CN111200470B - High-order modulation signal transmission control method suitable for being interfered by nonlinearity - Google Patents

High-order modulation signal transmission control method suitable for being interfered by nonlinearity Download PDF

Info

Publication number
CN111200470B
CN111200470B CN202010027487.8A CN202010027487A CN111200470B CN 111200470 B CN111200470 B CN 111200470B CN 202010027487 A CN202010027487 A CN 202010027487A CN 111200470 B CN111200470 B CN 111200470B
Authority
CN
China
Prior art keywords
signal
neural network
qam
bit
qam signal
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
Application number
CN202010027487.8A
Other languages
Chinese (zh)
Other versions
CN111200470A (en
Inventor
姜明
邹龙浩
赵春明
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.)
Southeast University
Original Assignee
Southeast University
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 Southeast University filed Critical Southeast University
Priority to CN202010027487.8A priority Critical patent/CN111200470B/en
Publication of CN111200470A publication Critical patent/CN111200470A/en
Application granted granted Critical
Publication of CN111200470B publication Critical patent/CN111200470B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • 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
    • H04L27/3405—Modifications of the signal space to increase the efficiency of transmission, e.g. reduction of the bit error rate, bandwidth, or average power

Abstract

The invention relates to a method suitable for a computerThe transmission control method of the linear interference high-order modulation signal realizes the demodulation function through the neural network to reduce the nonlinear interference caused by the power amplifier to the signal, thereby recovering the signal of the transmitting end, wherein the cost function of the neural network adopts a cross entropy function; the training method adopts a back propagation algorithm; the real part and the imaginary part of the signal after passing through the matched filter and the corresponding power back-off coefficient are used as the input of the neural network, and the output result of the neural network representsQAMA probability value of 1 for each bit in the signal; and the activation functions of the hidden layer and the output layer of the neural network are respectively adoptedtanhFunction sumsigmoidA function; compared with the demodulation algorithm of the high-order modulation signal in the existing nonlinear scene, the whole design scheme not only improves the performance, but also reduces the complexity of the algorithm.

Description

High-order modulation signal transmission control method suitable for being interfered by nonlinearity
Technical Field
The invention relates to a high-order modulation signal transmission control method suitable for being interfered by nonlinearity, and belongs to the technical field of mobile communication.
Background
With the increasing number of mobile users, the demand for communication technology is becoming higher and higher. In the past 40 years or so, mobile communication systems have evolved from the first generation mobile communication system (1G) of the last 70 th century to the fifth generation mobile communication system (5G) which is currently in widespread interest. The current research in the industry on 5G wireless access technology mainly centers on three application scenarios of 5G: the method comprises the following steps of enhancing mobile broadband (eMBB), massive machine type communication (mMTC) and ultra-high-reliability low-delay communication (URLLC), wherein the enhancing of the mobile broadband is the comprehensive enhancement of data transmission rate, time delay and user capacity in 4G at present, the massive machine type communication is used in the field of Internet of things, and the ultra-high-reliability low-delay communication is used in the fields of Internet of vehicles and the like. In order to realize the above scenario, the bandwidth and transmission rate required by 5G will be much greater than 4G, and compared with 4G, the transmission rate of 5G will be up to 1Gb/s, and in order to achieve such a high transmission rate, the most direct method is to use high-order QAM modulation. The high-order QAM modulation has the advantages of greatly improving the transmission rate of the communication system and increasing the utilization rate of the frequency band, however, the high-order QAM modulated signal is easily subjected to nonlinear interference during the transmission process, especially the nonlinear interference caused by the power amplifier to the signal, so that the signal may be distorted.
Around this technology, various technologies for overcoming the nonlinear interference of the power amplifier to the signal are proposed at home and abroad, and these technologies can be generally divided into a predistortion compensation technology and a post-distortion compensation technology. One common predistortion compensation technique is the Digital Predistortion (DPD) technique, which basically corrects the baseband signal input to the power amplifier according to the nonlinear characteristics of the power amplifier, and eliminates the nonlinear interference on the signal as much as possible. Due to the different non-linear characteristics of different power amplifiers, it is inefficient to perform linearization correction on each power amplifier in practical engineering applications. Different from the predistortion compensation technology, the post-distortion compensation technology can eliminate the interference of nonlinearity to signals as much as possible at a receiving end of a communication system through a specific demodulation algorithm, and the complexity of the realization of the current post-distortion compensation technology is high.
Disclosure of Invention
The technical problem to be solved by the present invention is to provide a transmission control method for high-order modulation signals subjected to nonlinear interference, which can reduce the nonlinear interference caused by a power amplifier to the signals, thereby improving the transmission capability of a communication system
The invention adopts the following technical scheme for solving the technical problems: the invention designs a high-order modulation signal transmission control method suitable for being interfered by nonlinearity, which is used for aiming at binary bit stream X { X1,x2,...,xNRealizing transmission control from a transmitting end to a receiving end; the method comprises the following steps:
step A, the transmitting end is according to the modulation order M, and M is log2M, obtaining the bit number M in single QAM signal
Figure BDA0002362989910000021
Obtaining a group number K, and then entering a step B; wherein the content of the first and second substances,
Figure BDA0002362989910000022
represents rounding up, N represents the number of bits in the binary bitstream X;
and step B, the transmitting end carries out sequential grouping on each bit in the binary bit stream X by taking the bit number in a single QAM signal as M to obtain K QAM signals with the modulation order of M to form an M-QAM signal S { S1,...,sk,...,sK}，skRepresenting the kth QAM signal in the M-QAM signal S; if the bit number in the last QAM signal in the sequence is less than m, then 0 is supplemented at the end until the bit number m is met, and then the step C is carried out;
step C, the transmitting end aims at the M-QAM signal S { S1,...,sk,...,sKApply a multiplier, by vk＝G×skRespectively processing and updating each QAM signal to obtain a signal V { V }1,...,vk,...,vK}，vkRepresenting the kth QAM signal in the signal V, and sending the signal to a receiving end by a transmitting end, wherein G represents a preset power back-off coefficient, and then entering step D;
d, the receiving end receives the signal from the transmitting end, the real part, the imaginary part and the power back-off coefficient G of each QAM signal in the received signal are used as the input of a trained neural network, the neural network processes each QAM signal to respectively obtain the probability that each bit in each QAM signal is 1, and then the step E is carried out;
e, the receiving end judges whether the probability of the bit being 1 is not less than a preset probability threshold value aiming at the processing result of the neural network and aiming at each bit in each QAM signal, if so, the value of the bit is judged to be equal to 1; otherwise, judging the value of the bit to be 0; after the judgment of each bit in each QAM signal is completed, sequencing the values of all the bits according to the sequence of each QAM signal and the sequence of each bit in the QAM signal, namely obtaining a result binary bit stream received by a receiving end; and B, if the operation of complementing 0 aiming at the last QAM signal in the sequence exists in the step B, deleting the last corresponding digit bit in the result binary bit stream.
As a preferred technical scheme of the invention: in the step B, the transmitting end sequentially groups each bit in the binary bit stream X by using the bit number m in a single QAM signal and applying a constellation mapping method.
As a preferred technical scheme of the invention: further comprising a step CD-1 of obtaining a signal V { V } in said step C as follows1,...,vk,...,vKAfter that, entering the step CD-1;
step CD-1. the transmitting end aims at the signal V { V1,...,vk,...,vK}, applying a square root raised cosine filter, passing u throughk＝Ι(vk) Respectively processing and updating each QAM signal to obtain a signal U { U }1,...,uk,...,uK}，ukRepresenting the kth QAM signal in the signal U, and sending the signal to the receiving end by the transmitting end, where i (·) represents the transform function of the square root raised cosine filter on the input signal, and then entering step D.
As a preferred technical scheme of the invention: further comprising a step CD-2 of obtaining a signal U { U } in said step CD-1 as follows1,...,uk,...,uKAfter that, the step CD-2 is carried out;
step CD-2. the transmitting end aims at the signal U { U }1,...,uk,...,uKApplying a power amplifier, passing through tk＝PA(uk) Respectively processing and updating each QAM signal to obtain a signal T { T }1,...,tk,...,tK}，tkRepresenting the kth QAM signal in the signal T and transmitting the signal from the transmitting end to the receiving end, wherein PA (-) represents the transformation function of the power amplifier to the input signal, and then proceeding to step D.
As a preferred technical solution of the present invention, the step D includes the steps of:
step D1. the receiving end receives the signal from the transmitting end and aims at the obtained signal R { R1,...,rk,...,rKApplying a matched filter, passing yk＝Ι(rk) Respectively processing and updating each QAM signal to obtain a signal Y { Y1,...,yk,...,yKThen step D2 is entered; wherein the transformation function in the matched filter and the transformation function in the square root raised cosine filterThe same number, rkRepresenting the k-th QAM signal, y, of the signal RkRepresenting the kth QAM signal in signal Y.
Step D2. the receiver side targets the signal Y Y1,...,yk,...,yKUsing each QAM signal y in the signalkThe real part, the imaginary part and the power back-off coefficient G are used as the input of a trained neural network, the neural network processes each QAM signal to respectively obtain each QAM signal ykThe probability that each bit is 1, and then step E is entered.
As a preferred technical scheme of the invention: the neural network is a fully-connected neural network with L hidden layers, wherein the number of neurons in each hidden layer is Z; output a of the l hidden layer in neural network applicationslOutput a of the l-1 hidden layer of the neural networkl-1The relationship between them is:
Figure BDA0002362989910000031
in the formula (I), the compound is shown in the specification,
Figure BDA0002362989910000032
representing the weight value from the ith neuron in the l-1 hidden layer to the jth neuron in the l hidden layer in the neural network;
Figure BDA0002362989910000033
representing the bias of the ith neuron in the ith hidden layer in the neural network, wherein the bias of each neuron in the same hidden layer in the neural network is the same as that of each neuron in the first hidden layer in the neural network; f (-) represents the activation function of the hidden layer in the neural network.
As a preferred technical solution of the present invention, the neural network training phase includes the following steps:
step I, aiming at the sample binary bit stream, obtaining a training sample psi through steps A to D, wherein the training sample psi comprises a signal Y ' { Y ' of a receiving end after passing through a matched filter '1,...,y′k,...,y′KAnd a power back-off coefficient G, thereby determining the input layer of the neural networkThe number of nodes is equal to 3 and is used for respectively receiving QAM signals y'kThe real part, the imaginary part and the power back-off coefficient G, and then entering step II;
step II, initializing the method by applying a xavier initialization method
Figure BDA0002362989910000041
And
Figure BDA0002362989910000042
and the bias of each neuron element in the same hidden layer in the neural network is the same with each other, and is respectively corresponding to each QAM signal Y 'in the signal Y'kThe input layers of the neural network receive QAM signals y'kAnd a power back-off coefficient G, and then obtaining the QAM signal y 'by the neural network output layer according to the number of nodes of the neural network output layer being equal to m'kThe probability that each bit in the sequence is 1; thus, the training of the neural network is realized through the steps A to E, the number of hidden layers in the neural network is determined to be L, the number of neurons of each hidden layer is determined to be Z, and the optimization is carried out
Figure BDA0002362989910000043
And
Figure BDA0002362989910000044
as a preferred technical scheme of the invention: the cost function in a cross entropy form is adopted in the neural network, and iterative optimization is carried out on the cost function in the neural network through a back propagation algorithm
Figure BDA0002362989910000045
And
Figure BDA0002362989910000046
as a preferred technical scheme of the invention: the activation function of the hidden layer in the neural network is a tanh function, and the activation function of the output layer is a sigmoid function.
Compared with the prior art, the transmission control method of the high-order modulation signal suitable for being interfered by the nonlinearity has the following technical effects by adopting the technical scheme:
the invention designs a high-order modulation signal transmission control method suitable for being interfered by nonlinearity, which realizes the demodulation function through a neural network so as to reduce the nonlinearity interference caused by a power amplifier to the signal and recover the signal of a transmitting end, wherein the cost function of the neural network adopts a cross entropy function; the training method adopts a back propagation algorithm; the real part and the imaginary part of the signal after passing through the matched filter and the corresponding power back-off coefficient are used as the input of a neural network, and the output result of the neural network represents the probability value that each bit in the QAM signal is 1; the activation functions of the hidden layer and the output layer of the neural network respectively adopt a tanh function and a Sigmoid function; compared with the demodulation algorithm of the high-order modulation signal in the existing nonlinear scene, the whole design scheme not only improves the performance, but also reduces the complexity of the algorithm.
Drawings
FIG. 1 is a system diagram of a high order modulation signal transmission control method suitable for non-linear interference according to the present invention;
FIG. 2 is a block diagram of a neural network in accordance with the present invention;
FIG. 3 is a graph of BER simulation of the first embodiment of the present invention;
fig. 4 is a graph showing BER simulation in the second embodiment of the present invention.
Detailed Description
The following description will explain embodiments of the present invention in further detail with reference to the accompanying drawings.
The invention designs a high-order modulation signal transmission control method suitable for being interfered by nonlinearity, which is used for aiming at binary bit stream X { X1,x2,...,xNRealizing transmission control from a transmitting end to a receiving end; as shown in fig. 1, the method includes the following steps a to E.
Step A, the transmitting end is according to the modulation order M, and M is log2M, obtaining the bit number M in single QAM signal
Figure BDA0002362989910000051
Obtaining a group number K, and then entering a step B; wherein the content of the first and second substances,
Figure BDA0002362989910000052
indicating rounding up and N indicates the number of bits in the binary bit stream X.
And step B, the transmitting terminal uses the bit number in a single QAM signal as M and applies a constellation mapping method to sequentially group each bit in the binary bit stream X to obtain K QAM signals with the modulation order of M to form an M-QAM signal S { S }1,...,sk,...,sK}，skRepresenting the kth QAM signal in the M-QAM signal S; if the bit number in the last QAM signal in the sequence is less than m, then 0 is supplemented at the end until the bit number m is met, and then the step C is carried out.
Step C, the transmitting end aims at the M-QAM signal S { S1,...,sk,...,sKApply a multiplier, by vk＝G×skRespectively processing and updating each QAM signal to obtain a signal V { V }1,...,vk,...,vK}，vkRepresenting the kth QAM signal in the signal V, and then entering step CD-1; wherein G represents a preset power back-off coefficient.
Step CD-1. the transmitting end aims at the signal V { V1,...,vk,...,vK}, applying a square root raised cosine filter, passing u throughk＝Ι(vk) Respectively processing and updating each QAM signal to obtain a signal U { U }1,...,uk,...,uK}，ukRepresenting the kth QAM signal in the signal U, and then entering the step CD-2; where i (-) denotes the transform function of the square root raised cosine filter on the input signal.
Step CD-2. the transmitting end aims at the signal U { U }1,...,uk,...,uKApplying a power amplifier, passing through tk＝PA(uk) Respectively processing and updating each QAM signal to obtain a signal T { T }1,...,tk,...,tK}，tkRepresenting the kth QAM signal in the signal T, and then proceeds to step D1;where PA (-) represents the transfer function of the power amplifier to the input signal.
Step D1. the receiving end receives the signal from the transmitting end and aims at the obtained signal R { R1,...,rk,...,rKApplying a matched filter, passing yk＝Ι(rk) Respectively processing and updating each QAM signal to obtain a signal Y { Y1,...,yk,...,yKThen step D2 is entered; wherein the transform function in the matched filter is the same as the transform function in the square-root raised cosine filter, rkRepresenting the k-th QAM signal, y, of the signal RkRepresenting the kth QAM signal in signal Y.
Step D2. As shown in FIG. 2, the receiving end targets the signal Y { Y }1,...,yk,...,yKUsing each QAM signal y in the signalkThe real part, the imaginary part and the power back-off coefficient G are used as the input of a trained neural network, the neural network processes each QAM signal to respectively obtain each QAM signal ykThe probability that each bit is 1, and then step E is entered.
In the specific application of the neural network, as shown in fig. 2, the neural network is a fully-connected neural network with L hidden layers, where the number of neurons in each hidden layer is Z; output a of the l hidden layer in neural network applicationslOutput a of the l-1 hidden layer of the neural networkl-1The relationship between them is:
Figure BDA0002362989910000061
in the formula (I), the compound is shown in the specification,
Figure BDA0002362989910000062
representing the weight value from the ith neuron in the l-1 hidden layer to the jth neuron in the l hidden layer in the neural network;
Figure BDA0002362989910000063
representing the bias of the ith neuron in the first hidden layer in the neural network, and the phases in the neural networkThe bias of each neuron in the hidden layer is the same with each other; f (-) represents the activation function of the hidden layer in the neural network.
And for the training of the neural network, the following steps I to II are included.
Step I, aiming at the sample binary bit stream, obtaining a training sample psi through steps A to D, wherein the training sample psi comprises a signal Y ' { Y ' of a receiving end after passing through a matched filter '1,...,y'k,...,y'KAnd a power backoff factor G, whereby it is determined that the number of input layer nodes of the neural network is equal to 3, for receiving the QAM signals y'kThe real part, the imaginary part, and the power back-off coefficient G, and then step II.
Step II, initializing the method by applying a xavier initialization method
Figure BDA0002362989910000064
And
Figure BDA0002362989910000065
and the bias of each neuron element in the same hidden layer in the neural network is the same with each other, and is respectively corresponding to each QAM signal Y 'in the signal Y'kThe input layers of the neural network receive QAM signals y'kAnd a power back-off coefficient G, and then obtaining the QAM signal y 'by the neural network output layer according to the number of nodes of the neural network output layer being equal to m'kThe probability that each bit in the sequence is 1; thus, the training of the neural network is realized through the steps A to E, the number of hidden layers in the neural network is determined to be L, the number of neurons of each hidden layer is determined to be Z, and the iterative optimization of the neural network through a back propagation algorithm
Figure BDA0002362989910000071
And
Figure BDA0002362989910000072
in practical application, a cost function in a cross entropy form is adopted in the neural network, an activation function of a hidden layer in the neural network is a tanh function, and an activation function of an output layer is a sigmoid function.
E, the receiving end judges whether the probability of the bit being 1 is not less than a preset probability threshold value or not aiming at each bit in each QAM signal aiming at the processing result of the neural network, wherein in the practical application, the probability threshold value is set to be 0.5, and if yes, the value of the bit is judged to be equal to 1; otherwise, judging the value of the bit to be 0; after the judgment of each bit in each QAM signal is completed, sequencing the values of all the bits according to the sequence of each QAM signal and the sequence of each bit in the QAM signal, namely obtaining a result binary bit stream received by a receiving end; and B, if the operation of complementing 0 aiming at the last QAM signal in the sequence exists in the step B, deleting the last corresponding digit bit in the result binary bit stream.
The method is applied to the reality, wherein in the first embodiment, a communication system simulation platform with nonlinear interference in the transmission process needs to be established, and the main physical layer parameters of the communication system simulation platform are shown in the following table 1.
Figure BDA0002362989910000073
TABLE 1
As shown in FIG. 1, the invention is applied to a neural network demodulation method of a high-order modulation signal subjected to nonlinear interference, and a binary bit stream X { X }1,x2,...,xN9220, i.e., N9220; the execution steps are as follows.
In step a, the modulation order M is set to 1024, and the number M of bits in a single QAM signal is 10 and the number K of packets is 922 are obtained.
In step B, the transmitting end applies a constellation mapping method to sequentially group each bit in the binary bit stream X to obtain 922 QAM signals to form a 1024-QAM signal S { S1,...,sk,...,sK}；
In step C, the transmitting terminal aims at 1024-QAM signal S { S1,...,sk,...,sKApply a multiplier, by vk＝G×skRespectively processing and updating each QAM signal to obtain a signal V { V }1,...,vk,...,vKAnd the power back-off coefficient G is selected according to different conditions, and needs to be set to be 0dB, -2dB, -4dB and-6 dB when the neural network is trained.
In step CD-1, the transmitting end is directed to the signal V { V }1,...,vk,...,vK}, applying a square root raised cosine filter, passing u throughk＝Ι(vk) Respectively processing and updating each QAM signal to obtain a signal U { U }1,...,uk,...,uKIn practical application, the roll-off coefficient of the square root raised cosine filter is 0.25, the number of taps is 201, and the expression of the square root raised cosine filter is as follows:
Figure BDA0002362989910000081
wherein v iskRepresenting the k-th QAM signal, u, of the signal VkRepresenting the k-th QAM signal, h, of the signal UkDenotes the filter coefficients and C denotes the length of the filter, which in this embodiment has a value of 201.
In step CD-2, the transmitting end is directed to the signal U { U }1,...,uk,...,uKApplying Rapp model power amplifier, specifically by tk＝PA(uk) Respectively processing and updating each QAM signal to obtain a signal T { T }1,...,tk,...,tK}。
Figure BDA0002362989910000082
Wherein PA (-) represents the transfer function of the power amplifier to the input signal, AinRepresenting the input signal ukAmplitude of (d), tkRepresenting the input signal ukOutput signals v, A after passing through Rapp model of power amplifier0And p takes the values of 1, 1 and 2 respectively.
In step D1, the receiving end receives the signal from the transmitting end and aims at the obtained signal R { R }1,...,rk,...,rKApplying a matched filter, passing yk＝Ι(rk) Respectively processing and updating each QAM signal to obtain a signal Y { Y1,...,yk,...,yKThen step D2 is entered; wherein the transform function in the matched filter is the same as the transform function in the square root raised cosine filter.
In step D2, Y { Y }1,y2,...,yKThe power back-off coefficient G and the corresponding power back-off coefficient G are used as the input of a neural network, and the number of output layer nodes of the neural network is equal to the QAM signal skThe bit number m carried on it is 10, the output result of the neural network represents QAM signal skProbability value P { d of 1 in each biti＝1|yk}。
In step E, when the probability value P { d }i＝1|ykSatisfy P { d }i＝1|ykIf the rate is greater than or equal to 0.5, the bit d is setiThe value of (1); when probability value P { di＝1|ykSatisfy P { d }i＝1|ykIf the rate is less than 0.5, the bit d is setiThe value of (1) is judged to be 0; after the above-mentioned judgment of each bit in each QAM signal is completed, the values of all bits are sorted according to the sequence of each QAM signal and the sequence of each bit in the QAM signal, that is, the resulting binary bit stream received by the receiving end is obtained.
It should be noted that, in the implementation of the above steps a to E, when training the neural network, the training sample Ψ is obtained through steps a to D, and includes the signal Y' { Y, which is obtained after the receiving end passes through the matched filter1',...,y'k,...,y'KSetting G as 0dB, -2dB, -4dB and-6 dB, obtaining signals Y 'under different power back-off coefficient G conditions through steps A to E, taking the signals Y' and the corresponding power back-off coefficient G as training samples psi, determining the number L of hidden layers of the neural network as 6, the number Z of neurons in each hidden layer as 64, initializing the weight and bias of the neural network, and selecting x in the initialization modeInitializing an avier, training the neural network through the steps A to E, setting the training times to be 600000, and selecting an Adam algorithm by a gradient descent method of a back propagation algorithm of the neural network.
Compared with the conventional demodulation algorithm under the existing nonlinear scene as a comparison object, the superiority of the neural network demodulation method is evaluated from the perspective of the Bit Error Rate (BER) of the system, as shown in fig. 3, the abscissa represents the normalized output power, the performance of the neural network demodulation method is basically the same as that of the existing best conventional demodulation algorithm, but the complexity of the algorithm is reduced, and the complexity of the neural network demodulation method in the embodiment is compared with that of the conventional demodulation algorithm, as shown in table 2 below.
Figure BDA0002362989910000091
TABLE 2
The method for controlling transmission of the high-order modulation signal subjected to the nonlinear interference is applied to practice, wherein in the second embodiment, a communication system simulation platform with the nonlinear interference in the transmission process needs to be set up, and the main physical layer parameters of the communication system simulation platform are shown in the following table 3.
Figure BDA0002362989910000092
Figure BDA0002362989910000101
TABLE 3
As shown in FIG. 1, the invention is applied to a neural network demodulation method of a high-order modulation signal subjected to nonlinear interference, and a binary bit stream X { X }1,x2,...,xN9220, i.e., N9220; the execution steps are as follows.
In step a, the modulation order M is set to 1024, and the number M of bits in a single QAM signal is 10 and the number K of packets is 922 are obtained.
In step B, the transmitting end applies a constellation mapping method to sequentially group each bit in the binary bit stream X to obtain 922 QAM signals to form a 1024-QAM signal S { S1,...,sk,...,sK}；
In step C, the transmitting terminal aims at 1024-QAM signal S { S1,...,sk,...,sKApply a multiplier, by vk＝G×skRespectively processing and updating each QAM signal to obtain a signal V { V }1,...,vk,...,vKAnd the power back-off coefficient G is selected according to different conditions, and needs to be set to be 0dB, -3dB, -6dB, -9dB, -12dB, -14dB and-16 dB when the neural network is trained.
In step CD-1, the transmitting end is directed to the signal V { V }1,...,vk,...,vK}, applying a square root raised cosine filter, passing u throughk＝Ι(vk) Respectively processing and updating each QAM signal to obtain a signal U { U }1,...,uk,...,uKIn practical application, the roll-off coefficient of the square root raised cosine filter is 0.25, the number of taps is 201, and the expression of the square root raised cosine filter is as follows:
Figure BDA0002362989910000102
wherein v iskRepresenting the k-th QAM signal, u, of the signal VkRepresenting the k-th QAM signal, h, of the signal UkDenotes the filter coefficients and C denotes the length of the filter, which in this embodiment has a value of 201.
In step CD-2, the transmitting end is directed to the signal U { U }1,...,uk,...,uKApplying a Saleh model power amplifier, specifically by tk＝PA(uk) Respectively processing and updating each QAM signal to obtain a signal T { T }1,...,tk,...,tK}。
Figure BDA0002362989910000111
Wherein PA (-) represents the transfer function of the power amplifier to the input signal, AinRepresenting the input signal ukAmplitude of (d), tkRepresenting the input signal ukOutput signal after passing through Saleh model of power amplifier, g0、A0And the values of alpha and beta are 2, 1, 2 and 1 respectively.
In step D1, the receiving end receives the signal from the transmitting end and aims at the obtained signal R { R }1,...,rk,...,rKApplying a matched filter, passing yk＝Ι(rk) Respectively processing and updating each QAM signal to obtain a signal Y { Y1,...,yk,...,yKThen step D2 is entered; wherein the transform function in the matched filter is the same as the transform function in the square root raised cosine filter.
In step D2, Y { Y }1,y2,...,yKThe power back-off coefficient G and the corresponding power back-off coefficient G are used as the input of a neural network, and the number of output layer nodes of the neural network is equal to the QAM signal skThe bit number m carried on it is 10, the output result of the neural network represents QAM signal skProbability value P { d of 1 in each biti＝1|yk}。
In step E, when the probability value P { d }i＝1|ykSatisfy P { d }i＝1|ykIf the rate is greater than or equal to 0.5, the bit d is setiThe value of (1); when probability value P { di＝1|ykSatisfy P { d }i＝1|ykIf the rate is less than 0.5, the bit d is setiThe value of (1) is judged to be 0; after the above-mentioned judgment of each bit in each QAM signal is completed, the values of all bits are sorted according to the sequence of each QAM signal and the sequence of each bit in the QAM signal, that is, the resulting binary bit stream received by the receiving end is obtained.
It should be noted that in the execution of the above steps a to E, when training the neural network, the training samples Ψ are obtained through steps a to D, and include the signal Y '{ Y'1,...,y'k,...,y'KSetting G as 0dB, -3dB, -6dB, -9dB, -12dB, -14dB and-16 dB, obtaining signals Y 'under different power backoff coefficient G conditions through steps A to E, taking the signals Y' and the corresponding power backoff coefficients G as training samples psi, determining the number L of hidden layers of the neural network as 6, the number Z of neurons in each hidden layer as 64, then initializing the weights and the bias of the neural network, selecting xavier initialization in an initialization mode, training the neural network through steps A to E, setting the training times as 600000 times, and selecting an Adam algorithm in a gradient descent method of a back propagation algorithm of the neural network.
Compared with the conventional demodulation algorithm under the existing nonlinear scene, the method evaluates the superiority of the neural network demodulation method from the aspect of the Bit Error Rate (BER) of the system, as shown in FIG. 4, the abscissa represents the normalized output power, and the performance of the neural network demodulation method is basically the same as that of the existing best conventional demodulation algorithm, but the complexity of the algorithm is reduced.
The technical scheme is designed to be suitable for a transmission control method of a high-order modulation signal subjected to nonlinear interference, the demodulation function is realized through the neural network, the nonlinear interference of the power amplifier on the signal is reduced, and the signal of a transmitting end is recovered.
The embodiments of the present invention have been described in detail with reference to the drawings, but the present invention is not limited to the above embodiments, and various changes can be made within the knowledge of those skilled in the art without departing from the gist of the present invention.

Claims (9)

1. A transmission control method for high-order modulation signal affected by nonlinear interference is used for aiming at binary bit stream X { X1,x2,...,xNRealizing transmission control from a transmitting end to a receiving end; the method is characterized by comprising the following steps:
step A. the transmitting terminal modulates according to the modulationOrder M, in terms of M ═ log2M, obtaining the bit number M in single QAM signal
Figure FDA0002362989900000011
Obtaining a group number K, and then entering a step B; wherein the content of the first and second substances,
Figure FDA0002362989900000012
represents rounding up, N represents the number of bits in the binary bitstream X;
and step B, the transmitting end carries out sequential grouping on each bit in the binary bit stream X by taking the bit number in a single QAM signal as M to obtain K QAM signals with the modulation order of M to form an M-QAM signal S { S1,...,sk,...,sK}，skRepresenting the kth QAM signal in the M-QAM signal S; if the bit number in the last QAM signal in the sequence is less than m, then 0 is supplemented at the end until the bit number m is met, and then the step C is carried out;
step C, the transmitting end aims at the M-QAM signal S { S1,...,sk,...,sKApply a multiplier, by vk＝G×skRespectively processing and updating each QAM signal to obtain a signal V { V }1,...,vk,...,vK}，vkRepresenting the kth QAM signal in the signal V, and sending the signal to a receiving end by a transmitting end, wherein G represents a preset power back-off coefficient, and then entering step D;
d, the receiving end receives the signal from the transmitting end, the real part, the imaginary part and the power back-off coefficient G of each QAM signal in the received signal are used as the input of a trained neural network, the neural network processes each QAM signal to respectively obtain the probability that each bit in each QAM signal is 1, and then the step E is carried out;
e, the receiving end judges whether the probability of the bit being 1 is not less than a preset probability threshold value aiming at the processing result of the neural network and aiming at each bit in each QAM signal, if so, the value of the bit is judged to be equal to 1; otherwise, judging the value of the bit to be 0; after the judgment of each bit in each QAM signal is completed, sequencing the values of all the bits according to the sequence of each QAM signal and the sequence of each bit in the QAM signal, namely obtaining a result binary bit stream received by a receiving end; and B, if the operation of complementing 0 aiming at the last QAM signal in the sequence exists in the step B, deleting the last corresponding digit bit in the result binary bit stream.
2. The method according to claim 1, wherein the method comprises the following steps: in the step B, the transmitting end sequentially groups each bit in the binary bit stream X by using the bit number m in a single QAM signal and applying a constellation mapping method.
3. The method according to claim 1 or 2, wherein the transmission control method of the higher order modulation signal subjected to the nonlinear interference comprises: further comprising a step CD-1 of obtaining a signal V { V } in said step C as follows1,...,vk,...,vKAfter that, entering the step CD-1;
step CD-1. the transmitting end aims at the signal V { V1,...,vk,...,vK}, applying a square root raised cosine filter, passing u throughk＝Ι(vk) Respectively processing and updating each QAM signal to obtain a signal U { U }1,...,uk,...,uK}，ukRepresenting the kth QAM signal in the signal U, and sending the signal to the receiving end by the transmitting end, where i (·) represents the transform function of the square root raised cosine filter on the input signal, and then entering step D.
4. The method of claim 3, wherein the step of controlling transmission of the higher order modulated signal is further characterized by comprising: further comprising a step CD-2 of obtaining a signal U { U } in said step CD-1 as follows1,...,uk,...,uKAfter that, the step CD-2 is carried out;
step CD-2. the transmitting end aims at the signal U { U }1,...,uk,...,uKApplying a power amplifier, passing through tk＝PA(uk) For eachRespectively processing and updating QAM signals to obtain signals T { T }1,...,tk,...,tK}，tkRepresenting the kth QAM signal in the signal T and transmitting the signal from the transmitting end to the receiving end, wherein PA (-) represents the transformation function of the power amplifier to the input signal, and then proceeding to step D.
5. The method according to claim 4, wherein said step D comprises the steps of:
step D1. the receiving end receives the signal from the transmitting end and aims at the obtained signal R { R1,...,rk,...,rKApplying a matched filter, passing yk＝Ι(rk) Respectively processing and updating each QAM signal to obtain a signal Y { Y1,...,yk,...,yKThen step D2 is entered; wherein the transform function in the matched filter is the same as the transform function in the square-root raised cosine filter, rkRepresenting the k-th QAM signal, y, of the signal RkRepresenting the kth QAM signal in signal Y;
step D2. the receiver side targets the signal Y Y1,...,yk,...,yKUsing each QAM signal y in the signalkThe real part, the imaginary part and the power back-off coefficient G are used as the input of a trained neural network, the neural network processes each QAM signal to respectively obtain each QAM signal ykThe probability that each bit is 1, and then step E is entered.
6. The method of claim 5, wherein the step of controlling transmission of the higher order modulated signal is further characterized by comprising: the neural network is a fully-connected neural network with L hidden layers, wherein the number of neurons in each hidden layer is Z; output a of the l hidden layer in neural network applicationslOutput a of the l-1 hidden layer of the neural networkl-1The relationship between them is:
Figure FDA0002362989900000031
in the formula (I), the compound is shown in the specification,
Figure FDA0002362989900000032
representing the weight value from the ith neuron in the l-1 hidden layer to the jth neuron in the l hidden layer in the neural network;
Figure FDA0002362989900000033
representing the bias of the ith neuron in the ith hidden layer in the neural network, wherein the bias of each neuron in the same hidden layer in the neural network is the same as that of each neuron in the first hidden layer in the neural network; f (-) represents the activation function of the hidden layer in the neural network.
7. The method according to claim 6, wherein the neural network training stage comprises the following steps:
step I, aiming at the sample binary bit stream, obtaining a training sample psi through steps A to D, wherein the training sample psi comprises a signal Y ' { Y ' of a receiving end after passing through a matched filter '1,...,y′k,...,y′KAnd a power backoff factor G, whereby it is determined that the number of input layer nodes of the neural network is equal to 3, for receiving the QAM signals y'kThe real part, the imaginary part and the power back-off coefficient G, and then entering step II;
step II, initializing the method by applying a xavier initialization method
Figure FDA0002362989900000034
And
Figure FDA0002362989900000035
and the bias of each neuron element in the same hidden layer in the neural network is the same with each other, and is respectively corresponding to each QAM signal Y 'in the signal Y'kThe input layers of the neural network receive QAM signals y'kReal part, imaginary part, and power back-off coefficient G, then according to the neural networkThe number of nodes of the network output layer is equal to m, and the QAM signal y 'is obtained by the neural network output layer'kThe probability that each bit in the sequence is 1; thus, the training of the neural network is realized through the steps A to E, the number of hidden layers in the neural network is determined to be L, the number of neurons of each hidden layer is determined to be Z, and the optimization is carried out
Figure FDA0002362989900000036
And
Figure FDA0002362989900000037
8. the method according to claim 7, wherein the transmission control method is adapted to the transmission of the higher order modulated signal interfered by the non-linearity, and comprises: the cost function in a cross entropy form is adopted in the neural network, and iterative optimization is carried out on the cost function in the neural network through a back propagation algorithm
Figure FDA0002362989900000038
And
Figure FDA0002362989900000039
9. the method according to claim 7, wherein the transmission control method is adapted to the transmission of the higher order modulated signal interfered by the non-linearity, and comprises: the activation function of the hidden layer in the neural network is a tanh function, and the activation function of the output layer is a sigmoid function.
CN202010027487.8A 2020-01-10 2020-01-10 High-order modulation signal transmission control method suitable for being interfered by nonlinearity Active CN111200470B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010027487.8A CN111200470B (en) 2020-01-10 2020-01-10 High-order modulation signal transmission control method suitable for being interfered by nonlinearity

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010027487.8A CN111200470B (en) 2020-01-10 2020-01-10 High-order modulation signal transmission control method suitable for being interfered by nonlinearity

Publications (2)

Publication Number Publication Date
CN111200470A CN111200470A (en) 2020-05-26
CN111200470B true CN111200470B (en) 2021-09-07

Family

ID=70747068

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010027487.8A Active CN111200470B (en) 2020-01-10 2020-01-10 High-order modulation signal transmission control method suitable for being interfered by nonlinearity

Country Status (1)

Country Link
CN (1) CN111200470B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112532548B (en) * 2020-12-23 2024-02-27 国网信息通信产业集团有限公司 Signal optimization method and device
CN113325375B (en) * 2021-05-25 2022-12-13 哈尔滨工程大学 Self-adaptive cancellation method based on deep neural network

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2019191099A1 (en) * 2018-03-26 2019-10-03 Zte Corporation Non-linear adaptive neural network equalizer in optical communication
US10944440B2 (en) * 2018-04-11 2021-03-09 Booz Allen Hamilton Inc. System and method of processing a radio frequency signal with a neural network
CN109889462B (en) * 2019-01-21 2021-05-28 东南大学 Neural network assisted iterative receiving method suitable for high-speed visible light communication
CN110120926B (en) * 2019-05-10 2022-01-07 哈尔滨工程大学 Communication signal modulation mode identification method based on evolution BP neural network

Also Published As

Publication number Publication date
CN111200470A (en) 2020-05-26

Similar Documents

Publication Publication Date Title
CN106877825B (en) Digital predistortion device and method based on band-limited simplified nonlinear filter
US5105445A (en) Predistortion arrangement for a digital transmission system
CN111200470B (en) High-order modulation signal transmission control method suitable for being interfered by nonlinearity
CN111245375B (en) Power amplifier digital predistortion method of complex-valued fully-connected recurrent neural network model
EP1523838A1 (en) Adaptive pre-equalization method and apparatus
Benvenuto et al. A neural network approach to data predistortion with memory in digital radio systems
CN111082757A (en) Method for applying adaptive band-limited digital predistortion in broadband satellite communication system
CN104410593B (en) Numerical chracter nonlinearity erron amendment equalization methods based on decision-feedback model
CN111064439B (en) System and method for improving short wave digital predistortion performance
CN114400979A (en) System and method for improving digital predistortion of short wave interaction and harmonic distortion
CN112491754B (en) Channel estimation and signal detection method based on DDST and deep learning
Logins et al. Block-structured deep learning-based OFDM channel equalization
Bauduin et al. Non-linear satellite channel equalization based on a low complexity echo state network
CN111988254A (en) Low-complexity peak-to-average ratio compression and predistortion joint optimization method
CN114598274B (en) Low-complexity lookup table construction method oriented to broadband predistortion
CN111884602B (en) Power amplifier predistortion method based on single-output-node neural network
Pupolin et al. Performance analysis of digital radio links with nonlinear transmit amplifier and data predistorter with memory
Qian et al. Neural network predistortion technique for nonlinear power amplifiers with memory
US20140126670A1 (en) Method and apparatus for nonlinear-channel identification and estimation of nonlinear-distorted signals
Li et al. Complex-valued pipelined chebyshev functional link recurrent neural network for joint compensation of wideband transmitter distortions and impairments
CN115913140B (en) Piecewise polynomial digital predistortion device and method for controlling operation precision
Kohonen et al. Start-up behaviour of a neural network assisted decision feedback equaliser in a two-path channel
Naskas et al. Adaptive baseband predistorter for radio frequency power amplifiers based on a multilayer perceptron
CN114978313B (en) Compensation method of visible light CAP system based on Bayesian neurons
CN114528925B (en) Time-varying channel OFDM signal equalization method based on deep classification network

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