Patent Grant
12176952
This application is a 371 U.S. National Phase of International Application No. PCT/JP2020/031435, filed on Aug. 20, 2020. The entire disclosure of the above application is incorporated herein by reference.
The present invention relates to a signal detection device, a signal detection method and a program.
With the start of the service of the fifth generation mobile communication system (5G), the distribution service of a high definition moving image, and the development of the IoT (Internet of Things) service, the demand of the communication traffic of an optical signal is increasing. Up to now, without changing the structure of an optical fiber provided in a transmission line, an increased demand has been dealt with by improving the function of an optical communication device installed at a terminal station of an optical network, introducing an optical amplifier and an optical switch, and the like.
At present, optical fibers provided in transmission lines are fundamental to a large-capacity optical network. A single mode fiber is widely used as an optical fiber of the transmission line, except for a local network for a short distance (for example, LAN (Local Area Network)). In the single mode fiber, a single core serving as a path for an optical signal is provided in a clad of the optical fiber.
In a wavelength band used in a large capacity optical network (for example, in a C band and an L band), the single mode fiber allows only transmission of a single mode optical signal. Thus, an optical network for stably transmitting a large amount of information reaching several terabits per second over a long distance is realized.
A transmission technology (a digital coherent transmission technology) in which a digital signal processing technology and a coherent transmission/reception technology are combined is introduced into an optical communication system of 100 gigabits per second.
Information independently placed on the amplitude and phase of the optical signal transmitted using the digital coherent transmission technology is extracted from the optical signal using the coherent reception technology. Thus, not only the reception sensitivity by coherent reception is improved but also the distortion generated in the waveform of the transmitted optical signal can be corrected with high accuracy.
As a first simple example, polarization multiplexing optical transmission using two modes of polarized wave orthogonal to each other in a single mode fiber will be described.
In the polarization multiplexing optical transmission, different information can be placed on polarized waves orthogonal to each other. In the transmission line, these polarized waves are mixed complicatedly. Since the orthogonal axes of these polarized waves fluctuate at high speed, it is difficult to follow the fluctuation by using an optical device. Therefore, a reception device corresponding to the polarization diversity structure receives the multiplexing optical signal of the mixed polarized wave. The reception device converts the multiplexing optical signal of the polarized wave into a digital signal. The communication device separates the mixed polarized waves by using digital signal processing.
Such a separation processing system can be modeled as a “2×2 MIMO (Multiple-Input Multiple-Output)” system used in a wireless communication system. Information placed on polarized waves orthogonal to each other is extracted from the separated signal for each polarized wave. Thus, communication between the transmission device and the reception device is enabled.
As a second simple example, mode multiplexing optical transmission using a plurality of modes in a multimode fiber will be described.
In the mode multiplexing optical transmission, the core diameter of the multimode fiber is longer than that of the single mode fiber. Thus, a plurality of modes can be excited in a wavelength band such as a C band. Different information can be placed on the optical signals of the respective modes.
In the mode multiplexing optical transmission, the mode-multiplexed optical signal (a mode signal) transmitted through the multimode fiber are mixed complicatedly in the transmission line, similarly to the polarization multiplexing optical transmission. The reception device corresponding to the mode diversity structure receives the mixed mode multiplexing optical signal. The reception device converts the mode multiplexing optical signal into a digital signal. The reception device separates each digital signal based on the mixed mode multiplexing optical signal by using MIMO digital signal processing of a scale corresponding to the number of excited modes.
As a more specific example, a few-mode fiber, which excites two LP (Linear Polarized) (2LP) modes, will be described.
In the few-mode fiber for the 2LP mode, an “LP01 mode” to be a base mode and an “LP11 mode” to be a high-order mode are excited. Further, by utilizing “LP11a” and “LP11b” which are two degraded modes of “LP11 mode” and “X polarized wave” and “Y polarized wave” which are polarization modes of the respective modes, it is possible to place different information on each of signals of a total six special mode of “LP01X”, “LP01Y”, “LP11aX”, “LP11aY”, “LP11bX”, and “LP11bY” in the few-mode fiber for the 2LP mode.
Therefore, when the nonlinear optical effect of the optical fiber is ignored, the transmission capacity of three times as large as that of the single mode fiber can be attained by the few-mode fiber for the 2 LP mode in a principle sense.
However, in mode multiplex transmission in which a multimode fiber or the few-mode fiber is used as a transmission line for an optical signal and independent information is placed on an optical signal in each spatial mode, there is a problem that a deviation in transmission characteristics for each spatial channel is large. Here, due to power loss of the optical signal transmitted through the optical fiber, power loss of the optical signal at a connection unit of the optical fiber, and deterioration of a high-order spatial mode compared to a base spatial mode, there is a case where a deviation of transmission characteristics for each spatial channel becomes large.
Such a phenomenon can be comprehensively evaluated as an influence of mode dependent loss. When the transmission characteristic of a specific spatial channel is deteriorated, it is difficult for the transmission capacity and the transmission distance to enjoy the benefit of mode multiplexing transmission.
The influence of mode dependent loss on transmission capacity and transmission distance in a linear MIMO channel will be described. In the following description, the number of modes excited in the multi-mode fiber is N (N is an integer of 1 or more). The N modes are used as a signal carrying mode. In the following description, the symbol “T” described at the shoulder of the matrix (vector) represents the transposition. A reception signal vector “y=[y1 y2 yN]T” corresponding to the transmission signal vector “x” is represented by formula (1).
[Math. 1]
y=Hx+n=Σi=1nhixi+n (1)
Here, “x=[x1 x2 . . . xN]T” represents a transmission signal vector. “H” represents a full-rank transmission channel matrix of “N×N” size. “n=[n1 n2 . . . nN]T-” represents a noise vector.
In the formula (1), a transmission line channel matrix “H” is developed as shown by “H=[h1 h2 . . . hN]”. Here, “hi(i is an integer of 1 or more and N or less)” represents a column vector. According to the formula (1), the reception signal vector “Y” is a vector obtained by adding a discrete point (lattice) in a linear space spanned by the column vector “hi” and the noise vector “n”.
In the following description, symbols added above characters in a mathematical formula or a function (referred to as “mathematical expression or the like” below) are written before the characters. For example, the symbol “{circumflex over ( )}” added above a character in a formula and the like is written before the character “x” below as “{circumflex over ( )}x”. For example, the symbol “˜” added above a character in a formula and the like is written before the character “x” below as “˜x”.
Next, a linear detection filter will be described.
The linear detection filter detects an estimation vector “˜x” of a transmission signal vector by multiplying a reception signal vector “y” by a linear weight matrix “WT”. As a design criterion of such the linear weight matrix “WT”, for example, maximum ratio combining, zero forcing, or minimum means squared error (MMSE) are used.
The linear detection filter designed by using maximum ratio combining exhibits good detection characteristics with respect to an estimation vector in a region of a low signal-to-noise ratio (SNR). The linear detection filter designed by using zero-forcing exhibits good detection characteristics for estimation vectors in a region of high SNR. The linear detection filter designed by using a minimum means squared error exhibits good detection characteristics compared to the linear detection filter designed by using maximum ratio combining or zero forcing.
In the following description, the linear weight matrix is expressed as “WT=[w1 w2 . . . wN]T”. Here, “wi” represents the i-th column of the linear weight matrix “WT”.
The linear weight matrix “wiT” is multiplied from the left to both sides of the formula (1) so that the i-th component “xi” of the transmission signal vector “x” is detected.
[Math. 2]
{tilde over (x)}i=wiTy=wiThixi+Σj=1,j≠inwiThjxj+wiTn (2)
Here, a second term of formula (2) represents a residual interference component. A third term of the formula (2) represents a noise component. When the orthogonality of the transmission channel matrix “H” is lost and the mode dependent loss is increased, the residual interference components and noise enhancement or the like occur. Thus, the detection accuracy of the i-th component “xi” of the transmission signal vector “x” is deteriorated.
Next, a linear detection (LD) based on Lattice reduction (LR) will be described.
The linear detection “LR-LD” based on the lattice reduction is a detection method for improving the detection accuracy of the signal. Here, the lattice reduction is to find out a basis vector (a reduced basis) which exhibits good characteristics for the basis vector which spans the given lattice from among other basis vectors which span the same lattice as that of the given lattice.
“Exhibit good characteristics” means that when the transmission channel matrix “H” has orthogonality, the lattice reduction can be applied to a signal detection method. For example, by solving a short vector problem, it is possible to find out a basis vector exhibiting good characteristics. As a method for efficiently solving the approximate solution of the short vector problem, “LLL (Lenstra-Lenstra-Lovasz) basis reduction” is known (refer to NPD 1).
However, the calculation amount required for executing “LLL base reduction” is large. If “LLL basis reduction” is executed on the transmission channel matrix “H”, the required calculation amount is estimated by “0 (N4 log B)” (refer to NPD 2). Here, the maximum value “B” represents the maximum value of the L2 norm of the column vector “hi”.
Further, a lattice reduction based on basic deformation in which the calculation amount is reduced is disclosed in NPD 3. Since application to a transmission channel having static characteristics is assumed in NPD 3, the lattice reduction is executed only once, and the basis vector (a system for increasing orthogonality) exhibiting good characteristics is not updated.
When the linear detection “LR-LD” based on the lattice reduction is applied to an optical communication system, it is necessary to use a basis reduction capable of following a transmission channel having dynamic characteristics (time variability). However, an error in information placed on an optical signal transmitted through a transmission line having time variability may not be reduced by a calculation amount (a low calculation amount) equal to or less than a predetermined threshold value.
In view of the above-mentioned circumstances, an object of the present invention is to provide a signal detection device, a signal detection method, and a program that can reduce errors in information placed on an optical signal transmitted through a transmission line having the time variability with a calculation amount equal to or less than a predetermined threshold.
An embodiment of the present invention is a signal detection device includes a signal detection unit that multiplies the reception signal vector by a separation matrix to derive an estimation vector of a transmission signal vector, a first conversion unit that converts the estimation vector of the transmission signal vector to the estimation vector of the transmission signal vector based on the reduced basis by using an inverse matrix of a unimodular matrix, a first determination unit that converts the estimation vector of the transmission signal vector based on the reduced basis to a determination value vector of the transmission signal vector by using the inverse matrix of the unimodular matrix and the unimodular matrix, a first update unit that updates the separation matrix by using a first error signal vector between the estimation vector of the transmission signal vector and the determination value vector of the transmission signal vector, a second conversion unit that converts the first error signal vector to a second error signal vector based on the reduced basis by using the inverse matrix of the unimodular matrix, a second update unit that updates an error covariance matrix based on the reduced basis by using the second error signal vector based on the reduced basis, a second determination unit that determines whether or not a predetermined condition is satisfied, and a third update unit that updates the unimodular matrix, the inverse matrix of the unimodular matrix, and the error covariance matrix based on the reduced basis when it is determined that the predetermined condition is satisfied.
The aspect of the present invention is the signal detection device includes the signal detection unit that multiplies the reception signal vector by the separation matrix based on a reduced basis to derive the estimation vector of the transmission signal vector based on the reduced basis, the first determination unit that derives the determination value vector of the transmission signal vector based on the reduced basis by using the inverse matrix of the unimodular matrix and derives the determination value vector of the transmission signal vector by using the determination value vector of the estimation vector of the transmission signal vector based on the reduced basis, and the unimodular matrix, an error derivation unit that derives an error signal vector between the determination value vector of the transmission signal vector based on the reduced basis and the estimation vector of the transmission signal vector based on the reduced basis as an error signal vector based on the reduced basis, the first update unit that updates the separation matrix based on the reduced basis using the error signal vector based on the reduced basis, the second update unit that updates the error covariance matrix based on the reduced basis by using the error signal vector based on the reduced basis, the second determination unit that determines whether or not the predetermined condition is satisfied, and the third update unit that updates the unimodular matrix, the inverse matrix of the unimodular matrix, the error covariance matrix based on the reduced basis and the separation matrix based on the reduced basis when it is determined that the predetermined condition is satisfied.
The aspect of the present invention is a signal detection method executed by the signal detection device, the signal detection method includes a signal detection step for multiplying the reception signal vector by the separation matrix to derive the estimation vector of the transmission signal vector, a first conversion step for converting the estimation vector of the transmission signal vector to the estimation vector of the transmission signal vector based on the reduced basis by using the inverse matrix of the unimodular matrix, a first determination step for converting the estimation vector of the transmission signal vector based on the reduced basis to the determination value vector of the transmission signal vector by using the inverse matrix of the unimodular matrix and the unimodular matrix, a first update step for updating the separation matrix by using the first error signal vector between the estimation vector of the transmission signal vector and the determination value vector of the transmission signal vector, a second conversion step for converting the first error signal vector to the second error signal vector based on the reduced basis by using the inverse matrix of the unimodular matrix, a second update step for updating the error covariance matrix based on the reduced basis by using the second error signal vector based on the reduced basis, a second determination step for determining whether or not the predetermined condition is satisfied, and a third update step for updating the unimodular matrix, the inverse matrix of the unimodular matrix, and the error covariance matrix based on the reduced basis when it is determined that the predetermined condition is satisfied.
The aspect of the present invention is the signal detection method executed by the signal detection device, the signal detection method includes the signal detection step for multiplying the reception signal vector by the separation matrix based on a reduced basis to derive the estimation vector of the transmission signal vector based on the reduced basis, the first determination step for deriving the determination value vector of the transmission signal vector based on the reduced basis by using the inverse matrix of the unimodular matrix and deriving the determination value vector of the transmission signal vector by using the determination value vector of the estimation vector of the transmission signal vector based on the reduced basis, and the unimodular matrix, an error derivation step for deriving an error signal vector between the determination value vector of the transmission signal vector based on the reduced basis and the estimation vector of the transmission signal vector based on the reduced basis as an error signal vector based on the reduced basis, the first update step for updating the separation matrix based on the reduced basis using the error signal vector based on the reduced basis, the second update step for updating the error covariance matrix based on the reduced basis by using the error signal vector based on the reduced basis, the second determination step for determining whether or not the predetermined condition is satisfied, and the third update step for updating the unimodular matrix, the inverse matrix of the unimodular matrix, the error covariance matrix based on the reduced basis and the separation matrix based on the reduced basis when it is determined that the predetermined condition is satisfied.
The aspect of the present invention is a program for causing a computer to function as the signal detection device described above.
According to the present invention, it is possible to reduce the error in information placed on the optical signal transmitted through the transmission line having the time variability with the calculation amount equal to or less than the predetermined threshold.
Embodiments of the present invention will be described in detail with reference to the drawings.
The optical communication system 1a is a system for performing communication using optical signals. Hereinafter, independent information sequences are referred to as “data sequences”. The communication system 1a transmits N data sequences (transmission signal vectors) from the transmission device 2 toward the reception device 4a.
The transmitter 20-1 to 20-N encodes the N data sequences into N electric signals. The transmitter 20-1 to 20-N converts the electric signals of N data sequences into optical signals of N data sequences. The multimode fiber 3 transmits the optical signals of the N data series to the reception device 4a as optical signals of each spatial mode.
The receiver 40-1 to 40-N receives the optical signals of the transmitted N data sequences. The receiver 40-1 to 40-N converts the optical signals of the N data sequences into electric signals (reception signal vectors) of the N data sequences. The signal detection device 41a removes distortion or the like generated in the waveform from the electric signals of the N data sequences by a digital signal processing. The signal detection device 41a corrects an error generated in each data sequences through transmission by the digital signal processing. Thus, the separation of N original data sequences (transmission signal vectors) and the extraction of N data sequences (reception signal vectors) are made possible.
Next, a signal separation processing in the signal detection device 41a will be described.
Hereinafter, an inverse matrix derived without performing an inverse matrix operation for the matrix “A” is referred to as “Ainv”. As a result, the inverse matrix “A−1” derived by executing the operation of the inverse matrix and the inverse matrix “Ainv” are distinguished from each other.
Hereinafter, a system which is transformed by using a unimodular matrix “T(k)” or the unimodular inverse matrix “Tinv(k)” (basis) and which enhances orthogonality of the transmission channel matrix “H” (reduced basis) is referred to as “LR domain”. The unimodular matrix is a matrix in which each component is an integer, and an absolute value of a matrix expression is 1.
In the following description, characters representing vectors or matrices based on the LR domain are described by attaching a tilde “˜” to the front of the characters.
The signal detection unit 411a executes the sampling processing to the electric signal (the reception signal) inputted from the receiver 40-1 to 40-N. Thus, the signal detection unit 411a detects the inputted electric signals (the reception signal vectors). At the sampling time point “k”, the estimation vector “x(k)” of the transmission signal vector “x” is represented by formula (3).
[Math. 3]
{circumflex over (x)}(k)=WT(k)y(k) (3)
Here, “W(k)” represents the separation matrix (the signal separation matrix).
The basis conversion unit 412a converts the estimation vector “{circumflex over ( )}x(k)” of the transmission signal vector to the estimation vector “{circumflex over ( )}˜x(k)” of the transmission signal vector based on the LR domain, as shown in formula (4), using the unimodular inverse matrix “Tinv(k)” of the transmission signal vector based on the LR domain.
[Math. 4]
{circumflex over ({tilde over (x)})}(k)=Tinv(k){circumflex over (x)}(k) (4)
The first determination unit 413a determines a signal in the LR domain (the system that increases orthogonality of the transmission channel matrix “H”) by using the estimation vector “{circumflex over ( )}˜x(k)” of the transmission signal vector based on the LR domain. That is, the first determination unit 413a derives the sampling value (the determination value vector) of the transmission signal vector (each data sequence from the first date to the N-th data) with respect to the estimation vector of the transmission signal vector based on the LR domain. The first determination unit 413a outputs the determination value vector (the output signal) to a predetermined device (not shown).
For example, the first determination unit 413a derives the determination value vector “xHD” of the transmission signal vector, as shown by formula (5), by using the unimodular matrix “T(k)”.
[Math. 5]
xHD(k)=T(k)Q({circumflex over ({tilde over (x)})}(k),Tinv(k)) (5)
Here, “Q(⋅)” represents a signal determination function. When the signal format is a QAM signal, for example, the signal determination function “Q(⋅)” is represented by formula (6) by using a given vector “a” and the unimodular inverse matrix “Tinv(k)”.
Here, for the in-phase component and the orthogonal component, the Euclidean distance between the respective symbol points is 2. Further, “1N×1” represents a column vector of “N×1” size. All components of “1N×1” are 1. The symbol surrounding the fraction in the first term of the formula (6) represents the rounding processing. The error derivation unit 414a derives the error signal vector by formula (7).
[Math. 7]
∈(k)=xHD(k)−{circumflex over (x)}(k) (7)
The separation matrix update unit 415a updates the separation matrix “WT(k)”. For example, the separation matrix update unit 415a derives the separation matrix “WT(k+1)” by formula (8), so as to reduce an error by sequentially using the LMS (Least mean square) method.
[Math. 8]
WT(k+1)=WT(k)+μ∈(k)yH(k) (8)
Here, “μ” is a parameter representing a step size.
The error basis conversion unit 416a converts the error signal vector shown in the formula (7) to the error signal vector based on the LR domain, as shown in the formula (9). That is, the error basis conversion unit 416a multiplies the error signal vector shown in the formula (7) by the unimodular inverse matrix “Tinv(k)” from the left as shown in formula (9).
[Math. 9]
{tilde over (∈)}(k)=Tinv(k)∈(k) 9)
The first covariance matrix update unit 417a updates the error covariance matrix “˜Re(k)” based on the LR domain, as shown in formula (10).
[Math. 10]
{tilde over (R)}e(k+1)=δ{tilde over (R)}e(k)+{tilde over (∈)}(k){tilde over (∈)}H(k) (10)
Here, “5” is a forgetting coefficient. In the following, “5” is an arbitrary real number of 0 or more and 1 or less.
The second determination unit 418 determines whether or not the sampling time point “k” is divisible by a predetermined number of times (a natural number) “K”. When itis determined that the sampling time point “k” is divisible by the predetermined number of times “K”, the unimodular matrix “T(k)”, the unimodular inverse matrix “Tinv(k)”, and the error covariance matrix “˜Re(k)” based on the LR domain are updated.
Next, the update of the unimodular matrix “T(k)”, the update of the unimodular inverse matrix “Tinv(k)”, and the update of the error covariance matrix “˜Re(k)” based on the LR domain will be described.
When it is determined that the sampling time point “k” is divisible by the predetermined number of times “K”, these matrices are updated by execution of replacement processing. When the predetermined number of times “K” is 1, these matrices are updated every sampling time point “k”.
The unimodular matrix update unit 419a derives an integer “λi, j(k)” as shown in formula (11).
Here, “˜ri, j(k)” represents the (i, j) component of the error covariance matrix “˜Re(k)” based on the LR domain. Therefore, in the pair of (i, j) in which “i=j” is satisfied, the value of “λi, j(k)” is “−1”.
When the value of the integer “λi, j(k)” is “other than 0” in any pair of (i, j) in which “i≠j” is satisfied, the unimodular matrix “T(k)”, the unimodular inverse matrix “Tinv(k)”, and the error covariance matrix “˜Re(k)” (a related statistical amount) based on the LR domain are updated to be capable of detecting the signal on the LR domain.
The condition in which the value of the integer “λi, j(k)” is “other than 0” in any pair of (i, j) that satisfies “i≠j” can be read to the condition a sum of an absolute values of the integer “λi, j(k)” for all of (i, j) pair is more than a number of modes “N”.
The unimodular matrix update unit 419a selects a pair of (i, j) which minimizes “Δi, j(k)” as the specified pair among pairs of (i, j) that satisfies “i≠j”, as shown in formula (12).
[Math. 12]
(i,j)=arg min Δi,j(k) (12)
Here, “Δi, j” (k)” is represented as shown in formula (13).
[Math. 13]
Δi,j(k)=λi,j(k){tilde over (r)}i,j(k)+Δi,j*(k){tilde over (r)}i,j(k)+|λi,j(k)|2{tilde over (r)}j,j(k) (13)
The process of selecting the pair of (i, j) which minimizes “Δi, j(k)” corresponds to the processing to select a pair of (i, j) which minimizes a root mean square error between on output signal and a desired signal.
Furthermore, when it is determined that the sum of the absolute values of the integer “λi, j(k)” for all of (i, j) pair is more than the number of modes “N”, each processing shown in formula (14) to (18) is performed to reduce the calculation amount once, for example.
A matrix “Λi, j(k)” having the “λi, j(k)” for (I, j) component, 1 for diagonal components, 0 for other than components is represented as shown in formula (14).
[Math. 14]
Λi,j(k)=I+λi,j(k)eiejH (14)
Here, “ep” represents a column vector of “N×1” size. The p-th component in “ep” is 1. The other components in “ep” are 0. “p” is a natural number equal to or larger than N.
The unimodular matrix update unit 419a derives an inverse matrix “Λi, jinv(k)” without executing the operation of the inverse matrix, as shown in formula (15).
[Math. 15]
Λi,jinv(k)=I−λi,j(k)eiejH (15)
Here, “Λi, jinv(k)” represents the inverse matrix “Λi, j −1(k)” of the matrix “Λi, j(k)”.
The unimodular matrix update unit 419a updates the unimodular matrix “T(k)” matrix based the integer “Λi, j(k)” shown in the formula (14) and the integer “Λi, jinv(k)” shown in the formula (15), as shown in formula (16).
[Math. 16]
T(k)←T(k)Λi,jinv(k) (16)
In the following description, the symbol “←” described in the formula represents the replacement processing. The unimodular inverse matrix update unit 420a updates the unimodular inverse matrix “Tinv(k)” as shown in formula (17).
[Math. 17]
Tinv(k)←Λi,j(k)Tinv(k) (17)
The second covariance matrix update unit 421a updates the error covariance matrix “˜Re(k)” based on the LR domain as shown in formula (18).
[Math. 18]
{tilde over (R)}e(k)←Λi,j(k){tilde over (R)}e(k)Λi,jH(k) (18)
The time point update unit 422 updates the sampling time point to “k+1”. At the sampling time point “k+1”, the signal detection device 41a executes the same processing as that executed at the sampling time point “k”.
Next, operation example of the signal detection device 41a will be described.
As shown in a third line and a fourth line, the first determination unit 413a derives the determination value vector of a transmission signal vector “xHD” shown in the formula (5) by using the unimodular matrix “T(k)”. As shown in a fifth line, the error derivation unit 414a derives the error signal vector as shown in the formula (7). As shown in a sixth line, the separation matrix update unit 415a derives the separation matrix “WT(k+1)” as shown in formula (8).
As shown in a seventh line, the error basis conversion unit 416a multiplies the error signal vector shown in the formula (7) by the unimodular inverse matrix “Tinv(k)” from the left as shown in the formula (9). As shown in a eighth line, the first covariance matrix update unit 417a updates the error covariance matrix “˜Re(k)” based on the LR domain, as shown in the formula (10).
As shown in a ninth line, the second determination unit 418 determines whether or not the sampling time point “k” is divisible by the predetermined number of times (a natural number) “K”. When it is determined that the sampling time point “k” is divisible by the predetermined number of times “K”, the unimodular matrix update unit 419a as shown in a tenth line, derives an integer “λi, j(k)” as shown in formula (11).
As shown in an eleventh line, the unimodular matrix update unit 419a determines whether or not the sum of absolute values of the integer “λi, j(k)” for all of the pair of (i, j) is more than the number of modes “N”. When it is determined that the sum of the absolute values is more than the number of modes “N”, as shown in a twelfth line, the unimodular matrix update unit 419a selects a pair of (i, j) which minimizes “Δi, j(k)” as the specific pair among pairs of (i, j) in which “i≠j” is satisfied in the formula (12).
As shown in a thirteenth line, the unimodular matrix update unit 419a updates the unimodular matrix “T(k)” based on the integer “Λi, j(k)” and the integer “Λi, jinv(k)” as shown in the formula (16). As shown in a fourteenth line, the unimodular inverse matrix update unit 420a updates the unimodular inverse matrix “Tinv(k)” as shown in the formula (17). As shown in a fifteenth line, the second covariance matrix update unit 421a updates the error covariance matrix “˜Re(k)” based on the LR domain as shown in formula (18).
The basis conversion unit 412a converts the estimation vector “{circumflex over ( )}x(k)” to the estimation vector of the transmission signal vector based on the LR domain “{circumflex over ( )}˜x(k)” as shown in the formula (4) by using the unimodular inverse matrix “Tinv(k)” (a step S102).
The first determination unit 413a determines the signal in the LR domain as shown in the formula (5) by using the estimation vector “{circumflex over ( )}˜x(k)” of the transmission signal vector based on the LR domain (a step S103). The error derivation unit 414a derives the error signal vector as shown in the formula (7) (a step S104). The separation matrix update unit 415a updates the separation matrix “WT(k)” as shown in the formula (8) (a step S105).
The error basis conversion unit 416a converts the error signal vector to the error signal vector based on the LR domain as shown in the formula (9) (a step S106). The first covariance matrix update unit 417a updates the error covariance matrix “˜Re(k)” based on the LR domain, as shown in the formula (10) (a step S107). The second determination unit 418 determines whether or not the sampling time point “k” is divisible by the predetermined number of times “K” (a step S108).
When it is determined that the sampling time point “k” is divisible by the predetermined number of times “K” (the step S108: YES), the unimodular matrix update unit 419a updates the unimodular matrix “T(k)” based on the integer “Λi, j(k)” and the integer “Λi, jinv(k)” as shown in the formula (16). The unimodular inverse matrix update unit 420a updates the unimodular inverse matrix “Tinv(k)” as shown in the formula (17). The second covariance matrix update unit 421a updates the error covariance matrix “˜Re(k)” based on the LR domain as shown in the formula (18). The time point update unit 422 updates the sampling time point to “k+1”. With respect to the sampling time point “k”, the signal detection device 41a ends the operation of the flow chart shown in
When it is determined that the sampling time point “k” is not divisible by the predetermined number of times “K” (the step S108: No), the time point update unit 422 updates the sampling time point to “k+1”. With respect to the sampling time point “k”, the signal detection device 41a ends the operation of the flow chart shown in
As described above, the signal detection unit 411a (signal detector) multiplies the reception signal vector “y” by the separation matrix “W” to derive the estimation vector “{circumflex over ( )}x” of the transmission signal vector “x” corresponding to the reception signal vector. The basis conversion unit 412a (the first conversion unit) (first converter) converts the estimation vector “{circumflex over ( )}x” of the transmission signal vector to the estimation vector “{circumflex over ( )}˜x” of the transmission signal vector based on an LR domain (reduced basis) by using the inverse matrix of the unimodular matrix “Tinv”. The first determination unit 413a converts the estimation vector “{circumflex over ( )}˜x” of a transmission signal vector based on the LR domain to the determination value vector “xHD” of the transmission signal vector by using the inverse matrix of the unimodular matrix “Tinv” and the unimodular matrix “T”.
The separation matrix update unit 415a (the first update unit) (first updater) updates the separation matrix “W” by using the first error signal vector between the estimation vector “{circumflex over ( )}x” of the transmission signal vector and the determination value vector “xHD” of the transmission signal vector. The error basis conversion unit 416a (the second conversion unit) (second converter) converts the first error signal vector to the second error signal vector based on the LR domain, as shown in the formula (9), by using the inverse matrix of the unimodular matrix “Tinv”. The first covariance matrix update unit 417a (the second update unit) (second updater) update the error covariance matrix “˜Re” based on the LR domain by using the second error signal vector based on the LR domain.
The second determination unit 418 (second determiner) determines whether or not the predetermined condition is satisfied. This predetermined condition is, for example, the condition that the sampling time point “k” is divisible by the predetermined number of times “K” and the sum of the absolute values of the integer “λi, j(k)” for all of the pair of (l, j) is more than the number of modes “N”.
The unimodular matrix update unit 419a (the third update unit) updates the unimodular matrix “T” when it is determined that the predetermined condition is satisfied. The unimodular inverse matrix update unit 420a (the third update unit) (third updater) updates the inverse matrix of the unimodular matrix “Tinv” when it is determined that the predetermined condition is satisfied. The second covariance matrix update unit 421a (the third update unit) updates the error covariance matrix “˜Re” based on the LR domain when it is determined that the predetermined condition is satisfied.
Thus, it is possible to reduce the error of the information placed on the optical signal transmitted through the transmission line having the time variability with the calculation amount equal to or less than the predetermined threshold.
The configuration of the communication system according to the second embodiment is the same as that of the communication system 1a according to the first embodiment. That is, the communication system of the second embodiment includes the transmission device 2, the multimode fiber 3, and the reception device.
The signal detection device 41b updates the unimodular matrix “T(k)”, the unimodular inverse matrix “Tinv(k)”, and the error covariance matrix “˜Re(k)” based on the LR domain without using each matrix operation shown from the formulas (16) to (18) in the first embodiment. That is, the signal detection device 41b updates the unimodular matrix “T(k)”, the unimodular inverse matrix “Tinv(k)”, and the error covariance matrix “˜Re(k)” based on the LR domain by using the row basic deformation or the column basic deformation of the matrix. This makes it possible to reduce the calculation amount as compared with the calculation amount in the first embodiment.
The unimodular matrix update unit 419b updates the unimodular matrix “T(k)” as shown in equation (19).
[Math. 19]
ti(k)←ti(k)−λi,j(k)tj(k) (19)
Here, “ti” represents the i-th column of the unimodular matrix “T(k)”.
The unimodular inverse matrix update unit 420b updates the unimodular inverse matrix “Tinv(k)” as shown in formula (20).
[Math. 20]
tinv(i)(k)←tinv(i)(k)+λi,j(k)tinv(i)(k) (20)
Here, “tiny (i)” represents the i-th row of the unimodular inverse matrix “Tinv(k)”.
The second covariance matrix update unit 421b updates the error covariance matrix “˜Re(k)” based on the LR domain, using the row basic deformation as shown in formula (21).
[Math. 21]
{tilde over (r)}i(k)←{tilde over (r)}i(k)+λi,j*(k){tilde over (r)}j(k) (21)
Here, “˜ri(k)” represents the i-th column of the error covariance matrix “˜Re(k)” based on the LR domain. “˜rj(k)” represents the j-th column of the error covariance matrix “˜Re(k)” based on the LR domain.
The second covariance matrix update unit 421b updates the error covariance matrix “˜Re(k)” based on the LR domain, using the column basic deformation as shown in formula (22).
[Math. 22]
{tilde over (r)}(i)(k)←{tilde over (r)}(i)(k)+λi,j(k){tilde over (r)}(j)(k) (22)
Here, “˜r(i)(k)” represents the i-th row of the error covariance matrix “˜Re(k)” based on the LR domain. “˜r(j) (k)” represents the j-th row of the error covariance matrix “˜Re(k)” based on the LR domain.
Next, operation example of the signal detection device 41b will be described.
As shown in a thirteenth line, the unimodular matrix update unit 419b updates the unimodular matrix “T(k)” as shown in the formula (19). As shown in a fourteenth line, the unimodular inverse matrix update unit 420b updates the unimodular inverse matrix “Tinv(k)” as shown in the formula (20).
As shown in a fifteenth line, the second covariance matrix update unit 421b updates the j-th column “˜rj(k)” of the error covariance matrix “˜Re(k)” based on the LR domain, as shown in formula (21). As shown in a sixteenth line, the second covariance matrix update unit 421b updates the i-th row “˜ri(k)” of the error covariance matrix “˜Re(k)” based on the LR domain, as shown in formula (22).
As described above, the signal detection unit 411b multiplies the reception signal vector “y” by the separation matrix “W” to derive the estimation vector “{circumflex over ( )}x” of the transmission signal vector “x” corresponding to the reception signal vector. The basis conversion unit 412b (the first conversion unit), by using the inverse matrix “Tiny” of the inverse matrix of the unimodular matrix, converts the estimation vector “{circumflex over ( )}x” of the transmission signal vector to the estimation vector “{circumflex over ( )}˜x” of the transmission signal vector based on the LR domain (reduced basis). The first determination unit 413b converts the estimation vector “{circumflex over ( )}˜x” of the transmission signal vector based on the LR domain to the determination value vector “xHD” of the transmission signal vector by using the inverse matrix “Tiny” of the unimodular matrix and the unimodular matrix “T”.
The separation matrix update unit 415b (the first update unit) updates the separation matrix “W” by using the first error signal vector between the estimation vector “{circumflex over ( )}x” of the transmission signal vector and the determination value vector “xHD” of the transmission signal vector. The error basis conversion unit 416b (the second conversion unit) converts the first error signal vector to the second error signal vector based on the LR domain by using the inverse matrix “Tiny” of the unimodular matrix as shown in the formula (9). The first covariance matrix update unit 417b (the second update unit) updates the error covariance matrix “˜Re” based on the LR domain by using the second error signal vector based on the LR domain.
The second determination unit 418 determines whether or not the predetermined condition is satisfied. The unimodular matrix update unit 419b (the third update unit) updates the unimodular matrix “T” when it is determined that the predetermined condition is satisfied. The unimodular inverse matrix update unit 420b (the third update unit) updates the inverse matrix “Tiny” of the unimodular matrix when it is determined that the predetermined condition is satisfied. The second covariance matrix update unit 421b (the third update unit) updates the error covariance matrix “˜Re” based on the LR domain when it is determined that the predetermined condition is satisfied.
Thus, it is possible to reduce the error of the information placed on the optical signal transmitted through the transmission line having the time variability with the calculation amount equal to or less than the predetermined threshold.
Next, an effect example (a simulation result of the two data sequences transmitted with the mode multiplexing) of the operation of the signal detection device 41b will be described.
The transmission device 2 modulates bit strings of a data sequences (transmission signal vectors) to be transmitted as the mode multiplexing signal as 16QAM signals. The mode dependent loss of the transmission channel is, for example, 20 dB. In the transmission channel, the mode multiplexing signals are mixed with each other. Therefore, the signal detector 41b separates the mode multiplexing signal by using the digital signal processing.
The signal detection device 41b executes a detection processing based on maximum likelihood (ML) decoding, a linear detection processing based on minimum mean squared error (MMSE), and a MIMO signal detection processing based on lattice reduction (LR) as separation detection methods for comparing accuracy with each other.
The detection processing based on the maximum likelihood decoding is known as an optimal decoding method in the case where prior knowledge about the frequency of signal generation cannot be obtained. However, the calculation amount required for the detection processing based on the maximum likelihood decoding is very large.
In
On the other hand, the MIMO signal detection processing based on the lattice reduction in the signal detection device 41b shows excellent detection accuracy compared with linear detection based on a minimum mean squared error. In addition, the detection accuracy by MIMO signal detection processing based on lattice reduction in the signal detection device 41b is gradually close to the detection accuracy by the detection processing based on the maximum likelihood decoding.
In the detection processing based on the maximum likelihood decoding, candidates of transmission signals are searched for in total. Therefore, in the detection processing based on the maximum likelihood decoding, the calculation amount becomes very large. In addition, the number of candidates for the transmission signal is increased in accordance with the increase of the number of multi-values of the signal. For this reason, for example, when “N=10”16QAM signals are detected, the complex multiplication of 1014 times is required.
The linear detection processing based on the minimum mean squared error is a simple detection method with a small calculation amount. However, as shown in
On the other hand, the calculation amount of the MIMO signal detection processing based on the lattice reduction in the signal detection device 41b is suppressed to about three times the calculation amount of the linear detection processing based on the minimum mean squared error. The calculation amount of the linear detection processing based on the minimum mean squared error and the calculation amount of the MIMO signal detection processing based on the lattice reduction do not depend on the format of the signal to be detected.
In this way, in the communication system 1a provided with the transmission channel having the mode dependent loss, the MIMO signal detection based on the lattice reduction can reduce the error of information with the operation amount equal to or less than the predetermined threshold.
The configuration of the communication system of the third embodiment is the same as that of the communication system 1a of the first embodiment. That is, the communication system of the third embodiment includes the transmission device 2, the multimode fiber 3, and the reception device.
A signal detection unit 411d detects the inputted electric signal (the reception signal). At the sampling time point “k”, the estimation vector of the transmission signal vector based on the LR domain “{circumflex over ( )}˜x(k)” is represented as shown in formula (23).
[Math. 23]
{circumflex over (x)}={tilde over (W)}T(k)y(k) (23)
Here, “˜W(k)” represents the separation matrix (the signal separation matrix) based on the LR domain.
The first determination unit 413c determines the signal in the LR domain (the system that increases orthogonality of the transmission channel matrix “H”) by using the estimation vector “{circumflex over ( )}˜x(k)” of a transmission signal vector based on the LR domain. That is, the first determination unit 413c derives a sampling value (a determination value vector) of the reception signal vector for the estimation vector of the transmission signal vector based on the LR domain. For example, the first determination unit 413c derives the determination value vector “˜xHD(k)” of the transmission signal vector based on the LR domain as shown in formula 24 by using the unimodular reverse matrix “Tinv(k)”.
[Math. 24]
{tilde over (x)}HD(k)=Q({circumflex over ({tilde over (x)})}(k),Tinv(k)) (24)
Further, the first determination unit 413c derives the determination value vector “xHD(k)” of the transmission signal vector by using the unimodular matrix “T(k)” and the determination value vector “{circumflex over ( )}˜xHD(k)” of the estimation value vector of the transmission signal vector based on the LR domain, as shown in formula (25).
[Math. 25]
xHD(k)=T(k){circumflex over ({tilde over (x)})}HD(k) (25)
The error derivation unit 414c derives the error signal vector based on the LR domain as shown in formula (26).
[Math. 26]
{tilde over (∈)}(k)={circumflex over ({tilde over (x)})}HD(k)−{circumflex over ({tilde over (x)})}(k) (26)
The separation matrix update unit 415c updates the separation matrix “˜W(k)” based on the LR domain. That is, the separation matrix update unit 415c updates the separation matrix “˜W(k)” based on the LR domain as shown in formula (27) to derive the separation matrix “˜W(k+1)” based on the LR domain.
[Math. 27]
{tilde over (W)}T(k+1)={tilde over (W)}T(k)+μ{tilde over (∈)}(k)yH(k) (27)
The first covariance matrix update unit 417c updates the error covariance matrix “˜Re(k)” based on the LR domain, as shown in formula (28).
[Math. 28]
{tilde over (R)}e(k+1)=δ{tilde over (R)}e(k)+{tilde over (∈)}(k){tilde over (∈)}H(k) (28)
A method for deriving the integer “Λi, j(k)” by the unimodular matrix update unit 419c, a method for selecting the pair of (i, j) by the unimodular matrix update unit 419c, and a method for updating the unimodular matrix “T(k)” by the unimodular matrix update unit 419c are the same as the method by the unimodular matrix update unit 419a described in the first embodiment.
The method for updating the unimodular inverse matrix “Tinv(k)” by the unimodular inverse matrix update unit 420c is the same as the method for updating by the unimodular inverse matrix update unit 420a described in the first embodiment.
The method for updating the error covariance matrix “˜Re(k)” based on the LR domain by the second covariance matrix update unit 421c is the same as the method for updating by the second covariance matrix update unit 421a described in the first embodiment.
The second separation matrix update unit 423c updates the separation matrix “˜W(k)” based on the LR domain as shown in formula (29).
[Math. 29]
{tilde over (W)}T(k)←Λi,j(k){tilde over (W)}T(k) (29)
The second separation matrix update unit 423c may update the separation matrix “˜W(k)” based on the LR domain as shown in formula (30).
[Math. 30]
{tilde over (W)}(k)←{tilde over (W)}(k)Λi,jT (30)
Next, operation example of the signal detection device 41c will be described.
As shown in a third line, the first determination unit 413c derives the error signal vector based on the LR domain as shown in the formula (26). As shown in a fourth line, the first determination unit 413c derives the determination value vector “xHD(k)” of the transmission signal vector, as shown in the formula (25), by using the unimodular matrix “T(k)”.
As shown in a fifth line, the separation matrix update unit 415c update the separation matrix “˜W(k)” based on the LR domain as shown in formula (27) to derive the separation matrix “˜W(k+1)” based on the LR domain. As shown in a sixth line, the first covariance matrix update unit 417c updates the error covariance matrix “˜Re(k)” based on the LR domain, as shown in the formula (28).
The operations of seventh to thirteenth lines shown in
As described above, the signal detection unit 411c (signal detector) multiplies the reception signal vector “y” by the separation matrix “˜W” based on the LR domain (reduced basis), thereby deriving the estimation vector “{circumflex over ( )} ˜x” of the transmitted signal vector based on the LR domain. The first determination unit 413c (first determiner) derives the determination value vector “˜xHD” of the transmission signal vector based on an LR domain by using the inverse matrix “Tinv” of the unimodular matrix as shown in the formula (24). The first determination unit 413c derives the determination value vector “xHD” of the transmission signal vector by using, as shown in the formula (25), the determination value vector “{circumflex over ( )}˜xHD” of the estimation vector of the transmission signal vector based on the LR domain, and the unimodular matrix “T”.
The error derivation unit 414c (error calculator) (error derivator) derives the error signal vector between the determination value vector “˜xHD” of the transmission signal vector based on the LR domain and an estimation vector “{circumflex over ( )}˜x” of the transmission signal vector based on the LR domain as the error signal vector based on the LR domain as shown in the formula (26). The separation matrix update unit 415c (the first update unit) (first updater) updates the separation matrix “˜W” based on the LR domain by using the error signal vector based on the LR domain. The first covariance matrix update unit 417c (the second update unit) (second updater) updates the error covariance matrix “˜Re” based on the LR domain by using the error signal vector based on the LR domain.
The second determination unit 418 (second determiner) determines whether or not the predetermined condition is satisfied. The unimodular matrix update unit 419c (the third update unit) (third updater) updates the unimodular matrix “T” when it is determined that the predetermined condition is satisfied. The unimodular inverse matrix update unit 420c (the third update unit) (third updater) updates the inverse matrix “Tinv” of the unimodular matrix when it is determined that the predetermined condition is satisfied. The second covariance matrix update unit 421c (the third update unit) (third updater) updates the error covariance matrix “˜Re” based on the LR domain when it is determined that the predetermined condition is satisfied. The second separation matrix update unit 423c (the third update unit) (third updater) updates the separation matrix “˜W” based on the LR domain when it is determined that the predetermined condition is satisfied.
Thus, it is possible to reduce the error of the information placed on the optical signal transmitted through the transmission line having the time variability with the calculation amount equal to or less than the predetermined threshold.
In a fourth embodiment,
the point where the signal separation is executed by the method described in the third embodiment and the update for the unimodular matrix, the unimodular inverse matrix, and the error covariance matrix based on the LR domain is performed by each method described in the second embodiment is different from the first embodiment. For the fourth embodiment, the differences from the first embodiment will be mainly described.
The configuration of the communication system of the fourth embodiment is the same as that of the communication system 1a of the first embodiment. That is, the communication system of the fourth embodiment includes the transmission device 2, the multimode fiber 3, and the reception device.
The separation matrix update unit 415c updates the separation matrix “˜W(k)” based on the LR domain as shown in formula (31).
[Math. 31]
{tilde over (w)}i(k)←{tilde over (w)}i(k)+λi,j(k){tilde over (w)}j(k) (31)
Here, “˜wi(k)” represents the i-th column of the separation matrix “˜W(k)” based on the LR domain.
Next, operation example of the signal detection device 41d will be described.
As described above, the signal detection unit 411d multiplies the reception signal vector “y” by the separation matrix “˜W” based on the LR domain (reduced basis), thereby deriving the estimation vector “{circumflex over ( )}˜x” of the transmission signal vector based on the LR domain. The first determination unit 413d derives the determination value vector “˜xHD” of the transmission signal vector based on the LR domain by using the inverse matrix “Tiny” of the unimodular matrix as shown in the formula (24). The first determination unit 413d derives the determination value vector “xHD” of the transmission signal vector by using, as shown in the formula (25), the determination value vector “{circumflex over ( )}˜xHD” of the estimation vector of the transmission signal vector based on the LR domain, and the unimodular matrix “T”.
The error derivation unit 414d derives the error signal vector between the determination value vector “˜xHD” of the transmission signal vector based on the LR domain and the estimation vector “{circumflex over ( )}˜x” of the transmission signal vector based on the LR domain as the error signal vector based on the LR domain as shown in the formula (26). The separation matrix update unit 415d (the first update unit) updates the separation matrix “˜W” based on the LR domain by using the error signal vector based on the LR domain. The first covariance matrix update unit 417d (the second update unit) updates the error covariance matrix “˜Re” based on the LR domain by using the error signal vector based on the LR domain.
The second determination unit 418 determines whether or not the predetermined condition is satisfied. The unimodular matrix update unit 419d (the third update unit) updates the unimodular matrix “T” when it is determined that the predetermined condition is satisfied. The unimodular inverse matrix update unit 420d (the third update unit) updates the inverse matrix “Tiny” of the unimodular matrix when it is determined that the predetermined condition is satisfied. The second covariance matrix update unit 421d (the third update unit) updates the error covariance matrix “˜Re” based on the LR domain when it is determined that the predetermined condition is satisfied. The second separation matrix update unit 423d (the third update unit) updates the separation matrix “˜W” based on the LR domain as shown in the formula (31) when it is determined that the predetermined condition is satisfied.
Thus, it is possible to reduce the error of the information placed on the optical signal transmitted through the transmission line having the time variability with the calculation amount equal to or less than the predetermined threshold.
Next, an example of a hardware configuration of the signal detection device 41 will be described.
A part or all of each functional units of the signal detection device 41 are implemented as a software by a processor 100 such as a CPU (Central Processing Unit) executing a program stored in a storage device 200 including a non-volatile recording medium (non-transitory recording medium) and in a memory 300. The program may be recorded on a computer-readable recording medium. The computer-readable recording medium is a portable medium such as a flexible disc, an optical magnetic disc, an ROM (Read Only Memory) or a CD-ROM (Compact Disc Read Only Memory) and a non-transitory recording medium such as a storage device including a hard disk built in computer system or the like, for example.
A part or the entirety of the functional units of the signal detection device 41 may be implemented by a hardware including an electronic circuit or a circuitry in which an LSI (Large Scale Integration circuit), an ASIC (Application Specific Integrated Circuit), a PLD (Programmable Logic Device), an FPGA (Field Programmable Gate Array), or the like is used.
Although the embodiments of the present invention have been described in detail with reference to the drawings, specific configurations are not limited to these embodiments, and designs and the like within a range that does not deviating from the gist of the present invention are also included.
In each of the above embodiments, the communication system performs multiplexing optical transmission using the spatial mode in the multimode fiber or the few-mode fiber, for example. The communication system is not limited to a system for executing a specific communication as long as it is a system for executing a communication. For example, the communication system may perform multiplexing optical transmission using spatial modes in the coupled multicore fiber. For example, the communication system may be any of a wireless communication system, a satellite communication system, a communication system between a recording medium and a chip, and the like.
The present invention is applicable to a device for separating and detecting a plurality of signals by the MIMO-type signal processing.
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/JP2020/031435 | 8/20/2020 | WO |
| Publishing Document | Publishing Date | Country | Kind |
|---|---|---|---|
| WO2022/038739 | 2/24/2022 | WO | A |
| Number | Name | Date | Kind |
|---|---|---|---|
| 9294146 | Mumtaz | Mar 2016 | B2 |
| 10985846 | Shibahara | Apr 2021 | B2 |
| 11641251 | Rekaya | May 2023 | B2 |
| 11689294 | Matsuda | Jun 2023 | B2 |
| 11770205 | Shibahara | Sep 2023 | B2 |
| 20200259544 | Demmer | Aug 2020 | A1 |
| Entry |
|---|
| A.K. Lenstra et al., “Factoring polynomials with rational coefficients”, Mathematische Annalen, vol. 261, No. 4, pp. 515-534, 1982. |
| J. Park et al., “Lattice reduction aided MMSE decision feedback equalizers”, IEEE transactions of signal processing, vol. 59, No. 1, pp. 436-441, 2011. |
| Q. Zhou et al., “Element-based lattice reduction algorithms for large MIMO detection”, IEEE Journal on Selected Areas in Communications, vol. 31, No. 2, pp. 274-286, 2013. |
| Number | Date | Country | |
|---|---|---|---|
| 20230216589 A1 | Jul 2023 | US |
