise subspace in the noisy space to zero to obtain the noise reduction space; The signal reconstruction module is used to restore the coherent received signal after noise reduction according to the coherent received signal, the singular values ​​in the singular value matrix of the coherent received signal, the orthogonal matrix V, and the noise reduction space.

6. A computer program product, comprising a computer program, characterized in that: When the program is executed by a processor, the steps of the optical fiber communication noise reduction method based on singular value decomposition according to any one of claims 1 to 4 are implemented.