A radio frequency self-interference cancellation method based on LASSO regression

Abstract

This invention discloses a radio frequency self-interference cancellation method based on LASSO regression, belonging to the field of wireless communication and radio frequency signal processing technology, and applied to simultaneous full-duplex communication devices on the same frequency. A reference radio frequency signal is coupled from the output of a power amplifier and input into a multi-tap delay structure. In the digital domain, the self-interference channel is modeled as an FIR filter. LASSO regression is applied to the self-interference signal acquired by the ADC and the reference signal to obtain sparse channel coefficients, and the N paths contributing the most to self-interference are selected. The index, amplitude, and phase of the non-zero coefficients are mapped to the delay, attenuation, and phase shift parameters of each tap, and the radio frequency hardware is adjusted in real time to reconstruct a copy of the self-interference signal. Finally, this copy signal is injected into the receiving link to achieve cancellation. This method achieves efficient cancellation with fewer taps, suppresses phase noise with distance-dependent effects, and reduces the residual self-interference power to within the dynamic range of the ADC, facilitating subsequent digital domain cancellation.

Description

Technical Field

[0001] This invention belongs to the field of wireless communication and radio frequency signal processing technology, and in particular relates to a radio frequency self-interference cancellation method based on LASSO regression. Background Technology

[0002] Simultaneous full-duplex communication on the same frequency band enables simultaneous transmission and reception of signals, significantly improving spectrum utilization. However, the transmitter signal can generate strong self-interference in the receiver, which must be suppressed below the receiver noise floor to achieve reliable full-duplex communication. Self-interference cancellation is typically implemented in multiple stages across the spatial, radio frequency (RF), and digital domains. RF domain cancellation, by reconstructing and injecting a cancellation signal with the opposite amplitude and phase to the self-interference signal at the RF front end, suppresses most of the interference energy. Its performance directly affects the complexity of subsequent digital domain cancellation and the overall system performance.

[0003] There are two main technical approaches to self-interference cancellation in the radio frequency domain: The first is DAC-based reconstruction, which reconstructs the RF cancellation signal using a DAC after estimating the self-interference channel response in the digital domain. However, this approach struggles to suppress interference components related to phase noise and introduces additional DAC quantization noise, resulting in a large amount of unstructured noise in the residual signal after RF cancellation, thus limiting the effectiveness of digital domain cancellation. The second approach is based on multi-tap delay structures, which uses fixed delay units and estimates tap weights through least squares or gradient descent. However, this approach struggles to accurately match actual sparse multipath channels and is insufficient in suppressing phase noise with distance-dependent effects. Furthermore, it requires a large number of taps in high-delay multipath environments, leading to high hardware complexity and increased costs.

[0004] Therefore, existing radio frequency cancellation methods still have shortcomings in terms of noise suppression performance and hardware efficiency. There is an urgent need for a solution that can accurately estimate sparse multipath self-interference channels and achieve efficient radio frequency cancellation with low hardware complexity. Summary of the Invention

[0005] In view of this, the present invention aims to propose a radio frequency self-interference cancellation method based on LASSO regression to at least solve one of the problems in the background art.

[0006] To achieve the above objectives, the technical solution of the present invention is implemented as follows:

[0007] Firstly, this solution discloses a radio frequency self-interference cancellation method based on LASSO regression, applicable to simultaneous full-duplex communication devices on the same frequency, including:

[0008] Obtain the reference RF signal coupled at the output of the power amplifier;

[0009] The reference radio frequency signal is input to a multi-tap delay structure to form an adjustable self-interference reconstruction signal. Each tap of the multi-tap delay structure includes a controllable delay unit, an attenuation unit, and a phase shifting unit.

[0010] The self-interference signal at the receiving end is coupled in the receiving link and then obtained as a digital sampling signal through down-conversion and analog-to-digital conversion.

[0011] In the digital domain, the self-interference channel is modeled as a finite-length impulse response filter. Based on the reference radio frequency signal and the digital sampled signal, LASSO regression is applied to solve the sparse channel coefficients, and the dominant self-interference path corresponding to the non-zero coefficients is selected.

[0012] The sequence number, amplitude, and phase information of the non-zero coefficients are mapped to the delay value, attenuation, and phase shift of each tap of the multi-tap delay structure, and the multi-tap delay structure is adjusted accordingly to reconstruct a copy of the interference signal.

[0013] The reconstructed signal and the received signal are combined at the radio frequency front end to achieve self-interference cancellation in the radio frequency domain.

[0014] Furthermore, the regularization parameter of the LASSO regression is configured such that the number of non-zero coefficients of the sparse channel coefficients is equal to the number of taps N of the multi-tap delay structure, so as to reconstruct the canceled signal based on N dominant self-interference paths.

[0015] Furthermore, the LASSO regression is solved using a subgradient-based coordinate descent algorithm, and the sparse channel coefficients are updated using a soft threshold function to obtain a sparse solution.

[0016] Furthermore, the controllable delay unit is implemented by a combination of radio frequency delay devices or delay lines with different delay values ​​and a switch matrix.

[0017] Furthermore, the attenuation unit and the phase shifting unit are combined into an orthogonal vector modulator to reduce the complexity of the radio frequency circuit.

[0018] Furthermore, when the power amplifier is significantly nonlinear or the baseband signal is difficult to obtain, a coupler is added at the output of the power amplifier, and the digital representation of the reference RF signal is obtained through analog-to-digital conversion as the input of the finite impulse response filter.

[0019] Furthermore, by increasing the analog-to-digital conversion sampling rate or using digital interpolation methods to reduce the sampling interval, the estimation accuracy of the delay, attenuation, and phase shift parameters of the dominant self-interference path can be improved.

[0020] Furthermore, after the radio frequency domain self-interference cancellation, the residual self-interference signal power is suppressed to below a preset threshold and within the dynamic range of the analog-to-digital converter, providing input for subsequent digital domain cancellation.

[0021] Furthermore, the radio frequency self-interference cancellation method based on LASSO regression disclosed in this scheme involves a cancellation system including:

[0022] A reference signal coupling unit is located at the output of the power amplifier;

[0023] A multi-tap delay structure connected to the reference signal coupling unit, wherein each tap of the multi-tap delay structure includes a controllable delay unit, an attenuation unit, and a phase shifting unit;

[0024] The coupling and combining unit, located in the receiving link, is used to combine the reconstructed signal output by the multi-tap delay structure with the received signal.

[0025] An analog-to-digital converter is used to sample the self-interference signal obtained through coupling.

[0026] The digital processing unit is used to model the self-interference channel as a finite-length impulse response filter and obtain sparse channel coefficients based on LASSO regression, and to map the index, amplitude and phase information corresponding to the non-zero coefficients to the delay, attenuation and phase shift control quantities of the multi-tap delay structure.

[0027] Furthermore, the controllable delay unit is implemented using radio frequency delay devices or delay lines and switch matrices with different delay values, and / or the attenuation unit and phase shift unit are combined into an orthogonal vector modulator.

[0028] Compared with existing technologies, the radio frequency self-interference cancellation method based on LASSO regression described in this invention has the following advantages:

[0029] (1) The radio frequency self-interference cancellation method based on LASSO regression proposed in this invention adopts an architecture that combines digital domain sparse channel estimation with radio frequency domain multi-tap delay structure. It automatically selects the N multipath components that contribute most significantly to self-interference through LASSO regression, and maps the sequence number, amplitude and phase corresponding to their non-zero coefficients to the delay, attenuation and phase shift control quantities of the radio frequency taps, so as to achieve precise control of the radio frequency cancellation circuit.

[0030] (2) Compared with the prior art, the present invention can reduce the number of unnecessary taps while ensuring the cancellation effect, thereby significantly reducing the configuration of devices such as delay lines, attenuators, and phase shifters, and reducing hardware complexity, power consumption and manufacturing costs.

[0031] (3) By accurately estimating the parameters of the main interference path, the present invention makes the reconstructed self-interference signal copy more matched with the real interference signal in amplitude and phase, which can more effectively suppress phase noise with distance correlation effect, reduce residual self-interference after radio frequency cancellation and other residual noise that is difficult to model, and provide better input conditions for subsequent digital domain cancellation.

[0032] (4) This invention utilizes the sparsity characteristics of self-interference channels and the path filtering capability of LASSO regression, and can control the number of selected paths by adjusting the regularization parameter, so that the system has better adaptability and practicality to different interference intensities and multipath richness scenarios. Attached Figure Description

[0033] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0034] Figure 1 This is a schematic diagram of the self-interference cancellation implementation structure based on LASSO regression as described in an embodiment of the present invention;

[0035] Figure 2 This is a schematic diagram of the radio frequency self-interference cancellation method based on LASSO regression as described in an embodiment of the present invention. Detailed Implementation

[0036] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0037] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0038] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0039] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0040] The method described in this invention is based on a typical full-duplex communication radio frequency front-end and digital domain collaborative signal processing mode, and is specifically implemented according to the steps of radio frequency domain acquisition, digital domain estimation, and radio frequency domain execution.

[0041] like Figure 1 As shown, the system structure of this scheme is as follows: a reference signal is coupled from the output of the power amplifier and input to a multi-tap delay structure. Each tap contains an independently controllable delay unit (delay unit), attenuation unit (attenuator), and phase shifting unit (phase shifter) for reconstructing a copy of the self-interference signal. A directional coupler is provided in the receiving link to feed back the synthesized signal to the digital domain for processing. In terms of the core algorithm, this scheme models the self-interference channel as a finite impulse response (FIR) filter in the digital domain and uses LASSO regression for sparse channel estimation. By minimizing the mean square error between the model output and the actual acquired signal and applying L1 regularization constraints, the dominant interference path is automatically selected, and its delay, attenuation, and phase shift parameters are accurately estimated. Regarding parameter mapping and control, this scheme maps the tap number, amplitude, and phase information corresponding to the non-zero coefficients obtained from LASSO regression to the delay, attenuation, and phase shift values ​​of each tap in the RF cancellation circuit. Based on this, the RF hardware is precisely controlled to generate a cancellation signal with opposite amplitude and phase to the real self-interference signal and matching delay. Furthermore, in the receiving link, the reconstructed cancellation signal is combined with the received signal through a coupler to achieve self-interference cancellation in the RF domain. After cancellation, the main interference path is significantly suppressed, the residual signal power is reduced to within the ADC dynamic range, and the phase noise characteristics are improved.

[0042] In a full-duplex communication system, the self-interference signals received by the receiving link are mainly superimposed from circulator leakage, antenna mismatch reflection, and reflections from near-end objects. These signals are all copies of the power amplifier's transmitted signal after delay, attenuation, and phase shift. The system function of the self-interference signal... It can be represented as:

[0043] ;

[0044] in, The unit impulse function, This represents the total number of multipaths. For the first Complex gain of the path, For the first The delay of the path.

[0045] Self-interference signal in the receiving channel This can be represented as the convolution of the power amplifier output signal and the system function:

[0046] ;

[0047] in, This indicates the signal output by the power amplifier.

[0048] Ignoring the nonlinear distortion introduced by the power amplifier, the power amplifier output Linear amplification of baseband signals:

[0049] ;

[0050] in, The radio frequency signal obtained after up-conversion and DAC conversion of the baseband signal sent by the transmitter. This represents the linear gain of the power amplifier.

[0051] Therefore, the self-interference signal received by the receiving channel This can be further expressed as:

[0052] ;

[0053] After sampling by the ADC, the received self-interference signal is represented in the digital domain as follows:

[0054] ;

[0055] in, For discrete-time indexing, Discrete baseband signal transmitted by the transmitter For delay The value after quantization.

[0056] In the digital domain, a self-interference signal system model is established using a transversely tapped FIR filter, and its output... for:

[0057] ;

[0058] Where M is the total number of taps in the FIR filter. These are the input data for each tap of the FIR filter. These are the tap coefficients of the FIR filter.

[0059] The input matrix is ​​constructed by collecting tap input data over a period of time. Model output vector Combined with the self-interference signal vector acquired by the ADC The model parameters are solved using LASSO regression, which optimizes the mean square error between the model output and the received self-interference signal, while applying L1 regularization constraints to the coefficient vector.

[0060] ;

[0061] in, This is a regularization hyperparameter used to balance fitting error and model sparsity;

[0062] Since the L1 regularization term is not differentiable, a coordinate descent method based on subgradients is used to solve it. (Coefficient vector) The update rules for each component are as follows:

[0063] ;

[0064] Iterate through all coefficients Until convergence, among which, This indicates the number of data points collected. The soft threshold function is expressed as:

[0065] ;

[0066] in, For the input variables of the soft threshold function, For value parameters, Indicates that an update is being made. When, with other coefficients fixed, the first... Prediction residuals for each sample:

[0067] ;

[0068] Represents the input matrix The Line number Column element, i.e., the first The sample at the th The value at the tap. express The first vector The element that is, the first element of the self-interference signal vector. The value of each sample express The k-th element of the vector is the coefficient value at the k-th tap of the FIR filter.

[0069] Sparse solutions are obtained through soft thresholding, making some coefficients zero, thereby achieving FIR coefficient screening. Hyperparameters are then adjusted. The number of controllable non-zero coefficients is equal to the number of RF taps N. The index of these non-zero coefficients corresponds to the delay value of the RF tap, and the coefficient value corresponds to the complex gain on the corresponding path.

[0070] Ignoring the effects of linear transformations of signals such as power amplifier gain, coupler coupling, and other link attenuation on the expression, the system function of the RF multi-tap structure can be expressed as:

[0071] ;

[0072] ;

[0073] in, The number of non-zero coefficients in the FIR filter. For the FIR filter One non-zero coefficient value, For the first The sequence number of each non-zero coefficient, where T is the clock cycle for digital domain signal processing. The output of the RF multi-tap structure is the self-interference signal reconstructed after signal estimation. After subtracting the reconstructed RF self-interference signal from the receiver link, the residual signal... for: ;in, This is the set of indices for the N selected paths.

[0074] Since the self-interference signals from the N paths with the largest amplitudes have been canceled, and the phase noise with distance correlation effects has also been significantly suppressed, the residual self-interference power is suppressed below a certain threshold, meeting the dynamic range requirements of the ADC and laying the foundation for digital domain cancellation. Further modeling and cancellation of the residual signals can then be performed in the digital domain, ultimately achieving full-duplex communication.

[0075] This invention can also be further optimized or replaced in the following ways:

[0076] To simplify the hardware, the variable delay unit can use radio frequency delay devices or delay lines with different delay values ​​and be implemented by combining switch matrices;

[0077] Variable attenuators and phase shifters can be combined using quadrature vector modulators to simplify the circuit.

[0078] If the power amplifier has significant nonlinearity or baseband signal If the sample is difficult to obtain, a coupler can be added to the output of the power amplifier to sample it using an ADC. The resulting digital baseband signal is then used as the input to the FIR model;

[0079] To improve the accuracy of parameter estimation, the ADC sampling rate can be increased or digital interpolation methods can be used to reduce the sampling interval.

[0080] like Figure 2 As shown, this embodiment provides a radio frequency self-interference cancellation method based on LASSO regression, applied to co-frequency full-duplex communication equipment, to suppress self-interference signals leaking from the transmit link to the receive link when the transmitter and receiver are operating on the same frequency. This method is executed online during the actual operation of the full-duplex node, and specifically includes the following steps:

[0081] Obtain the reference RF signal coupled at the output of the power amplifier;

[0082] The reference radio frequency signal is input to a multi-tap delay structure to form an adjustable self-interference reconstruction signal. Each tap of the multi-tap delay structure includes a controllable delay unit, an attenuation unit, and a phase shifting unit.

[0083] The self-interference signal at the receiving end is coupled in the receiving link and then obtained as a digital sampling signal through down-conversion and analog-to-digital conversion.

[0084] In the digital domain, the self-interference channel is modeled as a finite-length impulse response filter. Based on the reference radio frequency signal and the digital sampled signal, LASSO regression is applied to solve the sparse channel coefficients, and the dominant self-interference path corresponding to the non-zero coefficients is selected.

[0085] The sequence number, amplitude, and phase information of the non-zero coefficients are mapped to the delay value, attenuation, and phase shift of each tap of the multi-tap delay structure, and the multi-tap delay structure is adjusted accordingly.

[0086] The reconstructed signal and the received signal are combined at the radio frequency front end to achieve self-interference cancellation in the radio frequency domain.

[0087] Based on the above steps, the detailed steps in the specific implementation process are as follows:

[0088] First, in the transmit link, an RF coupling unit is set at the output of the power amplifier to couple the RF signal at the output of the power amplifier and obtain a reference RF signal. In this embodiment, considering that in some application scenarios the power amplifier has significant nonlinear distortion or the transmit baseband signal is difficult to obtain directly, it is preferable to add an RF coupler at the output of the power amplifier. The coupled reference RF signal is used as the input for subsequent RF self-interference reconstruction on the one hand, and its digital representation is obtained through down-conversion and analog-to-digital conversion circuits on the other hand, and used as the input signal sequence for the finite impulse response filter. This ensures that the self-interference excitation signal can still be accurately characterized when there is power amplifier nonlinearity or the baseband signal is unavailable.

[0089] Then, the reference RF signal is input to the multi-tap delay structure of the RF front end to form an adjustable self-interference reconstruction signal. Specifically, the multi-tap delay structure includes N taps, each tap sequentially containing a controllable delay unit, an attenuation unit, and a phase-shifting unit. The controllable delay unit is implemented by combining several RF delay devices or RF delay lines with different delay values ​​with a switch matrix. By controlling the conduction state of the switches, the delay can be switched between multiple discrete delay values, thereby achieving fine adjustment of the time delay parameters of each tap. To reduce the overall complexity and number of components in the RF circuit, in this embodiment, the originally separate attenuation unit and phase-shifting unit are preferably combined into an orthogonal vector modulator. By adjusting the amplitude and phase of the orthogonal components, attenuation and phase-shifting control are completed simultaneously, so that each tap comprehensively modulates the amplitude and phase of the reference RF signal, thereby constructing a reconstructed copy in the RF domain that is highly matched with the target self-interference signal in terms of time delay, amplitude, and phase.

[0090] Next, a coupling unit is set up in the receiving link to couple the receiving end RF signal containing self-interference components from the receiving end's front-end RF path. This coupled signal is then down-converted and analog-to-digital converted to obtain a digital sampled signal sequence of the receiving end's self-interference signal. This digital sampled signal includes multipath components from the transmitted signal reaching the receiving end after passing through the self-interference channel, as well as random interference such as environmental noise, and is used to characterize the comprehensive impact of the actual self-interference channel.

[0091] In the digital domain, the self-interference channel is modeled as a finite-length impulse response filter. The impulse response coefficients of the filter are used to characterize the amplitude and phase of each multipath self-interference component under different delays. In this embodiment, the digital representation of the reference RF signal is used as the input sequence of the filter, and the digital sampling signal of the self-interference at the receiving end is used as the observation output. A LASSO regression optimization problem is constructed. By introducing a ℓ1 regularization term to impose sparsity constraints on the channel coefficients, a sparse channel coefficient vector is obtained. In the solution process, a subgradient-based coordinate descent algorithm is used as an iterative solution strategy. The channel coefficients in each dimension are updated coordinate by coordinate. In the update step, a soft threshold function is used to shrink and truncate the intermediate results to suppress small-amplitude coefficients and strengthen a few large-amplitude components, ultimately obtaining a sparse solution with a limited number of non-zero coefficients. By adjusting the regularization parameter in the LASSO regression, the number of non-zero coefficients in the obtained sparse channel coefficient vector is equal to the number of taps N in the multi-tap delay structure. That is, only N dominant self-interference paths that contribute the most significant self-interference energy are retained, thereby limiting the number of hardware taps while ensuring cancellation performance.

[0092] After obtaining the sparse channel coefficients, path filtering and parameter extraction are performed on the non-zero coefficients. Specifically, the path delay is determined based on the index of the non-zero coefficient in the finite-length impulse response filter, and the gain and phase rotation of the path are determined based on the amplitude and phase of the non-zero coefficient. The index, amplitude, and phase information of each non-zero coefficient are mapped to the delay value, attenuation, and phase shift of the corresponding tap in the multi-tap delay structure: on the one hand, a suitable combination of RF delay lines is selected by controlling the switch matrix to align the tap delay with the corresponding path delay; on the other hand, the amplitude and phase of the tap output are made consistent with the estimated values ​​of the corresponding path by adjusting the control voltage or digital control word of the quadrature vector modulator. Through the above mapping and adjustment process, N taps jointly perform precise modulation of the time delay, attenuation, and phase of the reference RF signal, thereby generating a reconstructed signal that is highly matched with the self-interference signal in the multipath structure at the RF front end.

[0093] Subsequently, the self-interference reconstruction signal, after being modulated by the multi-tap delay structure, is synthesized with the received signal at the RF front-end. Specifically, synthesis coupling units can be set at appropriate positions before and after the low-noise amplifier in the receiving link. The reconstructed cancellation signal is injected into the receiving path as an inverted signal, superimposed on the original self-interference component under conditions of similar amplitude and opposite phase, thereby achieving active cancellation of self-interference energy in the RF domain. Through the precise control of the aforementioned multi-tap delay structure, this embodiment can suppress most of the self-interference signal power to below a preset threshold at the RF front-end, ensuring that the residual self-interference signal power after cancellation is within the dynamic range of the analog-to-digital converter, preventing ADC front-end overload or significant decrease in quantization accuracy. This provides good input conditions for further self-interference refinement and cancellation in the digital domain and conventional receiver signal processing.

[0094] Through the above steps, this embodiment utilizes LASSO regression to perform sparse modeling and dominant path selection for the self-interference channel, concentrating a limited number of RF tap resources on the self-interference path that contributes the most. This reduces RF hardware complexity and cost while improving the suppression capability for multipath self-interference and phase noise, achieving an efficient RF self-interference cancellation method suitable for simultaneous full-duplex communication devices on the same frequency. The above steps can be executed periodically or on demand in a practical system to adapt to the time-varying characteristics of the self-interference channel. The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A radio frequency self-interference cancellation method based on LASSO regression, applied to simultaneous full-duplex communication equipment on the same frequency, characterized in that, include: Obtain the reference RF signal coupled at the output of the power amplifier; The reference radio frequency signal is input to a multi-tap delay structure to form an adjustable self-interference reconstruction signal. Each tap of the multi-tap delay structure includes a controllable delay unit, an attenuation unit, and a phase shifting unit. The self-interference signal at the receiving end is coupled in the receiving link and then obtained as a digital sampling signal through down-conversion and analog-to-digital conversion. In the digital domain, the self-interference channel is modeled as a finite-length impulse response filter. Based on the reference RF signal and the digital sampled signal, LASSO regression is applied to solve for the sparse channel coefficients, and the dominant self-interference path corresponding to the non-zero coefficients is selected. Specifically: The input matrix is ​​constructed by collecting tap input data over a period of time. Model output vector Combined with the self-interference signal vector acquired by the ADC The model parameters are solved using LASSO regression, which optimizes the mean square error between the model output and the received self-interference signal, while applying L1 regularization constraints to the coefficient vector. ； in, This is a regularization hyperparameter used to balance fitting error and model sparsity; Since the L1 regularization term is not differentiable, a coordinate descent method based on subgradient is used to solve it, and the coefficient vector is... The update rules for each component are as follows: ； Iterate through all coefficients Until convergence, among which, This indicates the number of data points collected. The soft threshold function is expressed as: ； in, For the input variables of the soft threshold function, For value parameters, Indicates that an update is being made. When, with other coefficients fixed, the first... Prediction residuals for each sample: ； Represents the input matrix The Line number Column element, i.e., the first The sample at the th The value at the tap. express The first vector The element that is, the first element of the self-interference signal vector. The value of each sample express The k-th element of the vector is the coefficient value at the k-th tap of the FIR filter; A sparse solution is obtained by using a soft thresholding operation, making some coefficients zero, thus achieving FIR coefficient screening and adjusting hyperparameters. The number of control non-zero coefficients is equal to the number of RF taps N; The sequence number, amplitude, and phase information of the non-zero coefficients are mapped to the delay value, attenuation, and phase shift of each tap of the multi-tap delay structure, and the multi-tap delay structure is adjusted accordingly. The reconstructed signal and the received signal are combined at the radio frequency front end.

2. The radio frequency self-interference cancellation method based on LASSO regression according to claim 1, characterized in that, The regularization parameter of the LASSO regression is configured such that the number of non-zero coefficients of the sparse channel coefficients is equal to the number of taps N of the multi-tap delay structure, so as to reconstruct the cancellation signal based on N dominant self-interference paths.

3. The radio frequency self-interference cancellation method based on LASSO regression according to claim 1, characterized in that, The LASSO regression is solved using a subgradient-based coordinate descent algorithm, and the sparse channel coefficients are updated using a soft threshold function to obtain a sparse solution.

4. The radio frequency self-interference cancellation method based on LASSO regression according to claim 1, characterized in that, The controllable delay unit is implemented by a combination of radio frequency delay devices or delay lines with different delay values ​​and a switch matrix.

5. The radio frequency self-interference cancellation method based on LASSO regression according to claim 1, characterized in that, The attenuation unit and the phase shifting unit are combined into an orthogonal vector modulator to reduce the complexity of the radio frequency circuit.

6. The radio frequency self-interference cancellation method based on LASSO regression according to claim 1, characterized in that, After the radio frequency domain self-interference cancellation, the residual self-interference signal power is suppressed to below a preset threshold and within the dynamic range of the analog-to-digital converter, providing input for subsequent digital domain cancellation.

Images