CN111526103A - PCMA signal single-channel blind separation method - Google Patents

PCMA signal single-channel blind separation method Download PDF

Info

Publication number
CN111526103A
CN111526103A CN201910106383.3A CN201910106383A CN111526103A CN 111526103 A CN111526103 A CN 111526103A CN 201910106383 A CN201910106383 A CN 201910106383A CN 111526103 A CN111526103 A CN 111526103A
Authority
CN
China
Prior art keywords
particles
signal
particle
channel
value
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.)
Granted
Application number
CN201910106383.3A
Other languages
Chinese (zh)
Other versions
CN111526103B (en
Inventor
张珊珊
陈刚
鲁华祥
邓琪
边昳
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Institute of Semiconductors of CAS
University of Chinese Academy of Sciences
Original Assignee
Institute of Semiconductors of CAS
University of Chinese Academy of Sciences
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Institute of Semiconductors of CAS, University of Chinese Academy of Sciences filed Critical Institute of Semiconductors of CAS
Priority to CN201910106383.3A priority Critical patent/CN111526103B/en
Publication of CN111526103A publication Critical patent/CN111526103A/en
Application granted granted Critical
Publication of CN111526103B publication Critical patent/CN111526103B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • 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
    • 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
    • H04L25/00—Baseband systems
    • H04L25/02—Details ; arrangements for supplying electrical power along data transmission lines
    • H04L25/0202—Channel estimation
    • H04L25/024—Channel estimation channel estimation algorithms
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L27/00—Modulated-carrier systems
    • H04L27/18—Phase-modulated carrier systems, i.e. using phase-shift keying
    • H04L27/20—Modulator circuits; Transmitter circuits
    • 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

Abstract

The invention discloses a PCMA signal single-channel blind separation method based on a genetic improved particle filter algorithm, which comprises the following steps: determining the value range and distribution of the parameters, and performing initial estimation on the signals and the channel parameters; generating a prediction signal according to the estimation parameters and the generated prediction symbol sequence by establishing a plurality of state distributions; calculating an evaluation result according to the similarity coefficient of the prediction signal and the real signal corresponding to each particle to obtain a particle probability value, and introducing selective cross operation in a genetic algorithm to generate new particles; sorting all the current particles according to the evaluation values to generate preferred particles; continuously optimizing particles by adding a dichotomy subsequently, and outputting a separated symbol sequence optimal value; the signal is segmented symbol estimated to form a closed loop to reduce computational complexity. The method has stronger robustness to the parameter estimation error of the modulation channel and has higher separation accuracy rate compared with the prior method.

Description

PCMA signal single-channel blind separation method
Technical Field
The invention relates to Paired Carrier Multiple Access (PCMA) signal processing, in particular to a PCMA signal single-channel blind separation problem based on a genetic improved particle filter algorithm.
Background
Paired Carrier Multiple Access (PCMA) is an emerging communication technology for increasing the utilization of satellite channel frequency bands. In the PCMA satellite communication system, two different ground stations share the uplink and downlink, so uplink signals transmitted by two satellite terminals occupy the same frequency band, and the two uplink signals overlap each other in the time domain. In some special occasions, information sequences sent by two communication parties need to be acquired from PCMA signals, for example, in military communication, the information needs to be intercepted by the identity of a third party, so that the traditional PCMA technology cannot be continuously used in non-cooperative communication.
From the viewpoint of signal separation, the acquisition of the PCMA signal is to separate two uplink signals from the PCMA signal received by a single sensor, i.e., the single-channel blind separation problem, because the two uplink signals constituting the PCMA signal have the same frequency, time slot and symbol rate, and the power of the two uplink signals is equivalent, the single-channel blind separation problem of the PCMA signal cannot be solved by using a traditional single-channel blind separation algorithm based on different powers, symbol rates, roll-off coefficients and the like. In the prior art, a single-channel blind separation problem of a communication signal is solved by using per-survivor path processing (PSP), but a PSP algorithm requires that channel response handling is set accurately, otherwise separation is easy to fall into a local minimum point, and the calculation complexity of the PSP algorithm is exponential relative to a channel order and a modulation order. In 2006, the particle filtering algorithm is formally applied to single-channel blind separation of mixed signals with the same frequency, but the complexity of the algorithm increases exponentially along with modulation orders, which is not beneficial to practical application. In the prior art, the weight updating step in the algorithm is improved, the algorithm performance is improved, but the complexity is high, in addition, a QRD-M Gibbs blind separation method is also provided in the prior art, but the application conditions are limited, the QRD-M Gibbs blind separation method can only be applied to blind separation signals sampled in the whole period, and the calculation accuracy rate is also required to be further improved.
Disclosure of Invention
Technical problem to be solved
Aiming at the problem of blind separation of PCMA signals in non-cooperative communication, a PCMA signal single-channel blind separation method based on a genetic improved particle filtering algorithm is provided, and the problem of joint estimation of parameters and symbol sequences of the PCMA signals is solved. The method can obviously reduce the calculated amount, has good channel acquisition and tracking capability and better symbol sequence estimation performance, and can recover the symbol sequences of two paths of uplink signals with lower error rate.
(II) technical scheme
The invention discloses a PCMA signal single-channel blind separation method, which comprises the following steps: determining the value range and distribution of parameters according to the received PCMA signal, and performing initial estimation on the signal and channel parameters, wherein the signal and channel parameters comprise the amplitude of two paths of uplink signals, residual carriers, symbol rate, initial phase and estimated value range of time delay; based on the algorithm framework of particle filtering, approximating to a true value posterior probability density by establishing a plurality of state distributions, then generating a prediction signal according to the estimation parameters and a generated prediction symbol sequence, and scattering points in a state space to update particles; calculating an evaluation result according to the similarity coefficient of the prediction signal and the real signal corresponding to each particle to obtain a particle probability value; adding a selective cross mechanism of a genetic algorithm to complete resampling of the particle filter; repeating the particle updating step on the particles before cross mutation and the newly generated particles, and carrying out local optimization; sorting all the current particles according to the evaluation values, reserving the preferred particles according to the thought of the winner eating the particles in the neural calculation, subsequently adding the dichotomy optimization particles, and outputting the separated symbol sequence preferred values; and the signals are subjected to segmented symbol estimation to form a closed loop, so that the computational complexity is reduced.
The step of establishing a plurality of state distributions and approximating a posterior probability density of a true value comprises the following steps: first, a plurality of state equations are established:
Figure BDA0001966676720000021
wherein t is Ng＝1，L，N，N＞1，NgIn order to optimize the number of iterations,
Figure BDA0001966676720000022
in order to estimate the sequence of symbols,
Figure BDA0001966676720000023
to estimate the channel parameters. Code element sequence
Figure BDA0001966676720000024
The phases are randomly extracted according to the equal probability of a constellation diagram of QPSK modulation; to increase the particle diversity and to combat the particle degradation phenomenon of conventional particle filters, a plurality of Gaussian state distributions are established
Figure BDA0001966676720000025
Wherein i is more than 1, and i is the number of state distributions. Channel parameters
Figure BDA0001966676720000026
Obey mean
Figure BDA0001966676720000027
Variance σ2Gaussian distribution of (a)2And setting the size according to the initial estimated parameter value.
The step of adding the selective cross mechanism of the genetic algorithm to complete the resampling of the particle filter comprises selection operation and cross operation, wherein the selection operation is to select the high-quality particles at the current moment as a parent set in the genetic operation, the cross operation is to generate offspring particles through the cross operation of the high-quality particles, and the genetic resampling is finished after new particles are generated circularly by utilizing the selection and cross steps.
The method for acquiring the particle probability value comprises the steps of analyzing the similarity between the predicted signal particles and the real signals, then obtaining the evaluation value of the predicted signal particles, meanwhile, after the evaluation value of the signal particles is acquired, sequencing the particles, comparing the sizes of the evaluation values according to the thought of the winner in the neural calculation, and reserving the preferred particles.
And subsequently, a dichotomy is required to be added to optimize the particles, wherein the dichotomy comprises the following steps: and setting a certain optimization interval, dividing the channel parameters into two parts to continuously approach a true value, setting iteration times, and finally reserving and outputting the preferred particles according to the particle evaluation values.
Finally, the signal is segmented and symbol-estimated to form a closed loop to reduce the computational complexity, wherein the method for segmenting and symbol-estimating the signal is that the channel parameter ξ of the preferred particle is output by the separation of the signal of the previous segmentk-1Channel parameters ξ for later stage signal separationkThe initial value of the previous section of signal is fed back to the previous section of signal, and the blind separation process of the previous section of signal is guided to form closed-loop iterative optimization particles.
(III) advantageous effects
The invention uses the improved particle filtering method and genetic algorithm to jointly estimate the parameters and the symbol sequence of the single-channel PCMA signal, models the PCMA signal into a single-channel mixed signal consisting of two paths of same-frequency digital modulation signals, and compared with the traditional PCMA signal blind separation algorithm such as QRD-M Gibbs, the method can obviously reduce the calculated amount, and the improved method discards the particles with poor quality, thereby reducing the calculation precision required by the algorithm when the algorithm is implemented, having good channel capturing and tracking capability and realizing better symbol sequence estimation performance.
Drawings
FIG. 1 is a flow chart of a PCMA signal single-channel blind separation method according to an embodiment of the present invention;
figure 2 is a schematic diagram of a PCMA communication link according to an embodiment of the present invention;
FIG. 3 is a diagram of PCMA signal separation performance according to an embodiment of the present invention;
FIG. 4 is a comparison graph of the separation performance of the method of the present invention and the QRD-M Gibbs algorithm;
FIG. 5 is a schematic diagram of a dichotomy in an embodiment of the invention
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail below with reference to the accompanying drawings in combination with specific experiments.
Fig. 1 is a flowchart of a PCMA signal single-channel blind separation method according to an embodiment of the present invention, which includes the following specific steps: determining the value range and distribution of parameters of a received PCMA signal, initially estimating the signal and channel parameters, establishing a plurality of state distributions based on an algorithm framework of particle filtering to approximate to a true value posterior probability density, generating a prediction signal by using initialized modulation channel parameters and randomly generated code element information according to an observation equation in an initialization stage, approximating an importance sampling function in standard particle filtering by establishing Gaussian distribution in an updating stage, extracting new particles in a state space, generating the prediction signal according to the estimation parameters and a generated prediction symbol sequence, calculating an evaluation result according to the similarity coefficient of the prediction signal corresponding to each particle and the true signal to obtain a particle probability value, introducing a selection cross operation in a genetic algorithm to generate the new particles, increasing the particle diversity, and avoiding the particle exhaustion phenomenon possibly existing in the resampling process, and repeating the particle updating step on the particles before the cross mutation and the newly generated particles, and performing local optimization. Sorting all the current particles according to the evaluation values, generating preferred particles according to the thought of a winner eating the particles in the neural calculation, adding a dichotomy to continuously optimize the particles subsequently, outputting the preferred values of the separated symbol sequences, and carrying out segmented symbol estimation on the signals to form a closed loop so as to reduce the calculation complexity.
FIG. 2 is a schematic diagram of PCMA communication according to an embodiment of the present invention, in which communication stations 1, 2 transmit uplink signals using the same frequency, time slot and spreading code
Figure BDA0001966676720000041
And
Figure BDA0001966676720000042
respectively correspond to each other after passing through a transparent satellite transponder
Figure BDA0001966676720000043
And
Figure BDA0001966676720000044
therefore, when the noise is ignored, both the communication stations 1 and 2 receive the downlink signals
Figure BDA0001966676720000045
The complex baseband form expression of the PCMA mixed signal is
Figure BDA0001966676720000046
In formula (1): y is1(t) and y2(t) uplink signals transmitted from two different satellite stations, h1And h2Corresponds to y1(t) and y2Instantaneous amplitude, Vf, of (t)1And Vf2Corresponds to y1(t) and y2(t) residual carrier frequency, [ phi ]1And phi2Corresponds to y1(,) and y2(t) initial phase, v (t) variance σ2Additive white gaussian noise, x1(t) and x2(t) is two paths of baseband signals, and the expression is as follows:
Figure BDA0001966676720000051
wherein the content of the first and second substances,
Figure BDA0001966676720000052
the value of the nth information symbol of the ith uplink signal is related to a baseband digital modulation mode; t is a symbol period; tau isiFor the channel transmission time delay of the ith uplink signal, satisfy T/2 ≦ tauiT/2 is less than or equal to; baseband signal x in embodiments of the present invention1(t) and x2(t) all adopt raised cosine roll-off forming, so the impulse response function g of the forming filter adopted by the two paths of baseband signalsi(t) ofThe expression is as follows:
Figure BDA0001966676720000053
wherein, αiIs the rise cosine roll off coefficient.
For received PCMA mixed signal according to sampling period TsThe sampling is performed with a duration of-L due to the finite length of the shaping filter in practice1T，L2T]，L＝L1+L2+1 wherein L1、L2The non-causal and causal periods of the equivalent filter, so that the baseband signal x at the moment k of the sampled i-th signali，kCan be written as:
Figure BDA0001966676720000054
the over-sampling model of the received mixed signal is:
Figure BDA0001966676720000055
in the formula: y isk＝y(kT)，hi，k＝hi(kT)，τi，k＝τi(kT)，vk＝v(kT)。
Particle filtering (particle filtering) is a sequential monte carlo method for solving nonlinear non-gaussian state estimation. The core idea of the method is to iteratively estimate the posterior probability density distribution of unknown parameters by establishing a recursive Bayes filter, approximate the state posterior distribution by using discrete sample points extracted from the state posterior distribution, and convert the integral operation into the summation operation.
The basic particle filter algorithm can be summarized as follows:
(1) particle initialization: initializing particles according to parameter value range and distribution
Figure BDA0001966676720000056
Sum weight
Figure BDA0001966676720000057
(2) Particle updating: sampling function and particle trajectory before time k according to importance
Figure BDA0001966676720000058
Updating to obtain new particles
Figure BDA0001966676720000059
(3) Weight updating and resampling
And then aiming at the problems of particle degradation in the particle filter algorithm and particle exhaustion in the resampling process, improving the particle filter algorithm process based on the particle filter algorithm framework.
In the prior art, channel parameter estimation methods include a cyclic accumulation method, a point-by-point capture method and the like, limited conditions exist in practical application, and a certain amount of calculation is added to a blind separation process. The method provided by the invention has stronger robustness to modulation channel parameter estimation errors, and in order to simplify the operation amount, the embodiment of the invention directly predicts the channel parameter range value of the received PCMA signal and sets an initial parameter value interval according to the information such as the waveform of the observed received PCMA signal.
In an exemplary embodiment of the present invention, the step of establishing the state space model and the particle update comprises assuming that the symbol sequence of the i-th information sequence after code interleaving is Ci，n(n is 1, 2..) the complex modulation sequence at the ith path k after constellation mapping is si，k(k ═ 1, 2.). Then, without taking coding into account, the single-channel blind separation of the PCMA signal is based only on the received signal y, without knowledge of the channel parameters1：k＝{y1，y1，…，ykRecovering the code element sequence sent by two different communication stations as Cn，k(n＝1，2，...)。
Obtained from the formulae (4) and (5):
Figure BDA0001966676720000061
the amplitude and time delay parameters can be considered to be constant within one symbol interval and have hi，k≡hi，τi，k≡τiThen the observation equation can be expressed as:
y′k＝g1，ks1，k+g2，ks2，k+vk(7)
order to
Figure BDA0001966676720000062
xk＝[s1，k，s2，k]Then, equation (7) can be abbreviated as:
yk＝f(ξk，xk)+vk(8)
the state transition equation is:
Figure BDA0001966676720000063
wherein t is Ng＝1，L，N，N＞1，NgIn order to optimize the number of iterations,
Figure BDA0001966676720000064
in order to estimate the symbol complex modulation sequence,
Figure BDA0001966676720000065
to estimate the channel parameters.
Figure BDA0001966676720000066
The phases are randomly extracted according to the equal probability of a constellation diagram of QPSK modulation; to increase the particle diversity and to combat the particle degradation phenomenon of conventional particle filters, a plurality of Gaussian state distributions are established
Figure BDA0001966676720000067
Wherein i is more than 1, and i is the number of state distributions. Channel parameters
Figure BDA0001966676720000071
Obey mean
Figure BDA0001966676720000072
Variance σ2Gaussian distribution of (a)2And setting the size according to the initial estimated parameter value.
Equation (8) is an observation equation, and equation (9) is a state transition equation, which together form a state space for single-channel blind separation of PCMA signals. And in the initialization stage, a prediction signal is generated by using the initialized modulation channel parameters and the randomly generated code element information according to the observation equation.
And then, in an updating stage, new particles are extracted in a state space by establishing an importance sampling function in Gaussian distribution approximation standard particle filtering to serve as the number of the particles, so that the purpose of quickly and iteratively estimating the preferred particles is achieved. And carrying out likelihood estimation on the prediction signal corresponding to the updated particle and the actual received signal, using the similarity coefficient as an evaluation value of the particle, and sequencing the particle according to the evaluation value.
Figure BDA0001966676720000073
In an exemplary embodiment of the present invention, aiming at the problem of particle depletion in the resampling process of the standard particle filter, the method proposed by the present invention uses genetic evolution operation to complete the resampling of the particle filter. The main idea is to regard the symbol sequence and unknown parameters as chromosome samples, regard the evaluation value corresponding to each sample as a fitness function, replace the resampling process in the particle filter algorithm with the selection and crossover operations of the genetic algorithm, make the child samples inherit and change the parent through the selection and crossover operations, and make the change of the child to the parent evolve toward the optimal direction (in the problem of symbol and parameter estimation, it is proceeding toward the direction of the globally optimal particle) through the size of the evaluation value corresponding to each staining sample.
In the method provided by the invention, the genetic resampling step is as follows:
step 1 is selection operation, firstly, sorting the current particles according to the evaluation value size, and selecting the current previous NeffParticles ofAs a set of parents in genetic manipulation
Figure BDA0001966676720000074
Step 2 is a cross operation, and filial generation particles are generated through the cross operation of high-quality particles. Firstly, randomly generating 1-Neff2 i, j in between, obtained from a set of good-quality particles
Figure BDA0001966676720000075
And
Figure BDA0001966676720000076
reuse of the formula (11, 12) with a crossover probability PcGeneration of 2 novel particles
Figure BDA0001966676720000077
And
Figure BDA0001966676720000078
Figure BDA0001966676720000079
Figure BDA00019666767200000710
and (4) finishing genetic resampling after circularly generating new particles by utilizing the selection and crossing steps. And repeating the particle updating step on the particles before the cross mutation and the newly generated particles to perform local optimization.
And after the operation, sorting all the current particles according to the evaluation value to generate the preferred particles.
In the example embodiment provided by the present invention, a subsequent optimization of the channel parameters, i.e., the continuous amount of preferred particles remaining ξ, is also requiredkAnd (3) carrying out dichotomy optimization, specifically, setting a certain optimization interval, carrying out dichotomy on the channel parameters to continuously approach a true value, setting iteration times, and finally reserving and outputting the preferred particles according to the particle evaluation values.
In the exemplary embodiment provided by the present invention, it is also necessary to perform a segment estimation on the received signal, and the channel parameters ξ of the preferred particles output by the separation of the signals of the previous segment are outputk-1Channel parameters ξ for later stage signal separationkThe initial value of the previous segment of signal is fed back by the blind separation result of the next segment of signal, the blind separation process of the previous segment of signal is guided, closed-loop continuous iteration optimization of the optimal particles is formed, the optimal separation accuracy is achieved, meanwhile, due to continuous approximation of channel parameters of the optimal particles, the operation amount of the subsequent code element updating particles is simplified, and the calculation complexity of the whole separation process is reduced.
Combining the analysis of the above parts, the genetic improved particle filter-based PCMA signal blind separation algorithm comprises the following steps:
step 1, preprocessing a received signal;
step 2, initializing modulation channel parameters and generating a prediction signal according to randomly generated code element information;
step 3, establishing a state space model and updating particles;
step 4, carrying out likelihood estimation on the prediction signal and the actual receiving signal, taking the similarity coefficient as an evaluation value of the particles, and sequencing the particles according to the evaluation value;
step 5, selecting high-quality particles at the current moment, and completing resampling through selective cross operation of a genetic algorithm;
step 6, channel parameter subsequent optimization;
and 7, carrying out segmented estimation on the received signals to form closed-loop iterative optimization optimized particles and improve the separation accuracy.
In an exemplary embodiment of the present invention, the theoretical derivation of the present algorithm was verified by simulation experiments, where for a single sensor received 2-way QPSK modulated PCMA hybrid signal, the channel noise is white gaussian noise in the simulation, and the carrier-to-noise ratio CNR is used as a measure of the noise magnitude. The roll-off coefficients of the forming filter and the matched filter are both 0.35, and 2 paths of signal amplitude h are set1＝1.0，h2＝0.8，f1＝-f2＝10-3T (T is symbol period), timing offsetτ1＝0.20T，τ20.40T, phase offset
Figure BDA0001966676720000091
In the range of [ - π, π]Internal random generation of particle number Np100, iteration number N g10, the equivalent channel order L is 7 (L)1＝L2＝3)。
Fig. 3 shows the performance of separating the PCMA mixed signal modulated by two QPSK channels when the carrier-to-noise ratio CNR is between 5dB and 23dB, using the average Signal Error Rate (SER) of the separation result of the 2 channels as the performance evaluation index under the given experimental conditions. When the carrier-to-noise ratio CNR is 5dB, the separation accuracy reaches 90%, when the carrier-to-noise ratio CNR is 11dB, the separation accuracy reaches 99%, and when the carrier-to-noise ratio CNR is 21dB, the separation accuracy reaches 99.9%.
As can be seen from fig. 3, under the condition of 2 times oversampling, the separation performance is better and better along with the increase of the carrier-to-noise ratio, and under the condition of low carrier-to-noise ratio, the method provided by the present invention can also maintain a higher separation accuracy.
Fig. 4 is a performance comparison of the QRD-M Gibbs separation algorithm, and for the problem that the application condition restriction and the separation accuracy of the QRD-M Gibbs blind separation algorithm in the actual PCMA signal blind separation process are to be improved, the present document introduces genetic evolution operation to resample the excellent particle set by establishing a plurality of state distributions to approach the posterior probability density of the true value, and performs segmented symbol estimation to form a closed loop, thereby improving the separation accuracy and reducing the algorithm computation amount.
Experiments compare the separation performance of the GA-PF and QRD-M Gibbs blind separation algorithm under different carrier-to-noise ratios, fig. 4 shows the comparison performance curve of the algorithm and the QRD-M Gibbs blind separation algorithm under 2-time oversampling, and as can be seen from fig. 4, under given conditions, the separation performance of the two algorithms is better and better along with the increase of the carrier-to-noise ratio, and for the QRD-M Gibbs blind separation algorithm, the BER is required to reach 10-2Order of magnitude, the carrier-to-noise ratio CNR reaches at least 15dB, and the BER is 10-3Order of magnitude, the carrier-to-noise ratio CNR reaches at least 18dB, and the BER is 10-4The order of magnitude, the carrier-to-noise ratio CNR reaches at least 23 dB; for the method provided by the invention, the BER is 10-2Order of magnitude, the carrier-to-noise ratio CNR reaches at least 11dB, and the BER is 10-3Order of magnitude, the carrier-to-noise ratio CNR reaches at least 17dB, and the BER is 10-4And the magnitude order, the carrier-to-noise ratio CNR reaches at least 21dB, and under the same experimental condition, compared with a QRD-M Gibbs separation algorithm, the separation performance of the method provided by the invention is improved by at least 0.3 magnitude order.
The above-mentioned embodiments are intended to illustrate the objects, technical solutions and advantages of the present invention in further detail, and it should be understood that the above-mentioned embodiments are only exemplary embodiments of the present invention and are not intended to limit the present invention, and any modifications, equivalents, improvements and the like made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims (8)

1. A PCMA signal single-channel blind separation method is characterized by comprising the following steps:
determining the value range and distribution of parameters according to the received PCMA signal, and performing initial estimation on the signal and channel parameters;
based on the algorithm framework of particle filtering, approximating to a true value posterior probability density by establishing a plurality of state distributions, generating a prediction signal according to an estimation parameter and a generated prediction symbol sequence, and scattering points in a state space to update particles;
calculating an evaluation result according to the similarity coefficient of the prediction signal and the real signal corresponding to each particle to obtain a particle probability value;
adding a selective cross mechanism of a genetic algorithm to complete resampling of the particle filter;
repeating the particle updating step on the particles before cross mutation and the newly generated particles, and carrying out local optimization;
sorting all the current particles according to the evaluation values, reserving the preferred particles according to the thought of the winner eating the particles in the neural calculation, subsequently adding the dichotomy optimization particles, and outputting the separated symbol sequence preferred values;
and the signals are subjected to segmented symbol estimation to form a closed loop, so that the computational complexity is reduced.
2. The method of claim 1, wherein the channel parameters include the estimated ranges of amplitude, residual carrier, symbol rate, initial phase and delay of the two uplink signals.
3. The method of claim 1, wherein the step of establishing a plurality of state distributions to approximate the true posterior probability density comprises:
first, a plurality of state equations are established:
Figure FDA0001966676710000011
wherein t is Ng＝1，L，N，N＞1，NgIn order to optimize the number of iterations,
Figure FDA0001966676710000012
in order to estimate the sequence of symbols,
Figure FDA0001966676710000013
to estimate the channel parameters.
Code element sequence
Figure FDA0001966676710000014
The phases are randomly extracted according to the equal probability of a constellation diagram of QPSK modulation; to increase the particle diversity and to combat the particle degradation phenomenon of conventional particle filters, a plurality of Gaussian state distributions are established
Figure FDA0001966676710000015
Wherein i is more than 1, and i is the number of state distributions. Channel parameters
Figure FDA0001966676710000016
Obey mean
Figure FDA0001966676710000017
Variance σ2Gaussian distribution of (a)2And setting the size according to the initial estimated parameter value.
4. The method of claim 1, wherein the selective cross-over mechanism added to the genetic algorithm performs the resampling step of the particle filter as follows:
step 1 is selection operation, namely selecting the high-quality particles at the current moment as a parent set in genetic operation:
firstly, sorting the current particles according to the weight value, and selecting the top NeffTaking the individual particle as the current set of the good-quality particles
Figure FDA0001966676710000021
Step 2 is a crossover operation, namely, progeny particles are generated through the crossover operation of the high-quality particles:
first, 1 to N are randomly generatedeff2 i, j in between, obtained from a set of good-quality particles
Figure FDA0001966676710000022
And
Figure FDA0001966676710000023
and then the cross probability P is given bycGeneration of 2 novel particles
Figure FDA0001966676710000024
And
Figure FDA0001966676710000025
Figure FDA0001966676710000026
Figure FDA0001966676710000027
and (4) finishing genetic resampling after circularly generating new particles by utilizing the selection and crossing steps.
5. The method of claim 1, wherein the particle probability value is an evaluation value of the predicted signal particle obtained by analyzing a similarity between the predicted signal particle and a real signal.
6. The method according to claim 5, wherein the evaluation values of the signal particles are used for sorting the particles, and the evaluation values are compared according to the idea of eating by winners in the neural calculation, so as to retain the preferred particles.
7. The method of claim 1, wherein the bisection method is to set a certain optimization interval, divide channel parameters into two to approach a true value continuously, set iteration times, and finally retain and output preferred particles according to particle evaluation values.
8. The method of claim 1, wherein the segmented symbol estimation is ξ of channel parameters of preferred particles output by previous segment signal separationk-1Channel parameters ξ for later stage signal separationkThe initial value of the previous section of signal is fed back to the previous section of signal, and the blind separation process of the previous section of signal is guided to form closed-loop iterative optimization particles.
CN201910106383.3A 2019-02-01 2019-02-01 PCMA signal single-channel blind separation method Active CN111526103B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910106383.3A CN111526103B (en) 2019-02-01 2019-02-01 PCMA signal single-channel blind separation method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910106383.3A CN111526103B (en) 2019-02-01 2019-02-01 PCMA signal single-channel blind separation method

Publications (2)

Publication Number Publication Date
CN111526103A true CN111526103A (en) 2020-08-11
CN111526103B CN111526103B (en) 2021-06-18

Family

ID=71900085

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910106383.3A Active CN111526103B (en) 2019-02-01 2019-02-01 PCMA signal single-channel blind separation method

Country Status (1)

Country Link
CN (1) CN111526103B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112613222A (en) * 2021-01-04 2021-04-06 重庆邮电大学 Improved particle filter-based inclination detection ionosphere MUF short-term prediction method

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110268168A1 (en) * 2009-12-10 2011-11-03 The Aerospace Corporation Methods and systems for increased communication throughput
CN104320362A (en) * 2014-09-29 2015-01-28 电子科技大学 Method for PCMA signal blind separation
CN105791183A (en) * 2016-03-16 2016-07-20 重庆邮电大学 PCMA signal single-channel blind separation method based on improved particle filter

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110268168A1 (en) * 2009-12-10 2011-11-03 The Aerospace Corporation Methods and systems for increased communication throughput
CN104320362A (en) * 2014-09-29 2015-01-28 电子科技大学 Method for PCMA signal blind separation
CN105791183A (en) * 2016-03-16 2016-07-20 重庆邮电大学 PCMA signal single-channel blind separation method based on improved particle filter

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
YONG YANG ET AL: "Single-channel blind source separation for", 《IET SIGNAL PROCESSING》 *
付卫红等: "单通道盲源分离的研究现状与展望", 《北京邮电大学学报》 *
叶飞等: "基于独立分量分析的PCMA信号盲分离算法 ", 《电视技术》 *
路威等: "时变幅度LFM干扰下的单通道盲分离算法研究 ", 《信号处理》 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112613222A (en) * 2021-01-04 2021-04-06 重庆邮电大学 Improved particle filter-based inclination detection ionosphere MUF short-term prediction method
CN112613222B (en) * 2021-01-04 2023-09-15 重庆邮电大学 Method for short-term prediction of ionized layer MUF by tilt detection based on improved particle filtering

Also Published As

Publication number Publication date
CN111526103B (en) 2021-06-18

Similar Documents

Publication Publication Date Title
CN108650201B (en) Neural network-based channel equalization method, decoding method and corresponding equipment
Güner et al. Automatic digital modulation classification using extreme learning machine with local binary pattern histogram features
CN109657604B (en) Satellite amplitude-phase signal identification and demodulation method and device based on cyclic neural network
CN111464465B (en) Channel estimation method based on integrated neural network model
CN109246039B (en) Soft information iteration receiving method based on bidirectional time domain equalization
CN107038421A (en) Modulation Types recognition methods based on sparse storehouse own coding
CN109274624B (en) Carrier frequency offset estimation method based on convolutional neural network
CN103117964B (en) Signal detecting method and device under a kind of 60GHz millimeter-wave communication systems
CN109672639B (en) Signal demodulation method based on machine learning
CN109951214B (en) Signal detection method suitable for large-scale MIMO system
Zhang et al. A data preprocessing method for automatic modulation classification based on CNN
CN112887239B (en) Method for rapidly and accurately identifying underwater sound signal modulation mode based on deep hybrid neural network
CN109412725B (en) Radio communication PCMA signal blind demodulation method and device
CN107947761B (en) Variable threshold value proportion updating self-adaptive filtering method based on fourth order of least mean square
CN112910812A (en) Modulation mode identification method for deep learning based on space-time feature extraction
CN111526103B (en) PCMA signal single-channel blind separation method
CN112215335A (en) System detection method based on deep learning
CN104980202B (en) Uplink detection method based on MCMC technologies in a kind of extensive mimo system
CN107534622A (en) Merge the joint identification of signal in non-cooperation digital telecommunication
CN114499601A (en) Large-scale MIMO signal detection method based on deep learning
CN110233808B (en) FTN system signal detection method
CN109167744B (en) Phase noise joint estimation method
CN116405158A (en) End-to-end communication system based on deep learning under non-Gaussian noise
WO2023121659A1 (en) Communication device and method for performing communication signal processing
Fang et al. Non-cooperative MPSK Modulation Identification in SIMO Underwater Acoustic Multipath Channel

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