CN-121462017-B - Radio frequency self-interference cancellation method based on LASSO regression

CN121462017BCN 121462017 BCN121462017 BCN 121462017BCN-121462017-B

Abstract

The invention discloses a radio frequency self-interference cancellation method based on LASSO regression, which belongs to the technical field of wireless communication and radio frequency signal processing and is applied to common-frequency and full-duplex communication equipment. The method comprises the steps of obtaining a reference radio frequency signal through coupling of the output end of a power amplifier, inputting the reference radio frequency signal into a multi-tap delay structure, modeling a self-interference channel as an FIR filter in a digital domain, obtaining sparse channel coefficients through application of LASSO regression to the self-interference signal and the reference signal acquired by an ADC, screening N paths with the largest contribution to the self-interference, mapping sequence numbers, amplitude and phase of non-zero coefficients into delay, attenuation and phase shift parameters of each tap, regulating radio frequency hardware in real time, reconstructing a copy of the self-interference signal, and finally injecting the copy signal into a receiving link to realize cancellation. The method realizes high-efficiency offset by fewer taps, suppresses phase noise with distance correlation effect, reduces residual self-interference power to the ADC dynamic range, and facilitates the offset of subsequent digital domains.

Inventors

YANG YANG
LI YANBO
WU CHUNJIE
LI GUANG
LI GUOQIANG
ZHANG JIANJUN

Assignees

天津七一二通信广播股份有限公司
天津电子信息职业技术学院

Dates

Publication Date: 20260505
Application Date: 20260107

Claims (6)

1. The radio frequency self-interference cancellation method based on LASSO regression is applied to the same-frequency and full-duplex communication equipment and is characterized by comprising the following steps of: acquiring a reference radio frequency signal coupled with the output end of the power amplifier; inputting the reference radio frequency signal into a multi-tap delay structure to form an adjustable self-interference reconstruction signal, wherein each tap of the multi-tap delay structure comprises a controllable delay unit, an attenuation unit and a phase shift unit; Coupling in a receiving link to obtain a self-interference signal of a receiving end, and obtaining a digital sampling signal through down-conversion and analog-to-digital conversion; Modeling a self-interference channel as a finite-length impulse response filter in a digital domain, solving a sparse channel coefficient by applying LASSO regression based on the reference radio frequency signal and the digital sampling signal, and screening a dominant self-interference path corresponding to a non-zero coefficient, wherein the dominant self-interference path comprises the following specific steps: collecting tap input data in a period of time to form an input matrix Model output vector Self-interference signal vector acquired by combining ADC Solving model parameters by adopting LASSO regression, namely optimizing the mean square error between model output and a received self-interference signal, and simultaneously applying the constraint of L1 regularization of coefficient vectors: ; Wherein, the The regularized super parameters are used for balancing fitting errors and model sparsity; because the L1 regular term is not differentiable, solving is carried out by adopting a coordinate descent method based on a secondary gradient, and the coefficient vector is calculated The updating rule of each component is as follows: ; Traversing all coefficients Until convergence, wherein, The number of data acquired is represented by the number of data acquired, The soft threshold function is expressed as: ; Wherein, the Is an input variable of the soft threshold function, As a value parameter, a value of the parameter, The representation is updated Other coefficients are fixed when Prediction residuals for individual samples: ; Representing an input matrix Is the first of (2) Line 1 Column elements, i.e. the first The first sample is at The value on the one tap is set to, Representation of Vector number The elements, i.e. the first of the self-interference signal vectors The value of the individual samples is calculated, Representation of The kth element of the vector is the coefficient value on the kth tap of the FIR filter; the soft threshold operation is used for obtaining the thin fluffy, so that part of coefficients are zero, FIR coefficient screening is realized, and the super-parameters are regulated Controlling the number of non-zero coefficients to be equal to the number N of radio frequency taps; Mapping the sequence number, amplitude and phase information of the non-zero coefficient into delay values, attenuation amounts and phase shift amounts of all taps of the multi-tap delay structure, and regulating and controlling the multi-tap delay structure accordingly; and synthesizing the regulated reconstruction signal and the received signal at the radio frequency front end.
2. The method of claim 1, wherein the regularization parameters of the LASSO regression are configured to equate the number of non-zero coefficients of the sparse channel coefficients to the number of taps N of the multi-tap delay structure to reconstruct the cancellation signal based on the N dominant self-interference paths.
3. The method of claim 1, wherein the LASSO regression is solved using a secondary gradient-based coordinate descent algorithm, and the sparse channel coefficients are updated by a soft threshold function to obtain a sparse solution.
4. The method of claim 1, wherein the controllable delay units are implemented by radio frequency delay devices or delay lines with different delay values and switch matrix combinations.
5. The method of claim 1, wherein the attenuation unit and the phase shift unit are combined into a quadrature vector modulator to reduce complexity of the radio frequency circuit.
6. The method of claim 1, wherein after the self-interference cancellation in the radio frequency domain, the residual self-interference signal power is suppressed to below a predetermined threshold and within the dynamic range of the analog-to-digital converter, and an input is provided for the cancellation in the subsequent digital domain.

Description

Radio frequency self-interference cancellation method based on LASSO regression Technical Field The invention belongs to the technical field of wireless communication and radio frequency signal processing, and particularly relates to a radio frequency self-interference cancellation method based on LASSO regression. Background The same-frequency simultaneous full duplex technology can realize simultaneous receiving and transmitting of signals on the same frequency band, thereby remarkably improving the frequency spectrum utilization rate. The transmitter signal will create strong self-interference to the receiver and must be suppressed below the receive noise floor to achieve reliable full duplex communication. Self-interference cancellation is generally achieved from multiple stages of the spatial domain, the radio frequency domain and the digital domain, wherein the radio frequency domain cancellation achieves suppression of most of interference energy by reconstructing and injecting cancellation signals opposite to the self-interference signal amplitude at the radio frequency front end, and the performance of the cancellation signals directly affects the complexity and the overall system performance of the subsequent digital domain cancellation. The existing radio frequency domain self-interference cancellation mainly comprises two technical routes, namely a cancellation method based on DAC reconstruction, wherein after the self-interference channel response is estimated in the digital domain, a radio frequency cancellation signal is reconstructed through the DAC, but interference components related to phase noise are difficult to restrain, and extra DAC quantization noise is introduced, so that a large amount of unstructured noise is contained in a residual signal after the radio frequency cancellation, and the cancellation effect of the digital domain is restrained. Secondly, a cancellation method based on a multi-tap delay structure adopts a fixed delay unit and estimates tap weights through least square or gradient descent, but is difficult to accurately match with an actual sparse multipath channel, and has insufficient phase noise suppression with distance correlation effect, and a large number of taps are needed in a large-delay multipath environment, so that hardware complexity is high and cost is increased. Therefore, the existing radio frequency cancellation method still has the defects in terms of noise suppression performance and hardware efficiency, and a solution capable of accurately estimating sparse multipath self-interference channels and realizing efficient radio frequency cancellation with lower hardware complexity is needed. Disclosure of Invention In view of the above, the present invention is directed to a radio frequency self-interference cancellation method based on LASSO regression, so as to solve at least one problem in the background art. In order to achieve the above purpose, the technical scheme of the invention is realized as follows: in a first aspect, the present solution discloses a radio frequency self-interference cancellation method based on LASSO regression, which is applied to a common-frequency and full-duplex communication device, and includes: acquiring a reference radio frequency signal coupled with the output end of the power amplifier; inputting the reference radio frequency signal into a multi-tap delay structure to form an adjustable self-interference reconstruction signal, wherein each tap of the multi-tap delay structure comprises a controllable delay unit, an attenuation unit and a phase shift unit; Coupling in a receiving link to obtain a self-interference signal of a receiving end, and obtaining a digital sampling signal through down-conversion and analog-to-digital conversion; modeling a self-interference channel as a finite-length impulse response filter in a digital domain, solving a sparse channel coefficient by applying LASSO regression based on the reference radio frequency signal and the digital sampling signal, and screening a dominant self-interference path corresponding to a non-zero coefficient; Mapping the sequence number, amplitude and phase information of the non-zero coefficient into delay values, attenuation amounts and phase shift amounts of all taps of the multi-tap delay structure, regulating and controlling the multi-tap delay structure accordingly, and reconstructing copies of self-interference signals; and synthesizing the regulated reconstruction signal and the received signal at the radio frequency front end to realize the self-interference cancellation of the radio frequency domain. Further, the regularization parameters of the LASSO regression are configured to equate the number of non-zero coefficients of the sparse channel coefficients to the number of taps N of the multi-tap delay structure to reconstruct the cancellation signal based on N dominant self-interference paths. Further, the LASSO regression is solved by adopting a