State estimation system, confidential signal generation device, state estimation method, and confidential signal generation program

Abstract

The present invention is provided with a secret signal generation unit (12) that: acquires training constellation data and identification constellation data outputted from a signal processing circuit for optical communications; by means of random projection, reduces and conceals the number of pieces of data from each kind of constellation data; and generates a training secret signal and an identification secret signal on the basis of each kind of constellation data after reducing and concealing the number of pieces of data. In addition, the present invention is provided with: a sparse dictionary learning unit (21) which, on the basis of the training secret signal, learns a secret sparse dictionary by using an algorithm for sparse dictionary learning; and an identification unit (22) which, on the basis of the identification secret signal, estimates the state of optical communications by using the secret sparse dictionary.

Description

[Technical field]

[0001] The present invention relates to a state estimation system, a concealed signal generation device, a state estimation method, and a concealed signal generation program. [Background technology]

[0002] In digital coherent communication, constellation data is generated, which represents the transmission data in a polar coordinate diagram of amplitude and phase, and the quality of communication is confirmed by analyzing the deviation of this constellation data from the theoretical value. By checking the quality of communication, it is possible to quickly identify the cause of degradation of communication quality and take measures to improve the quality of communication.

[0003] For example, Non-Patent Document 1 describes a state estimation method for optical communications using sparse coding. [Prior art documents] [Non-patent literature]

[0004] [Non-Patent Document 1] Takayuki Nakachi, Yitu Wang, Tetsuro Inui, Takaaki Tanaka, Takahiro Yamaguchi, and Katsuhiro Shimano,"State Estimation in Optical Communications Using Sparse Coding,"Technical Report of the Institute of Electronics, Information and Communication Engineers, Optical Communication Systems Research Committee, vol. 119, no. 229,OCS2019-42, pp. 77-82. Summary of the Invention [Problem to be solved by the invention]

[0005] However, the analysis of constellation data is highly dependent on the experience of experts. In addition, in order to estimate the cause of quality degradation in optical communication by statistical approaches or deep learning, it is necessary to acquire a large amount of constellation data, which causes a problem of a huge amount of calculation. However, the technology disclosed in the above-mentioned Non-Patent Document 1 does not mention reducing the amount of data, and does not solve the problem of a huge amount of calculation.

[0006] The present invention has been made in consideration of the above circumstances, and an object of the present invention is to provide a state estimation system, a state estimation method, and a enciphered signal generation program that are capable of estimating the cause of quality degradation in optical communications without increasing the amount of calculations. [Means for solving the problem]

[0007] A state estimation system according to one embodiment of the present invention includes a concealment signal generation unit that acquires training constellation data and identification constellation data output from an optical communication signal processing circuit, reduces the number of data items from each of the constellation data by random projection and conceals the data, and generates a training concealment signal and an identification concealment signal based on each of the constellation data after the reduction in the number of data items and concealment, a sparse dictionary learning unit that learns a concealment sparse dictionary using a sparse dictionary learning algorithm based on the training concealment signal, and an identification unit that estimates a state of optical communication using the concealment sparse dictionary based on the identification concealment signal.

[0008] An encipherment signal generation device according to one aspect of the present invention includes a data acquisition unit that acquires training constellation data and identification constellation data output from an optical communication signal processing circuit, and a encipherment signal generation unit that reduces the number of data items from each of the constellation data items by random projection and enciphers the data, and generates training encipherment signals and identification encipherment signals based on each of the constellation data items after the reduction in the number of data items and the encipherment.

[0009] A state estimation method according to one embodiment of the present invention includes the steps of acquiring training constellation data output from a signal processing circuit for optical communications, reducing the number of data items from the training constellation data by random projection and concealing the data, and generating a training concealed signal based on the training constellation data after the data reduction and concealment, learning a concealed sparse dictionary using a sparse dictionary learning algorithm based on the training concealed signal, acquiring identification constellation data output from the signal processing circuit for optical communications, reducing the number of data items from the identification constellation data by random projection and concealing the data, and generating an identification concealed signal based on the identification constellation data after the data reduction and concealment, and estimating a state of optical communications using the concealed sparse dictionary based on the identification concealed signal.

[0010] One aspect of the present invention is a concealment signal generating program for causing a computer to function as the concealment signal generating device. Effect of the Invention

[0011] According to the present invention, it is possible to estimate the cause of quality degradation in optical communications without increasing the amount of calculation. [Brief description of the drawings]

[0012] [Figure 1] FIG. 1 is an explanatory diagram showing 16QAM constellation data. [Figure 2A] FIG. 2A is an explanatory diagram showing constellation data when a modulator parent bias phase error occurs. [Figure 2B] FIG. 2B is an explanatory diagram showing constellation data in an I / Q gain imbalance state. [Figure 2C] FIG. 2C is an explanatory diagram showing constellation data in an I / Q skew imbalance state. [Diagram 3]FIG. 3 is a block diagram showing the configuration of the state estimation system according to this embodiment. [Figure 4] FIG. 4 is a block diagram showing a detailed configuration of the confidential signal generating unit. [Diagram 5] FIG. 5 is an explanatory diagram showing a flow of generating a concealed signal by random projection from the number of constellation data or the histogram I(s, t). [Figure 6] FIG. 6 is an explanatory diagram showing a sparse model when constellation data is not concealed. [Figure 7] FIG. 7 is an explanatory diagram showing a sparse model in which the dimensions are reduced and concealed by random projection. [Figure 8] FIG. 8 is a block diagram showing a hardware configuration. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0013] A state estimation system according to this embodiment will be described below. In this embodiment, the amount of constellation data is reduced and concealed by random projection, and the state of optical communication is estimated using the constellation data after the amount of data reduction and concealment. A "state estimation method using constellation data" and a "state estimation method based on concealment calculation using random projection" will be described below.

[0014] [State estimation method using constellation data] Constellation data can represent data transmitted in digital coherent communication on a complex plane, which makes it possible to visually represent the phase and amplitude information of a coherent communication signal.

[0015] Fig. 1 is an explanatory diagram showing constellation data of 16QAM (quadrature amplitude modulation). For the positions of the 16 points shown in Fig. 1, the direction of rotation about an axis indicates phase information, and the distance from the origin indicates amplitude information.

[0016] For example, a 16QAM signal can transmit 16 points (4 bits) of information per symbol. Note that constellation data represents the integral state over a certain period of time, and at any given time the signal state indicates one of the 16 points.

[0017] Since the constellation data indicates the phase and amplitude states of the signal, the shape of the data makes it possible to estimate the state of the transmission path and the optical transmitter. Here, specific examples of optical transmitter state estimation are shown for three errors: "modulator parent bias phase error," "I / Q gain imbalance state," and "I / Q skew imbalance state." The main cause of each of the above errors is insufficient adjustment of the optical IQ modulation module of the optical communication device. The specific causes of the expected error occurrence are shown below in (a) to (c).

[0018] (a) Modulator parent bias phase error Fig. 2A is an explanatory diagram showing constellation data when a modulator parent bias phase error occurs. As shown in Fig. 2A, the constellation data is distorted into a diamond shape due to the phase shift. From this result, it can be determined that there is a possibility that the bias of the phase modulator is shifted.

[0019] (b) I / Q gain imbalance condition 2B is an explanatory diagram showing constellation data when there is an I / Q gain imbalance. In this case, the I axis is longer than expected or the Q axis is shorter than expected. From this result, it can be determined that there is a possibility that a deviation has occurred in the drive amplitude of "I" or "Q" of the optical I / Q modulator.

[0020] (c) I / QSkew imbalance condition FIG. 2C is an explanatory diagram showing constellation data when there is an I / Q skew imbalance state. The four corner points of the constellation map are located at the expected positions. However, the trajectory of the transition from one point to another is different from the expected one. From this result, it can be determined that there is a possibility that a deviation has occurred in the correction value of the signal delay.

[0021] In this way, by using the constellation data, it is possible to estimate the communication quality of the digital coherent communication. In this embodiment, by adopting random projection, the constellation data is concealed, and the data amount and the calculation load are reduced to estimate the state of the optical communication.

[0022] [State estimation method based on secret computation using random projection] A specific example of a state estimation method based on a secret operation using random projection will be described below. [1. Overall overview of the state estimation system], [2. Generation of secret signals using random projection], [3. General sparse dictionary learning and identification], and [4. Secret sparse dictionary learning and secret identification] will be described below.

[0023] [1. Overview of the state estimation system] An overview of a state estimation system that estimates a state of optical communication using confidential sparse coding will be described below. Fig. 3 is a block diagram showing a configuration of a state estimation system 100 according to this embodiment.

[0024] 3, the state estimation system 100 according to this embodiment includes a plurality of concealment signal generation devices 1 (1-1 to 1-N) and a calculation device 2. Each of the concealment signal generation devices 1 and the calculation device 2 are connected via a network 3.

[0025] Each of the encryption signal generation devices 1 (1-1 to 1-N) includes a data acquisition unit 11 and an encryption signal generation unit 12. The calculation device 2 includes a sparse dictionary learning unit 21 and a classification unit 22.

[0026] The data acquisition unit 11 is connected to a digital coherent signal processing circuit 10 (hereinafter referred to as "DSP 10").

[0027] The DSP 10 processes signals transmitted and received in digital coherent communication. The data acquisition unit 11 acquires constellation data from the DSP 10.

[0028] The encrypting signal generating unit 12 generates an encrypting signal (learning encrypting signal) based on the learning constellation data acquired by the data acquiring unit 11. The encrypting signal generating unit 12 also generates an encrypting signal (identification encrypting signal) based on the identification constellation data acquired by the data acquiring unit 11. The encrypting signal generating unit 12 will be described in detail later with reference to FIG. 4.

[0029] The sparse dictionary learning unit 21 performs learning of the private sparse dictionary using the "Label Consistent K-SVD algorithm" (hereinafter, referred to as the "LC K-SVD algorithm"), which is a sparse dictionary learning method.

[0030] The identification unit 22 calculates sparse coefficients (described in detail later) using the private sparse dictionary learned by the sparse dictionary learning unit 21 and an orthogonal matching pursuit (OMP) method, which is an example of a greedy algorithm.

[0031] In the state estimation system 100 according to this embodiment, the confidentiality signal generating device 1 and the calculation device 2 execute the confidentiality sparse dictionary learning step and the confidentiality identification step described below to calculate sparse coefficients and estimate the confidentiality state of optical communication.

[0032] In the private sparse dictionary learning step, learning is performed based on information on whether the learning constellation data is normal or erroneous, and if erroneous, what the error state is, and parameters such as the sparse dictionary are determined.

[0033] In the step of stealth sparse dictionary learning, in the stealth signal generation device 1, the constellation data is acquired from the DSP 10, and a stealth signal is generated. Then, the generated stealth signal is transmitted to the arithmetic device 2 via the network 3.

[0034] In the sparse dictionary learning unit 21 of the arithmetic device 2, learning of the stealth sparse dictionary is performed using the "LC K-SVD" algorithm of the sparse dictionary learning method.

[0035] In the step of stealth identification, in the stealth signal generation devices 1 (1-1 to 1-N) installed at each site, the identification constellation data is acquired from the DSP 10, and a stealth signal is generated. Next, using the stealth sparse dictionary estimated in the above-described step of stealth sparse dictionary learning, a process of identifying whether the identification constellation data is normal or in an error state is executed.

[0036] [2. Generation of Stealth Signal Using Random Projection] Next, the generation of the stealth signal using random projection will be described in detail. The generation of the stealth signal is common to the above-described steps of stealth sparse dictionary learning and stealth identification. FIG. 4 is a block diagram showing a detailed configuration of the stealth signal generation unit 12. As shown in FIG. 4, the stealth signal generation unit 12 includes a random sampling unit 121, a distribution calculation unit 122, and a random projection unit 123. Details will be described below.

[0037] [2-1. Random Sampling Unit 121] The random sampling unit 121 performs a process of reducing the number of sample data by random projection. In random projection, the original d-dimensional data "X d×N " is projected onto a k-dimensional (k << d) subspace using a random matrix "R k×d " whose column unit length is random. That is, the data "X RP k×N " after random projection is calculated by the following equation (1).

[0038]

number

[0039] If each element of the randomization matrix "R" is "rij", each component "rij" is set as shown in the following formula (2). Therefore, by performing random projection, it is possible to reduce the amount of data. Note that formula (2) is an example of the randomization matrix "R", and a randomization matrix "R" for general random projection can be used.

[0040] [ 2-3 .Random Projection Section 123] The random projection unit 123 reduces the number of pieces of data using random projection that can simultaneously achieve both dimensionality reduction and confidentiality processing, based on the number of pieces of constellation data or the histogram “I(s,t)” calculated by the distribution calculation unit 122.

[0041] [2-2. Distribution calculation section 122] The distribution calculation unit 122 calculates the number of constellation data belonging to each coordinate s, t (s=1, . . . , S, t=1, . . . , T) or a histogram “I(s, t)” from the constellation data acquired from the DSP 10.

[0042] [2-2. Random Projection Section 123] The random projection unit 123 reduces the number of pieces of data using random projection that can simultaneously achieve both dimensionality reduction and confidentiality processing, based on the number of pieces of constellation data or the histogram “I(s,t)” calculated by the distribution calculation unit 122.

[0043] Fig. 5 is an explanatory diagram showing a flow of generating a secret signal by random projection from the number of constellation data or the histogram "I(s,t)". As shown in Fig. 5, first, the histogram "I(s,t)" is sorted in lexicographic order by the column vector "yi" shown in the following equation (3) as shown in the following equation (4).

[0044]

number

[0045]

Number

[0046] In equation (3), "M" is a non - negative integer defined as "S×T", "i" indicates the sample index of the learning data or the discrimination data and "i = 1, ···, N". Also, "N" indicates the number of data.

[0047] Random projection is a linear transformation by a random matrix and can be used for dimensionality reduction of high - dimensional data. Random projection multiplies a matrix R shown in the following equation (5) with elements being random numbers by an M - dimensional vector "yi" to convert it into a low - dimensional vector "yi(^)" of dimension "M(^)" (where M(^)<M). The vector "yi(^)" can be expressed by the following equation (6).

[0048]

Number

[0049]

Number

[0050] If the elements of the random matrix R are random numbers with an average of "0" and a variance of "1 / M(^)", when any N pieces of learning data or discrimination data "yi(i = 1, ···, N)" are randomly projected into the dimension of the following equation (7), the distance between the data before and after the random projection is approximately preserved with a high probability as shown in the following equation (8).

[0051]

Number

[0052]

Number

[0053] In equation (8), "ε" (where 0<ε<1) is a coefficient. This theorem shows that when mapping from an "M"-dimensional space to a lower-dimensional "M(^)"-dimensional space, the Euclidean distance between two points is preserved with extremely high probability. Furthermore, it is known that this random projection can be obtained by any random value. The above process generates a secret signal by random projection.

[0054] [3. General sparse dictionary learning and classification] Next, we will explain sparse dictionary learning and classification when the vector "yi" is not concealed. Sparse dictionary learning is performed by supervised sparse dictionary learning "LC K-SVD". Since unsupervised sparse dictionary learning "K-SVD" is required for the processing of "LC K-SVD", we will first explain "K-SVD".

[0055] [3-1. Unsupervised sparse dictionary learning: K-SVD] The set “Y” of vectors “yi” is expressed as the following equation (9).

[0056]

number

[0057] Fig. 6 is a diagram showing a sparse model when constellation data is not concealed. The blank areas shown in Fig. 6 indicate that the data is "0". Here, as shown in Fig. 6, it is assumed that a set "Y" can be expressed by a linear combination of K bases. That is, it is assumed as shown in the following formula (10).

[0058]

number

[0059] However, "D" in formula (10) can be expressed as formula (11) below. Also, "X" in formula (10) can be expressed as formula (12) below.

[0060]

number

[0061]

number

[0062] In equations (11) and (12), “D” is a dictionary matrix whose elements are the basis “dk” (M-dimensional column vector), and “X” is a matrix whose elements are the sparse coefficients “xi” (K-dimensional column vector).

[0063] In general, the number of bases is larger than the dimension of the observed signal (i.e., "K>M"), and an overcomplete dictionary matrix is ​​used as the dictionary matrix. In the expression "Y=DX" (the above-mentioned formula (10)) using a basis with more dimensions than the observed signal, the uniqueness of "X" cannot be guaranteed.

[0064] For this reason, the basis used to express the observed signal "Y" is usually restricted to a portion of "D". In other words, a constraint is imposed that only a small number of "T0" coefficients take nonzero values, and most of the remaining coefficients take zero values. This state in which the number of nonzero elements is a small percentage of the total is called sparse. The optimization problem with the sparse constraint is formulated as the following equation (13), which minimizes the reconstruction error.

[0065]

number

[0066] However, the following equation (14) shown in equation (13) indicates the L0 norm (the number of non-zero elements in a vector).

[0067]

number

[0068] The following equation (15) indicates the Frobenius norm, which is defined by the following equation (16).

[0069]

number

[0070]

number

[0071] Generally, dictionary learning solves the optimization problem of the above-mentioned equation (13) by alternately repeating the following two steps, steps S1 and S2. In the following step S1, sparse coefficients are calculated, and in step S2, the dictionary is updated.

[0072] (Step S1; Calculate sparse coefficients) In step S1, the problem is to fix the dictionary D and obtain the sparse coefficients xi, which can be rewritten as the following equation (17).

[0073]

number

[0074] However, this problem is a combinatorial optimization problem in which the optimal solution cannot be obtained unless all combinations of bases are tested, and is known to be computationally difficult (NP-hard). Many algorithms have been proposed to solve this problem, including a greedy method and a method of solving the problem after relaxing the l0 constraint with the l1 constraint. As an example, this invention uses the orthogonal matching pursuit method (OMP), which is an approximate solution method based on the l0 constraint.

[0075] (Step S2: Dictionary update) In step S2, "X (a matrix whose elements are sparse coefficients xi)" obtained in step S1 is fixed, and the dictionary "D" is updated. "K-SVD" is considered a generalization of the "k-means" method. In the "k-means" method, clusters and samples correspond one-to-one, whereas in "K-SVD", samples are represented as linear combinations of cluster centroids (basis in K-SVD). In "K-SVD", the residual "Ek" between the linear prediction value excluding the basis "dk" from the set of observed signals "Y" is subjected to singular value decomposition (SVD), and "dk" and "x k T The residual "Ek" is expressed by the following equation (18).

[0076]

number

[0077] However, the solution obtained does not necessarily satisfy the sparse constraint, so in "K-SVD", the "x k T Update only the non-zero elements in E R k By applying “SVD” to “u”, “v” and decomposing it into orthogonal matrices “U” and “V” and diagonal matrix “Σ”, we obtain the following equation (19).

[0078]

number

[0079] In equation (19), “ui” and “vj” are the i-th column vectors of “U” and “V”, respectively, and “σi” is the i-th diagonal component of “Δ”. In “K-SVD”, the components “u1” and “σ1v T 1" to obtain an approximate solution for the basis and row vector of sparse coefficients as shown in the following equation (20).

[0080]

number

[0081] [3-2. Supervised Sparse Dictionary Learning: LC K-SVD] The supervised sparse dictionary learning is executed by the sparse dictionary learning unit 21 shown in Fig. 3. While "K-SVD" obtains a sparse representation so as to minimize the reconstruction error, "LC K-SVD" sets a cost function as a weighted sum of (a) the reconstruction error, (b) the classification sparse code error, and (c) the classification error for class classification, and learns the sparse representation by the following equation (21).

[0082]

number

[0083] The first term on the right-hand side of equation (21) is the same reconstruction error as in "K-SVD". "Q" in the second term on the right-hand side of equation (21), shown in equation (22) below, is a discriminant sparse code for classifying the observed signal vector "yi", and imposes a constraint that the observed signal vectors "yi" classified into the same class share the same basis "dk".

[0084]

number

[0085] The third term on the right side of equation (21) is the classification error for class classification. "W" is the projection matrix for classification, and "H" in equation (23) below is the class label of input "Y".

[0086]

number

[0087] "hi" in equation (23) can be expressed by the following equation (24): "hi" is a label vector of a class corresponding to the observed signal vector "yi", where "l" indicates the corresponding class and "m" indicates the number of classes.

[0088]

number

[0089] The "α" and "β" in the above formula (21) are parameters that adjust the contribution rate. Note that formula (21) can be rewritten as the following formula (25). This has the same format as the above formula (13), and the dictionary can be trained using the same algorithm as "K-SVD".

[0090]

number

[0091] [3-3.Identification] The classification process is executed by the classification unit 22 shown in Fig. 3. In the classification step, for a vector "yi" of an observation signal shaped from the classification constellation data, a dictionary "D" estimated by "LC K-SVD" is used to solve the following equation (26) using "OMP" or the like to calculate a sparse coefficient "xi".

[0092]

number

[0093] Next, the calculated sparse coefficients “xi” are projected using the matrix “W” according to the following equation (27).

[0094]

number

[0095] Based on the projected estimate hi(^), we identify whether the concealed signal vector yi(^) belongs to one of the m classes. The vector yi(^) is classified into the class corresponding to the element of hi(^) that is closest to l.

[0096] [4. Private sparse dictionary learning and private identification] The processes of private sparse dictionary learning and private identification are executed by the sparse dictionary learning unit 21 and the identification unit 22 shown in Fig. 3. In this process, private sparse learning and identification are performed using a vector "yi(^)" of a private signal generated by using random projection based on constellation data.

[0097] [4-1. Supervised Private Sparse Dictionary Learning: LC K-SVD] The supervised private sparse dictionary learning will be explained below. The set of vectors “yi(̂)” of private signals is expressed by the following equation (28).

[0098]

number

[0099] At this time, the secret dictionary “D(^)” and the projection matrix “W” shown in the following equation (29) are obtained using “LC K-SVD”.

[0100]

number

[0101] The cost function for learning a private sparse dictionary using “LC K-SVD” can be expressed as the following equation (30), and can be solved using the same algorithm as “K-SVD”.

[0102]

number

[0103] Here, "Q" in the following equation (31) is a discriminant sparse code for classifying the observed signal vector "yi", and "W" in the following equation (30) is a projection matrix for classification. "H" in the following equation (32) is the class label of the input "Y".

[0104]

number

[0105]

number

[0106] In the following equation (33), "hi" is a label vector of the class corresponding to the observed signal vector "yi," where "l" indicates the corresponding class and "m" indicates the number of classes.

[0107]

number

[0108] The “α” and “β” in the formula (30) are parameters that adjust the contribution rate. It is assumed that the dictionary “D” is concealed by random projection according to the relationship “D(^)=RD(^)”.

[0109] Here, as shown in the above-mentioned equation (8), since the distance between data before and after random projection is approximately preserved with a high probability, the optimal solution of the private sparse dictionary learning in the above-mentioned equation (30) is calculated as a value close to the optimal solution in the case of no anonymization in the equation (25).

[0110] At this time, the encryption dictionary "D(^)" can be updated and obtained in the domain of the encryption signal by the "LC K-SVD" algorithm. FIG. 6, mentioned above, is a diagram showing a sparse model when the constellation data is not encrypted. FIG. 7 is a diagram showing a sparse model when the dimensions are reduced and encrypted by random projection. As can be understood by comparing FIG. 6 and FIG. 7, it can be seen that the dimensions of the training / identification data and the dictionary are reduced and encrypted by reducing the dimensions "Y" and "D" by random projection.

[0111] [4-2. Secret Identification] The process of the secret identification is executed by the identification unit 22 shown in Fig. 3. In the step of the secret identification, for the secret signal "y(^)i" generated by random projection from the constellation data for identification, the secret dictionary "D(^)" estimated by the secret sparse dictionary learning is used to calculate the sparse coefficient "xi" by solving the following equation (34) using "OMP" or the like.

[0112]

number

[0113] Next, the calculated sparse coefficient “xi” is projected by the following equation (35) using the projection matrix “W” estimated by the private sparse dictionary learning.

[0114]

number

[0115] Based on the projected estimate "h(?)i", it is determined whether the concealed signal vector "yi(^)" belongs to one of the "m" classes. Then, the concealed signal vector "yi(^)" can be classified into the class corresponding to the element of "h(^)i" that is closest to "l".

[0116] [Effects of this embodiment] As described above, the state estimation system 100 according to this embodiment includes a concealment signal generation unit 12 that acquires the training constellation data and the identification constellation data output from the optical communication signal processing circuit, reduces the number of data from each constellation data by random projection and conceals it, and generates a training concealment signal and an identification concealment signal based on each constellation data after the reduction in the number of data and concealment, a sparse dictionary learning unit 21 that learns a concealment sparse dictionary using a sparse dictionary learning algorithm based on the training concealment signal, and a discrimination unit 22 that estimates the state of the optical communication using the concealment sparse dictionary based on the identification concealment signal.

[0117] In this embodiment, the constellation data can be reduced while being kept confidential. Therefore, it is possible to estimate the state of a transmission path or an optical transmitter in optical communication with a small amount of calculation while maintaining high confidentiality. As a result, it is possible to estimate the cause of quality degradation in optical communication without increasing the amount of calculation.

[0118] Moreover, even if data is leaked due to human error, it is possible to prevent data leakage.

[0119] The encipherment signal generating unit 12 also reduces the number of constellation data by a random sampling unit. It then performs distribution calculation on the constellation data after the data reduction, and performs random projection that simultaneously achieves dimensionality reduction and encipherment processing, thereby generating a training encipherment signal and an identification encipherment signal. This makes it possible to generate the training encipherment signal and the identification encipherment signal with high accuracy.

[0120] Furthermore, the sparse dictionary learning unit 21 updates the sparse coefficients “xi”, the private dictionary “D(^)”, and the projection matrix “W” by learning the private sparse dictionary. Therefore, it is possible to always adopt new data to perform sparse dictionary learning.

[0121] In addition, the identification unit 22 updates the sparse coefficients “xi”, the secret dictionary “D(^)”, and the projection matrix “W” estimated by the sparse dictionary learning unit 21, thereby making it possible to estimate the state of optical communication with high accuracy.

[0122] Furthermore, as shown in FIG. 3, by providing a plurality of (N in the figure) encipherment signal generating devices 1, a plurality of encipherment signal generating units 12 are provided, so that it is possible to perform state estimation on the constellation data obtained from the DSP 11 provided at each point.

[0123] The above-described encipherment signal generation device 1 of the present embodiment may be, for example, a general-purpose computer system including a CPU (Central Processing Unit, processor) 901, a memory 902, a storage 903 (HDD: Hard Disk Drive, SSD: Solid State Drive), a communication device 904, an input device 905, and an output device 906, as shown in Fig. 8. The memory 902 and the storage 903 are storage devices. In this computer system, the CPU 901 executes a predetermined program loaded on the memory 902, thereby realizing each function of the encipherment signal generation device 1.

[0124] The encipherment signal generation device 1 may be implemented in one computer or in multiple computers, and may also be a virtual machine implemented in a computer.

[0125] The program for the confidential signal generating device 1 can be stored in a computer-readable recording medium such as a HDD, SSD, USB (Universal Serial Bus) memory, CD (Compact Disc), or DVD (Digital Versatile Disc), or can be distributed via a network.

[0126] The present invention is not limited to the above-described embodiment, and various modifications are possible within the scope of the present invention. [Explanation of symbols]

[0127] 1(1-1~1-N) Secret signal generator 2 Arithmetic unit 3. Network 10 Digital coherent signal processing circuit (DSP) 11 Data Acquisition Section 12. Secret signal generator 21 Sparse Dictionary Learning Unit 22 Identification section 100 State Estimation System 121 Random Sampling Section 122 Distribution calculation section 123 Random Projection Section

Claims

1. a concealment signal generation unit that acquires the training constellation data and the identification constellation data output from the optical communication signal processing circuit, performs distribution calculation, calculates the number or histogram of constellation data belonging to predetermined coordinates among the training constellation data and the identification constellation data, and performs random projection that simultaneously achieves dimensionality reduction and concealment processing based on the calculated number or histogram of constellation data, thereby generating a training concealment signal and an identification concealment signal; a sparse dictionary learning unit that learns a secret sparse dictionary using a sparse dictionary learning algorithm based on the learning secret signal; an identification unit that estimates a state of optical communication based on the identification secret signal by using the secret sparse dictionary; A state estimation system comprising:

2. The state estimation system according to claim 1 , wherein the sparse dictionary learning unit updates a sparse coefficient, a private dictionary, and a projection matrix by learning the private sparse dictionary.

3. The state estimation system according to claim 2 , wherein the identification unit estimates the state of the optical communication based on the sparse coefficients estimated by the sparse dictionary learning unit and the secret dictionary.

4. 4. The state estimation system according to claim 1, wherein a plurality of the concealment signal generating units are provided.

5. a data acquisition unit that acquires learning constellation data and identification constellation data output from a signal processing circuit for optical communication; a concealment signal generation unit that performs a distribution calculation, calculates the number or a histogram of constellation data belonging to a predetermined coordinate among the learning constellation data and the identification constellation data, performs random projection based on the calculated number or the histogram of constellation data to simultaneously achieve dimensionality reduction and concealment processing, reduces the number of data from each of the constellation data and conceals it, and generates a learning concealment signal and an identification concealment signal based on each of the constellation data after the data reduction and concealment; A confidential signal generating device.

6. A concealment signal generating unit acquires training constellation data output from an optical communication signal processing circuit, performs a distribution calculation, and calculates the number or histogram of constellation data belonging to specified coordinates, performs random projection based on the calculated number or histogram of constellation data to simultaneously achieve dimensionality reduction and concealment processing, thereby reducing the number of data and concealing the training constellation data, and generates a training concealment signal based on the training constellation data after the number of data reduction and concealment; a sparse dictionary learning unit learning a secret sparse dictionary using a sparse dictionary learning algorithm based on the learning secret signal; a concealment signal generating unit acquiring the identification constellation data output from the optical communication signal processing circuit, performing the distribution calculation and the random projection to reduce the number of data from the identification constellation data and conceal it, and generating an identification concealment signal based on the identification constellation data after the data reduction and concealment; An identification unit estimates a state of optical communication based on the identification secret signal by using the secret sparse dictionary; A state estimation method comprising:

7. A confidential signal generating program that causes a computer to function as the confidential signal generating device described in claim 5.

Images