CN109618311B - Blind detection algorithm based on M2M communication spectrum sharing and coexistence - Google Patents
Blind detection algorithm based on M2M communication spectrum sharing and coexistence Download PDFInfo
- Publication number
- CN109618311B CN109618311B CN201910042795.5A CN201910042795A CN109618311B CN 109618311 B CN109618311 B CN 109618311B CN 201910042795 A CN201910042795 A CN 201910042795A CN 109618311 B CN109618311 B CN 109618311B
- Authority
- CN
- China
- Prior art keywords
- matrix
- signal
- communication
- user
- channel
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
- H04W4/70—Services for machine-to-machine communication [M2M] or machine type communication [MTC]
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B17/00—Monitoring; Testing
- H04B17/30—Monitoring; Testing of propagation channels
- H04B17/391—Modelling the propagation channel
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L1/00—Arrangements for detecting or preventing errors in the information received
- H04L1/0001—Systems modifying transmission characteristics according to link quality, e.g. power backoff
- H04L1/0036—Systems modifying transmission characteristics according to link quality, e.g. power backoff arrangements specific to the receiver
- H04L1/0038—Blind format detection
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L25/00—Baseband systems
- H04L25/02—Details ; arrangements for supplying electrical power along data transmission lines
- H04L25/0202—Channel estimation
- H04L25/0238—Channel estimation using blind estimation
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L27/00—Modulated-carrier systems
- H04L27/32—Carrier systems characterised by combinations of two or more of the types covered by groups H04L27/02, H04L27/10, H04L27/18 or H04L27/26
- H04L27/34—Amplitude- and phase-modulated carrier systems, e.g. quadrature-amplitude modulated carrier systems
- H04L27/345—Modifications of the signal space to allow the transmission of additional information
- H04L27/3461—Modifications of the signal space to allow the transmission of additional information in order to transmit a subchannel
- H04L27/3483—Modifications of the signal space to allow the transmission of additional information in order to transmit a subchannel using a modulation of the constellation points
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Power Engineering (AREA)
- Quality & Reliability (AREA)
- Physics & Mathematics (AREA)
- Electromagnetism (AREA)
- Mobile Radio Communication Systems (AREA)
- Cable Transmission Systems, Equalization Of Radio And Reduction Of Echo (AREA)
Abstract
The invention discloses a blind detection algorithm based on sharing and coexistence of an M2M communication frequency spectrum, which comprises the following steps: step SS 1: constructing a traditional user overcomplete model and an M2M communication sparse model; step SS 2: solving the transmission signal of M2M by convex optimization; step SS 3: constructing a traditional user receiving data matrix; step SS 4: performing singular value decomposition on the received data matrix in the step SS 3; step SS 5: setting a weight matrix; step SS 6: and selecting an activation function of the Hopfield neural network, and performing Hopfield neural network iterative operation. The method recovers the recovery of the traditional user in the M2M communication by utilizing the blind detection of the Hopfield neural network for the first time, obtains the M2M equipment transmission signal by utilizing convex optimization according to the overcomplete model of the traditional user and the sparse model of the M2M equipment, and iterates the state equation: and when iteration is performed each time, the Hopfield neural network is entered, and the simulation verifies that under the same condition, the bit error rate of the method is superior to that of a method for recovering the traditional user signal under the condition of assuming that a channel is known.
Description
Technical Field
The invention relates to a blind detection algorithm based on M2M communication spectrum sharing and coexistence, and belongs to the technical field of communication signal processing.
Background
The support of M2M communication will be an important component of communication systems, especially 5G cellular systems, which puts higher demands on the adaptability of communication signals.
A new SPECTRUM sharing scheme is described in the literature [ Xiaoohua Li, Jian Zheng, Mingjian Zhuang, Compound sensitive BASED Spectrum SHARING AND COEXISTENCE FOR MACHINE-TO-MACHINE COMMUNICATIONS, "International Conference on Acoustics Speech and Signal Processing (ICASSP), IEEE,2017: 3604-: a large number of dispersed M2M devices share the same channel with legacy communication users, such as cellular users. When legacy users continue to transmit, M2M devices transmit directly on this channel without much channel scheduling and handshaking overhead. M2M communication is through simulating PMU communication in the smart power grids, and 16 bit A/D converter quantization data are passed through to the PMU data package that obtains, and add UDP package and 20% training sequence again. The literature creates a sparse signal model using redundancies in the transmitted signal, such as training symbols, pilots, multiple access control overhead and related data, and then can separate the mixed signals using the redundancies in the transmitted signal under the compressed sensing technique. The feasibility of the sparse signal model and the spectrum sharing scheme is verified by simulating an M2M communication scheme in the smart grid. Reference is made to a generic sparsity control outlier framework described in the documents g.b. giannakis, g.math, s.farahmand, v.kekatos, and h.zhu, "usparor: Universal space-controlling outlier rejection," proc.ieee int.conf.on Acoustics, spech and Signal Processing (ICASSP' 11), Prague, May 2011 ], constructing a legacy user Signal with an overcomplete model and a Signal with a sparse model M2M using sparsity in the sparse adjustment parameter lambda control estimator in Lasso and the number of frame rejection outliers, and separating the M2M Signal and the legacy user Signal from the mixed Signal using Signal redundancy using compressive sensing techniques. The compressed sensing is firstly applied to multi-user detection by utilizing the sparsity of user behaviors in a document [ ZHHUH, GIANNAKISGB. Exploiting partial activity in multi-user detection [ J ]. IEEE Transactions on Communications,2011,59(2): 454-465 ], and the complexity of a CDMA multi-user detection system is reduced by solving the convex optimization problem.
The document [ zhang man kong, epi-zhi, xu ping.m 2M communication system based on compressed sensing multi-user detection technology, electronic measurement technology, 2018,8 ] indicates that M2M communication is sporadic, and in order to alleviate the problem of a large amount of control signal overhead brought by large-scale M2M communication, discusses that the MMV model based on compressed sensing is utilized under a frame transmission structure to further improve the behavior detection performance in multi-user detection. However, the above scheme assumes that the transmission channel is known at the time of system design, so that the error rate of recovering the conventional user signal is high when the influence of the channel is removed. While the actual communication system transmission channel is time-varying and mostly unknown. The blind detection technique is applicable to a case where a channel is unknown, and can detect a transmission signal using only a reception signal itself, thereby removing inter-symbol interference (ISI) to improve an information transmission rate and reliability. Many documents have begun to study the signal blind detection problem using the Hopfield neural network. The Hopfield Neural Network (HNN) blind detection algorithm is not limited by whether a channel contains a common zero point or not, the required sent data is shorter, and compared with a second-order statistic blind algorithm and a high-order statistic blind algorithm, the method can better meet the requirement of high-speed data transmission in a modern communication system.
Disclosure of Invention
The technical problem to be solved by the present invention is to provide a blind detection algorithm based on M2M communication spectrum sharing and coexistence in order to improve the defects and shortcomings of the prior art. The algorithm of the invention adopts a blind detection algorithm to recover the signals sent by the traditional users on the basis of the M2M communication spectrum sharing and coexistence scheme, and further improves the reliability of processing the communication signals by utilizing the Hopfield neural network. The method aims to provide an accurate and reliable detection method for recovering the traditional user signal in the M2M communication.
The invention adopts the following technical scheme: the blind detection algorithm based on the sharing and coexistence of the M2M communication spectrum is characterized by comprising the following steps:
step SS 1: constructing a traditional user overcomplete model and an M2M communication sparse model;
step SS 2: solving the transmission signal of M2M by convex optimization;
step SS 3: constructing a traditional user receiving data matrix;
step SS 4: performing singular value decomposition on the received data matrix in the step SS 3;
step SS 6: and selecting an activation function of the Hopfield neural network, and performing Hopfield neural network iterative operation.
As a preferred embodiment, step SS1 specifically includes: the receiving end receives the transmission signals of a single traditional user and M2M devices, and the receiving end obtains a receiving equation after OFDM modulation and demodulation and sampling:
where q is an oversampling factor, P represents the order of the channel, M is the number of M2M communication devices, si(m) is a transmission signal matrix, s when i is 00(m)=U0a0(m)=U0[a0(m),...,a0(m+N-1)]TSignalling matrix a for legacy users0(M) a vector-converted transmission matrix, i ═ 1, L, M times si(m)=Uiai(m)=Ui[ai(m),...,ai(m+N-1)]TSignalling matrix a for M2M devicesi(m) vector-converted Transmission matrix, U0,UiAre all NxN vector transformation unitary matrixes; [ h ] ofi0,…,hiP]q×(P+1)Is the impulse response of the communication channel between the traditional user and the M2M device, and is the channel matrix under the condition that the assumed channel is known when the sampling factor q takes 1; vector v(m) is determined by a mean of zero and a variance ofThe additive white Gaussian noise component, the sending signal and the additive noise are independent. Index function Ii(m) ═ 1 denotes that user/device I transmits in OFDM block m, and Ii(m) ═ 0 indicates that user/device i has not transmitted a signal; suppose UiAs is known at system design, neglecting the noise effect, the above equation can be written as follows:
RN=SiΓT;
Si=[sL+P(m),L,sL+P(m+N-1)]T=[sN(k),L,sN(m-P-L)]N×(L+P+1)is a transmit signal matrix; (R)N)N×(L+1)q=[rL(m),L,rL(m+N-1)]TIs the receive matrix of the legacy users; (gamma)(L+1)q×(L+P+1)Is formed byijJ is 0,1, L, P, where L is a parameter of the equalizer;
considering the similarity of data packets and the similarity of adjacent OFDM blocks, from ai(m) minus ai(m-1) enhancing sparsity using such similarity; order to
xi(m)=ai(m)-ai(m-1),i=1,L,M;
Suppose that all a have been addressedi(l) The block is analyzed and known, where l ≦ m-1, then only a needs to be detectedi(M), i ═ 0, L, M; thus subtracting the detected receiver-side signal from the received signal vector r (m)Obtaining an overcomplete model of the traditional user and an M2M communication sparse model:
the first term is an overcomplete model of the traditional user signal, and the second term is an M2M communication signalSparse model of numbers, q is an oversampling factor, P represents the order of the channel, M is the number of M2M communication devices, Ii(m)=1,UiAre all NxN vector transform unitary matrices, NxK0Matrix arrayIs U0Corresponding submatrix, (h)0j)q×1Is the impulse response of the conventional user signal, (h)ij)q×1Is the impulse function of M2M communication, and the vector v (M) is formed by the mean value of zero and the variance ofThe additive white Gaussian noise component, the sending signal and the additive noise are independent.
As a preferred embodiment, step SS2 specifically includes:
estimating x by solving for optimization using compressed sensing techniques0(m) and xi(m):
Adjusting the weighting factor lambdaiAnd | | | xi||0Matching, | | · | luminance0Is represented by0A norm;
one common practice for compressive sensing is to use a set of bumps l1Norm instead of l0Norm, which makes the above formula become:
Is obtained byThe M2M device transmission signal may then be estimated using the following equation:
where q is an oversampling factor, P represents the order of the channel, M is the number of M2M communication devices, si(m) is a transmission signal matrix, s when i is 00(m)=U0a0(m)=U0[a0(m),...,a0(m+N-1)]TSignalling matrix a for legacy users0(M) a vector-converted transmission matrix, i ═ 1, L, M times si(m)=Uiai(m)=Ui[ai(m),...,ai(m+N-1)]TSignalling matrix a for M2M devicesi(m) vector-converted Transmission matrix, U0,UiAre all NxN vector transformation unitary matrixes; [ h ] ofi0,…,hiP]q×(P+1)Is the impulse response of the communication channel between the legacy user and the M2M device, [ h ]i0,…,hiP]q×(P+1)Is the impulse response of the communication channel between the legacy user and the M2M device; vector v (m) is mean zero and varianceAdditive white gaussian noise of (1); index function Ii(m) ═ 1 denotes that user/device I transmits in OFDM block m, and Ii(m) ═ 0 means that user/device i does not transmit in OFDM.
As a preferred embodiment, step SS3 specifically includes: substituting the M2M transmission signal obtained in the step SS2 into the conventional user overcomplete model and the M2M communication sparse model to obtain the reception data of only the conventional user:
under the noiseless condition, the receiving end receives the signal sent by a single traditional user, and the receiving equation of the above formula obtained by sampling can be abbreviated as follows:
yN=SHT;
wherein, it is madeS=[sL+P(m),L,sL+P(m+N-1)]T=[sN(k),L,sN(m-P-L)]N×(L+P+1)Is a traditional user sending signal matrix; (y)N)N×(L+1)qIs the receive matrix of the legacy users; (H)(L+1)q×(L+P+1)is the channel impulse response h00,Kh0P]q×(P+1)A constructed block Toeplitz matrix; (.)TRepresenting a matrix transposition;
wherein, the legacy user sends a signal matrix:
s=[sN(m),L,sN(m-P-L)]N×(L+P+1);
K0P + L, where P is the channel order, L is the equalizer order, and N is the required data length;
sL+P(m)=[s(m),L,s(m-L-P)]T(ii) a Wherein s belongs to { +/-1 +/-i }, and the time m is a natural number;
H0corresponding to H is the impulse response of the channel [ H (0), H (1), Kh (P)]The block Toeplitz matrix formed:
p represents the order of the channel, L represents the order of the equalizer, and q is an oversampling factor and takes the value of a positive integer;
XN=[xL(m),L,xL(m+N-1)]Ta legacy user receiving data matrix that is N x (L +1) q,
wherein xL(m)=Γ·sL+P(m)。
As a preferred embodiment, step SS4 specifically includes: singular value decomposition of a received data matrix:
in the formula (I), the compound is shown in the specification,
(·)His the Hermitian transpose;
u is nx (L + P +1) unitary matrix in singular value decomposition;
0 is an (N- (L + P +1)) × (L +1) q zero matrix;
v is (L +1) qx (L +1) q unitary matrix;
Ucis N × (N- (L + P +1)) unitary matrix;
d is a (L + P +1) × (L +1) q singular value array;
the performance function and optimization problem is constructed accordingly:
wherein s belongs to { +/-1 +/-i }NIs an N-dimensional vector, the character set belongs to +/-1 +/-i },representing an estimate of the signal. argmin () represents the variable value at which the target function is minimized, d is the delay factor, d is 0, L, P + L.
As a preferred embodiment, step SS6 specifically includes:
for QPSK transmission signals, the complex Hopfield neural network state equation is:
the state equation is as follows:
an output equation: y (m) ═ ws (m);
wherein s ═ s1,s2,L,sN]A transmit vector for the neural network; w is the connection weight matrix of each neuron, and W is CN×N,WHW; σ (-) is the activation function of the neural network; when the neural network reaches the final balance, approximately considering that s (m) ═ s (m +1) of each neuron, and s (m) is the acquired sending signal;
and (3) performing iterative operation on the state equation and the output equation, then substituting the result of each iteration into an energy function E (m) of the Hopfield neural network, and when the energy function E (m) reaches a minimum value, namely s (m) ═ s (m +1), the Hopfield neural network reaches equilibrium and the iteration is ended.
As a preferred embodiment, step SS6 further includes:
for adapting to complex channel and complex constellation signal QPSK, the activation function of the blind detection algorithm based on spectrum sharing and coexistence of M2M communication is:
σ(x)=tanh(Re(a·x))+itanh(Im(a·x));
wherein a is a gradient coefficient and mainly controls the gradient of the activation function;
the energy function based on blind detection of M2M communication spectrum coexistence and shared Hopfield neural network is expressed as:
wherein: e is the energy function of the network, the energy function being one and the iteration timeRelated variables, weight matrix W satisfying W ═ WHS (m) is the output of the neuron, σ-1(x) Is the inverse of the Sigmoid function σ (x) of the neuron.
The invention achieves the following beneficial effects: : the method recovers the recovery of the traditional user in the M2M communication by utilizing the blind detection of the Hopfield neural network for the first time, obtains the M2M equipment transmission signal by utilizing convex optimization according to the overcomplete model of the traditional user and the sparse model of the M2M equipment, and iterates the state equation: and when iteration is performed each time, the Hopfield neural network is entered, and the simulation verifies that under the same condition, the error rate of the method is superior to that of a method for recovering the traditional user signal under the condition of assuming that a channel is known.
Drawings
Fig. 1 is a schematic diagram of blind detection algorithm based on M2M communication spectrum sharing and coexistence.
Fig. 2 is a system diagram of a blind detection algorithm for recovering a conventional user signal according to the present invention.
Fig. 3 is a graph comparing the bit error rate of the M2M communication spectrum sharing and coexistence based blind detection algorithm with the recovery signal algorithm assuming a known channel. The HNN (Hopfield neural network) algorithm is a classical Hopfield neural network algorithm.
Fig. 4 shows four different neural networks under the blind detection method of the present invention based on M2M communication spectrum sharing and coexistence: error rate maps for classical Hopfield Neural Networks (HNN), Hopfield type Positive Feedback Neural networks (PFHNN), and Complex-System Hopfield type Neural networks (CSHNN).
Fig. 5 is a comparison of bit error rates for fixed transmission sequence lengths using two different classical channels under the blind detection algorithm of the present invention. Channel one: and adopting a synthesized channel with fixed weight and delay and no common zero point. And a second channel: and randomly synthesizing channels with variable weights and delay degrees.
Detailed Description
The invention is further described below with reference to the accompanying drawings. The following examples are only for illustrating the technical solutions of the present invention more clearly, and the protection scope of the present invention is not limited thereby.
As shown in fig. 1, the present invention provides a blind detection algorithm based on M2M communication spectrum sharing and coexistence, which is characterized by comprising the following steps:
step SS 1: constructing a traditional user overcomplete model and an M2M communication sparse model;
step SS 2: solving the transmission signal of M2M by convex optimization;
step SS 3: constructing a traditional user receiving data matrix;
step SS 4: performing singular value decomposition on the received data matrix in the step SS 3;
step SS 6: and selecting an activation function of the Hopfield neural network, and performing Hopfield neural network iterative operation.
As a preferred embodiment, step SS1 specifically includes: the receiving end receives the transmission signals of a single traditional user and M2M devices, and the receiving end obtains a receiving equation after OFDM modulation and demodulation and sampling:
where q is an oversampling factor, P represents the order of the channel, M is the number of M2M communication devices, si(m) is a transmission signal matrix, s when i is 00(m)=U0a0(m)=U0[a0(m),...,a0(m+N-1)]TSignalling matrix a for legacy users0(M) a vector-converted transmission matrix, i ═ 1, L, M times si(m)=Uiai(m)=Ui[ai(m),...,ai(m+N-1)]TSignalling matrix a for M2M devicesi(m) vector-converted Transmission matrix, U0,UiAre all NxN vectorsTransforming a unitary matrix; [ h ] ofi0,…,hiP]q×(P+1)Is the impulse response of the communication channel between the traditional user and the M2M device, and is the channel matrix under the condition that the assumed channel is known when the sampling factor q takes 1; vector v (m) is formed by a mean of zero and a variance ofThe additive white Gaussian noise component, the sending signal and the additive noise are independent. Index function Ii(m) ═ 1 denotes that user/device I transmits in OFDM block m, and Ii(m) ═ 0 indicates that user/device i has not transmitted a signal; suppose UiAs is known at system design, neglecting the noise effect, the above equation can be written as follows:
RN=SiΓT;
Si=[sL+P(m),L,sL+P(m+N-1)]T=[sN(k),L,sN(m-P-L)]N×(L+P+1)is a transmit signal matrix; (R)N)N×(L+1)q=[rL(m),L,rL(m+N-1)]TIs the receive matrix of the legacy users; (gamma)(L+1)q×(L+P+1)Is formed byijJ is 0,1, L, P, where L is a parameter of the equalizer;
considering the similarity of data packets and the similarity of adjacent OFDM blocks, from ai(m) minus ai(m-1) enhancing sparsity using such similarity; order to
xi(m)=ai(m)-ai(m-1),i=1,L,M;
Suppose that all a have been addressedi(l) The block is analyzed and known, where l ≦ m-1, then only a needs to be detectedi(M), i ═ 0, L, M; thus subtracting the detected receiver-side signal from the received signal vector r (m)Obtaining an overcomplete model of the traditional user and an M2M communication sparse model:
the first term is an overcomplete model of the traditional user signal, the second term is a sparse model of the M2M communication signal, q is an oversampling factor, P represents the order of the channel, M is the number of M2M communication devices, and Ii(m)=1,UiAre all NxN vector transform unitary matrices, NxK0Matrix arrayIs U0Corresponding submatrix, (h)0j)q×1Is the impulse response of the conventional user signal, (h)ij)q×1Is the impulse function of M2M communication, and the vector v (M) is formed by the mean value of zero and the variance ofThe additive white Gaussian noise component, the sending signal and the additive noise are independent.
As a preferred embodiment, step SS2 specifically includes:
estimating x by solving for optimization using compressed sensing techniques0(m) and xi(m):
Adjusting the weighting factor lambdaiAnd | | | xi||0Matching, | | · | luminance0Is represented by0A norm;
one common practice for compressive sensing is to use a set of bumps l1Norm instead of l0Norm, which makes the above formula become:
Is obtained byThe M2M device transmission signal may then be estimated using the following equation:
where q is an oversampling factor, P represents the order of the channel, M is the number of M2M communication devices, si(m) is a transmission signal matrix, s when i is 00(m)=U0a0(m)=U0[a0(m),...,a0(m+N-1)]TSignalling matrix a for legacy users0(M) a vector-converted transmission matrix, i ═ 1, L, M times si(m)=Uiai(m)=Ui[ai(m),...,ai(m+N-1)]TSignalling matrix a for M2M devicesi(m) vector-converted Transmission matrix, U0,UiAre all NxN vector transformation unitary matrixes; [ h ] ofi0,…,hiP]q×(P+1)Is the impulse response of the communication channel between the legacy user and the M2M device, [ h ]i0,…,hiP]q×(P+1)Is the impulse response of the communication channel between the legacy user and the M2M device; vector v (m) is mean zero and varianceAdditive white gaussian noise of (1); index function Ii(m) ═ 1 denotes that user/device I transmits in OFDM block m, and Ii(m) ═ 0 means that user/device i does not transmit in OFDM.
As a preferred embodiment, step SS3 specifically includes: substituting the M2M transmission signal obtained in the step SS2 into the conventional user overcomplete model and the M2M communication sparse model to obtain the reception data of only the conventional user:
under the noiseless condition, the receiving end receives the signal sent by a single traditional user, and the receiving equation of the above formula obtained by sampling can be abbreviated as follows:
yN=SHT;
wherein, it is madeS=[sL+P(m),L,sL+P(m+N-1)]T=[sN(k),L,sN(m-P-L)]N×(L+P+1)Is a traditional user sending signal matrix; (y)N)N×(L+1)qIs the receive matrix of the legacy users; (H)(L+1)q×(L+P+1)is the channel impulse response h00,Kh0P]q×(P+1)A constructed block Toeplitz matrix; (.)TRepresenting a matrix transposition;
wherein, the legacy user sends a signal matrix:
s=[sN(m),L,sN(m-P-L)]N×(L+P+1);
K0P + L, where P is the channel order, L is the equalizer order, and N is the required data length;
sL+P(m)=[s(m),L,s(m-L-P)]T(ii) a Wherein s belongs to { +/-1 +/-i }, and the time m is a natural number;
H0corresponding to H is the impulse response of the channelShould be [ h (0), h (1), Kh (P)]The block Toeplitz matrix formed:
p represents the order of the channel, L represents the order of the equalizer, and q is an oversampling factor and takes the value of a positive integer;
XN=[xL(m),L,xL(m+N-1)]Ta legacy user receiving data matrix that is N x (L +1) q,
wherein xL(m)=Γ·sL+P(m)。
As a preferred embodiment, step SS4 specifically includes: singular value decomposition of a received data matrix:
in the formula (I), the compound is shown in the specification,
(·)His the Hermitian transpose;
u is nx (L + P +1) unitary matrix in singular value decomposition;
0 is an (N- (L + P +1)) × (L +1) q zero matrix;
v is (L +1) qx (L +1) q unitary matrix;
Ucis N × (N- (L + P +1)) unitary matrix;
d is a (L + P +1) × (L +1) q singular value array;
the performance function and optimization problem is constructed accordingly:
wherein s belongs to { +/-1 +/-i }NIs an N-dimensional vector, the character set belongs to +/-1 +/-i },an estimate value representing the signal; argmin () represents the variable value at which the target function is minimized, d is the delay factor, d is 0, L, P + L.
As a preferred embodiment, step SS6 specifically includes:
for QPSK transmission signals, the complex Hopfield neural network state equation is:
the state equation is as follows:
an output equation: y (m) ═ ws (m);
wherein s ═ s1,s2,L,sN]A transmit vector for the neural network; w is the connection weight matrix of each neuron, and W is CN×N,WHW; σ (-) is the activation function of the neural network; when the neural network reaches the final balance, approximately considering that s (m) ═ s (m +1) of each neuron, and s (m) is the acquired sending signal;
and (3) performing iterative operation on the state equation and the output equation, then substituting the result of each iteration into an energy function E (m) of the Hopfield neural network, and when the energy function E (m) reaches a minimum value, namely s (m) ═ s (m +1), the Hopfield neural network reaches equilibrium and the iteration is ended.
As a preferred embodiment, step SS6 further includes:
for adapting to complex channel and complex constellation signal QPSK, the activation function of the blind detection algorithm based on spectrum sharing and coexistence of M2M communication is:
σ(x)=tanh(Re(a·x))+itanh(Im(a·x));
wherein a is a gradient coefficient and mainly controls the gradient of the activation function;
the energy function based on blind detection of M2M communication spectrum coexistence and shared Hopfield neural network is expressed as:
wherein: e is the energy function of the network, the energy function is a variable related to the iteration time, and the weight matrix W satisfies that W is WHS (m) is the output of the neuron, σ-1(x) Is the inverse of the Sigmoid function σ (x) of the neuron.
In summary, the network enters the Hopfield neural network each cycle until the network reaches equilibrium.
In order to realize signal blind detection by using the Hopfield neural network and solve the problem of signal blind detection of a performance function, the minimum point of the energy function is required to correspond to the minimum point of the blind detection performance function. Because the Euler formula can make the interconversion between continuous time and discrete time, when the network is stable, s (m) s (m +1) can be approximately considered, and the difference between the first part of the energy function and the performance function formula can be seen as a negative sign by comparing the first part of the energy function with the performance function formula, the weight matrix for designing the double-Sigmoid hysteresis noise chaotic neural network can be considered as a projection operator form W (I)N-Q, wherein INIs a unit matrix of dimension N x N,thus, the minimum point of the energy function E (m) is corresponding to the minimum point of the blind detection performance function (2), thereby realizing the blind detection of the signal by using the Hopfield neural network.
Fig. 3, fig. 4 and fig. 5 are simulation experiment diagrams of the blind detection method based on M2M communication spectrum sharing and coexistence according to the present invention, respectively. The simulation transmission signal here is a QPSK signal, the fixed data length N is 320, and the noise is additive white gaussian noise, and all simulation results are obtained through 100 monte carlo experiments.
Fig. 3 is a graph comparing the error rates of a blind detection algorithm based on spectrum sharing and coexistence of M2M communications with a recovery signal algorithm assuming a known channel. The HNN (Hopfield neural network) algorithm is shown as a Hopfield neural network algorithm.
Fig. 4 shows four different neural networks under the blind detection method of the present invention based on M2M communication spectrum sharing and coexistence: error rate maps for classical Hopfield Neural Networks (HNN), Hopfield type Positive Feedback Neural networks (PFHNN), and Complex-System Hopfield type Neural networks (CSHNN).
Fig. 5 is a comparison of bit error rates for fixed transmission sequence lengths using two different classical channels under the blind detection algorithm of the present invention. Channel one: and adopting a synthesized channel with fixed weight and delay and no common zero point. And a second channel: and randomly synthesizing channels with variable weights and delay degrees.
Simulation experiments show that: compared with the algorithm for recovering the signal under the condition of assuming the known channel, the blind detection algorithm under the condition of the unknown channel has better error code performance under the same initial condition, can be well suitable for the classical channel, and successfully realizes the blind detection.
The above description is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, several modifications and variations can be made without departing from the technical principle of the present invention, and these modifications and variations should also be regarded as the protection scope of the present invention.
Claims (1)
1. The blind detection algorithm based on the sharing and coexistence of the M2M communication spectrum is characterized by comprising the following steps:
step SS 1: constructing a traditional user overcomplete model and an M2M communication sparse model;
step SS 2: solving the transmission signal of M2M by convex optimization;
step SS 3: constructing a traditional user receiving data matrix;
step SS 4: performing singular value decomposition on the received data matrix in the step SS 3;
step SS 6: selecting a Hopfield neural networkPerforming Hopfield neural network iterative operation on the activation function; the step SS1 specifically includes: a receiving end receives a signal sent by a Single traditional user and M2M devices, the signal is modulated and demodulated by OFDM, sampled and received by a Single-input Multi-output (SIMO) system receiving end, and a signal vector r (M) (r (M)),. once, r (M + N-1)]TThe reception equation:
where q is the oversampling factor, P represents the order of the channel, M is the number of M2M communication devices, si(m) is a transmission signal matrix, s when i is 00(m)=U0a0(m)=U0[a0(m),...,a0(m+N-1)]TSignalling matrix a for legacy users0(M) a vector-converted transmission matrix, i 1, …, M time si(m)=Uiai(m)=Ui[ai(m),...,ai(m+N-1)]TSignalling matrix a for M2M devicesi(m) vector-converted Transmission matrix, U0,UiAre all NxN vector transformation unitary matrixes; [ h ] ofi0,…,hiP]q×(P+1)Is the impulse response of the communication channel between the traditional user and the M2M device, and is the channel matrix under the condition that the assumed channel is known when the sampling factor q takes 1; vector v (m) is formed by a mean of zero and a variance ofThe additive white Gaussian noise component is formed, and a sending signal and the additive noise are independent; index function Ii(m) ═ 1 denotes that user/device I transmits in OFDM block m, and Ii(m) ═ 0 indicates that user/device i has not transmitted a signal; suppose UiAs is known at system design, neglecting the noise effect, the above equation can be written as follows:
RN=SiΓT;
Si=[sL+P(m),…,sL+P(m+N-1)]T=[sN(k),…,sN(m-P-L)]N×(L+P+1)is a transmit signal matrix; (R)N)N×(L+1)q=[rL(m),…,rL(m+N-1)]TIs the receive matrix of the legacy users; (gamma)(L+1)q×(L+P+1)Is formed byijJ is 0,1, …, P, where L is the equalizer parameter;
considering the similarity of data packets and the similarity of adjacent OFDM blocks, from ai(m) minus ai(m-1) enhancing sparsity using such similarity; order to
xi(m)=ai(m)-ai(m-1),i=1,…,M;
Suppose that all a have been addressedi(l) The block is analyzed and known, where l ≦ m-1, then only a needs to be detectedi(M), i ═ 0, …, M; thus subtracting the detected receiver-side signal from the received signal vector r (m)Obtaining an overcomplete model of the traditional user and an M2M communication sparse model:
the first term is an overcomplete model of the traditional user signal, the second term is a sparse model of the M2M communication signal, q is an oversampling factor, P represents the order of the channel, M is the number of M2M communication devices, and Ii(m)=1,UiAre all NxN vector transform unitary matrices, NxK0Matrix arrayIs U0Corresponding submatrix, (h)0j)q×1Is the impulse response of the conventional user signal, (h)ij)q×1Is the impulse function of M2M communication, and the vector v (M) is formed by the mean value of zero and the variance ofThe additive white Gaussian noise component is formed, and a sending signal and the additive noise are independent; the step SS2 specifically includes:
estimating x by solving for optimization using compressed sensing techniques0(m) and xi(m):
Adjusting the weighting factor lambdaiAnd | | | xi||0Matching, | | · | luminance0Is represented by0A norm;
one common practice for compressive sensing is to use a set of bumps l1Norm instead of l0Norm, which makes the above formula become:
Is obtained byThe M2M device transmission signal may then be estimated using the following equation:
where q is an oversampling factor, P represents the order of the channel, M is the number of M2M communication devices, si(m) is the moment of the transmitted signalWhen i is 0, s0(m)=U0a0(m)=U0[a0(m),...,a0(m+N-1)]TSignalling matrix a for legacy users0(M) a vector-converted transmission matrix, i 1, …, M time si(m)=Uiai(m)=Ui[ai(m),...,ai(m+N-1)]TSignalling matrix a for M2M devicesi(m) vector-converted Transmission matrix, U0,UiAre all NxN vector transformation unitary matrixes; [ h ] ofi0,…,hiP]q×(P+1)Is the impulse response of the communication channel between the legacy user and the M2M device, [ h ]i0,…,hiP]q×(P+1)Is the impulse response of the communication channel between the legacy user and the M2M device; vector v (m) is mean zero and varianceAdditive white gaussian noise of (1); index function Ii(m) ═ 1 denotes that user/device I transmits in OFDM block m, and Ii(m) ═ 0 denotes that user/device i does not transmit in OFDM; the step SS3 specifically includes: substituting the M2M transmission signal obtained in the step SS2 into the conventional user overcomplete model and the M2M communication sparse model to obtain the reception data of only the conventional user:
under the noiseless condition, the receiving end receives the signal sent by a single traditional user, and the receiving equation of the above formula obtained by sampling can be abbreviated as follows:
yN=SHT;
wherein, it is madeS=[sL+P(m),…,sL+P(m+N-1)]T=[sN(k),…,sN(m-P-L)]N×(L+P+1)Is a matrix of legacy user transmitted signals;(yN)N×(L+1)qIs the receive matrix of the legacy users; (H)(L+1)q×(L+P+1)is the channel impulse response h00,...h0P]q×(P+1)A constructed block Toeplitz matrix; (.)TRepresenting a matrix transposition;
wherein, the legacy user sends a signal matrix:
s=[sN(m),…,sN(m-P-L)]N×(L+P+1);
K0P + L, where P is the channel order, L is the equalizer order, and N is the required data length;
sL+P(m)=[s(m),…,s(m-L-P)]T(ii) a Wherein s belongs to { +/-1 +/-i }, and the time m is a natural number;
H0corresponding to H is the channel impulse response [ H (0), H (1) ]. H (P)]The block Toeplitz matrix formed:
p represents the order of the channel, L represents the order of the equalizer, and q is an oversampling factor and takes the value of a positive integer;
XN=[xL(m),…,xL(m+N-1)]Ta legacy user receiving data matrix that is N x (L +1) q,
wherein xL(m)=Γ·sL+P(m); the step SS4 specifically includes: singular value decomposition of a received data matrix:
in the formula (I), the compound is shown in the specification,
(·)His the Hermitian transpose;
u is nx (L + P +1) unitary matrix in singular value decomposition;
0 is an (N- (L + P +1)) × (L +1) q zero matrix;
v is (L +1) qx (L +1) q unitary matrix;
Ucis N × (N- (L + P +1)) unitary matrix;
d is a (L + P +1) × (L +1) q singular value array;
the performance function and optimization problem is constructed accordingly:
wherein s belongs to { +/-1 +/-i }NIs an N-dimensional vector, the character set belongs to +/-1 +/-i },an estimate value representing the signal; argmin () represents the variable value at which the target function takes the minimum value, d is the delay factor, d is 0, …, P + L; the step SS6 specifically includes:
for QPSK transmission signals, the complex Hopfield neural network state equation is:
an output equation: y (m) ═ ws (m);
wherein s ═ s1,s2,…,sN]A transmit vector for the neural network; w is the connection weight matrix of each neuron, and W is CN ×N,WHW; σ (-) is the activation function of the neural network; when the neural network reaches the final balance, approximately considering that s (m) ═ s (m +1) of each neuron, and s (m) is the acquired sending signal;
performing iterative operation on the state equation and the output equation, and substituting the result of each iteration into an energy function E (m) of the Hopfield neural network, when the energy function E (m) reaches a minimum value, namely s (m) ═ s (m +1), the Hopfield neural network reaches equilibrium, and the iteration is ended
Step SS6 specifically includes:
for adapting to complex channel and complex constellation signal QPSK, the activation function of the blind detection algorithm based on spectrum sharing and coexistence of M2M communication is:
σ(x)=tanh(Re(a·x))+itanh(Im(a·x));
wherein a is a gradient coefficient and mainly controls the gradient of the activation function;
the energy function based on blind detection of M2M communication spectrum coexistence and shared Hopfield neural network is expressed as:
wherein: e is the energy function of the network, the energy function is a variable related to the iteration time, and the weight matrix W satisfies that W is WHS (m) is the output of the neuron, σ-1(x) Is the inverse of the Sigmoid function σ (x) of the neuron.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910042795.5A CN109618311B (en) | 2019-01-17 | 2019-01-17 | Blind detection algorithm based on M2M communication spectrum sharing and coexistence |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910042795.5A CN109618311B (en) | 2019-01-17 | 2019-01-17 | Blind detection algorithm based on M2M communication spectrum sharing and coexistence |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109618311A CN109618311A (en) | 2019-04-12 |
CN109618311B true CN109618311B (en) | 2021-12-14 |
Family
ID=66019847
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910042795.5A Active CN109618311B (en) | 2019-01-17 | 2019-01-17 | Blind detection algorithm based on M2M communication spectrum sharing and coexistence |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109618311B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112257773B (en) * | 2020-10-19 | 2022-06-10 | 重庆邮电大学 | Mechanical equipment fault diagnosis method based on multiple measurement vectors of wireless sensor network |
CN112804049B (en) * | 2021-04-14 | 2021-07-06 | 南京信息工程大学 | Hexagonal CAP optical transmission method based on dynamic compression and multi-chaos encryption |
CN114362881B (en) * | 2021-12-16 | 2023-06-06 | 中国人民解放军战略支援部队信息工程大学 | Space-time block code blind identification method and system based on space-time correlation matrix |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106056141A (en) * | 2016-05-27 | 2016-10-26 | 哈尔滨工程大学 | Target recognition and angle coarse estimation algorithm using space sparse coding |
WO2016179303A1 (en) * | 2015-05-04 | 2016-11-10 | Kamama, Inc. | System and method of vehicle sensor management |
CN106462951A (en) * | 2014-06-10 | 2017-02-22 | 特拉维夫大学拉莫特有限公司 | Method and system for processing an image |
CN103916344B (en) * | 2014-03-07 | 2017-03-08 | 南京邮电大学 | Wireless sense network virtual multiple input multiple output signal blind checking method |
CN108768904A (en) * | 2018-04-19 | 2018-11-06 | 南京邮电大学 | Signal blind checking method based on the width facies pattern discrete hopfield neural network with disturbance |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9648444B2 (en) * | 2014-01-06 | 2017-05-09 | Brian G. Agee | Physically secure digital signal processing for wireless M2M networks |
-
2019
- 2019-01-17 CN CN201910042795.5A patent/CN109618311B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103916344B (en) * | 2014-03-07 | 2017-03-08 | 南京邮电大学 | Wireless sense network virtual multiple input multiple output signal blind checking method |
CN106462951A (en) * | 2014-06-10 | 2017-02-22 | 特拉维夫大学拉莫特有限公司 | Method and system for processing an image |
WO2016179303A1 (en) * | 2015-05-04 | 2016-11-10 | Kamama, Inc. | System and method of vehicle sensor management |
CN106056141A (en) * | 2016-05-27 | 2016-10-26 | 哈尔滨工程大学 | Target recognition and angle coarse estimation algorithm using space sparse coding |
CN108768904A (en) * | 2018-04-19 | 2018-11-06 | 南京邮电大学 | Signal blind checking method based on the width facies pattern discrete hopfield neural network with disturbance |
Non-Patent Citations (2)
Title |
---|
Abnormal signal detection in grid monitoring system based on overcomplete-basis independent component analysis;Tianshu Zhu,Jingshu Zhang;《2018 International Conference on Electronics Technology (ICET)》;20180527;全文 * |
基于压缩感知的跳频信号盲检测技术研究;吴俊;《中国优秀硕士学位论文全文数据库信息科技辑》;20131215;全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN109618311A (en) | 2019-04-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Chen et al. | Impulsive synchronization of reaction–diffusion neural networks with mixed delays and its application to image encryption | |
CN109618311B (en) | Blind detection algorithm based on M2M communication spectrum sharing and coexistence | |
US5461640A (en) | Method and system for optimizing an equalizer in a data transmission system | |
Hanna et al. | Signal processing-based deep learning for blind symbol decoding and modulation classification | |
CN103888391B (en) | Signal blind detection method based on double Sigmoid chaotic neural network | |
CN112637094A (en) | Multi-user MIMO receiving method based on model-driven deep learning | |
CN112491442B (en) | Self-interference elimination method and device | |
CN114826832B (en) | Channel estimation method, neural network training method, device and equipment | |
EP4262121A1 (en) | Neural network training method and related apparatus | |
US20030048861A1 (en) | Dynamic diversity combiner with associative memory model for recovering signals in communication systems | |
Hassan et al. | Performance Evaluation of Machine Learning‐Based Channel Equalization Techniques: New Trends and Challenges | |
TWI639319B (en) | Method, communication device and communication system for mimo communication | |
JP6980136B2 (en) | Communication system, communication method and storage medium | |
Azari et al. | Automated deep learning-based wide-band receiver | |
Feres et al. | Wirtinger flow meets constant modulus algorithm: Revisiting signal recovery for grant-free access | |
CN109391315B (en) | Data model dual-drive MIMO receiver | |
Li et al. | Blind source separation of signals with known alphabets using/spl epsi/-approximation algorithms | |
CN109688076B (en) | Blind detection method of noise chaotic neural network based on discrete multi-level hysteresis | |
CN116128070B (en) | Federal learning method based on wireless air calculation and multi-bit quantization compressed sensing | |
CN108574653B (en) | Signal blind detection method based on double Sigmoid hysteresis noise chaotic neural network | |
Vía et al. | Deterministic CCA-based algorithms for blind equalization of FIR-MIMO channels | |
Causey et al. | Blind multiuser detection using linear prediction | |
CN113852434B (en) | LSTM and ResNet-assisted deep learning end-to-end intelligent communication method and system | |
Diamantaras | Blind channel identification based on the geometry of the received signal constellation | |
CN113037409B (en) | Large-scale MIMO system signal detection method based on deep learning |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |