A method and system for suppressing simultaneous, same-frequency, full-duplex self-interference

Abstract

This invention relates to the field of communication technology, specifically disclosing a method and system for suppressing self-interference in simultaneous full-duplex communication. The method includes: in the analog domain, using a radio frequency coupling self-interference suppression method, adjusting the amplitude and time delay of the radio frequency transmitted signal to reconstruct the analog domain self-interference signal; subtracting the reconstructed analog domain interference signal from the radio frequency received signal to complete self-interference suppression in the analog domain; and then feeding the filtered signal into an ADC (analog-to-digital converter). In the digital domain, using a deep learning network self-interference suppression method, the digital domain self-interference signal is reconstructed; and the reconstructed digital domain self-interference signal is subtracted from the near-end signal and the radio frequency received signal processed by the ADC to complete self-interference suppression in the digital domain. This method can effectively reduce the impact of self-interference on the signal of simultaneous full-duplex communication technology.

Description

Technical Field

[0001] This invention relates to the field of communication technology, and specifically to a method and system for suppressing simultaneous, same-frequency, full-duplex self-interference. Background Technology

[0002] With the rapid commercialization of fifth-generation mobile communication (5G) internationally, simultaneous full-duplex communication on the same frequency, offering higher communication speeds and throughput, has become a research hotspot. There are three common duplex modes in mobile communication systems: Time Division Duplex (TDD), Frequency Division Duplex (FDD), and Simultaneous Full-Duplex on the Same Frequency (CCFD). Simultaneous full-duplex (hereinafter referred to as full-duplex) breaks through the limitations of existing frequency division duplex and time division duplex modes, ensuring that the transmitter and receiver of wireless communication equipment can work simultaneously when occupying the same frequency resources. It is one of the key technologies of 5G. In full-duplex systems, the signal isolation between the transmit and receive links is limited. Typically, the signal power leaked from the near-end transmit signal to the receiver is much greater than the signal power received by the receiver from the far-end transmitter. This leaked signal can overwhelm the received far-end transmit signal, and may even saturate and block the receive link; this phenomenon is called self-interference.

[0003] Simultaneous transmission and reception of signals by the transmitter and receiver on the same frequency band will cause the local transmitted signal to enter the receiver through electromagnetic coupling, resulting in severe self-interference at the receiver front end and deterioration of the receiver's demodulation performance. Therefore, the key issue in achieving simultaneous full-duplex transmission on the same frequency is how to effectively suppress self-interference. Summary of the Invention

[0004] To address the shortcomings of the aforementioned technologies, the problem this invention aims to solve is: how to provide a method and system for suppressing self-interference in simultaneous full-duplex communication at the same frequency, so as to better reduce the impact of self-interference on the signal of simultaneous full-duplex technology.

[0005] To solve the above problems, the present invention adopts the following technical solution:

[0006] A method for suppressing simultaneous, same-frequency, full-duplex self-interference includes:

[0007] S1. In the analog domain, the self-interference suppression method of radio frequency coupling is used to reconstruct the analog domain self-interference signal after adjusting the amplitude and time delay of the radio frequency transmitted signal. The self-interference suppression in the analog domain is completed by subtracting the reconstructed analog domain interference signal from the radio frequency received signal, and the filtered signal is transferred to the ADC analog-to-digital converter.

[0008] S2. In the digital domain, the self-interference signal is reconstructed using a deep learning network self-interference suppression method. The reconstructed digital domain self-interference signal is then subtracted from the near-end signal and the RF received signal processed by the ADC digital-to-analog converter, thus completing the self-interference suppression in the digital domain.

[0009] Furthermore, the self-interference suppression method for deep learning networks in step S2 includes:

[0010] S21. Establish a self-interference channel model based on the Hammerstein model, which conforms to...

[0011]

[0012] In the formula, K and M represent the polynomial order and memory depth, respectively, r[n] is the input of the dynamic nonlinear system, and g[n] is the output of the dynamic nonlinear system.

[0013] S22. A deep learning network model is established using the self-interference channel model, and the digital domain self-interference signal is reconstructed by the deep learning network model. conform to:

[0014]

[0015] In the formula: K and M represent the polynomial order and memory depth, respectively, and x[n] is the signal transmitted by the near-end device.

[0016] More preferably, the deep learning network model is trained offline and inferred online to obtain the reconstructed self-interference signal.

[0017] Furthermore, the deep learning network model is optimized using the Adam algorithm, which conforms to the following:

[0018] m t =β1m t-1 +(1-β1)g t ,

[0019]

[0020]

[0021]

[0022]

[0023] In the formula: and To correct for the estimation of the first and second moments of the gradient, β1 is set to 0.9, β2 to 0.999, and ε to le-8.

[0024] Furthermore, in the deep learning network model, the fully connected layer performs nonlinear combination of input data features, and the long short-term memory structure layer extracts time-related characteristics from the input data features.

[0025] Specifically, in step S1, the analog domain self-interference reconstruction signal s C (t) satisfies:

[0026] s C (t)=α C s(t-τ C ),

[0027] In the formula: α C τ is the amplitude adjustment factor. C For time delay;

[0028] A system for the method of suppressing simultaneous, same-frequency, full-duplex self-interference includes:

[0029] The radio frequency coupling self-interference suppression module is used to process the radio frequency transmitted signal, reconstruct the analog domain self-interference signal, subtract the reconstructed analog domain self-interference signal from the radio frequency received signal to complete the self-interference suppression in the analog domain, and then transfer the filtered signal to the ADC analog-to-digital converter.

[0030] The deep learning network self-interference suppression module uses a deep learning network self-interference suppression method to reconstruct the digital domain self-interference signal. The reconstructed digital domain self-interference signal is then subtracted from the near-end signal and the RF received signal processed by the ADC digital-to-analog converter to complete the self-interference suppression in the digital domain.

[0031] Furthermore, the radio frequency coupling self-interference suppression module includes an amplitude delay adjustment unit, an analog domain self-interference reconstruction unit, and an analog domain self-interference cancellation unit;

[0032] The amplitude and delay adjustment unit is used to adjust the amplitude and delay of the radio frequency transmitted signal;

[0033] The analog domain self-interference reconstruction unit is used to receive the signal adjusted by the amplitude delay adjustment unit and reconstruct the analog domain self-interference signal.

[0034] The analog domain self-interference cancellation unit, based on the analog domain self-interference signal information reconstructed by the analog domain self-interference reconstruction unit, subtracts the received radio frequency signal from the reconstructed analog domain self-interference signal to complete self-interference suppression in the analog domain.

[0035] Furthermore, the deep learning network self-interference suppression module includes a reference signal preprocessing unit, a signal feature reconstruction unit, a deep learning network unit, and a digital domain self-interference cancellation unit;

[0036] The reference signal preprocessing unit is used for receiving and preprocessing the reference signal, and for obtaining feature reconstruction parameters;

[0037] The signal feature reconstruction module is used to reconstruct the feature reconstruction parameters obtained by the reference signal processing unit;

[0038] Deep learning network units are used for analysis and reconstruction of self-interference signals in the digital domain;

[0039] The digital domain self-interference cancellation unit, based on the reconstructed digital domain self-interference signal derived from the deep learning network unit, subtracts it from the near-end signal and the RF received signal processed by the ADC digital-to-analog converter, thus completing the self-interference suppression in the digital domain.

[0040] Specifically, the deep learning network unit includes a long short-term memory (LSTM) sub-unit and a fully connected sub-unit; the LSM sub-unit is used to reconstruct the temporal information features in the digital domain self-interference signal; and the fully connected sub-unit is used to analyze the nonlinear components in the digital domain self-interference signal.

[0041] The beneficial effects of this invention are as follows: This invention reconstructs the analog domain self-interference signal in the analog domain, subtracts the reconstructed analog domain interference signal from the RF received signal, thus achieving self-interference suppression in the analog domain. The filtered signal is then fed into an ADC (Analog-to-Digital Converter), and a deep learning network self-interference suppression method is used to reconstruct the digital domain self-interference signal. The reconstructed digital domain self-interference signal is then subtracted from the near-end signal and the RF received signal processed by the ADC, thus achieving self-interference suppression in the digital domain. By first suppressing self-interference in the far-end signal (RF received signal) in the analog domain, and then optimizing the digital domain self-interference reconstruction using a deep learning network, the self-interference in the digital domain can be accurately subtracted, further reducing the impact of self-interference on simultaneous full-duplex communication signals at the same frequency. Attached Figure Description

[0042] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will now be described in further detail with reference to the accompanying drawings, wherein:

[0043] Figure 1 This is a flowchart illustrating the steps of the method for suppressing simultaneous full-duplex self-interference at the same frequency according to the present invention.

[0044] Figure 2 This is a schematic diagram illustrating the principle of the method for suppressing simultaneous full-duplex self-interference at the same frequency according to the present invention.

[0045] Figure 3 This is a schematic diagram of the simultaneous, same-frequency, full-duplex self-interference suppression system of the present invention. Detailed Implementation

[0046] The present invention will be further described in detail below with reference to specific embodiments.

[0047] It should be noted that these embodiments are only used to illustrate the present invention and are not intended to limit the present invention. Simple improvements to the method under the premise of the present invention are all within the scope of protection claimed by the present invention.

[0048] See attached document Figure 1 , 2 This invention provides a method for suppressing simultaneous, same-frequency, full-duplex self-interference, comprising:

[0049] S1. In the analog domain, the RF coupling self-interference suppression method is used to reconstruct the analog domain self-interference signal after amplitude and time delay adjustment of the RF transmitted signal. The self-interference signal in the analog domain is then reconstructed by subtracting the reconstructed analog domain interference signal from the RF received signal. The filtered signal is then fed into the ADC analog-to-digital converter. The reconstructed analog domain self-interference signal s C (t) satisfies,

[0050] s C (t)=α C s(t-τ C ),

[0051] In the formula: α C τ is the amplitude adjustment factor. C For time delay.

[0052] S2. In the digital domain, the self-interference signal is reconstructed using a deep learning network self-interference suppression method. The reconstructed digital domain self-interference signal is then subtracted from the near-end signal and the RF received signal processed by the ADC digital-to-analog converter, thus completing the self-interference suppression in the digital domain.

[0053] The self-interference suppression methods for deep learning networks in step S2 include:

[0054] S21. Establish a self-interference channel model based on the Hammerstein model, which conforms to...

[0055]

[0056] In the formula, K and M represent the polynomial order and memory depth, respectively, r[n] is the input of the dynamic nonlinear system, and g[n] is the output of the dynamic nonlinear system. The Volterra series is one of the models that can accurately model dynamic nonlinear systems. The Hammerstein model is obtained by simplifying the Volterra series model.

[0057] S22. Establish a deep learning network model through a self-interference channel model, and reconstruct the digital domain self-interference signal using the deep learning network model. conform to:

[0058]

[0059] In the formula: K and M represent the polynomial order and memory depth, respectively, and x[n] is the signal transmitted by the near-end device.

[0060] The deep learning network model is trained offline and inferred online to reconstruct the self-interference signal in the digital domain. Offline training is performed in a full-duplex communication system where the near-end device does not receive useful signals (RF signals) from the far end, only receiving interference signals from itself. The signal received by the near-end device in this scenario is simply the residual self-interference signal. The near-end device's transmitted signal and the residual self-interference signal are used as training samples. Other parameters, such as transmit power, are adjusted to ensure the training data covers a wide range of noise ratios or adapts to more dynamic channel scenarios, thereby guaranteeing a relatively stable network model under various signal-to-noise ratios or channel conditions.

[0061] The deep learning network model is optimized using the Adam algorithm, which conforms to the following:

[0062] m t =β1m t-1 +(1-β1)g t ,

[0063]

[0064]

[0065]

[0066]

[0067] In the formula: and To correct for the estimation of the first and second moments of the gradient, β1 is set to 0.9, β2 to 0.999, and ε to le-8.

[0068] In deep learning network models, fully connected layers nonlinearly combine input data features, while long short-term memory (LSTM) layers extract time-related characteristics from these features. Higher-order nonlinear mappings can be modeled using fully connected layers with nonlinear activation function units in deep feedforward network structures, while time-related higher-order nonlinear function mappings can be modeled using LSM layers with time-information memory effects.

[0069] See attached document Figure 3 The present invention also provides a system for suppressing simultaneous, same-frequency, full-duplex self-interference, comprising:

[0070] The radio frequency coupling self-interference suppression module 1 processes the radio frequency transmitted signal to reconstruct the self-interference signal. It then subtracts the reconstructed analog domain self-interference signal from the received radio frequency signal to suppress self-interference in the analog domain. The filtered signal is then fed into the ADC (Analog-to-Digital Converter). The reconstructed analog domain self-interference signal s... C(t) satisfies,

[0071] s C (t)=α C s(t-τ C ),

[0072] In the formula: α C τ is the amplitude adjustment factor. C For time delay.

[0073] Deep learning network self-interference suppression module 2 employs a deep learning network self-interference suppression method to reconstruct the digital domain self-interference signal. The reconstructed digital domain self-interference signal is then subtracted from the near-end signal and the RF received signal processed by the ADC (Digital-to-Analog Converter) to complete self-interference suppression in the digital domain. The deep learning network self-interference suppression method specifically includes:

[0074] A self-interference channel model is established based on the Hammerstein model, which conforms to...

[0075]

[0076] In the formula, K and M represent the polynomial order and memory depth, respectively, r[n] is the input of the dynamic nonlinear system, and g[n] is the output of the dynamic nonlinear system;

[0077] A deep learning network model is established by using a self-interference channel model, and the digital domain self-interference signal is reconstructed based on the deep learning network model. conform to:

[0078]

[0079] In the formula: K and M represent the polynomial order and memory depth, respectively, and x[n] is the signal transmitted by the near-end device.

[0080] The radio frequency coupling self-interference suppression module includes an amplitude delay adjustment unit 11, an analog domain self-interference reconstruction unit 12, and an analog domain self-interference cancellation unit 13;

[0081] The amplitude and delay adjustment unit 11 is used to adjust the amplitude and delay of the radio frequency transmitted signal;

[0082] The analog domain self-interference reconstruction unit 12 is used to receive the signal adjusted by the amplitude delay adjustment unit 11 and reconstruct the analog domain self-interference signal.

[0083] The analog domain self-interference cancellation unit 13, based on the analog domain interference signal information reconstructed by the analog domain self-interference reconstruction unit 12, subtracts the received radio frequency signal from the reconstructed analog domain self-interference signal to complete the self-interference suppression in the analog domain.

[0084] The deep learning network self-interference suppression module 2 includes a reference signal preprocessing unit 21, a signal feature reconstruction unit 22, a deep learning network unit 23, and a digital domain self-interference cancellation unit 24;

[0085] The reference signal preprocessing unit 21 is used for receiving and preprocessing the reference signal, and for obtaining feature reconstruction parameters;

[0086] The signal feature reconstruction module 22 is used to reconstruct the feature reconstruction parameters obtained by the reference signal processing unit 21;

[0087] The deep learning network unit 23 is used for self-interference signal analysis and reconstruction; the deep learning network unit 23 includes a long short-term memory network subunit 231 and a fully connected layer subunit 232; the long short-term memory network subunit 231 is used to reconstruct the temporal information features in the self-interference signal; the fully connected layer subunit 232 reconstructs the nonlinear components in the self-interference signal.

[0088] Self-interference cancellation unit 24, based on the reconstructed self-interference signal obtained from the deep learning network unit, subtracts the self-interference signal from the near-end signal and the radio frequency received signal processed by the ADC digital-to-analog converter to complete self-interference suppression in the digital domain.

[0089] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described with reference to preferred embodiments, those skilled in the art should understand that various changes in form and detail can be made without departing from the spirit and scope of the invention as defined in the appended claims.

Claims

1. A method of suppressing self-interference for simultaneous transmit- receive full-duplex, the method comprising: Comprise: S1, in the analog domain, the self-interference suppression method is used to adjust the amplitude and time delay of the radio frequency transmission signal, the analog domain self-interference signal is reconstructed, the radio frequency receiving signal is subtracted from the reconstructed analog domain interference signal, the self-interference suppression in the analog domain is completed, and the filtered signal is converted into an ADC analog-to-digital converter; S2, in the digital domain, the self-interference suppression method is used to reconstruct the digital domain self-interference signal, and the reconstructed digital domain self-interference signal is subtracted from the near-end signal and the radio frequency receiving signal processed by the ADC digital-analog converter, so that the self-interference suppression in the digital domain is completed; The self-interference suppression method of the deep learning network in the step S2 comprises: S21, a self-interference channel model is established based on Hammerstein model, which meets In the formula, K and M represent the polynomial order and the memory depth respectively, r[n] is the input of the dynamic nonlinear system, and g[n] is the output of the dynamic nonlinear system; S22, a deep learning network model is established through the self-interference channel model, and a digital domain self-interference signal reconstructed through the deep learning network model Conformity: In the formula: K and M represent the polynomial order and the memory depth respectively, and x[n] is the signal sent by the near-end device; The step S1 analog domain self-interference reconstructed signal s C (t) complies with: s C (t) = a C s(t - τ C ), In the formula, α C is an amplitude adjustment factor, τ C is a time delay.

2. The method of claim 1, wherein, The deep learning network model is trained in an offline training mode and performs network inference in an online mode to obtain the reconstructed digital domain self-interference signal.

3. The method of suppressing self-interference for simultaneous transmit- receive full-duplex communications of claim 2, wherein, The deep learning network model is optimized by using Adam algorithm, and the Adam algorithm meets: m t = β1m t-1 + (1 - β1)g t , where: and are corrections for the first and second moment estimates of the gradient, β1 takes 0.9, β2 takes 0.999, and ε takes le-8.

4. The method of suppressing self-interference for simultaneous transmit- receive full-duplex communications of claim 2, wherein, The full connection layer in the deep learning network model performs nonlinear combination on the input data features, and the long short-term memory structure layer extracts the time information related characteristics in the input data features.

5. System for use in the method of suppression of self-interference for simultaneous- same-frequency full-duplex of any of claims 1 - 4, characterized in that, Comprise: The radio frequency coupling self-interference suppression module is used to process the radio frequency transmission signal, reconstruct the analog domain self-interference signal, subtract the radio frequency receiving signal from the reconstructed analog domain self-interference signal, complete the self-interference suppression in the analog domain, and convert the filtered signal into an ADC analog-to-digital converter; The deep learning network self-interference suppression module uses the deep learning network self-interference suppression method to reconstruct the digital domain self-interference signal, subtracts the reconstructed digital domain self-interference signal from the radio frequency receiving signal processed by the ADC digital-analog converter, and completes the self-interference suppression in the digital domain.

6. The system for the mitigation of self-interference for simultaneous- in-frequency full-duplex of claim 5, wherein, The radio frequency coupling self-interference suppression module comprises an amplitude time delay adjustment unit, an analog domain self-interference reconstruction unit and an analog domain self-interference cancellation unit; The amplitude time delay adjustment unit is used for adjusting the amplitude and time delay of the radio frequency transmission signal; The analog domain self-interference reconstruction unit is used for receiving the signal adjusted by the amplitude time delay adjustment unit and reconstructing the analog domain self-interference signal; The analog domain self-interference cancellation unit subtracts the received radio frequency signal from the reconstructed analog domain self-interference signal based on the analog domain self-interference signal information reconstructed by the analog domain self-interference reconstruction unit, and completes the self-interference suppression in the analog domain.

7. The system for the mitigation of self-interference for simultaneous- in-frequency full-duplex of claim 5, wherein, The deep learning network self-interference suppression module comprises a reference signal preprocessing unit, a signal feature reconstruction unit, a deep learning network unit and a digital domain self-interference cancellation unit; The reference signal preprocessing unit is used for receiving and preprocessing the reference signal and obtaining feature reconstruction parameters; The signal feature reconstruction module is used for reconstructing the feature reconstruction parameters obtained by the reference signal processing unit; The deep learning network unit is used for analyzing and reconstructing the digital domain self-interference signal. The digital domain self-interference cancellation unit subtracts the reconstructed digital domain self-interference signal obtained by the deep learning network unit from the near-end signal and the ADC digital-analog converter processed radio frequency receiving signal to complete the self-interference suppression in the digital domain.

8. The system for the mitigation of self-interference for simultaneous- in-frequency full-duplex of claim 7, wherein, The deep learning network unit comprises a long short-term memory network layer subunit and a fully connected layer subunit; the long short-term memory network layer subunit is used for reconstructing time information features in the digital domain self-interference signal; and the fully connected layer subunit is used for analyzing nonlinear components in the digital domain self-interference signal.

Images