WO2022042736A1 - 一种信号补偿处理方法及装置 - Google Patents

一种信号补偿处理方法及装置 Download PDF

Info

Publication number
WO2022042736A1
WO2022042736A1 PCT/CN2021/115480 CN2021115480W WO2022042736A1 WO 2022042736 A1 WO2022042736 A1 WO 2022042736A1 CN 2021115480 W CN2021115480 W CN 2021115480W WO 2022042736 A1 WO2022042736 A1 WO 2022042736A1
Authority
WO
WIPO (PCT)
Prior art keywords
signal
modeling
channel
model parameters
data
Prior art date
Application number
PCT/CN2021/115480
Other languages
English (en)
French (fr)
Inventor
黄新刚
迟楠
邹鹏
马壮
Original Assignee
中兴通讯股份有限公司
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 中兴通讯股份有限公司 filed Critical 中兴通讯股份有限公司
Priority to EP21860568.1A priority Critical patent/EP4207697A4/en
Publication of WO2022042736A1 publication Critical patent/WO2022042736A1/zh

Links

Images

Classifications

    • 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
    • H04L27/38—Demodulator circuits; Receiver circuits
    • H04L27/3845—Demodulator circuits; Receiver circuits using non - coherent demodulation, i.e. not using a phase synchronous carrier
    • H04L27/3854—Demodulator circuits; Receiver circuits using non - coherent demodulation, i.e. not using a phase synchronous carrier using a non - coherent carrier, including systems with baseband correction for phase or frequency offset
    • H04L27/3863—Compensation for quadrature error in the received signal
    • 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
    • H04L27/36—Modulator circuits; Transmitter circuits
    • H04L27/362—Modulation using more than one carrier, e.g. with quadrature carriers, separately amplitude modulated
    • H04L27/364—Arrangements for overcoming imperfections in the modulator, e.g. quadrature error or unbalanced I and Q levels
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B10/00—Transmission systems employing electromagnetic waves other than radio-waves, e.g. infrared, visible or ultraviolet light, or employing corpuscular radiation, e.g. quantum communication
    • H04B10/25—Arrangements specific to fibre transmission
    • H04B10/2507—Arrangements specific to fibre transmission for the reduction or elimination of distortion or dispersion
    • H04B10/2543—Arrangements specific to fibre transmission for the reduction or elimination of distortion or dispersion due to fibre non-linearities, e.g. Kerr effect
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B10/00—Transmission systems employing electromagnetic waves other than radio-waves, e.g. infrared, visible or ultraviolet light, or employing corpuscular radiation, e.g. quantum communication
    • H04B10/60—Receivers
    • H04B10/66—Non-coherent receivers, e.g. using direct detection
    • H04B10/69—Electrical arrangements in the receiver
    • H04B10/697—Arrangements for reducing noise and distortion
    • H04B10/6971—Arrangements for reducing noise and distortion using equalisation

Definitions

  • the embodiments of the present application relate to the field of communications, and in particular, to a signal compensation processing method and apparatus.
  • IQ optical signal is a common optical modulation method in communication.
  • the two channels of orthogonal signals can be separated by up-converting and then superimposing, and then performing corresponding down-conversion at the receiving end.
  • the superposition of the two-channel signals will greatly deteriorate the signal-to-noise ratio due to power mismatch, clock synchronization, nonlinear effects and other problems, thereby reducing the system performance. capacity.
  • SISO-LMS minimum mean square error
  • RLS Recursive, Least Square
  • DNN Deep Neural Network
  • Embodiments of the present application provide a signal compensation processing method and apparatus, so as to at least solve the problem of high bit error rate of mutual influence between IQ signals in optical communication in the related art.
  • a signal compensation processing method including:
  • the data characteristic signal is compensated according to the predetermined target model parameters to obtain an IQ equalized signal.
  • a signal compensation processing apparatus including:
  • the first acquisition module is configured to acquire the I-channel received data signal obtained by receiving the I-channel transmission data signal, and the Q-channel received data signal obtained by receiving the Q-channel transmission data signal;
  • the first cascading module is configured to perform cascade transformation on the received data signal of the I channel and the received data signal of the Q channel to obtain a data characteristic signal;
  • the compensation module is configured to compensate the characteristic signal according to the predetermined target model parameters to obtain an IQ equalized signal.
  • a computer-readable storage medium is also provided, and a computer program is stored in the computer-readable storage medium, wherein the computer program is configured to execute any one of the above methods when running steps in the examples.
  • an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute any one of the above Steps in Method Examples.
  • the I-channel received data signal obtained by receiving the I-channel transmit data signal, and the Q-channel received data signal obtained by receiving the Q-channel transmit data signal are obtained; the I-channel received data signal and the Q-channel received data signal are obtained.
  • the data characteristic signal obtained by the cascade transformation of the IQ received data signal is compensated to obtain an IQ equalized signal, which reduces the bit error rate of the IQ signal.
  • FIG. 1 is a block diagram of a hardware structure of a mobile terminal of a signal compensation processing method according to an embodiment of the present application
  • FIG. 2 is a flowchart of a signal compensation processing method according to an embodiment of the present application.
  • FIG. 3 is a flowchart of a signal compensation processing method according to an embodiment of the present application.
  • FIG. 4 is a flowchart of a method for receiving and compensating an IQ signal according to an embodiment of the present application
  • FIG. 5 is a schematic diagram of a multi-input multi-output multi-branch neural network equalizer of a heterogeneous deep neural network according to an embodiment of the present application
  • FIG. 6 is a schematic diagram of a verification experimental platform according to an embodiment of the present application.
  • FIG. 7 is a structural block diagram of a signal compensation processing apparatus according to an embodiment of the present application.
  • FIG. 8 is a structural block diagram of a signal compensation processing apparatus according to a preferred embodiment of the present application.
  • FIG. 1 is a block diagram of a hardware structure of a mobile terminal of a signal compensation processing method according to an embodiment of the present application.
  • the mobile terminal may include one or more (only shown in FIG. 1 ).
  • a processor 102 may include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA, etc.
  • a memory 104 for storing data
  • the above-mentioned mobile terminal may also include a communication function
  • the transmission device 106 and the input and output device 108 can understand that the structure shown in FIG. 1 is only for illustration, and it does not limit the structure of the above-mentioned mobile terminal.
  • the mobile terminal may also include more or fewer components than those shown in FIG. 1 , or have a different configuration than that shown in FIG. 1 .
  • the memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as computer programs corresponding to the data processing methods in the embodiments of the present application. A functional application and data processing are implemented, namely, the above-mentioned method is implemented.
  • Memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory located remotely from the processor 102, and these remote memories may be connected to the mobile terminal through a network. Examples of such networks include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.
  • Transmission means 106 are used to receive or transmit data via a network.
  • the specific example of the above-mentioned network may include a wireless network provided by a communication provider of the mobile terminal.
  • the transmission device 106 includes a network adapter (Network Interface Controller, NIC for short), which can be connected to other network devices through a base station so as to communicate with the Internet.
  • the transmission device 106 may be a radio frequency (Radio Frequency, RF for short) module, which is used to communicate with the Internet in a wireless manner.
  • RF Radio Frequency
  • FIG. 2 is a flow chart of the signal compensation processing method according to the embodiment of the present application. As shown in FIG. 2 , the flow includes the following step:
  • Step S202 acquiring the I-channel received data signal obtained by receiving the I-channel transmit data signal, and the Q-channel received data signal obtained by receiving the Q-channel transmit data signal;
  • Step S204 performing cascade transformation on the received data signal of the I channel and the received data signal of the Q channel to obtain a data characteristic signal
  • performing cascade transformation on the I-channel received data signal and the Q-channel received data signal, and obtaining the data characteristic signal includes:
  • the above step S204 may specifically include: respectively sampling the I-channel received data signal and the Q-channel received data signal according to the window length taps to obtain N groups of I-channel sampling signals and N groups of Q-channel sampling signals , wherein, for the center signal of each group of sampling signals in the N groups of I-channel sampling signals and the N groups of Q-channel sampling signals is the current sampling point, the left side of the center signal is (taps-1)/2 points are the current sampling point intercepted forward (taps-1)/2 sampling points, and the right side of the center signal (taps-1)/2 points are the current sampling point intercepted backward (taps -1)/2 sampling points, if the forward or backward sampling points are less than (taps-1)/2, fill with 0, the taps is an odd number, and the N is the received data signal of the I channel and the The length of the data signal received by the Q channel; the N groups of the I-channel sampling signals with the length of the taps are respectively cascaded with the N groups of the Q-channel sampled signals
  • Step S206 Compensate the data characteristic signal according to the predetermined target model parameters to obtain an IQ equalized signal.
  • the I-channel received data signal obtained by receiving the I-channel transmission data signal, and the Q-channel received data signal obtained by receiving the Q-channel transmitted data signal are obtained;
  • the received data signal is cascaded and transformed to obtain a characteristic signal;
  • the data characteristic signal is compensated according to the predetermined target model parameters to obtain an IQ equalized signal, which can solve the error of the mutual influence between the IQ signals in the optical communication in the related art.
  • the data characteristic signal obtained by the cascade transformation of the IQ received data signal is compensated to obtain an IQ equalized signal, which reduces the bit error rate of the IQ signal.
  • the foregoing step S206 may specifically include:
  • mapping the data characteristic signal according to the target model parameters to obtain an I-line linear data signal, a Q-line linear data signal, an I-line nonlinear data signal, and a Q-line nonlinear data signal;
  • step S2061 may specifically include:
  • the linear crosstalk of the I-path optical signal of the data feature signal is processed according to the target model parameters to obtain the I-line data signal, which is obtained by linearly mapping the data feature signal according to the target model parameters in the following manner
  • the I line-line data signal Wherein, 1 ⁇ t ⁇ N, N is the length of the data characteristic signal;
  • the linear crosstalk of the Q-path optical signal of the data characteristic signal is processed according to the target model parameters to obtain the Q-path linear data signal, which is obtained by linearly mapping the data characteristic signal according to the target model parameters in the following manner
  • the Q-Linear data signal is processed according to the target model parameters to obtain the Q-path linear data signal, which is obtained by linearly mapping the data characteristic signal according to the target model parameters in the following manner
  • the Q-Linear data signal is processed according to the target model parameters to obtain the Q-path linear data signal, which is obtained by linearly mapping the data characteristic signal according to the target model parameters in the following manner
  • the Q-Linear data signal is processed according to the target model parameters to obtain the Q-path linear data signal, which is obtained by linearly mapping the data characteristic signal according to the target model parameters in the following manner
  • the Q-Linear data signal is processed according to the target model parameters to obtain the Q-path linear data signal, which is obtained by linearly mapping the data characteristic signal according to the target model parameters in the following manner
  • the crosstalk between the IQ optical signals of the data characteristic signal is processed according to the target model parameters, and the I-channel nonlinear data signal and the Q-channel nonlinear data signal are obtained.
  • the model parameters perform nonlinear mapping on the data characteristic signal to obtain the I-channel nonlinear data signal and the Q-channel nonlinear data signal: in, is the target model parameter, f(x) is the activation function;
  • S2062 Determine the sum of the I line linear data signal and the I line nonlinear data signal as the I line equalized signal, and determine the sum of the Q line linear data signal and the Q line nonlinear data signal as The Q channel equalized signal, wherein the IQ equalized signal includes the I channel equalized signal and the Q channel equalized signal.
  • FIG. 3 is a flowchart of a signal compensation processing method according to the present preferred embodiment. As shown in FIG. 3 , before the IQ received data signal obtained by receiving the IQ transmit data signal is subjected to cascade transformation to obtain a data characteristic signal, the method also includes:
  • Step S302 acquiring the I-channel receiving modeling signal obtained by receiving the I-channel transmitting modeling signal, and the Q-channel receiving modeling signal obtained by receiving the Q-channel transmitting modeling signal;
  • Step S304 performing cascade transformation on the received modeling signal of the I channel and the received modeling signal of the Q channel to obtain a modeling characteristic signal
  • the above step S304 may specifically include: respectively sampling the I-channel received modeling signal and the Q-channel receiving modeling signal according to the window length taps, to obtain N groups of I-channel sampling signals and N groups of Q-channel sampling signals, wherein, For the N groups of I-channel sampling signals and the N groups of Q-channel sampling signals, the center signal of each group of sampling signals is the current sampling point, and the (taps-1)/2 points on the left side of the center signal are The current sampling point is intercepted forward (taps-1)/2 sampling points, and the (taps-1)/2 points on the right side of the central signal are the current sampling point intercepted backward (taps-1) /2 sampling points, if the forward or backward sampling points are less than (taps-1)/2, fill with 0, the taps is an odd number, and the N is the received modeling signal of the I channel and the Q channel Receive the length of the modeled signal; respectively concatenate the N groups of I-way sampling signals with the N groups of Q-way sampling signals with the length of the taps, to
  • Step S306 updating model parameters according to the modeling feature signal to obtain the target model parameters.
  • step S306 may specifically include:
  • a nonlinear modeling signal further, the linear crosstalk of the I-path optical signal of the modeling characteristic signal is processed according to the model parameters to obtain the I-line linear modeling signal.
  • the model parameters perform linear mapping on the modeling characteristic signal to obtain the linear modeling signal: Wherein, 1 ⁇ t ⁇ N, N is the length of the modeling feature; the linear crosstalk of the Q-path optical signal of the modeling feature signal is processed according to the model parameters, and the Q-path linear modeling signal is obtained.
  • the Q linear modeling signal can be obtained by linearly mapping the modeling characteristic signal according to the model parameters in the following manner:
  • the crosstalk between the IQ optical signals of the modeling characteristic signal is processed according to the model parameters to obtain the nonlinear modeling signal of the I channel and the nonlinear modeling signal of the Q channel.
  • the method performs nonlinear mapping on the modeling characteristic signal according to the model parameters to obtain the I-channel nonlinear modeling signal and the Q-channel nonlinear modeling signal: in, is the model parameter, f(x) is the activation function;
  • the I-line linear modeling signal and the I-line nonlinear modeling signal add and process to obtain the I-line modeling output signal, and according to the Q-line linear modeling signal and the Q-line nonlinear modeling signal
  • the modulo signal is added and processed to obtain the Q-channel modeling output signal
  • the model parameters are updated according to the I-channel transmission modeling signal, the Q-channel transmission modeling signal, the I-channel modeling output signal, and the Q-channel modeling output signal to obtain updated model parameters , wherein the IQ transmission modeling signal includes the I channel transmission modeling signal and the Q channel transmission modeling signal;
  • the processing process is divided into two stages: model establishment or training and signal reception compensation.
  • the switching between the two stages can be indicated by a flag bit. For example, when the flag bit is 0 or false, model establishment or training is performed, and the flag bit is 1 or If true, signal reception compensation is performed.
  • the above preset iterative conditions may be implemented by flag bits, or may be implemented by other methods, such as a fixed-length training sequence, parameters converging within a certain error, and so on.
  • the above-mentioned updating of the model parameters, and obtaining the updated model parameters may specifically include:
  • the signal output error is determined according to the I-channel transmission modeling signal, the Q-channel transmission modeling signal, the I-channel modeling output signal, and the Q-channel modeling output signal.
  • Channel emission modeling signal, the Q channel emission modeling signal, the I channel modeling output signal, and the Q channel modeling output signal determine the signal output error:
  • r(t) is the signal output error
  • p I (a) is the modeling output signal of the I channel
  • p Q (a) is the modeling output signal of the Q channel
  • s I (a) is the modeling output signal of the I channel signal
  • s Q (a) is the Q-channel emission modeling signal
  • N is the length of the modeling characteristic signal
  • the model parameters are updated according to the signal output errors to obtain the updated model parameters, and further, the partial derivatives of the signal output errors with respect to the model parameters are respectively determined to obtain the same dimension as the model parameters.
  • the target matrix, the model parameters are updated according to the target matrix, and the updated model parameters are obtained, and the update parameters are determined in the following manner:
  • the input signal in this embodiment has memory, and joint compensation and estimation are performed by several symbols before and after recording, and three branches are used, one branch handles the linear crosstalk of the I-path optical signal, and one branch handles the Q-path optical signal.
  • the two linear branches can be the linear weighting network of MLP; there is another branch to deal with the crosstalk between the two optical signals of IQ. Since the mutual crosstalk of IQ optical signals is not very complicated, a two-layer hidden layer network can be used to train nonlinearity, which greatly reduces the complexity of the system.
  • the linear output and nonlinear output of I, and the linear output of Q and the nonlinear output are added respectively, and the obtained compensation signal equalizes both the linear distortion and the nonlinear distortion, and achieves a very good effect.
  • This method is also applicable to wireless signals with crosstalk between two IQ channels of the same or similar architecture, and details are not repeated here. With this embodiment, the performance of the system can be effectively improved.
  • the equalization method proposed in this embodiment can increase the working range by 2.33 times to more than 8 times when the powers of two transmitters are present. At the same time, up to 3.63dB of Q-value improvement can be achieved.
  • Fig. 4 is the flow chart of the receiving compensation method of the IQ signal of the embodiment of the application, as shown in Fig. 4, may comprise the following steps:
  • the number of hidden layers of the model is H 1
  • the number of nodes in the second hidden layer is H 2
  • the input layer I, Q channel signal windows are taps respectively
  • hidden Layer number of nodes per layer weight matrix and the bias vector
  • the dimensions of the weight matrix and bias vector are shown in Table 1.
  • Step 403 Receive and record the I-channel received modeling signal, wherein the transmitting end loads the original transmitting modeling signal s(t), specifically, the original transmitting modeling signal s(t) is loaded by the transmitting modeling signal loading device.
  • the original transmission modeling signal is modulated and recorded as an I-channel transmission modeling signal. Specifically, s(t) is modulated by an I-channel modeling signal transmitting device and an I-channel transmission modeling signal x I (t) is recorded.
  • the original transmission modeling signal is modulated and recorded as the Q channel transmission modeling signal, specifically, s(t) is modulated by the Q channel modeling signal transmitting device and the Q channel transmission modeling signal x Q (t) is recorded.
  • x I (t) is received and recorded by the I-channel modeling signal receiving device, and the I-channel received modeling signal y I (t) is obtained.
  • Step 404 receiving and recording the Q channel received modeling signal. specifically:
  • the Q channel modeling signal receiving device receives and records x I (t) to obtain the Q channel received modeling signal y I (t).
  • Step 405 receive the modeling signal according to the recorded I channel, and receive the modeling signal in the Q channel, convert them in cascade and record them as the modeling characteristic signal.
  • y I (t), y I (t) are cascade-transformed into the modeling feature signal k(t) by the modeling cascade transforming means.
  • the I-channel received data signal and the Q-channel received data signal are sampled according to the window length taps, respectively, to obtain N groups of I-channel sampled signals and N groups of Q-channel sampled signals, wherein, for the N groups of I-channel sampled signals
  • the center signal of each group of sampling signals in the N groups of Q-way sampling signals is the current sampling point, and the (taps-1)/2 points on the left side of the center signal are the forward interception of the current sampling point ( taps-1)/2 sampling points, the (taps-1)/2 points on the right side of the central signal are the current sampling point intercepted backward (taps-1)/2 sampling points, if forward Or the backward sampling point is insufficient (taps-1)/2, fill with 0, the taps is an odd number, and the N is the length of the received data signal of the I channel and the received data signal of the Q
  • each k(t) is a column vector of length 2*taps.
  • the signal at this time has memory, that is, each signal input to the next stage includes each taps/2-1 signal before and after. Modules that use this signal as output have memory.
  • Step 406 performing linear mapping on the modeling feature signal to obtain an I linear modeling signal. specifically:
  • Step 407 Perform linear mapping on the modeling feature signal to obtain a Q linear modeling signal. specifically:
  • Step 408 Perform nonlinear mapping on the modeling characteristic signal to obtain the I-channel nonlinear modeling signal and the Q-channel nonlinear modeling signal. specifically:
  • the I-channel nonlinear modeling signal and the Q-channel nonlinear modeling signal are obtained through the nonlinear mapping module and k(t):
  • f(a) is the activation function, and its expression can be
  • x can be a scalar, vector, matrix, and f(x) has the same dimension as x.
  • f(x) means using the above formula for each element in the matrix or vector.
  • Step 409 according to the recorded I-line linear modeling signal and I-line nonlinear modeling signal, add and record the I-line modeling output signal.
  • Step 410 according to the recorded Q-path linear modeling signal and Q-path nonlinear modeling signal, add and record the Q-path modeling output signal.
  • Step 411 according to the recorded I-channel modeling output signal, Q-channel modeling output signal, I-channel transmitting the modeling signal, Q-channel transmitting the modeling signal, calculate the output error and compare the model parameters. to update.
  • model parameters are updated through the error calculation and model parameter update module, and the specific steps are:
  • V(t) ⁇ 2 V(t-1)+(1- ⁇ 2 )G 2 (t), where G 2 (t) represents the dot product of the elements of the matrix G(t);
  • step 412 If all model parameters have been updated, go to step 412, otherwise go to step (1) in step 411.
  • step 412 it is judged whether the iteration condition is terminated. If the judgment result is yes, step 413 is executed;
  • Step 414 receive and record the I-channel received data signal.
  • the transmitting end loads the original transmission data signal s'(t), specifically, the original transmission data signal s'(t) is loaded by the transmission data signal loading device.
  • the original transmission data signal is modulated and recorded as the I-channel transmission data signal, specifically, s'(t) is modulated by the I-channel data signal transmitting device and the I-channel transmission modeling signal x I '(t) is recorded.
  • the original transmit data signal is modulated and recorded as the Q channel transmit data signal, specifically, s'(t) is modulated by the Q channel data signal transmitting device and the I channel transmit modeling signal x Q '(t) is recorded.
  • x I '(t) is received and recorded by the I-channel data signal receiving device, and the I-channel received modeling signal y I '(t) is obtained.
  • Step 415 Receive and record the Q channel received data signal. specifically:
  • the Q channel data signal receiving device receives and records x Q '(t) to obtain the I channel received modeling signal y Q '(t).
  • Step 416 According to the recorded data signal of channel I, channel Q receives the data signal, cascade-converted and recorded as a data characteristic signal. Specifically, y I '(t), y Q '(t) are cascade-transformed into the data characteristic signal k'(t) by the data cascade transforming device.
  • the specific transformation method is similar to that of step 405, and details are not repeated here.
  • Step 417 according to the recorded model parameters and data characteristic signals, equalize and record the I-channel equalized data signals and the Q-channel equalized data signals q I '(t) and q Q '(t).
  • the equalization device is used to equalize the data characteristic signal to obtain the I channel equalized data signal and the Q channel equalized data signal q I '(t) and q Q '(t), and the specific manner is:
  • FIG. 5 is the MIMO multi-branch neural network equalization of the heterogeneous deep neural network according to the embodiment of the present application Schematic diagram of the device, as shown in Figure 5, the output end respectively adds the linear and nonlinear outputs of the I channel and the linear kernel nonlinear output of the Q channel through two summing modules (the I channel summing module 211 and the Q channel summing module 212). The linear outputs are summed and output. When the modeling process is complete, the stored parameters are not updated anymore.
  • FIG. 6 is a schematic diagram of a verification experiment platform according to an embodiment of the present application. As shown in FIG.
  • the transmission modeling signal loading device (202) loads the original transmission modeling signal, and uses the I channel and the Q channel to model the signal
  • the transmitting device (203, 204) generates I-channel and Q-channel transmission modeling signals, and after the I-channel and Q-channel signals are superimposed at the receiving end, they are received by the I-channel and Q-channel modeling signal receiving devices (205, 206) to obtain I-channel and Q-channel Receive modeling signals.
  • the modeling cascade transformation device (207) in the MIMO multi-branch neural network performs the modeling cascade transformation in step 402 on the I channel and Q channel received modeling signals to obtain modeling feature signals.
  • the modeled characteristic signal is transformed by the linear mapping modules (208, 209) and the nonlinear mapping module (210) of the I and Q lines to obtain the I and Q line linear modeling signals and nonlinear modeling signals, and then go through the I and Q lines and
  • the Q-path summing module obtains the I-path modeling output signal and the Q-path modeling output signal.
  • the modeling signals and the modeling output signals of the I and Q channels are transmitted through the I channel and the Q channel, and the parameters in the model are updated through the algorithm in step 412 until the maximum number of iterations is reached.
  • the model parameters are then fixed and the current model parameters are loaded by the equalization device (220). Then, it starts with the transmission data signal loading means (214) in FIG. 6 .
  • the I and Q data signal devices (217, 218) receive and record the I and Q data signals, and then cascade the signals through the data cascade transformation device (219) and the equalization device (220). Transform and equalize, and finally get the I-channel and Q-channel equalized data signals.
  • the equalization method proposed in this embodiment has fewer parameters, so the time and space required for training are smaller. Although the parameters of SISO-LMS are the least among the three, its performance is the worst.
  • the reception equalization method proposed in this embodiment can bring a performance gain of Q_Factor value of up to 3.63 dB to the system.
  • the model is trained, it is only necessary to input the transformed received signal into the equalization device, and then the equalized signal can be obtained, and no other additional operations are required.
  • This embodiment effectively solves the problems that the existing equalization mode has poor performance and weak resistance to power mismatch in the IQ optical communication system. Compared with the traditional equalization method, the communication system capacity and system performance are effectively improved. This embodiment is also relatively easy to implement.
  • the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is better implementation.
  • the technical solution of the present application can be embodied in the form of a software product in essence or in a part that contributes to the prior art, and the computer software product is stored in a storage medium (such as ROM/RAM, magnetic disk, CD-ROM), including several instructions to make a terminal device (which may be a mobile phone, a computer, a server, or a network device, etc.) execute the methods described in the various embodiments of this application.
  • a storage medium such as ROM/RAM, magnetic disk, CD-ROM
  • a signal compensation processing apparatus is also provided, and the apparatus is used to implement the above-mentioned embodiments and preferred implementations, and what has been described will not be repeated.
  • the term "module” may be a combination of software and/or hardware that implements a predetermined function.
  • the apparatus described in the following embodiments is preferably implemented in software, implementations in hardware, or a combination of software and hardware, are also possible and contemplated.
  • FIG. 7 is a structural block diagram of a signal compensation processing apparatus according to an embodiment of the present application, as shown in FIG. 7 , including:
  • the first acquisition module 72 is configured to acquire the I-channel received data signal obtained by receiving the I-channel transmission data signal, and the Q-channel received data signal obtained by receiving the Q-channel transmission data signal;
  • the first cascading module 74 is configured to perform cascade transformation on the received data signal of the I channel and the received data signal of the Q channel to obtain a data characteristic signal;
  • the compensation module 76 is configured to compensate the characteristic signal according to the predetermined target model parameters to obtain an IQ equalized signal.
  • the compensation module 76 includes:
  • a mapping submodule configured to map the data characteristic signal according to the target model parameters to obtain an I-line linear data signal, a Q-line linear data signal, an I-line nonlinear data signal, and a Q-line nonlinear data signal;
  • a determination submodule configured to determine the sum of the I-line linear data signal and the I-line nonlinear data signal as the I-channel equalized signal, and to determine the Q-line linear data signal and the Q-line nonlinear data signal The sum is determined as a Q channel equalized signal, wherein the IQ equalized signal includes the I channel equalized signal and the Q channel equalized signal.
  • mapping submodule includes:
  • a first processing unit configured to process the linear crosstalk of the I-path optical signal of the data characteristic signal according to the target model parameter, to obtain the I-line linear data signal
  • a second processing unit configured to process the linear crosstalk of the Q-path optical signal of the data characteristic signal according to the target model parameter, to obtain the Q-path linear data signal
  • the third processing unit is configured to process the crosstalk between the IQ optical signals of the data characteristic signal according to the target model parameters, and obtain the I-channel nonlinear data signal and the Q-channel nonlinear data signal.
  • the first processing unit is further configured to perform linear mapping on the data characteristic signal according to the target model parameter in the following manner to obtain the I linear data signal: Wherein, 1 ⁇ t ⁇ N, N is the length of the data characteristic signal; and/or
  • the second processing unit is further configured to perform linear mapping on the data characteristic signal according to the target model parameter in the following manner to obtain the Q linear data signal: and / or
  • f(x) is the activation function
  • the first cascade module includes:
  • the first sampling sub-module is configured to sample the I-channel received data signal and the Q-channel received data signal according to the window length taps, to obtain N groups of I-channel sampled signals and N groups of Q-channel sampled signals, wherein, for The center signal of each group of sampling signals in the N groups of I-way sampling signals and the N groups of Q-way sampling signals is the current sampling point, and the (taps-1)/2 points on the left side of the center signal are all The current sampling point is intercepted forward (taps-1)/2 sampling points, and the (taps-1)/2 points on the right side of the center signal are the current sampling point intercepted backward (taps-1)/ 2 sampling points, if the forward or backward sampling point is less than (taps-1)/2, fill with 0, the taps is an odd number, and the N is the received data signal of the I channel and the received data of the Q channel the length of the signal;
  • the first cascading sub-module is configured to cascade the N groups of I-way sampling signals and the N groups of Q-way sampling signals with a length of the taps respectively, to obtain the N groups of N signals with a length of 2 taps.
  • Data characteristic signal is configured to cascade the N groups of I-way sampling signals and the N groups of Q-way sampling signals with a length of the taps respectively, to obtain the N groups of N signals with a length of 2 taps.
  • FIG. 8 is a structural block diagram of a signal compensation processing apparatus according to a preferred embodiment of the present application. As shown in FIG. 8 , the apparatus further includes:
  • the second acquisition module 82 is configured to acquire the I-channel receiving modeling signal obtained by receiving the I-channel transmitting modeling signal, and the Q-channel receiving modeling signal obtained by receiving the Q-channel transmitting modeling signal;
  • the second cascading module 84 is configured to perform cascade transformation on the received modeling signal of the I channel and the received modeling signal of the Q channel to obtain a modeling characteristic signal;
  • the updating module 86 is configured to update the model parameters according to the modeling feature signal to obtain the target model parameters.
  • the second cascade module 84 includes:
  • the second sampling sub-module is configured to sample the I-channel received modeling signal and the Q-channel received modeling signal according to the window length taps, to obtain N groups of I-channel sampling signals and N groups of Q-channel sampling signals, wherein , for the center signal of each group of sampled signals in the N groups of I-way sampling signals and the N groups of Q-way sampled signals is the current sampling point, and the left side of the center signal is (taps-1)/2 points (taps-1)/2 sampling points are intercepted forward for the current sampling point, and (taps-1)/2 points on the right side of the central signal are intercepted (taps-1) backward for the current sampling point. )/2 sampling points, if the forward or backward sampling points are less than (taps-1)/2, fill with 0, the taps is an odd number, and the N is the received modeling signal of the I channel and the Q The length of the channel to receive the modeled signal;
  • the second cascading sub-module is configured to cascade the N groups of I-channel sampling signals and the N groups of Q-channel sampling signals with a length of the taps, respectively, to obtain the N groups of signals with a length of 2 taps. Model the characteristic signal.
  • the update module 86 includes:
  • the execution submodule is set to repeatedly perform the following steps until the number of iterations t satisfies the preset iteration conditions, and the updated model parameters are determined to be the target model parameters:
  • the summing unit is configured to perform summation processing according to the I-line linear modeling signal and the I-line nonlinear modeling signal to obtain an I-line modeling output signal, and according to the Q-line linear modeling signal and the I-line modeling output signal;
  • the nonlinear modeling signal of the Q channel is added and processed to obtain the modeling output signal of the Q channel;
  • mapping unit includes:
  • a first processing subunit configured to process the linear crosstalk of the I-path optical signal of the modeling characteristic signal according to the model parameter, to obtain the I-line linear modeling signal
  • a second processing subunit configured to process the linear crosstalk of the Q-path optical signal of the modeling characteristic signal according to the model parameters, to obtain the Q-path linear modeling signal
  • a third processing subunit configured to process the crosstalk between the IQ two-path optical signals of the modeled characteristic signal according to the model parameters, to obtain the I-path nonlinear modeling signal and the Q-path nonlinear modeling signal Signal.
  • the first processing subunit is further configured to perform linear mapping on the modeling feature signal according to the model parameters in the following manner to obtain the linear modeling signal: where 1 ⁇ t ⁇ N, where N is the length of the modeled feature; and/or
  • the second processing subunit is further configured to linearly map the modeling feature signal according to the model parameters in the following manner to obtain the Q linear modeling signal: and / or
  • the third processing subunit is also configured to perform nonlinear mapping on the modeling feature signal according to the model parameters in the following manner to obtain an I-channel nonlinear modeling signal and a Q-channel nonlinear modeling signal: in, are the model parameters, and f(x) is the activation function.
  • the update unit includes:
  • determining a subunit configured to determine a signal output error according to the I-channel transmission modeling signal, the Q-channel transmission modeling signal, the I-channel modeling output signal, and the Q-channel modeling output signal;
  • An update subunit configured to update the model parameters according to the signal output error to obtain the updated model parameters.
  • the update subunit is further set to
  • the model parameters are updated according to the target matrix to obtain the updated model parameters.
  • the update subunit is further set to
  • the update parameters are determined by:
  • the difference between the model parameter and the updated parameter is determined as the updated model parameter.
  • the determining subunit is further configured to transmit the modeled signal according to the I channel, the Q channel transmit modeling signal, the I channel modeling output signal, and the Q channel in the following manner Path modeling output signal to determine signal output error:
  • r(t) is the signal output error
  • p I (a) is the modeling output signal of the I channel
  • p Q (a) is the modeling output signal of the Q channel
  • s I (a) is the modeling output signal of the I channel signal
  • s Q (a) is the Q channel transmission modeling signal
  • N is the length of the modeling characteristic signal.
  • the above modules can be implemented by software or hardware, and the latter can be implemented in the following ways, but not limited to this: the above modules are all located in the same processor; or, the above modules can be combined in any combination The forms are located in different processors.
  • Embodiments of the present application further provide a computer-readable storage medium, where a computer program is stored in the computer-readable storage medium, wherein the computer program is configured to execute the steps in any one of the above method embodiments when running.
  • the above-mentioned computer-readable storage medium may include, but is not limited to, a USB flash drive, a read-only memory (Read-Only Memory, referred to as ROM for short), and a random access memory (Random Access Memory, referred to as RAM for short) , mobile hard disk, magnetic disk or CD-ROM and other media that can store computer programs.
  • ROM Read-Only Memory
  • RAM Random Access Memory
  • Embodiments of the present application further provide an electronic device, including a memory and a processor, where a computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.
  • the above-mentioned electronic device may further include a transmission device and an input-output device, wherein the transmission device is connected to the above-mentioned processor, and the input-output device is connected to the above-mentioned processor.
  • modules or steps of the present application can be implemented by a general-purpose computing device, and they can be centralized on a single computing device, or distributed in a network composed of multiple computing devices
  • they can be implemented in program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, can be performed in a different order than shown here.
  • the described steps, or they are respectively made into individual integrated circuit modules, or a plurality of modules or steps in them are made into a single integrated circuit module to realize.
  • the present application is not limited to any particular combination of hardware and software.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Physics & Mathematics (AREA)
  • Electromagnetism (AREA)
  • Nonlinear Science (AREA)
  • Optical Communication System (AREA)

Abstract

本申请实施例提供了一种信号补偿处理方法及装置，其中，该方法包括：获取接收I路发射数据信号得到的I路接收数据信号，以及接收Q路发射数据信号得到的Q路接收数据信号；对所述I路接收数据信号与所述Q路接收数据信号进行级联变换，得到数据特征信号；根据预先确定的目标模型参数对所述数据特征信号进行补偿，得到IQ均衡信号，可以解决相关技术中针对光通信中IQ信号之间相互影响的误码率高的问题，基于预先确定的目标模型参数对IQ接收数据信号级联变换得到的数据特征信号进行补偿，得到IQ均衡信号，降低了IQ信号的误码率。

Description

一种信号补偿处理方法及装置
相关申请的交叉引用
本申请基于2020年8月31日提交的发明名称为“一种信号补偿处理方法及装置”的中国专利申请CN202010899442.X，并且要求该专利申请的优先权，通过引用将其所公开的内容全部并入本申请。
技术领域
本申请实施例涉及通信领域，具体而言，涉及一种信号补偿处理方法及装置。
背景技术
IQ光信号是通信中一种常用的光调制手段。通过两路正交信号上变换后叠加生成，再在接收端进行对应的下变换，可以对两路信号进行分离。然而，对于接入网系统，当I路和Q路信号不同源的时候，两路信号叠加会由于功率失配，时钟同步，非线性效应等问题而极大恶化信噪比，从而降低系统的容量。
此时则需要利用光通信中常用的均衡算法，如最小均方误差(SISO-LMS)[1]，递归最小二乘等算法(Recursive,Least Square，简称为RLS)对信号失真进行补偿。然而，即使使用了基于Volterra级数的非线性项，对于较强的非线性效应，尤其是I，Q两路互相影响的时候，LMS(Least Mean Square)和RLS的补偿作用也是相当有限的。因此，具有复杂非线性映射能力的深度神经网络(SISO-DNN)[2]就被引入到了通信系统中作为后均衡器而存在。已有研究表明，深度神经网络(Deep Neural Network，简称为DNN)对于信号的后均衡等应用场合均具有良好的表现。然而，当系统非线性较强的时候，使用传统深度神经网络需要大幅度提升系统的复杂度以提供更好的均衡性能。因此，这对于实时性要求极高的通信网络来说是非常不利的因素。而且，目前已发表的大多数神经网络算法均是基于单入单出(Single Input Single Output，简称为SISO)系统，而针对两路信号，如IQ两路信号之间相互影响的均衡网络研究较少。因此，开发适用于光通信的IQ信号的神经网络就成为了当前一个急需解决的问题，采用传统的深度网络误码率高，且算法复杂度高。
发明内容
本申请实施例提供了一种信号补偿处理方法及装置，以至少解决相关技术中针对光通信中IQ信号之间相互影响的误码率高的问题。
根据本申请的一个实施例，提供了一种信号补偿处理方法，包括：
获取接收I路发射数据信号得到的I路接收数据信号，以及接收Q路发射数据信号得到的Q路接收数据信号；
对所述I路接收数据信号与所述Q路接收数据信号进行级联变换，得到数据特征信号；
根据预先确定的目标模型参数对所述数据特征信号进行补偿，得到IQ均衡信号。
根据本申请的另一个实施例，还提供了一种信号补偿处理装置，包括：
第一获取模块，设置为获取接收I路发射数据信号得到的I路接收数据信号，以及接收 Q路发射数据信号得到的Q路接收数据信号；
第一级联模块，设置为对所述I路接收数据信号与所述Q路接收数据信号进行级联变换，得到数据特征信号；
补偿模块，设置为根据预先确定的目标模型参数对所述特征信号进行补偿，得到IQ均衡信号。
根据本申请的又一个实施例，还提供了一种计算机可读存储介质，所述计算机可读存储介质中存储有计算机程序，其中，所述计算机程序被设置为运行时执行上述任一项方法实施例中的步骤。
根据本申请的又一个实施例，还提供了一种电子装置，包括存储器和处理器，所述存储器中存储有计算机程序，所述处理器被设置为运行所述计算机程序以执行上述任一项方法实施例中的步骤。
通过本申请，获取接收I路发射数据信号得到的I路接收数据信号，以及接收Q路发射数据信号得到的Q路接收数据信号；对所述I路接收数据信号与所述Q路接收数据信号进行级联变换，得到特征信号；根据预先确定的目标模型参数对所述数据特征信号进行补偿，得到IQ均衡信号，可以解决相关技术中针对光通信中IQ信号之间相互影响的误码率高的问题，基于预先确定的目标模型参数对IQ接收数据信号级联变换得到的数据特征信号进行补偿，得到IQ均衡信号，降低了IQ信号的误码率。
附图说明
图1是本申请实施例的信号补偿处理方法的移动终端的硬件结构框图；
图2是根据本申请实施例的信号补偿处理方法的流程图；
图3是根据本申请实施例的信号补偿处理方法的流程图；
图4为本申请实施例的IQ信号的接收补偿方法的流程图；
图5是根据本申请实施例的异构深度神经网络的多入多出多分支神经网络均衡器的示意图；
图6是根据本申请实施例的验证性实验平台的示意图；
图7是根据本申请实施例的信号补偿处理装置的结构框图；
图8是根据本申请优选实施例的信号补偿处理装置的结构框图。
具体实施方式
下文中将参考附图并结合实施例来详细说明本申请的实施例。
需要说明的是，本申请的说明书和权利要求书及上述附图中的术语“第一”、“第二”等是用于区别类似的对象，而不必用于描述特定的顺序或先后次序。
本申请实施例中所提供的方法实施例可以在移动终端、计算机终端或者类似的运算装置中执行。以运行在移动终端上为例，图1是本申请实施例的信号补偿处理方法的移动终端的硬件结构框图，如图1所示，移动终端可以包括一个或多个(图1中仅示出一个)处理器102(处理器102可以包括但不限于微处理器MCU或可编程逻辑器件FPGA等的处理装置)和用于存储数据的存储器104，其中，上述移动终端还可以包括用于通信功能的传输设备106以及输入输出设备108。本领域普通技术人员可以理解，图1所示的结构仅为示意，其并不对上 述移动终端的结构造成限定。例如，移动终端还可包括比图1中所示更多或者更少的组件，或者具有与图1所示不同的配置。
存储器104可用于存储计算机程序，例如，应用软件的软件程序以及模块，如本申请实施例中的数据处理方法对应的计算机程序，处理器102通过运行存储在存储器104内的计算机程序，从而执行各种功能应用以及数据处理，即实现上述的方法。存储器104可包括高速随机存储器，还可包括非易失性存储器，如一个或者多个磁性存储装置、闪存、或者其他非易失性固态存储器。在一些实例中，存储器104可进一步包括相对于处理器102远程设置的存储器，这些远程存储器可以通过网络连接至移动终端。上述网络的实例包括但不限于互联网、企业内部网、局域网、移动通信网及其组合。
传输装置106用于经由一个网络接收或者发送数据。上述的网络具体实例可包括移动终端的通信供应商提供的无线网络。在一个实例中，传输装置106包括一个网络适配器(Network Interface Controller，简称为NIC)，其可通过基站与其他网络设备相连从而可与互联网进行通讯。在一个实例中，传输装置106可以为射频(Radio Frequency，简称为RF)模块，其用于通过无线方式与互联网进行通讯。
在本实施例中提供了一种运行于上述移动终端或网络架构的信号补偿处理方法，图2是根据本申请实施例的信号补偿处理方法的流程图，如图2所示，该流程包括如下步骤：
步骤S202，获取接收I路发射数据信号得到的I路接收数据信号，以及接收Q路发射数据信号得到的Q路接收数据信号；
步骤S204，对所述I路接收数据信号与所述Q路接收数据信号进行级联变换，得到数据特征信号；
在一实施例中，对所述I路接收数据信号与所述Q路接收数据信号进行级联变换，得到所述数据特征信号包括：
本实施例中，上述步骤S204具体可以包括：分别对所述I路接收数据信号与所述Q路接收数据信号按照窗口长度taps进行采样，得到N组I路采样信号与N组Q路采样信号，其中，对于所述N组I路采样信号与所述N组Q路采样信号中每组采样信号的中心信号为当前的采样点，所述中心信号的左侧的(taps-1)/2个点为所述当前采样点向前截取(taps-1)/2个采样点，所述中心信号的右侧的(taps-1)/2个点为所述当前采样点向后截取(taps-1)/2个采样点，若向前或向后的采样点不足(taps-1)/2，补0，所述taps为奇数，所述N为所述I路接收数据信号与所述Q路接收数据信号的长度；分别将长度为所述taps的所述N组I路采样信号与所述N组Q路采样信号进行级联，得到包括长度为2taps的N个信号的所述数据特征信号。
步骤S206，根据预先确定的目标模型参数对所述数据特征信号进行补偿，得到IQ均衡信号。
通过上述步骤S202至S206，获取接收I路发射数据信号得到的I路接收数据信号，以及接收Q路发射数据信号得到的Q路接收数据信号；对所述I路接收数据信号与所述Q路接收数据信号进行级联变换，得到特征信号；根据预先确定的目标模型参数对所述数据特征信 号进行补偿，得到IQ均衡信号，可以解决相关技术中针对光通信中IQ信号之间相互影响的误码率高的问题，基于预先确定的目标模型参数对IQ接收数据信号级联变换得到的数据特征信号进行补偿，得到IQ均衡信号，降低了IQ信号的误码率。
在一实施例中，上述步骤S206具体可以包括：
S2061，根据所述目标模型参数对所述数据特征信号进行映射得到I路线性数据信号、Q路线性数据信号、I路非线性数据信号以及Q路非线性数据信号；
在一实施例中，上述步骤S2061具体可以包括：
根据所述目标模型参数处理所述数据特征信号的I路光信号的线性串扰，得到所述I路线性数据信号，具体通过以下方式根据所述目标模型参数对所述数据特征信号进行线性映射得到所述I路线性数据信号：
Figure PCTCN2021115480-appb-000001
其中，1≤t≤N，N为所述数据特征信号的长度；
根据所述目标模型参数处理所述数据特征信号的Q路光信号的线性串扰，得到所述Q路线性数据信号，具体通过以下方式根据所述目标模型参数对所述数据特征信号进行线性映射得到所述Q路线性数据信号：
Figure PCTCN2021115480-appb-000002
根据所述目标模型参数处理所述数据特征信号的IQ两路光信号之间的串扰，得到所述I路非线性数据信号和所述Q路非线性数据信号，具体通过以下方式根据所述目标模型参数对所述数据特征信号进行非线性映射得到所述I路非线性数据信号和所述Q路非线性数据信号：
Figure PCTCN2021115480-appb-000003
其中，
Figure PCTCN2021115480-appb-000004
Figure PCTCN2021115480-appb-000005
为所述目标模型参数，f(x)为激活函数；
S2062，将所述I路线性数据信号与所述I路非线性数据信号之和确定为I路均衡信号，并将所述Q路线性数据信号与所述Q路非线性数据信号之和确定为Q路均衡信号，其中，所述IQ均衡信号包括所述I路均衡信号与所述Q路均衡信号。
图3是根据本优选实施例的信号补偿处理方法的流程图，如图3所示，在对接收IQ发射数据信号得到的IQ接收数据信号进行级联变换，得到数据特征信号之前，所述方法还包括：
步骤S302，获取接收I路发射建模信号得到的I路接收建模信号，以及接收Q路发射建模信号得到的Q路接收建模信号；
步骤S304，对所述I路接收建模信号与所述Q路接收建模信号进行级联变换，得到建模特征信号；
上述步骤S304具体可以包括：分别对所述I路接收建模信号与所述Q路接收建模信号按照窗口长度taps进行采样，得到N组I路采样信号与N组Q路采样信号，其中，对于所述N 组I路采样信号与所述N组Q路采样信号中每组采样信号的中心信号为当前的采样点，所述中心信号的左侧的(taps-1)/2个点为所述当前采样点向前截取(taps-1)/2个采样点，所述中心信号的右侧的(taps-1)/2个点为所述当前采样点向后截取(taps-1)/2个采样点，若向前或向后的采样点不足(taps-1)/2，补0，所述taps为奇数，所述N为所述I路接收建模信号与所述Q路接收建模信号的长度；分别将长度为所述taps的所述N组I路采样信号与所述N组Q路采样信号进行级联，得到包括长度为2taps的N个信号的所述建模特征信号。
步骤S306，根据所述建模特征信号对模型参数进行更新，得到所述目标模型参数。
本实施例中，上述步骤S306具体可以包括：
重复执行以下步骤，直到迭代次数t满足预设迭代条件，确定更新后的模型参数为所述目标模型参数：
当t＝0,初始化所述模型参数，根据所述模型参数对所述建模特征信号进行映射得到I路线性建模信号、Q路线性建模信号、I路非线性建模信号以及Q路非线性建模信号；进一步的，根据所述模型参数处理所述建模特征信号的I路光信号的线性串扰，得到所述I路线性建模信号，具体的，可以通过以下方式根据所述模型参数对所述建模特征信号进行线性映射得到所述I路线性建模信号：
Figure PCTCN2021115480-appb-000006
其中，1≤t≤N，N为所述建模特征的长度；根据所述模型参数处理所述建模特征信号的Q路光信号的线性串扰，得到所述Q路线性建模信号，具体地，可以通过以下方式根据所述模型参数对所述建模特征信号进行线性映射得到所述Q路线性建模信号：
Figure PCTCN2021115480-appb-000007
根据所述模型参数处理所述建模特征信号的IQ两路光信号之间的串扰，得到所述I路非线性建模信号和所述Q路非线性建模信号，具体的，可以通过以下方式根据所述模型参数对所述建模特征信号进行非线性映射得到I路非线性建模信号和Q路非线性建模信号：
Figure PCTCN2021115480-appb-000008
其中，
Figure PCTCN2021115480-appb-000009
Figure PCTCN2021115480-appb-000010
为所述模型参数，f(x)为激活函数；
根据所述I路线性建模信号与所述I路非线性建模信号进行加和处理，得到I路建模输出信号，并根据所述Q路线性建模信号与所述Q路非线性建模信号进行加和处理，得到Q路建模输出信号；
根据所述I路发射建模信号、所述Q路发射建模信号、所述I路建模输出信号以及所述Q路建模输出信号对所述模型参数进行更新，得到更新后的模型参数，其中，所述IQ发射建模信号包括所述I路发射建模信号与所述Q路发射建模信号；
t＝t+1。
本实施例中处理过程分成模型建立或训练和信号接收补偿两个阶段，两个阶段的切换，可以通过标志位指示，例如标志位为0或false，执行模型建立或训练，标志位为1或true，执行信号接收补偿，上述预设迭代条件可以通过标志位实现，也可以通过其他方式，例如固 定长度训练序列，参数收敛到一定误差内等等。
在一实施例中，上述对所述模型参数进行更新，得到所述更新后的模型参数具体可以包括：
根据所述I路发射建模信号、所述Q路发射建模信号、所述I路建模输出信号以及所述Q路建模输出信号确定信号输出误差，具体可以通过以下方式根据所述I路发射建模信号、所述Q路发射建模信号、所述I路建模输出信号以及所述Q路建模输出信号确定信号输出误差：
Figure PCTCN2021115480-appb-000011
其中，r(t)为所述信号输出误差，p I(a)为I路建模输出信号，p Q(a)为Q路建模输出信号，s I(a)为I路发射建模信号，s Q(a)为Q路发射建模信号，N为所述建模特征信号的长度；
根据所述信号输出误差对所述模型参数进行更新，得到所述更新后的模型参数，进一步的，分别确定所述信号输出误差对所述模型参数的偏导数，得到与所述模型参数维度相同的目标矩阵，根据所述目标矩阵对所述模型参数进行更新，得到所述更新后的模型参数，具体通过以下方式确定更新参数：
Figure PCTCN2021115480-appb-000012
Figure PCTCN2021115480-appb-000013
Figure PCTCN2021115480-appb-000014
M(t)＝β 1M(t-1)+(1-β 1)G(t)，
V(t)＝β 2V(t-1)+(1-β 2)G 2(t)，其中，H(t)为所述更新参数，M(t)为动量矩阵，V(t)为速度矩阵，G(t)为所述目标矩阵，α＝0.001，β 1＝0.9，β 2＝0.999，ε＝10 -8，M(-1)＝0，V(-1)＝0，0≤t≤t max，t max为所述预设迭代次数阈值，将所述模型参数与所述更新参数的差值确定为所述更新后的模型参数。
本实施例的输入信号带有记忆性，通过记录前后的几个符号进行联合的补偿和估计，采用3分支，一个分支处理I路光信号的线性串扰，一个分支处理Q路光信号的线性串扰，两个线性分支可以是MLP的线性加权网络；还有一个分支处理IQ两路光信号之间的串扰。由于IQ光信号相互串扰并不是很复杂，可以采用两层隐藏层网络来训练非线性，极大的降低了系统的复杂度。在输出端分别将I路线性输出和非线性输出，Q路线性输出和非线性输出加和，得到的补偿信号既均衡了线性失真，又均衡了非线性失真，达到了很好的效果。对同样或类似架构的IQ两路存在串扰的无线信号，此方法同样适用，在此不再赘述。采用本实施例，可以有效提升系统的性能。本实施例提出的均衡方法在两个发射端功率的情况下，可以将工作区间提升2.33倍-8倍以上。同时，可以实现最多3.63dB Q值的提升。
下面以标志位的方式指示从模型建立与信号补偿进行切换为例，对本实施例进行详细说明。
图4为本申请实施例的IQ信号的接收补偿方法的流程图，如图4所示，可以包括以下步 骤：
步骤401，初始化模型参数，并初始化标志位为flag＝False，具体为：
设置模型隐藏层的层数为2，第一隐藏层节点数为H 1,第二隐藏层节点数为H 2，激活函数f(x)，输入层I，Q路信号窗口分别为taps，隐藏层每层节点数，权重矩阵
Figure PCTCN2021115480-appb-000015
和偏置向量
Figure PCTCN2021115480-appb-000016
的初始值为符合0均值高斯分布的随机矩阵或向量，并初始化迭代次数epoch＝0及迭代终止值epoch_max，标志位flag＝False。其中，权重矩阵和偏置向量的维度如表1所示。
表1
Figure PCTCN2021115480-appb-000017
步骤402，判断是否flag＝true，如果flag＝False,转到步骤403；如果flag＝True,转到步骤414。
步骤403，接收并记录I路接收建模信号，其中，发射端加载原始发射建模信号s(t)，具体通过发射建模信号加载装置加载原始发射建模信号s(t)。调制原始发射建模信号并记录为I路发射建模信号，具体通过I路建模信号发射装置对s(t)进行调制并记录I路发射建模信号x I(t)。调制原始发射建模信号并记录为Q路发射建模信号，具体通过Q路建模信号发射装置对s(t)进行调制并记录Q路发射建模信号x Q(t)。
具体地，通过I路建模信号接收装置接收并记录x I(t)，得到I路接收建模信号y I(t)。
步骤404，接收并记录Q路接收建模信号。具体地：
通过Q路建模信号接收装置接收并记录x I(t)，得到Q路接收建模信号y I(t)。
步骤405，根据记录的I路接收建模信号，Q路接收建模信号，级联变换并记录为建模特征信号。
具体地，通过建模级联变换装置将y I(t)，y I(t)级联变换为建模特征信号k(t)。分别对所述I路接收数据信号与所述Q路接收数据信号按照窗口长度taps进行采样，得到N组I路采样信号与N组Q路采样信号，其中，对于所述N组I路采样信号与所述N组Q路采样信号中每组采样信号的中心信号为当前的采样点，所述中心信号的左侧的(taps-1)/2个点为所述当 前采样点向前截取(taps-1)/2个采样点，所述中心信号的右侧的(taps-1)/2个点为所述当前采样点向后截取(taps-1)/2个采样点，若向前或向后的采样点不足(taps-1)/2，补0，所述taps为奇数，所述N为所述I路接收数据信号与所述Q路接收数据信号的长度，分别将长度为所述taps的所述N组I路采样信号与所述N组Q路采样信号进行级联，得到包括长度为2taps的N个信号的所述数据特征信号，具体变换方式为：
Figure PCTCN2021115480-appb-000018
其中，每个k(t)都是一个列向量，其长度为2*taps。经过这一步的变换，此时的信号已经具有记忆性，即输入到下一级的每路信号都包括前后各taps/2-1个信号。采用该信号作为输出的模块都带有记忆性。
步骤406，对建模特征信号进行线性映射得到I路线性建模信号。具体地：
通过I路线性映射模块和k(t)得到的I路线性建模信号：
Figure PCTCN2021115480-appb-000019
步骤407，对建模特征信号进行线性映射得到Q路线性建模信号。具体地：
通过Q路线性映射模块和k(t)得到的Q路线性建模信号：
Figure PCTCN2021115480-appb-000020
步骤408，对建模特征信号进行非线性映射得到I路非线性建模信号和Q路非线性建模信号。具体地：
通过非线性映射模块和k(t)得到I路非线性建模信号和Q路非线性建模信号：
Figure PCTCN2021115480-appb-000021
其中，f(a)为激活函数，其表达式可以为
Figure PCTCN2021115480-appb-000022
这里x可以为标量，向量，矩阵，且f(x)的维度与x相同。当x为矩阵或者向量时，f(x)表示对矩阵或向量中的每个元素使用上式。
步骤409，根据记录的I路线性建模信号和I路非线性建模信号，加和并记录I路建模输出信号。
具体地，通过I路加和模块对I路线性建模信号和I路非线性建模信号进行加和，得到I路建模输出信号p I(t)＝l I(t)+n I(t),1≤t≤N。
步骤410，根据记录的Q路线性建模信号和Q路非线性建模信号，加和并记录Q路建模输出信号。
具体地，通过Q路加和模块对Q路线性建模信号和Q路非线性建模信号进行加和，得到Q路建模输出信号p Q(t)＝l Q(t)+n Q(t),1≤t≤N。
步骤411，根据记录的I路建模输出信号，Q路建模输出信号，I路发射建模信号，Q路发射建模信号，计算输出的误差并对模型参数
Figure PCTCN2021115480-appb-000023
进行更新。
具体地，通过误差计算及模型参数更新模块对模型参数进行更新，具体步骤为：
(1)初始化参数α＝0.001，β 1＝0.9，β 2＝0.999，ε＝10 -8，动量矩阵M(-1)＝0，速度矩阵V(-1)＝0，t＝0，设置最大迭代次数为t max。
(2)对步骤(1)中所有模型参数进行如下更新，下面以
Figure PCTCN2021115480-appb-000024
为例进行说明。
令
Figure PCTCN2021115480-appb-000025
计算r(t)对
Figure PCTCN2021115480-appb-000026
的偏导数，得到与
Figure PCTCN2021115480-appb-000027
维度相同的矩阵G(t)；
M(t)＝β 1M(t-1)+(1-β 1)G(t)；
V(t)＝β 2V(t-1)+(1-β 2)G 2(t),其中G 2(t)代表矩阵G(t)元素的点积；
Figure PCTCN2021115480-appb-000028
Figure PCTCN2021115480-appb-000029
更新
Figure PCTCN2021115480-appb-000030
如果t＝t max,则转到步骤411(3),否则t＝t+1重复上述步骤。
(3)如果已更新完所有的模型参数，则转到步骤412，否则转到步骤411中的步骤(1)。
步骤412，判断迭代条件是否终止，在判断结果为是的情况下，执行步骤413，否则迭代次数加1，返回步骤406、407、408。
步骤413，设置标志位flag＝True，记录模型参数，并转到步骤402。
如果迭代次数达到最大迭代次数epoch_max，则通过均衡装置记录模型参数
Figure PCTCN2021115480-appb-000031
分别为
Figure PCTCN2021115480-appb-000032
Figure PCTCN2021115480-appb-000033
步骤414，接收并记录I路接收数据信号。其中，发射端加载原始发射数据信号s'(t)，具体通过发射数据信号加载装置加载原始发射数据信号s'(t)。调制原始发射数据信号并记录为I路发射数据信号，具体通过I路数据信号发射装置对s'(t)进行调制并记录I路发射建模信号x I'(t)。调制原始发射数据信号并记录为Q路发射数据信号，具体通过Q路数据信号发射装置对s'(t)进行调制并记录I路发射建模信号x Q'(t)。
具体地，通过I路数据信号接收装置接收并记录x I'(t)，得到I路接收建模信号y I'(t)。
步骤415，接收并记录Q路接收数据信号。具体地：
通过Q路数据信号接收装置接收并记录x Q'(t)，得到I路接收建模信号y Q'(t)。
步骤416，根据记录的I路接收数据信号，Q路接收数据信号，级联变换并记录为数据特征信号。具体地，通过数据级联变换装置将y I'(t)，y Q'(t)级联变换为数据特征信号k'(t)。具体变换方式与步骤405类似，在此不再赘述。
步骤417，根据记录的模型参数和数据特征信号，均衡并记录得到I路均衡数据信号和Q路均衡数据信号q I'(t)和q Q'(t)。具体地，通过均衡装置对数据特征信号进行均衡，得到I路均衡数据信号和Q路均衡数据信号q I'(t)和q Q'(t)，具体方式为：
Figure PCTCN2021115480-appb-000034
Figure PCTCN2021115480-appb-000035
Figure PCTCN2021115480-appb-000036
Figure PCTCN2021115480-appb-000037
基于本实施例提供的均衡方法和异构深度神经网络的多入多出多分支神经网络均衡器，图5是根据本申请实施例的异构深度神经网络的多入多出多分支神经网络均衡器的示意图，如图5所示，输出端通过两个加和模块(I路加和模块211、Q路加和模块212)分别将I路的线性和非线性输出以及Q路的线性核非线性输出加和后输出。当建模过程完成后，存储参数并不再更新。此时多入多出多分支神经网络变换为均衡装置(220)，用以对步骤417中的特征信号进行均衡得到I路和Q路均衡信号。针对传统的后均衡器最小均方误差均衡器(SISO-LMS)，深度神经网络(SISO-DNN)均衡器以及本实施例提出的异构深度神经网络均衡器，在单接收机IQ光通信系统中进行了验证性实验。图6是根据本申请实施例的验证性实验平台的示意图，如图6所示，首先，发射建模信号加载装置(202)加载原始发射建模信号，并通过I路和Q路建模信号发射装置(203,204)生成I路和Q路发射建模信号，I路和Q路信号在接收端叠加后，通过I路和Q路建模信号接收装置(205,206)接收，得到I路和Q路接收 建模信号。随后，多入多出多分支神经网络中的建模级联变换装置(207)对I路和Q路接收建模信号进行步骤402的建模级联变换，得到建模特征信号。然后建模特征信号经过I路和Q路的线性映射模块(208,209)和非线性映射模块(210)的变换得到I路和Q路线性建模信号和非线性建模信号，再经过I路和Q路加和模块得到I路建模输出信号和Q路建模输出信号。随后，通过I路和Q路发射建模信号以及I路和Q路建模输出信号，通过步骤412中的算法对模型中的参数进行更新，直到达到最大迭代次数。随后固定模型参数并由均衡装置(220)加载当前模型参数。随后，从图6中的发射数据信号加载装置(214)开始。重复上述流程直到通过I路和Q路数据信号装置(217，218)接收并记录I路和Q路数据信号后，通过数据级联变换装置(219)和均衡装置(220)对信号进行级联变换和均衡，最后得到I路和Q路均衡数据信号。
上述均衡器的初始化参数如表2所示。
表2
Figure PCTCN2021115480-appb-000038
表中’--’表示无法实现或未采用该结构。可以看出相比于传统的SISO-DNN均衡方法，本实施例提出的均衡方法参数更少，因此训练所需的时间和空间就更小。虽然SISO-LMS的参数是三个中最少的，但是其性能是最差的。
根据表2中的参数设置的均衡器，在单接收机IQ光通信平台上进行了实验。在I路信号峰峰值为350mV时，改变Q路信号峰峰值所测得的三种均衡方法的Q_Factor结果图。Q_Factor值越大，系统性能越好。实验结果表明，在Q路信号峰峰值为700mV时，基于本申请所提出均衡方法实现的MIMO-MBNN的工作范围510mV大幅度优于采用传统均衡算法SISO-LMS和SISO-DNN的系统工作范围170mV。同时，本实施例提出的接收均衡方法相比于传统的SISO-LMS和SISO-DNN可为系统带来最多3.63dB的Q_Factor值性能增益。本申请实施例中，当模型训练好后，只需将变换后的接收信号输入到均衡装置中，即可得到均衡后的信号，无需其它额外的操作。
通过本实施例，有效解决了IQ光通信系统中现有均衡方式性能较差，且抗功率失配能力弱的问题。相比于传统的均衡方法，有效提升了通信系统容量和系统性能。本实施例实现也相对容易，一旦模型训练好，只要不改变信号的强度和系统状态，即可使用均衡装置对接收 信号进行均衡，无需额外的操作，当建模完成后，只要将I路接收数据和Q路接收数据经过级联变换后带入均衡装置即可得到均衡后的信号，并利用均衡后的信号进行解调，提升系统的性能。
通过以上的实施方式的描述，本领域的技术人员可以清楚地了解到根据上述实施例的方法可借助软件加必需的通用硬件平台的方式来实现，当然也可以通过硬件，但很多情况下前者是更佳的实施方式。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质(如ROM/RAM、磁碟、光盘)中，包括若干指令用以使得一台终端设备(可以是手机，计算机，服务器，或者网络设备等)执行本申请各个实施例所述的方法。
在本实施例中还提供了一种信号补偿处理装置，该装置用于实现上述实施例及优选实施方式，已经进行过说明的不再赘述。如以下所使用的，术语“模块”可以实现预定功能的软件和/或硬件的组合。尽管以下实施例所描述的装置较佳地以软件来实现，但是硬件，或者软件和硬件的组合的实现也是可能并被构想的。
图7是根据本申请实施例的信号补偿处理装置的结构框图，如图7所示，包括：
第一获取模块72，设置为获取接收I路发射数据信号得到的I路接收数据信号，以及接收Q路发射数据信号得到的Q路接收数据信号；
第一级联模块74，设置为对所述I路接收数据信号与所述Q路接收数据信号进行级联变换，得到数据特征信号；
补偿模块76，设置为根据预先确定的目标模型参数对所述特征信号进行补偿，得到IQ均衡信号。
在一实施例中，所述补偿模块76包括：
映射子模块，设置为根据所述目标模型参数对所述数据特征信号进行映射得到I路线性数据信号、Q路线性数据信号、I路非线性数据信号以及Q路非线性数据信号；
确定子模块，设置为将所述I路线性数据信号与所述I路非线性数据信号之和确定为I路均衡信号，并将所述Q路线性数据信号与所述Q路非线性数据信号之和确定为Q路均衡信号，其中，所述IQ均衡信号包括所述I路均衡信号与所述Q路均衡信号。
在一实施例中，所述映射子模块包括：
第一处理单元，设置为根据所述目标模型参数处理所述数据特征信号的I路光信号的线性串扰，得到所述I路线性数据信号；
第二处理单元，设置为根据所述目标模型参数处理所述数据特征信号的Q路光信号的线性串扰，得到所述Q路线性数据信号；
第三处理单元，设置为根据所述目标模型参数处理所述数据特征信号的IQ两路光信号之间的串扰，得到所述I路非线性数据信号和所述Q路非线性数据信号。
在一实施例中，所述第一处理单元，还设置为通过以下方式根据所述目标模型参数对所 述数据特征信号进行线性映射得到所述I路线性数据信号：
Figure PCTCN2021115480-appb-000039
其中，1≤t≤N，N为所述数据特征信号的长度；和/或
所述第二处理单元，还设置为通过以下方式根据所述目标模型参数对所述数据特征信号进行线性映射得到所述Q路线性数据信号：
Figure PCTCN2021115480-appb-000040
和/或
所述第三处理单元，还设置为通过以下方式根据所述目标模型参数对所述数据特征信号进行非线性映射得到所述I路非线性数据信号和所述Q路非线性数据信号：
Figure PCTCN2021115480-appb-000041
其中，
Figure PCTCN2021115480-appb-000042
为所述目标模型参数，f(x)为激活函数。
在一实施例中，所述第一级联模块包括：
第一采样子模块，设置为分别对所述I路接收数据信号与所述Q路接收数据信号按照窗口长度taps进行采样，得到N组I路采样信号与N组Q路采样信号，其中，对于所述N组I路采样信号与所述N组Q路采样信号中每组采样信号的中心信号为当前的采样点，所述中心信号的左侧的(taps-1)/2个点为所述当前采样点向前截取(taps-1)/2个采样点，所述中心信号的右侧的(taps-1)/2个点为所述当前采样点向后截取(taps-1)/2个采样点，若向前或向后的采样点不足(taps-1)/2，补0，所述taps为奇数，所述N为所述I路接收数据信号与所述Q路接收数据信号的长度；
第一级联子模块，设置为分别将长度为所述taps的所述N组I路采样信号与所述N组Q路采样信号进行级联，得到包括长度为2taps的N个信号的所述数据特征信号。
图8是根据本申请优选实施例的信号补偿处理装置的结构框图，如图8所示，所述装置还包括：
第二获取模块82，设置为获取接收I路发射建模信号得到的I路接收建模信号，以及接收Q路发射建模信号得到的Q路接收建模信号；
第二级联模块84，设置为对所述I路接收建模信号与所述Q路接收建模信号进行级联变换，得到建模特征信号；
更新模块86，设置为根据所述建模特征信号对模型参数进行更新，得到所述目标模型参数。
在另一实施例中，所述第二级联模块84包括：
第二采样子模块，设置为分别对所述I路接收建模信号与所述Q路接收建模信号按照窗口长度taps进行采样，得到N组I路采样信号与N组Q路采样信号，其中，对于所述N组I路采样信号与所述N组Q路采样信号中每组采样信号的中心信号为当前的采样点，所述中心 信号的左侧的(taps-1)/2个点为所述当前采样点向前截取(taps-1)/2个采样点，所述中心信号的右侧的(taps-1)/2个点为所述当前采样点向后截取(taps-1)/2个采样点，若向前或向后的采样点不足(taps-1)/2，补0，所述taps为奇数，所述N为所述I路接收建模信号与所述Q路接收建模信号的长度；
第二级联子模块，设置为分别将长度为所述taps的所述N组I路采样信号与所述N组Q路采样信号进行级联，得到包括长度为2taps的N个信号的所述建模特征信号。
在另一实施例中，所述更新模块86包括：
执行子模块，设置为重复执行以下步骤，直到迭代次数t满足预设迭代条件，确定更新后的模型参数为所述目标模型参数：
映射单元，设置为当t＝0,初始化所述模型参数，根据所述模型参数对所述建模特征信号进行映射得到I路线性建模信号、Q路线性建模信号、I路非线性建模信号以及Q路非线性建模信号；
加和单元，设置为根据所述I路线性建模信号与所述I路非线性建模信号进行加和处理，得到I路建模输出信号，并根据所述Q路线性建模信号与所述Q路非线性建模信号进行加和处理，得到Q路建模输出信号；
更新单元，设置为根据所述I路发射建模信号、所述Q路发射建模信号、所述I路建模输出信号以及所述Q路建模输出信号对所述模型参数进行更新，得到更新后的模型参数，其中，所述IQ发射建模信号包括所述I路发射建模信号与所述Q路发射建模信号；t＝t+1。
在另一实施例中，所述映射单元包括：
第一处理子单元，设置为根据所述模型参数处理所述建模特征信号的I路光信号的线性串扰，得到所述I路线性建模信号；
第二处理子单元，设置为根据所述模型参数处理所述建模特征信号的Q路光信号的线性串扰，得到所述Q路线性建模信号；
第三处理子单元，设置为根据所述模型参数处理所述建模特征信号的IQ两路光信号之间的串扰，得到所述I路非线性建模信号和所述Q路非线性建模信号。
在另一实施例中，所述第一处理子单元，还设置为通过以下方式根据所述模型参数对所述建模特征信号进行线性映射得到所述I路线性建模信号：
Figure PCTCN2021115480-appb-000043
其中，1≤t≤N，N为所述建模特征的长度；和/或
所述第二处理子单元，还设置为通过以下方式根据所述模型参数对所述建模特征信号进行线性映射得到所述Q路线性建模信号：
Figure PCTCN2021115480-appb-000044
和/或
所述第三处理子单元，还设置为通过以下方式根据所述模型参数对所述建模特征信号进行非线性映射得到I路非线性建模信号和Q路非线性建模信号：
Figure PCTCN2021115480-appb-000045
其中，
Figure PCTCN2021115480-appb-000046
Figure PCTCN2021115480-appb-000047
为所述模型参数，f(x)为激活函数。
在另一实施例中，所述更新单元包括：
确定子单元，设置为根据所述I路发射建模信号、所述Q路发射建模信号、所述I路建模输出信号以及所述Q路建模输出信号确定信号输出误差；
更新子单元，设置为根据所述信号输出误差对所述模型参数进行更新，得到所述更新后的模型参数。
在另一实施例中，所述更新子单元，还设置为
分别确定所述信号输出误差对所述模型参数的偏导数，得到与所述模型参数维度相同的目标矩阵；
根据所述目标矩阵对所述模型参数进行更新，得到所述更新后的模型参数。
在另一实施例中，所述更新子单元，还设置为
通过以下方式确定更新参数：
Figure PCTCN2021115480-appb-000048
Figure PCTCN2021115480-appb-000049
Figure PCTCN2021115480-appb-000050
M(t)＝β 1M(t-1)+(1-β 1)G(t)，
V(t)＝β 2V(t-1)+(1-β 2)G 2(t)，其中，H(t)为所述更新参数，M(t)为动量矩阵，V(t)为速度矩阵，G(t)为所述目标矩阵，α＝0.001，β 1＝0.9，β 2＝0.999，ε＝10 -8，M(-1)＝0，V(-1)＝0，0≤t≤t max，t max为所述预设迭代次数阈值；
将所述模型参数与所述更新参数的差值确定为所述更新后的模型参数。
在另一实施例中，所述确定子单元，还设置为通过以下方式根据所述I路发射建模信号、所述Q路发射建模信号、所述I路建模输出信号以及所述Q路建模输出信号确定信号输出误差：
Figure PCTCN2021115480-appb-000051
其中，r(t)为所述信号输出误差，p I(a)为I路建模输出信号，p Q(a)为Q路建模输出信号，s I(a)为I路发射建模信号，s Q(a)为Q路发射建模信号，N为所述建模特征信号的长度。
需要说明的是，上述各个模块是可以通过软件或硬件来实现的，对于后者，可以通过以下方式实现，但不限于此：上述模块均位于同一处理器中；或者，上述各个模块以任意组合的形式分别位于不同的处理器中。
本申请的实施例还提供了一种计算机可读存储介质，该计算机可读存储介质中存储有计算机程序，其中，该计算机程序被设置为运行时执行上述任一项方法实施例中的步骤。
在一个示例性实施例中，上述计算机可读存储介质可以包括但不限于：U盘、只读存储器(Read-Only Memory，简称为ROM)、随机存取存储器(Random Access Memory，简称为RAM)、移动硬盘、磁碟或者光盘等各种可以存储计算机程序的介质。
本申请的实施例还提供了一种电子装置，包括存储器和处理器，该存储器中存储有计算机程序，该处理器被设置为运行计算机程序以执行上述任一项方法实施例中的步骤。
在一个示例性实施例中，上述电子装置还可以包括传输设备以及输入输出设备，其中，该传输设备和上述处理器连接，该输入输出设备和上述处理器连接。
本实施例中的具体示例可以参考上述实施例及示例性实施方式中所描述的示例，本实施例在此不再赘述。
显然，本领域的技术人员应该明白，上述的本申请的各模块或各步骤可以用通用的计算装置来实现，它们可以集中在单个的计算装置上，或者分布在多个计算装置所组成的网络上，它们可以用计算装置可执行的程序代码来实现，从而，可以将它们存储在存储装置中由计算装置来执行，并且在某些情况下，可以以不同于此处的顺序执行所示出或描述的步骤，或者将它们分别制作成各个集成电路模块，或者将它们中的多个模块或步骤制作成单个集成电路模块来实现。这样，本申请不限制于任何特定的硬件和软件结合。
以上所述仅为本申请的优选实施例而已，并不用于限制本申请，对于本领域的技术人员来说，本申请可以有各种更改和变化。凡在本申请的原则之内，所作的任何修改、等同替换、改进等，均应包含在本申请的保护范围之内。

Claims (17)

  1. 一种信号补偿处理方法，包括：
    获取接收I路发射数据信号得到的I路接收数据信号，以及接收Q路发射数据信号得到的Q路接收数据信号；
    对所述I路接收数据信号与所述Q路接收数据信号进行级联变换，得到数据特征信号；
    根据预先确定的目标模型参数对所述数据特征信号进行补偿，得到IQ均衡信号。
  2. 根据权利要求1所述的方法，其中，根据预先确定的目标模型参数对所述数据特征信号进行补偿，得到所述IQ均衡信号包括：
    根据所述目标模型参数对所述数据特征信号进行映射得到I路线性数据信号、Q路线性数据信号、I路非线性数据信号以及Q路非线性数据信号；
    将所述I路线性数据信号与所述I路非线性数据信号之和确定为I路均衡信号，并将所述Q路线性数据信号与所述Q路非线性数据信号之和确定为Q路均衡信号，其中，所述IQ均衡信号包括所述I路均衡信号与所述Q路均衡信号。
  3. 根据权利要求2所述的方法，其中，根据所述目标模型参数对所述数据特征信号进行映射得到I路线性数据信号、Q路线性数据信号、I路非线性数据信号以及Q路非线性数据信号包括：
    根据所述目标模型参数处理所述数据特征信号的I路光信号的线性串扰，得到所述I路线性数据信号；
    根据所述目标模型参数处理所述数据特征信号的Q路光信号的线性串扰，得到所述Q路线性数据信号；
    根据所述目标模型参数处理所述数据特征信号的IQ两路光信号之间的串扰，得到所述I路非线性数据信号和所述Q路非线性数据信号。
  4. 根据权利要求3所述的方法，其中，
    根据所述目标模型参数处理所述数据特征信号的I路光信号的线性串扰，得到所述I路线性数据信号包括：
    通过以下方式根据所述目标模型参数对所述数据特征信号进行线性映射得到所述I路线性数据信号：
    Figure PCTCN2021115480-appb-100001
    其中，1≤t≤N，N为所述数据特征信号的长度；和/或
    根据所述目标模型参数处理所述数据特征信号的Q路光信号的线性串扰，得到所述Q路线性数据信号包括：
    通过以下方式根据所述目标模型参数对所述数据特征信号进行线性映射得到所述Q路线性数据信号：
    Figure PCTCN2021115480-appb-100002
    和/或
    根据所述目标模型参数处理所述数据特征信号的IQ两路光信号之间的串扰，得到所述I路非线性数据信号和所述Q路非线性数据信号包括：
    通过以下方式根据所述目标模型参数对所述数据特征信号进行非线性映射得到所述I路非线性数据信号和所述Q路非线性数据信号：
    Figure PCTCN2021115480-appb-100003
    其中，
    Figure PCTCN2021115480-appb-100004
    为所述目标模型参数，f(x)为激活函数。
  5. 根据权利要求1所述的方法，其中，对所述I路接收数据信号与所述Q路接收数据信号进行级联变换，得到所述数据特征信号包括：
    分别对所述I路接收数据信号与所述Q路接收数据信号按照窗口长度taps进行采样，得到N组I路采样信号与N组Q路采样信号，其中，对于所述N组I路采样信号与所述N组Q路采样信号中每组采样信号的中心信号为当前的采样点，所述中心信号的左侧的(taps-1)/2个点为所述当前采样点向前截取(taps-1)/2个采样点，所述中心信号的右侧的(taps-1)/2个点为所述当前采样点向后截取(taps-1)/2个采样点，若向前或向后的采样点不足(taps-1)/2，补0，所述taps为奇数，所述N为所述I路接收数据信号与所述Q路接收数据信号的长度；
    分别将长度为所述taps的所述N组I路采样信号与所述N组Q路采样信号进行级联，得到包括长度为2taps的N个信号的所述数据特征信号。
  6. 根据权利要求1至5中任一项所述的方法，其中，在对接收IQ发射数据信号得到的IQ接收数据信号进行级联变换，得到数据特征信号之前，所述方法还包括：
    获取接收I路发射建模信号得到的I路接收建模信号，以及接收Q路发射建模信号得到的Q路接收建模信号；
    对所述I路接收建模信号与所述Q路接收建模信号进行级联变换，得到建模特征信号；
    根据所述建模特征信号对模型参数进行更新，得到所述目标模型参数。
  7. 根据权利要求6所述的方法，其中，对所述I路接收建模信号与所述Q路接收建模信号进行级联变换，得到所述建模特征信号包括：
    分别对所述I路接收建模信号与所述Q路接收建模信号按照窗口长度taps进行采样，得到N组I路采样信号与N组Q路采样信号，其中，对于所述N组I路采样信号与所述N组Q路采样信号中每组采样信号的中心信号为当前的采样点，所述中心信号的左侧的(taps-1)/2个点为所述当前采样点向前截取(taps-1)/2个采样点，所述中心信号的右侧的(taps-1)/2个点为所述当前采样点向后截取(taps-1)/2个采样点，若向前或向后的采样点不足(taps-1)/2，补0，所述taps为奇数，所述N为所述I路接收建模信号与所述Q路接收建模信号的长度；
    分别将长度为所述taps的所述N组I路采样信号与所述N组Q路采样信号进行级联，得 到包括长度为2taps的N个信号的所述建模特征信号。
  8. 根据权利要求6所述的方法，其中，根据所述建模特征信号对所述模型参数进行更新，得到所述目标模型参数包括：
    重复执行以下步骤，直到迭代次数t满足预设迭代条件，确定更新后的模型参数为所述目标模型参数：
    当t＝0,初始化所述模型参数，根据所述模型参数对所述建模特征信号进行映射得到I路线性建模信号、Q路线性建模信号、I路非线性建模信号以及Q路非线性建模信号；
    根据所述I路线性建模信号与所述I路非线性建模信号进行加和处理，得到I路建模输出信号，并根据所述Q路线性建模信号与所述Q路非线性建模信号进行加和处理，得到Q路建模输出信号；
    根据所述I路发射建模信号、所述Q路发射建模信号、所述I路建模输出信号以及所述Q路建模输出信号对所述模型参数进行更新，得到更新后的模型参数，其中，所述IQ发射建模信号包括所述I路发射建模信号与所述Q路发射建模信号；
    t＝t+1。
  9. 根据权利要求8所述的方法，其中，根据所述模型参数对所述建模特征信号进行映射得到所述I路线性建模信号、所述Q路线性建模信号、所述I路非线性建模信号以及所述Q路非线性建模信号包括：
    根据所述模型参数处理所述建模特征信号的I路光信号的线性串扰，得到所述I路线性建模信号；
    根据所述模型参数处理所述建模特征信号的Q路光信号的线性串扰，得到所述Q路线性建模信号；
    根据所述模型参数处理所述建模特征信号的IQ两路光信号之间的串扰，得到所述I路非线性建模信号和所述Q路非线性建模信号。
  10. 根据权利要求9所述的方法，其中，
    根据所述模型参数处理所述建模特征信号的I路光信号的线性串扰，得到所述I路线性建模信号包括：
    通过以下方式根据所述模型参数对所述建模特征信号进行线性映射得到所述I路线性建模信号：
    Figure PCTCN2021115480-appb-100005
    其中，1≤t≤N，N为所述建模特征的长度；和/或
    根据所述模型参数处理所述建模特征信号的Q路光信号的线性串扰，得到所述Q路线性建模信号包括：
    通过以下方式根据所述模型参数对所述建模特征信号进行线性映射得到所述Q路线性建模信号：
    Figure PCTCN2021115480-appb-100006
    和/或
    根据所述模型参数处理所述建模特征信号的IQ两路光信号之间的串扰，得到所述I路非线性建模信号和所述Q路非线性建模信号包括：
    通过以下方式根据所述模型参数对所述建模特征信号进行非线性映射得到I路非线性建模信号和Q路非线性建模信号：
    Figure PCTCN2021115480-appb-100007
    其中，
    Figure PCTCN2021115480-appb-100008
    Figure PCTCN2021115480-appb-100009
    为所述模型参数，f(x)为激活函数。
  11. 根据权利要求8所述的方法，其中，根据所述I路发射建模信号、所述Q路发射建模信号、所述I路建模输出信号以及所述Q路建模输出信号对所述模型参数进行更新，得到所述更新后的模型参数包括：
    根据所述I路发射建模信号、所述Q路发射建模信号、所述I路建模输出信号以及所述Q路建模输出信号确定信号输出误差；
    根据所述信号输出误差对所述模型参数进行更新，得到所述更新后的模型参数。
  12. 根据权利要求11所述的方法，其中，根据所述信号输出误差对所述模型参数进行更新，得到更新后的模型参数包括：
    分别确定所述信号输出误差对所述模型参数的偏导数，得到与所述模型参数维度相同的目标矩阵；
    根据所述目标矩阵对所述模型参数进行更新，得到所述更新后的模型参数。
  13. 根据权利要求12所述的方法，其中，根据所述目标矩阵对所述模型参数进行更新，得到所述更新后的模型参数包括：
    通过以下方式确定更新参数：
    Figure PCTCN2021115480-appb-100010
    Figure PCTCN2021115480-appb-100011
    Figure PCTCN2021115480-appb-100012
    M(t)＝β 1M(t-1)+(1-β 1)G(t)，
    V(t)＝β 2V(t-1)+(1-β 2)G 2(t)，其中，H(t)为所述更新参数，M(t)为动量矩阵，V(t)为速度矩阵，G(t)为所述目标矩阵，α＝0.001，β 1＝0.9，β 2＝0.999，ε＝10 -8，M(-1)＝0，V(-1)＝0，0≤t≤t max，t max为所述预设迭代次数阈值；
    将所述模型参数与所述更新参数的差值确定为所述更新后的模型参数。
  14. 根据权利要求11所述的方法，其中，所述方法还包括：
    通过以下方式根据所述I路发射建模信号、所述Q路发射建模信号、所述I路建模输出信号以及所述Q路建模输出信号确定信号输出误差：
    Figure PCTCN2021115480-appb-100013
    其中，r(t)为所述信号输出误差，p I(a)为I路建模输出信号，p Q(a)为Q路建模输出信号，s I(a)为I路发射建模信号，s Q(a)为Q路发射建模信号，N为所述建模特征信号的长度。
  15. 一种信号补偿处理装置，包括：
    第一获取模块，设置为获取接收I路发射数据信号得到的I路接收数据信号，以及接收Q路发射数据信号得到的Q路接收数据信号；
    第一级联模块，设置为对所述I路接收数据信号与所述Q路接收数据信号进行级联变换，得到数据特征信号；
    补偿模块，设置为根据预先确定的目标模型参数对所述特征信号进行补偿，得到IQ均衡信号。
  16. 一种计算机可读存储介质，所述计算机可读存储介质中存储有计算机程序，其中，所述计算机程序被设置为运行时执行所述权利要求1至14任一项中所述的方法。
  17. 一种电子装置，包括存储器和处理器，所述存储器中存储有计算机程序，所述处理器被设置为运行所述计算机程序以执行所述权利要求1至14任一项中所述的方法。
PCT/CN2021/115480 2020-08-31 2021-08-30 一种信号补偿处理方法及装置 WO2022042736A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
EP21860568.1A EP4207697A4 (en) 2020-08-31 2021-08-30 SIGNAL COMPENSATION PROCESSING METHOD AND APPARATUS

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010899442.X 2020-08-31
CN202010899442.XA CN114124640A (zh) 2020-08-31 2020-08-31 一种信号补偿处理方法及装置

Publications (1)

Publication Number Publication Date
WO2022042736A1 true WO2022042736A1 (zh) 2022-03-03

Family

ID=80354701

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/115480 WO2022042736A1 (zh) 2020-08-31 2021-08-30 一种信号补偿处理方法及装置

Country Status (3)

Country Link
EP (1) EP4207697A4 (zh)
CN (1) CN114124640A (zh)
WO (1) WO2022042736A1 (zh)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101272373A (zh) * 2008-05-07 2008-09-24 北京北方烽火科技有限公司 一种自适应模拟正交调制失衡补偿方法和装置
CN105471784A (zh) * 2016-01-13 2016-04-06 中国人民解放军国防科学技术大学 一种联合补偿iq不平衡和pa非线性的数字预失真方法
CN110380789A (zh) * 2018-04-12 2019-10-25 中兴通讯股份有限公司 一种信号处理方法和装置
CN110460402A (zh) * 2019-07-15 2019-11-15 哈尔滨工程大学 一种基于深度学习的端到端通信系统建立方法
CN110830112A (zh) * 2019-10-16 2020-02-21 青岛海信电器股份有限公司 一种可见光通信的方法及设备

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110505020B (zh) * 2019-07-15 2022-03-15 电子科技大学 基于迁移学习的光非线性均衡方法
CN110598859B (zh) * 2019-08-01 2022-12-13 北京光锁科技有限公司 基于门控循环神经网络的非线性均衡方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101272373A (zh) * 2008-05-07 2008-09-24 北京北方烽火科技有限公司 一种自适应模拟正交调制失衡补偿方法和装置
CN105471784A (zh) * 2016-01-13 2016-04-06 中国人民解放军国防科学技术大学 一种联合补偿iq不平衡和pa非线性的数字预失真方法
CN110380789A (zh) * 2018-04-12 2019-10-25 中兴通讯股份有限公司 一种信号处理方法和装置
CN110460402A (zh) * 2019-07-15 2019-11-15 哈尔滨工程大学 一种基于深度学习的端到端通信系统建立方法
CN110830112A (zh) * 2019-10-16 2020-02-21 青岛海信电器股份有限公司 一种可见光通信的方法及设备

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP4207697A4 *

Also Published As

Publication number Publication date
CN114124640A (zh) 2022-03-01
EP4207697A4 (en) 2024-02-28
EP4207697A1 (en) 2023-07-05

Similar Documents

Publication Publication Date Title
CN108650201B (zh) 基于神经网络的信道均衡方法、译码方法及对应设备
Uncini et al. Complex-valued neural networks with adaptive spline activation function for digital-radio-links nonlinear equalization
CN1689292B (zh) 具有虚拟并联均衡器的通信接收机
CN114172767B (zh) 基于分数阶多模的信道盲均衡方法
CN112217545B (zh) 一种大规模mimo系统检测模型构建方法
JPS6346623B2 (zh)
CN104410593B (zh) 基于判决反馈模型的数字符号非线性误差修正均衡方法
CN114726702B (zh) 信道频偏的估计和补偿方法及装置
CN106161304A (zh) 一种联合信道估计的发射端iq不平衡补偿方法
WO2022042736A1 (zh) 一种信号补偿处理方法及装置
CN109818891B (zh) 一种格基约减辅助的低复杂度贪心球形译码检测方法
WO2023236608A1 (zh) 信号检测方法及终端设备
CN107438985A (zh) 处理快于奈奎斯特的信令接收信号
CN111200470A (zh) 一种适用于受非线性干扰的高阶调制信号传输控制方法
CN114666012B (zh) 一种基于mpa的多采样点联合检测的scma检测方法
JP3424723B2 (ja) 適応等化器
CN105471783B (zh) 基于输入序列的mimo系统发射端数字预失真优化方法
CN111800356B (zh) 并行变步长cma均衡算法、装置、电子设备及存储介质
Feng et al. The design of satellite equalizer based on CNN
US6693568B2 (en) Apparatus, and an associated method, for detecting digital data using MLSE detection with a dynamically-sized trellis
Wei et al. Single-channel demodulation algorithm for non-cooperative PCMA signals based on neural network
CN111884602A (zh) 一种基于单输出节点神经网络的功率放大器预失真方法
CN113973041B (zh) 一种太赫兹信号解调方法、装置和电子设备
CN107659523A (zh) 一种无线移动通信中bpsk调制的均衡系统和方法
CN113660021B (zh) 一种偏移空间调制中的低复杂度天线选择方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21860568

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 2021860568

Country of ref document: EP

Effective date: 20230331