A digital predistortion method and device for a common sense integrated OFDM baseband signal

Abstract

The application provides a digital pre-distortion method for a sensing-integrated OFDM baseband signal, comprising: obtaining a multi-subband OFDM baseband signal and a current state vector; wherein the current state vector at least includes a current service mode and a current working condition state; inputting the multi-subband OFDM baseband signal into a trained hybrid pre-distorter to generate a pre-distortion signal; wherein the hybrid pre-distorter is composed based on a set of generalized memory polynomial basis functions and a light network as a gating weight and / or a coefficient increment of each basis function; wherein the light network takes the current state vector as input to output the gating weight and / or the coefficient increment of each basis function; and outputting the pre-distortion signal to a radio frequency power amplifier. The problems that the prior art still takes communication linearization as a single target and lacks a mechanism for explicitly including sensing side compression quality into the training / update target of the digital pre-distortion method are solved.

Description

Technical Field

[0001] This application belongs to the technical field of 6G integrated sensing system and broadband communication system, and relates to a radio frequency transmission link linearization technology, and in particular to a digital predistortion method, device, electronic device and medium for integrated sensing OFDM baseband signals. Background Technology

[0002] With the development of applications such as low-altitude economy, vehicle-to-everything (V2X) communication, and satellite internet, transmitters often need to operate under greater bandwidth, higher power, and stricter spectral mask constraints. The nonlinearity and memory effect of power amplifiers can cause in-band amplitude and phase distortion, out-of-band spectral regeneration, and fluctuations in time-frequency response, leading to an increase in the error vector magnitude (EVM) on the communication side, a deterioration in the adjacent channel leakage ratio (ACPR), and a worsening of the normalized mean square error (NMSE) in the in-band.

[0003] In 6G integrated sensing scenarios, transmitted waveforms are used for sensing and detection in addition to carrying communication data. Multi-subband Orthogonal Frequency Division Multiplexing (OFDM) baseband signals typically carry communication and sensing pilots / sequences in parallel across multiple subbands. Their matched filtering or range-Doppler processing relies on known reference symbols at the transmitter to maintain coherence. The nonlinearity and memory effects introduced by the power amplifier (PA) disrupt amplitude-phase consistency and effective group delay flatness, leading to deterioration of the main and sidelobe structure of the matched filter output and an increase in the sidelobes of the range image / range-Doppler map. This manifests as increased false alarms and decreased detectability of weak targets.

[0004] The existing Digital Pre-Distortion (DPD) technology is a technique used to improve the nonlinear distortion of radio frequency power amplifiers. It models the nonlinear distortion of the power amplifier based on the input and output signals, and then reverses the original waveform on the digital baseband side to make the output signal of the power amplifier as close as possible to the original signal, thereby improving the signal quality.

[0005] However, existing technologies have the following problems:

[0006] 1) Digital predistortion methods based on memory polynomials (MP), generalized memory polynomials (GMP), or Volterra series. These methods are interpretable and easy to deploy, but under ultra-wideband and strong memory conditions, the number of model terms tends to expand, training stability and generalization ability are limited, and it is difficult to achieve rapid adaptation when switching between different business models and operating conditions.

[0007] 2) End-to-end digital predistortion methods based on deep learning. These methods can improve compensation capabilities under certain conditions, but they often face problems such as large network size, high inference latency, insufficient engineering interpretability, and difficulties in deployment on platforms such as DSP (Digital Signal Processor) / FPGA (Field-Programmable Gate Array).

[0008] 3) Research on integrated sensing systems often focuses on waveform design or receiver processing. However, when there is nonlinear distortion in the RF front end, existing solutions still take communication linearization as the sole objective. They lack a mechanism to explicitly incorporate the compression quality of the sensing side (such as sidelobe energy, peak sidelobe ratio, integral sidelobe ratio, etc.) into the training / update objectives of digital predistortion methods, making it difficult to achieve a controllable, switchable, and verifiable trade-off between communication and sensing. Summary of the Invention

[0009] This application provides a digital predistortion method for integrated sensing OFDM baseband signals, which addresses the problems of existing digital predistortion methods under ultra-wideband and strong memory conditions, such as easy expansion of model terms, limited training stability and generalization ability, and the fact that most existing solutions still take communication linearization as the single objective and lack a mechanism to explicitly incorporate the compression quality of the sensing side into the training / update objectives of the digital predistortion method.

[0010] In a first aspect, this application provides a digital predistortion method for integrated sensing OFDM baseband signals, comprising: acquiring a multi-subband OFDM baseband signal and a current state vector; wherein the current state vector includes at least the current service mode and the current operating condition; inputting the multi-subband OFDM baseband signal into a trained hybrid predistorter to generate a predistorted signal; wherein the hybrid predistorter is constructed based on a set of generalized memory polynomial basis functions and introduces a lightweight network as the gate weights and / or coefficient increments of each basis function; wherein the lightweight network takes the current state vector as input and outputs the gate weights and / or coefficient increments of each basis function; and outputting the predistorted signal to an RF power amplifier.

[0011] In this application, the pre-trained hybrid predistorter is constructed based on a set of generalized memory polynomial basis functions and incorporates a lightweight network as the gating weights and / or coefficient increments for each basis function. The generalized memory polynomial part provides interpretable, low-complexity nonlinear memory compensation capabilities, while the lightweight network outputs gating weights and / or coefficient increments according to the current business mode and operating condition, enabling adaptive selection and rapid fine-tuning of different basis functions, thereby maintaining stable performance under different communication and sensing task requirements.

[0012] In one implementation of the first aspect, the hybrid predistorter is trained as follows: A multi-subband OFDM baseband signal is constructed, and communication data and sensing reference symbols are configured in each subband; a hybrid predistorter is constructed, wherein the hybrid predistorter uses a set of generalized memory polynomial basis functions as the main compensation structure, and a lightweight network is introduced as a gating weight and / or coefficient increment generator; the lightweight network takes a state vector as input and outputs the gating weights and / or coefficient increments of each basis function; input and output data of the RF power amplifier are collected to construct a training set; a multi-objective optimization loss function is constructed; wherein the multi-objective optimization loss function includes at least a communication-side loss term, a sensing-side loss term, and a phase / group delay consistency constraint term; based on the training set and the multi-objective optimization loss function, the hybrid predistorter is trained to obtain trained hybrid predistorter parameters.

[0013] In one implementation of the first aspect, the communication-side loss term is formed by a combination of one or more of the following: normalized mean square error, error vector magnitude, and adjacent channel power ratio, and is calculated separately for each sub-band and then weighted and summed.

[0014] In one implementation of the first aspect, the step of constructing the sensing-side loss term includes: using the sensing reference symbol as a reference, performing matched filtering or range-Doppler processing on the output signal of the radio frequency power amplifier to generate a range-Doppler map; dividing the main lobe region and side lobe region in the range-Doppler map; and statistically analyzing the energy in the side lobe region and / or calculating the peak sidelobe ratio or integral sidelobe ratio as the sensing-side loss term.

[0015] In one implementation of the first aspect, the construction step of the phase / group delay consistency constraint includes: transforming the output signal of the RF power amplifier to the frequency domain and calculating its equivalent transfer function relative to the input signal; calculating the group delay fluctuation and / or residual phase fluctuation within the operating bandwidth based on the equivalent transfer function; and constructing the phase / group delay consistency constraint using the group delay fluctuation and / or the residual phase fluctuation.

[0016] In one implementation of the first aspect, the hybrid predistorter is trained using a two-stage training strategy, including: the first stage mainly uses the communication-side loss term to train the basis function coefficients of the generalized memory polynomial; the second stage introduces the perception-side loss term and the phase / group delay consistency constraint term to train the lightweight network so that it can generate gating weights and / or coefficient increments that are adapted to each basis function based on the state vector.

[0017] In one implementation of the first aspect, the following steps are also included: after updating the gating weight and / or coefficient increment, if the communication index and / or perception index deteriorate beyond a preset threshold, the parameters before the update are rolled back.

[0018] Secondly, this application provides a digital predistortion device for integrated sensing OFDM baseband signals, comprising: a signal acquisition and state perception module for acquiring multi-subband OFDM baseband signals and a current state vector; wherein the current state vector includes at least the current service mode and the current operating condition; a predistortion signal generation module for inputting the multi-subband OFDM baseband signals into a trained hybrid predistorter to generate a predistortion signal; wherein the hybrid predistorter is constructed based on a set of generalized memory polynomial basis functions and introduces a lightweight network as the gate weights and / or coefficient increments of each basis function; wherein the lightweight network takes the current state vector as input and outputs the gate weights and / or coefficient increments of each basis function; and a predistortion signal transmission module for outputting the predistortion signal to an RF power amplifier.

[0019] Thirdly, this application provides an electronic device, including: a processor and a memory; the memory is used to store a computer program; the processor is used to execute the computer program stored in the memory, so that the electronic device performs the above-described digital predistortion method for integrated inductive OFDM baseband signals.

[0020] Fourthly, this application provides a computer-readable storage medium storing a computer program thereon, which, when executed by an electronic device, implements the above-described digital predistortion method for integrated inductive OFDM baseband signals.

[0021] As described above, the digital predistortion method, apparatus, device, and medium for integrated inductive OFDM baseband signals described in this application have the following beneficial effects:

[0022] 1) It adopts a generalized memory polynomial as the main structure and a lightweight network as the gating weight / coefficient increment generator, which balances the compensation capability and the complexity of engineering implementation, and is easy to deploy in embedded systems on platforms such as DSP / FPGA.

[0023] 2) Introduce group delay and residual phase consistency constraints to suppress the adverse effects of phase / group delay fluctuations on the range-Doppler sidelobe structure and improve the stability of sensing results.

[0024] 3) Communication-sensing joint optimization: The sensing compression quality (sidelobe energy, peak sidelobe ratio, integral sidelobe ratio, etc.) is explicitly incorporated into the multi-objective optimization loss function of the digital predistortion method, thereby improving the output quality of sensing processing while meeting the requirements of communication linearization.

[0025] 4) Supports communication priority, perception priority and balanced mode, and adaptively schedules weight and parameter updates based on threshold rules, and achieves long-term stable operation with the help of the rollback mechanism. Attached Figure Description

[0026] Figure 1 The diagram shown is a flowchart illustrating a digital predistortion method for OFDM baseband signals with integrated inductive and analog signals, as provided in an embodiment of this application.

[0027] Figure 2 The diagram shown is a schematic of multi-subband OFDM resource allocation in a digital predistortion method for integrated inductive OFDM baseband signals provided in an embodiment of this application.

[0028] Figure 3 The diagram shown is a flowchart illustrating the training method of the hybrid predistorter in a digital predistortion method for integrated inductive OFDM baseband signals provided in an embodiment of this application.

[0029] Figure 4 The diagram shown is a schematic of the construction of a transmission link and feedback measurement environment for an integrated inductive multi-subband OFDM baseband signal provided in an embodiment of this application.

[0030] Figure 5 The diagram shown is a schematic diagram of the sensing index calculation method in a digital predistortion method for OFDM baseband signals with integrated sensing and induction provided in an embodiment of this application.

[0031] Figure 6 The figure shown is a comparison of experimental results of the communication-side spectrum diagram ACPR of a digital predistortion method for OFDM baseband signals provided in this application embodiment and an existing conventional method.

[0032] Figure 7 The figure shows a comparison of experimental results of EVM between a digital predistortion method for OFDM baseband signals provided in this application embodiment and existing traditional methods.

[0033] Figure 8 The diagram shown is a functional module schematic of a digital predistortion device for OFDM baseband signals with integrated induction and sensing provided in this embodiment. Detailed Implementation

[0034] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, unless otherwise specified, the following embodiments and features in the embodiments can be combined with each other.

[0035] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this application. Therefore, the drawings only show the components related to this application and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0036] The technical solutions in the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0037] like Figure 1 The diagram shown is a flowchart illustrating a digital predistortion method for OFDM baseband signals with integrated inductive and analog signals, provided by an embodiment of this application.

[0038] refer to Figure 1 The digital predistortion method includes the following steps:

[0039] Step S101: Obtain the multi-subband OFDM baseband signal and the current state vector; wherein the current state vector includes at least the current service mode and the current operating condition.

[0040] Step S102: Input the multi-subband OFDM baseband signal into the trained hybrid predistorter to generate a predistorted signal; wherein, the hybrid predistorter is constructed based on a set of generalized memory polynomial basis functions and introduces a lightweight network as the gating weights and / or coefficient increments of each basis function; wherein, the lightweight network takes the current state vector as input and outputs the gating weights and / or coefficient increments of each basis function.

[0041] Step S103: Output the predistorted signal to the radio frequency power amplifier.

[0042] Those skilled in the art will understand that digital predistortion (DPD) is a technique used to improve the nonlinear distortion of radio frequency power amplifiers. It models the nonlinear distortion of the radio frequency power amplifier based on its input and output signals, and then reverses the original waveform on the digital baseband side so that the output signal of the radio frequency power amplifier is as close as possible to the original signal, thereby improving signal quality.

[0043] As described in step S101, the multi-subband OFDM baseband signal and the current state vector are obtained.

[0044] In this embodiment, the multi-subband OFDM baseband signal can be formed by superimposing several subbands in parallel. A subband refers to a continuous spectrum interval, and each subband contains multiple subcarriers and multiple OFDM symbols; wherein, an OFDM symbol is the smallest transmission unit in the time domain of the OFDM baseband signal. Guard bands can be set between subbands to meet the requirements of spectrum masking or service isolation. The transmitter (integrated sensing signal source) configures communication data and sensing reference symbols (such as pilots or training sequences) in each subband and saves the reference symbol index and values ​​for subsequent sensing processing.

[0045] like Figure 2 The diagram shown is a schematic of multi-subband OFDM resource allocation in the digital predistortion method for integrated inductive OFDM baseband signals provided in the embodiments of this application.

[0046] refer to Figure 2 The aggregation of three sub-bands (including guard bands between sub-bands) is presented in the form of a two-dimensional time-frequency grid. In this grid, the horizontal axis represents frequency (subcarrier), and the vertical axis represents time (OFDM symbol). Each sub-band is further divided into multiple resource blocks (RBs), which are the smallest units of resource scheduling. An RB consists of a single subcarrier and a single OFDM symbol. Multiple OFDM symbols constitute a time slot, and the system can dynamically adjust the time-based resource allocation ratio of each sub-band according to service requirements.

[0047] For example, resource allocation in a multi-subband configuration can be:

[0048] Subband 1 (used for communication data) means that all RBs are used for data transmission, aiming for the highest spectral efficiency.

[0049] Subband 2 (used for communication + sensing sharing or for pilot signals) carries both communication data and sensing signals simultaneously within the same subband (e.g., through time division or code division multiplexing) to achieve integrated communication and sensing; or it is used to transmit pilot signals (sensing reference symbols) to help the receiver complete the correct demodulation and processing of the integrated communication and sensing signals.

[0050] Sub-band 3 (for sensing enhancement) is specifically designed for high-precision environmental sensing. It can transmit continuous waves or special sequences to achieve better detection performance.

[0051] A protective strip is also installed between adjacent sub-bands to prevent interference between sub-bands and ensure the stability and reliability of signal transmission.

[0052] In this embodiment, the current state vector includes at least the current service mode and the current operating condition. The current service mode includes, but is not limited to, communication priority mode, sensing priority mode, and equalization mode. The current operating condition includes sub-band configuration parameters (e.g., number of sub-bands, center frequency / bandwidth of each sub-band), power allocation parameters (e.g., grid sub-band power allocation, average power, power back-off), signal characteristic parameters (e.g., peak-to-average power ratio statistics, instantaneous PAPR peak-to-average power ratio), operating temperature or ambient temperature of the RF power amplifier, time slice identifier, etc.; and group delay fluctuations, residual phase fluctuations, sensing sidelobe energy statistics, and nonlinear distortion estimations obtained from the feedback signal.

[0053] As described in step S102, the multi-subband OFDM baseband signal is input into the trained hybrid predistorter to generate a predistorted signal.

[0054] In this embodiment, the hybrid predistorter is constructed based on a set of generalized memory polynomial basis functions and introduces a lightweight network as the gating weights and / or coefficient increments of each basis function.

[0055] As those skilled in the art will know, the Generalized Memory Polynomial (GMP) is one of the mainstream digital predistortion (DPD) models, particularly suitable for describing the complex memory effects of radio frequency power amplifiers (PAs) due to their extremely wide signal bandwidth. Its basic principle is to accurately describe how the output of the radio frequency power amplifier is affected by the input signals at the current and past times by constructing a large, structured set of basis functions.

[0056] Basis function sets typically contain three main categories of basis functions: 1) Aligned Terms: These describe how the output signal at the current time is related not only to the input signal at the current time, but also to the nonlinearity of the input signal envelope (such as square, cube, etc.), and to the input signals at past times. These terms constitute a description of the basic nonlinear characteristics and short-term memory of the PA. 2) Lagging Cross Terms: To more accurately describe the long-term memory effect of the PA, these terms describe how the output signal at the current time is affected by both the "past input envelope" and the "even more past input signal". 3) Leading Cross Terms: These terms describe how the output signal at the current time may even be affected by the "future input envelope".

[0057] For example, a set of basis functions can be constructed based on the nonlinear order p, the memory depth m, and the cross-memory depth g. A typical basis function formula is shown below:

[0058] φ p,m (x) = x[nm] |x[nm]| p - 1

[0059] φ p,m,g (x) = x[nm] |x[nmg]| p - 1

[0060] Furthermore, based on the GMP-based main structure, a lightweight network (such as a two- or three-layer fully connected network) is introduced as a gate weight / coefficient increment generator for each basis function. Its input is the current state vector, and its output is the gate weights and / or coefficient increments for each basis function. Through the joint modulation of the gate weights and coefficient increments, the hybrid predistorter can quickly adapt to different operating conditions and modes while maintaining the stability of the main structure.

[0061] In this embodiment, if the input baseband signal is x[n], the GMP basis function set {φ} is constructed. k (x[n]}, the predistortion output signal u[n] can be expressed as the following formula:

[0062] u[n] = ∑K (a k + Δa k (s)) g k (s) φ k (x[n])

[0063] Among them, a k The base coefficient, g k (s)∈[0,1] is the gate weight, Δak (s) represents the coefficient increment; s is the state vector, which includes at least the identifier of the current business mode and the current operating condition, such as subband configuration and power allocation, average signal power / back-off, temperature or time slice identifier, and phase fluctuations, group delay fluctuations and sensing sidelobe statistics estimated from the feedback signal.

[0064] In this embodiment, the generalized memory polynomial is used to provide interpretable and low-complexity nonlinear memory compensation capability for the main structure, while the lightweight network outputs gating weights and / or coefficient increments according to the current business mode and current working condition, so as to achieve adaptive selection and rapid fine-tuning of different basis functions, thereby maintaining stable performance under different communication and sensing task requirements.

[0065] As described in step S103, the predistorted signal is output to the radio frequency power amplifier.

[0066] Specifically, in this embodiment, the hybrid predistorter is cascaded with a radio frequency power amplifier (PA). When the original input signal baseband x[n] passes through the hybrid predistorter, it becomes x[n] with inverse distortion. DPD [n]. For example, if the PA compresses the signal peaks, the hybrid predistorter "inflates" the peaks in advance; if the PA introduces a phase shift, the hybrid predistorter shifts in the opposite direction in advance. Then, x with inverse distortion... DPD [n] is then fed into the radio frequency power amplifier (PA) to obtain the output signal u[n]. Since the nonlinear characteristics of the PA are exactly opposite to those of the hybrid predistorter, after cascading, the input signal x[n] and the output signal u[n] of the system exhibit an ideal linear relationship from an overall perspective.

[0067] The scope of protection of the digital predistortion method for OFDM baseband signals with integrated inductive and acousto-inductive communication described in this application is not limited to the order of steps listed in this embodiment. Any solution implemented by adding, subtracting, or replacing steps in the prior art based on the principles of this application is included within the scope of protection of this application.

[0068] The training method of the hybrid predistorter is described in detail below with reference to specific embodiments.

[0069] like Figure 3 The diagram shown is a flowchart illustrating the training method of the hybrid predistorter in a digital predistortion method for integrated inductive OFDM baseband signals provided in an embodiment of this application.

[0070] refer to Figure 3 The training method for the hybrid predistorter includes the following steps:

[0071] Step S301: Construct a multi-subband OFDM baseband signal and configure communication data and sensing reference symbols in each subband;

[0072] Step S302: Construct a hybrid predistorter, wherein the hybrid predistorter uses a set of generalized memory polynomial basis functions as the main compensation structure and introduces a lightweight network as a gating weight and / or coefficient increment generator; the lightweight network takes the state vector as input and outputs the gating weight and / or coefficient increment of each basis function.

[0073] Step S303: Collect the input and output data of the RF power amplifier to construct a training set;

[0074] Step S304: Construct a multi-objective optimization loss function; wherein the multi-objective optimization loss function includes at least a communication-side loss term, a sensing-side loss term, and a phase / group delay consistency constraint term;

[0075] Step S305: Based on the training set and the multi-objective optimization loss function, train the hybrid predistorter to obtain the trained hybrid predistorter parameters.

[0076] The implementation methods for steps S301 to S302 above can be found in the above text. Figure 1 The implementation methods of the corresponding steps in the embodiments described herein will not be repeated here.

[0077] As described in steps S303-S305, input and output data of the RF power amplifier are collected to construct a training set. A multi-objective optimization loss function is constructed; wherein the multi-objective optimization loss function includes at least a communication-side loss term, a sensing-side loss term, and a phase / group delay consistency constraint term. Based on the training set and the multi-objective optimization loss function, the hybrid predistorter is trained to obtain the trained hybrid predistorter parameters.

[0078] The following is combined with, for example Figure 4 The schematic diagram shown in this application illustrates the construction of a transmission link and feedback measurement environment for an integrated inductive multi-subband OFDM baseband signal, which describes the specific implementation of steps S303 to S305 above.

[0079] refer to Figure 4 The launch link and feedback measurement environment includes the forward launch link 41, feedback acquisition and synchronous calibration 42, index evaluation 43, and multi-target control and mode management 44.

[0080] In the forward transmit link 41, a multi-subband OFDM baseband signal is first constructed and input into a constructed hybrid predistorter to obtain a predistorted output signal. Then, the predistorted output signal is up-converted by the RF transmit link and then processed by the RF power amplifier before being transmitted from the transmitter through the antenna / channel.

[0081] In the feedback acquisition and synchronization calibration 42, the output signal y[n] after being processed by the RF power amplifier is synchronized and calibrated (including gain / delay alignment). Then, the output signal y[n] after being processed by the RF power amplifier is fed back and sampled (as the output data of the RF power amplifier) ​​through feedback coupling and downconversion processing. This, together with the predistorted output signal (as the input data of the RF power amplifier), serves as the dataset for model training and validation of the hybrid predistorter.

[0082] Then, a multi-objective optimization loss function is constructed; wherein, the multi-objective optimization loss function includes at least a communication-side loss term, a sensing-side loss term, and a phase / group delay consistency constraint term. The output signal y[n] after synchronization and calibration processing is then evaluated for communication metrics, sensing metrics, and phase / group delay consistency, respectively.

[0083] Specifically, in this embodiment, during the model training and updating process of the hybrid predistorter, a communication-side loss term L is simultaneously constructed. comm With the perception-side loss term L sense .

[0084] Communication-side loss term L comm It can be formed by one or more of the following indicators: Normalized Mean Square Error (NMSE), Error Vector Magnitude (EVM), Adjacent Channel Power Ratio (ACPR), etc., and can be calculated separately for each sub-band and then weighted and summed.

[0085] The communication performance index is calculated as follows: After synchronizing and calibrating the PA's output signal y[n] with the reference signal (gain / delay alignment), demodulation and error statistics can be performed separately for each subband to obtain the subband EVM or subband NMSE, and the overall or subband ACPR can be calculated. To improve robustness, a stable communication-side loss term L can be formed by using methods such as sliding window statistics, abnormal frame removal, or weighted averaging. comm .

[0086] Perception-side loss term L senseThe output signal of the PA is subjected to matched filtering or range-Doppler processing with the reference symbol, and the energy is statistically analyzed or the peak sidelobe ratio (PSLR) / integral sidelobe ratio (ISLR) is calculated in the sidelobe region to characterize the sensing compression quality. To avoid the main lobe being suppressed during the optimization process, the embodiments of this application adopt a gating method of "main lobe protection + sidelobe suppression", which applies strong constraints only to the sidelobe region and can set different sensing weights for different subbands.

[0087] The method for calculating the sensing index (range-Doppler) is as follows: using the sensing reference symbol as a reference, the output signal of the RF power amplifier is subjected to matched filtering or range-Doppler processing to generate a range-Doppler map; the main lobe region and the side lobe region are divided in the range-Doppler map; the energy in the side lobe region is statistically analyzed and / or the peak sidelobe ratio or integral sidelobe ratio is calculated as the sensing-side loss term.

[0088] Combination Figure 5 The diagram shown is a schematic of the sensing index calculation method in a digital predistortion method for OFDM baseband signals with integrated sensing and communication provided in an embodiment of this application.

[0089] refer to Figure 5 The left-hand figure shows a schematic diagram of the division between the main lobe region Ω_M and the side lobe region Ω_S in the distance-Doppler image, and the right-hand figure is the distance-Doppler image R[τ, ν].

[0090] Specifically, using the known reference symbol sensed at the transmitting end as a reference, the received / feedback signal undergoes cyclic prefix (CP) removal and Fast Fourier Transform (FFT) to obtain the equivalent complex response of each subcarrier on each OFDM symbol. Subsequently, an Inverse Fast Fourier Transform (IFFT) is performed in the subcarrier dimension to form a range profile, and a Fast Fourier Transform (FFT) is performed in the symbol dimension to form a Doppler spectrum, thus obtaining the range-Doppler map. The calculation formula is shown below:

[0091] R[τ, ν] = ∑ n y[n] x * [n-τ]e −j2πν n T_s

[0092] Define the main lobe region Ω_M (the neighborhood surrounding the target peak) and the side lobe region Ω_S (the remaining region excluding the main lobe or a designated gated region) within the R domain. The side lobe energy E... S It can be calculated using the following formula:

[0093] E S = ∑ (τ, ν)∈Ω_S |R[τ, ν]| 2

[0094] Further calculations of metrics such as peak sidelobe ratio (PSLR) / integral sidelobe ratio (ISLR) are used as the sensing-side loss term L. sense To avoid main lobe loss, a main lobe preservation term can be added to the loss or no reinforcement constraint can be applied to the main lobe region. For multi-subband OFDM, R[τ, ν] can be obtained in each subband and then weighted and fused according to business requirements.

[0095] The method for phase / group delay consistency constraints is as follows: transform the output signal of the RF power amplifier to the frequency domain and calculate its equivalent transfer function relative to the input signal; calculate the group delay fluctuation and / or residual phase fluctuation within the operating bandwidth based on the equivalent transfer function; construct the phase / group delay consistency constraint term using the group delay fluctuation and / or the residual phase fluctuation.

[0096] Specifically, the output signal y[n] is transformed to the frequency domain, and the output equivalent transfer function is defined as: H(f) = Y(f) / X(f), whose group delay estimate τ g (f) can be calculated using the following formula:

[0097] τ g (f) = -1 / (2π) d / df ∠H(f)

[0098] Within the operating bandwidth, a phase / group delay consistency constraint L is constructed using the mean square error of the group delay relative to its mean and the fluctuation of the residual phase after de-meaning. tp This constraint is used to constrain coherence and time-frequency consistency. It can work in conjunction with sensing loss to reduce the adverse effects of phase / group delay fluctuations on the range-Doppler sidelobe structure. For multi-subband OFDM baseband signals, estimation can be performed separately within each subband and then aggregated.

[0099] The multi-objective optimization and mode adaptation process is as follows:

[0100] Considering the above constraints, the minimum weighted total loss in this embodiment is as follows:

[0101] L = λ c L comm + λ s L sense + λ tp L tp

[0102] Where, λ c , λ s , λ tp These are the weights of the communication-side loss term, the perception-side loss term, and the phase / group delay consistency constraint, respectively. These weights can be adaptively adjusted according to the business model or threshold rules.

[0103] This application employs a two-stage training approach. Specifically, the first stage primarily uses the communication-side loss term to train the basis function coefficients of the generalized memory polynomial. The second stage introduces the perception-side loss term and the phase / group delay consistency constraint term to train the lightweight network, enabling it to generate gating weights and / or coefficient increments adapted to each basis function based on the state vector.

[0104] The first phase focuses on training the basic coefficients a of GMP, primarily based on communication metrics. k (An indirect learning architecture or least squares / gradient method can be used). The second stage introduces perceptual loss and phase / group delay consistency constraints to train a lightweight network, enabling it to learn to output appropriate gating and increments under different modes / operating conditions. During online operation, the lightweight network output is updated based on the business mode or indicator threshold. Furthermore, to ensure stability, a step size limit and rollback mechanism can be set; that is, if the communication indicator and / or perceptual indicator deteriorates beyond a preset threshold after updating the gating weights and / or coefficient increments, the parameters are rolled back to their previous state.

[0105] The following comparison of experimental results between a digital predistortion method for OFDM baseband signals based on embodiments of this application and existing traditional methods further illustrates the beneficial effects of embodiments of this application.

[0106] like Figure 6 The figure shown is a comparison of experimental results of the adjacent channel power ratio of the communication side spectrum diagram of a digital predistortion method for OFDM baseband signals provided in this application embodiment and an existing traditional method.

[0107] This experiment compares and evaluates "no DPD", "communication target-only DPD", and "inductively coupled DPD" under a set of multi-subband OFDM baseband signal test conditions. The test conditions are as follows: carrier frequency 2.4 GHz, aggregate bandwidth approximately 200 MHz, single subband bandwidth approximately 20 MHz, PAPR approximately 9.6 dB.

[0108] refer to Figure 6 As shown, without DPD, the PA's output signal exhibits significant out-of-band spectral regeneration and substantial adjacent channel leakage (blue spectrum curve). Introducing DPD only for the communication target significantly reduces adjacent channel leakage (orange spectrum line). Furthermore, using the inductively coordinated DPD provided in this application embodiment further reduces adjacent channel leakage without sacrificing communication quality (green spectrum line), demonstrating stronger broadband linearization capabilities.

[0109] Taking the measured adjacent channel power ratio (ACPR) as an example: without DPD, the ACPR (L / R) is approximately -32.74 / -31.71 dBc; introducing only the communication target DPD can improve it to approximately -42.86 / -41.85 dBc. However, using the digital predistortion method of this application embodiment, the inductively coupled DPD can be further improved to approximately -44.69 / -44.47 dBc, an improvement of approximately 12 dB compared to without DPD, and an additional improvement of approximately 1.8~2.6 dB compared to introducing only the communication target DPD.

[0110] like Figure 7 The figure shown is a comparison of experimental results of the error vector magnitude between a digital predistortion method for OFDM baseband signals provided in this application embodiment and existing traditional methods.

[0111] refer to Figure 7 As shown, the digital predistortion method of this application improves the adjacent channel power ratio (ACPR) while maintaining the error vector magnitude (EVM) unchanged. Taking the measured EVM as an example, without DPD, the EVM is approximately -27.36 dB (approximately 4.29%); with only the communication target DPD introduced, the EVM is approximately -35.02 dB (approximately 1.77%); using the digital predistortion method of this application, the EVM is approximately -35.22 dB (approximately 1.73%), which is no worse or slightly better than with the introduction of the communication target DPD. Furthermore, using DPD-NMSE to measure the in-band fitting error, the digital predistortion method of this application can achieve approximately -41.85 dB, which is an improvement of approximately 2.22 dB compared to only with the introduction of the communication target DPD (approximately -39.63 dB). Therefore, the above results show that the present invention can achieve synergistic optimization of "communication linearization + sensing quality + coherent consistency" under broadband multi-subband OFDM conditions.

[0112] This application also provides a digital predistortion device for integrated inductive OFDM baseband signals, which can implement the digital predistortion method for integrated inductive OFDM baseband signals described in this application.

[0113] like Figure 8 The diagram shown is a functional module schematic of a digital predistortion device for integrated sensing and inductive OFDM baseband signals provided in this embodiment.

[0114] refer to Figure 8The digital predistortion device 8 includes: a signal acquisition and state awareness module 801, used to acquire a multi-subband OFDM baseband signal and a current state vector; wherein the current state vector includes at least the current service mode and the current operating condition. A predistortion signal generation module 802, used to input the multi-subband OFDM baseband signal into a trained hybrid predistorter to generate a predistortion signal; wherein the hybrid predistorter is constructed based on a set of generalized memory polynomial basis functions and introduces a lightweight network as the gate weights and / or coefficient increments of each basis function; wherein the lightweight network takes the current state vector as input and outputs the gate weights and / or coefficient increments of each basis function. A predistortion signal transmission module 803, used to output the predistortion signal to an RF power amplifier.

[0115] In this embodiment, the specific implementation methods of each functional module in the digital predistortion device 8 can be referred to the implementation methods of the corresponding steps in the above method embodiment, and will not be repeated here.

[0116] It should be noted that the implementation apparatus of the digital predistortion method for integrated inductive OFDM baseband signals described in this application includes, but is not limited to, the structure of the digital predistortion apparatus for integrated inductive OFDM baseband signals listed in this embodiment. All structural modifications and substitutions of the prior art made based on the principles of this application are included within the protection scope of this application.

[0117] In the embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, or methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of modules / units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or units may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of apparatuses or modules or units may be electrical, mechanical, or other forms.

[0118] The modules / units described as separate components may or may not be physically separate. The components shown as modules / units may or may not be physical modules; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules / units can be selected to achieve the objectives of the embodiments of this application, depending on actual needs. For example, the functional modules / units in the various embodiments of this application may be integrated into one processing module, or each module / unit may exist physically separately, or two or more modules / units may be integrated into one module / unit.

[0119] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0120] This application also provides an electronic device, which includes a processor and a memory; the memory is used to store a computer program; the processor is used to execute the computer program stored in the memory, so that the electronic device performs the above-described digital predistortion method for integrated inductive OFDM baseband signals.

[0121] This application also provides a computer-readable storage medium storing a computer program that, when executed by an electronic device, implements the aforementioned digital predistortion method for OFDM baseband signals with integrated sensing and induction. Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing a processor. The program can be stored in a computer-readable storage medium, which is a non-transitory medium, such as random access memory, read-only memory, flash memory, hard disk, solid-state drive, magnetic tape, floppy disk, optical disk, and any combination thereof. The aforementioned storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., digital video disc (DVD)), or a semiconductor medium (e.g., solid-state disk (SSD)).

[0122] The descriptions of the processes or structures corresponding to the above figures each have their own emphasis. For parts of a process or structure that are not described in detail, please refer to the relevant descriptions of other processes or structures.

[0123] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of this application. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in this application should still be covered by the claims of this application.

Claims

1. A digital predistortion method for integrated inductive OFDM baseband signals, characterized in that, include: Acquire the multi-subband OFDM baseband signal and the current state vector; wherein, the current state vector includes at least the current service mode and the current operating condition. The multi-subband OFDM baseband signal is input into a trained hybrid predistorter to generate a predistorted signal; wherein the hybrid predistorter is constructed based on a set of generalized memory polynomial basis functions and introduces a lightweight network as the gate weights and / or coefficient increments of each basis function; wherein the lightweight network takes the current state vector as input and outputs the gate weights and / or coefficient increments of each basis function; The predistorted signal is output to the radio frequency power amplifier.

2. The digital predistortion method for OFDM baseband signals with integrated inductive and analog signal processing according to claim 1, characterized in that, The hybrid predistorter is trained in the following manner: Construct a multi-subband OFDM baseband signal and configure communication data and sensing reference symbols in each subband; A hybrid predistorter is constructed, wherein the hybrid predistorter uses a set of generalized memory polynomial basis functions as the main compensation structure and introduces a lightweight network as a gating weight and / or coefficient increment generator; the lightweight network takes the state vector as input and outputs the gating weight and / or coefficient increment of each basis function. The input and output data of the RF power amplifier are collected to construct a training set; Construct a multi-objective optimization loss function; wherein, the multi-objective optimization loss function includes at least a communication-side loss term, a sensing-side loss term, and a phase / group delay consistency constraint term; Based on the training set and the multi-objective optimization loss function, the hybrid predistorter is trained to obtain the trained hybrid predistorter parameters.

3. The digital predistortion method for integrated inductive OFDM baseband signals according to claim 2, characterized in that, The communication-side loss term is formed by one or more of the following: normalized mean square error, error vector magnitude, and adjacent channel power ratio index, and is calculated separately for each sub-band and then weighted and summed.

4. The digital predistortion method for OFDM baseband signals with integrated inductive and analog signal processing according to claim 2, characterized in that, The steps for constructing the perception-side loss term include: Using the aforementioned sensing reference symbol as a reference, the output signal of the radio frequency power amplifier is subjected to matched filtering or range-Doppler processing to generate a range-Doppler image; The distance-Doppler image is divided into a main lobe region and a side lobe region; The energy within the sidelobe region is statistically analyzed and / or the peak sidelobe ratio or integral sidelobe ratio is calculated as a loss term on the sensing side.

5. The digital predistortion method for OFDM baseband signals with integrated inductive and analog signal processing according to claim 2, characterized in that, The steps for constructing the phase / group delay consistency constraint include: The output signal of the RF power amplifier is transformed to the frequency domain, and its equivalent transfer function relative to the input signal is calculated. Calculate the group delay fluctuation and / or residual phase fluctuation within the operating bandwidth based on the equivalent transfer function. The phase / group delay consistency constraint term is constructed using the group delay fluctuation and / or the residual phase fluctuation.

6. The digital predistortion method for integrated inductive OFDM baseband signals according to claim 2, characterized in that, The hybrid predistorter is trained using a two-stage training strategy, including: The first stage focuses on the communication-side loss term to train the basis function coefficients of the generalized memory polynomial. The second stage introduces a perception-side loss term and a phase / group delay consistency constraint term to train the lightweight network so that it can generate gating weights and / or coefficient increments that are adapted to each basis function based on the state vector.

7. The digital predistortion method for integrated inductive OFDM baseband signals according to claim 6, characterized in that, It also includes the following steps: after updating the gating weights and / or coefficient increments, if the communication indicators and / or perception indicators deteriorate beyond a preset threshold, then roll back to the parameters before the update.

8. A digital predistortion device for integrated sensing and inductive OFDM baseband signals, characterized in that, include: The signal acquisition and state awareness module is used to acquire multi-subband OFDM baseband signals and current state vectors; wherein, the current state vector includes at least the current service mode and the current operating condition. A predistortion signal generation module is used to input the multi-subband OFDM baseband signal into a trained hybrid predistorter to generate a predistortion signal; wherein, the hybrid predistorter is constructed based on a set of generalized memory polynomial basis functions and introduces a lightweight network as the gate weights and / or coefficient increments of each basis function; wherein, the lightweight network takes the current state vector as input and outputs the gate weights and / or coefficient increments of each basis function; A predistortion signal transmission module is used to output the predistortion signal to an RF power amplifier.

9. An electronic device, characterized in that, The electronic device includes: a processor and a memory; the memory is used to store a computer program; the processor is used to execute the computer program stored in the memory to cause the electronic device to perform the digital predistortion method for inductive OFDM baseband signals as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by an electronic device, the program implements the digital predistortion method for inductive OFDM baseband signals as described in any one of claims 1 to 7.

Images