Self-interference cancellation method and system based on hyperbolic tangent function variable step length LMS algorithm

Abstract

The invention provides a self-interference cancellation method and system based on a hyperbolic tangent function variable step length LMS algorithm, the method proposes a digital domain-based simultaneous and same-frequency self-interference cancellation model, uses an improved hyperbolic tangent function to establish a non-linear relationship between an error and a step length, and sets a step length control parameter. And the value of the step size of the improved hyperbolic tangent function is adjusted according to the error value, so that dynamic parameter adjustment is realized: when the error in the initial stage of system convergence is large, the larger step size value is adopted to improve the convergence speed, and when the error in the steady-state working interval is small, the smaller step size value is adopted to reduce the steady-state error. According to the method, different synchronization length control parameters can be selected according to different engineering scenes, so that the convergence speed is effectively improved, a relatively small steady-state error is achieved, and thus digital domain self-interference cancellation is realized.

Description

Technical Field

[0001] The present invention relates to the fields of digital signal processing technology and wireless communication transmission, and particularly relates to a self-interference cancellation method and system based on a hyperbolic tangent function variable step-size LMS algorithm. Background Art

[0002] With the rapid development of wireless communication technology, ensuring the security and reliability of information transmission has become increasingly crucial. At the same time, co-frequency communication technology, with its significant spectrum utilization advantage, can effectively alleviate the contradiction between limited spectrum resources and the explosive growth of communication service demands. However, co-frequency communication has a serious self-interference problem in implementation. Digital domain self-interference cancellation schemes achieve precise cancellation by constructing a self-interference signal model. Among them, most common schemes are based on adaptive filtering technologies such as the Least Mean Square (LMS) algorithm and the Recursive Least Square (RLS) algorithm. The LMS algorithm aims to minimize the Mean Square Error (MSE), combines the Wiener filter theory and the steepest gradient descent method optimization strategy to find the optimal value, and has advantages such as low algorithm operation complexity and strong robustness. It has been widely applied to digital domain self-interference cancellation in co-frequency communication systems.

[0003] In the process of iterative update of the weight coefficients of the LMS algorithm, the filter weight coefficients are dynamically adjusted along the negative gradient direction of the error surface, and the estimation accuracy of its gradient value directly affects the accuracy of the algorithm. The existing LMS algorithm adopts a design method with a fixed step-size parameter, which is difficult to effectively cope with the challenges brought by the time-varying characteristics of the self-interference channel: although a larger step-size setting can improve the convergence rate, the accuracy will be greatly reduced; a smaller step-size is beneficial to improving the convergence accuracy, but it will prolong the convergence process. Especially when the step-size parameter exceeds the reasonable range, the gradient estimation error will cause the divergence phenomenon of the weight coefficients, resulting in the inability of the weight coefficients of the adaptive filter to converge. It can be seen that the existing LMS algorithm has the defect that the two important performances of convergence speed and steady-state error cannot be improved simultaneously. Therefore, when selecting the step size of the LMS algorithm, the convergence accuracy and speed need to be comprehensively considered. Although many variable step-size LMS algorithms have been proposed to improve the problem of poor self-interference cancellation performance caused by the fixed step size of the LMS algorithm, most of these variable step-size LMS algorithms currently have problems of unreliable performance or too high computational complexity in practical scenario applications. The main reason is that these algorithms rely on many parameters that need to be manually adjusted, and manual parameter adjustment requires a large amount of practice and has poor adaptability to different environments. Summary of the Invention

[0004] In order to overcome the defect that the two important performances of the existing LMS algorithm, namely the convergence speed and the steady-state error, cannot be improved simultaneously, and the step-size control parameter of the existing variable-step LMS algorithm cannot be adaptively adjusted, the present invention provides a self-interference cancellation method and system based on a variable-step LMS (HT-VSS-LMS) algorithm with a hyperbolic tangent function. This method proposes a simultaneous co-frequency self-interference cancellation model based on the digital domain, establishes a non-linear relationship between the error and the step size by using an improved hyperbolic tangent function, sets the step-size control parameter and the principle of adaptive update of the step-size control parameter, and adjusts the value of the step size of the improved hyperbolic tangent function according to the error value. The present invention can either manually select different step-size control parameters according to different engineering scenarios, or adaptively select appropriate step-size control parameters through the adaptive update strategy of the step-size control parameter to effectively improve the convergence speed and have a small steady-state error, thereby realizing self-interference cancellation in the digital domain.

[0005] The technical solution of the present invention is as follows:

[0006] A self-interference cancellation method based on a variable-step LMS algorithm with a hyperbolic tangent function, comprising the following steps:

[0007] Step 1: Obtain the transmitted signal of simultaneous co-frequency communication and the received signal , indicating the th point;

[0008] Step 2: Calculate the self-interference signal

[0009]

[0010] according to the formula , where is the weight coefficient;

[0011] Step 3: Calculate the error signal

[0012]

[0013] according to the formula which is the received signal after self-interference cancellation;

[0014] Step 4: Update and iterate the weight coefficient according to the weight coefficient update formula

[0015]

[0016] until the iteration termination condition is reached; where is the weight coefficient at the th point, is the step size, and according to the formula

[0017]

[0018] Obtained, where , , are all control parameters for controlling the step size change.

[0019] Furthermore, adjust the step size formula to:

[0020]

[0021] where is the error signal of the th point.

[0022] Furthermore, adjust the step size formula to:

[0023]

[0024] where is the step size feedback factor.

[0025] Furthermore, design the step size control parameters , , adaptive update strategy:

[0026] Adopt the exponential moving average method to achieve increase when the error is large, and accelerate convergence; when the error tends to be stable, decrease to enhance the steady-state performance:

[0027]

[0028] where, is the smoothing coefficient, ρ ∈ [ 0 , 1 ] , is the scaling factor;

[0029] Use the error variance to adaptively adjust the curvature parameter :

[0030]

[0031] where, is the mapping ratio, , is the sliding window length, L ∈ [ 10 , 30 ] , is the average error within the sliding window range, is the mapping ratio control parameter, is the set minimum value to prevent ;

[0032] Optimize the response speed, accelerate the exponential contraction effect, and improve the dynamic adaptability. Set it as:

[0033]

[0034] After substituting the above step-size control parameters to update the adaptive update strategy, the step-size formula expression can be obtained as:

[0035] .

[0036] Furthermore, the step-size feedback factor is calculated according to the formula

[0037]

[0038] where is a set coefficient, is the step-size feedback factor at the th point.

[0039] Furthermore, adjust the weight coefficient update formula to

[0040]

[0041] where is a set constant.

[0042] In addition, the present invention also proposes a self-interference cancellation system based on the hyperbolic tangent function variable step-size LMS algorithm, where the weight coefficient of the adaptive filter is , and the update formula of

[0043]

[0044] where is the transmitted signal of the co-frequency communication, is the received signal of the co-frequency communication is the error signal between and the self-interference signal output by the adaptive filter, that is, the received signal after self-interference cancellation, is the weight coefficient at the th point, is the step size, and is obtained according to the formula

[0045]

[0046] where , , are all control parameters for controlling the step-size change.

[0047] Furthermore, adjust the step-size formula to:

[0048]

[0049] wherein is the error signal of the th point.

[0050] Furthermore, adjust the step size formula to:

[0051]

[0052] wherein is the step size feedback factor.

[0053] Furthermore, design the step size control parameters , , adaptive update strategy:

[0054] Adopt the exponential moving average method to achieve increase when the error is large, and accelerate convergence; when the error tends to be stable, decrease to enhance the steady-state performance:

[0055]

[0056] wherein, is the smoothing coefficient, ρ ∈ [ 0 , 1 ] , is the scaling factor;

[0057] Use the error variance to adaptively adjust the curvature parameter of the function:

[0058]

[0059] wherein, is the mapping ratio, , is the sliding window length, L ∈ [ 10 , 30 ] , is the average error within the sliding window range, is the mapping ratio control parameter, is the set minimum value to prevent ;

[0060] Optimize the response speed to accelerate the exponential contraction effect and improve the dynamic adaptability, set it to:

[0061]

[0062] After substituting the above step size control parameters to update the adaptive update strategy, the step size formula expression can be obtained as:

[0063] 。

[0064] Furthermore, the step feedback factor is calculated according to the formula

[0065]

[0066] wherein is a set coefficient, is the th step feedback factor of the point.

[0067] Furthermore, the weight coefficient update formula is adjusted to

[0068]

[0069] wherein is a set constant.

[0070] Beneficial effects:

[0071] Compared with the existing LMS algorithm self-interference cancellation technology, the present invention has the following advantages:

[0072] 1. First, the present invention uses an improved hyperbolic tangent function to construct a non-linear relationship between the error signal and the step size . The improved hyperbolic tangent function has the characteristic of being symmetric about relative to a certain value. When the value of the error signal is large (i.e., at the initial stage of the algorithm), the corresponding step size is also large, and at this time, the convergence speed of the algorithm can be accelerated. When the value of the error signal tends to (i.e., the algorithm converges and stably approaches the actual communication signal to be received), the corresponding step size value is small, and at this time, the steady-state error of the algorithm can be reduced. The improved hyperbolic tangent function conforms to the step size selection principle for optimizing the LMS algorithm, enabling the present invention to overcome the defect that the two important performances of the existing LMS algorithm, namely the convergence speed and the steady-state error, cannot be improved simultaneously, and at the same time having a high convergence speed and a small steady-state error.

[0073] 2. Secondly, the present invention uses a step control parameter to control the improved hyperbolic tangent function, so as to better adjust the value of the step of the improved hyperbolic tangent function according to the error value (i.e., the received signal after self-interference cancellation), taking into account both the convergence and stability of the algorithm. In addition, in a further preferred embodiment of the present invention, the step is further optimized, the influence of random variables is considered, the relationship between the step and the input signal is strengthened, and the tracking effect of the algorithm is improved; at the same time, when updating the weight coefficient, a normalization algorithm is used to limit the step, reducing the influence of the sudden increase of the received signal on the algorithm.

[0074] 3. Aiming at the problem that the step parameter in the common variable-step self-interference cancellation algorithm needs to be manually adjusted, resulting in unreliable algorithm or too high complexity in self-interference cancellation, the present invention designs a step control parameter by using the exponential moving average method and the error variance method 、 、 and an adaptive update strategy, which reduces the complexity of manual participation and hyperparameter tuning, has stronger robustness and fast adaptability, and enhances the generality and engineering implementation of the algorithm.

[0075] The present invention provides strong theoretical support for the engineering implementation of self-interference cancellation and can be used to improve the self-interference cancellation ability of the co-frequency communication system.

[0076] Some of the additional aspects and advantages of the present invention will be given in the following description, some will become obvious from the following description, or be understood through the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] The above and / or additional aspects and advantages of the present invention will become obvious and easy to understand from the description of the embodiments in conjunction with the following drawings, wherein:

[0078] Figure 1 : Block diagram of the co-frequency self-interference cancellation system;

[0079] Figure 2 : Influence of the step parameter in the HT-VSS-LMS algorithm on the step on the step;

[0080] Figure 3 : Influence of the step parameter in the HT-VSS-LMS algorithm on the step on the step;

[0081] Figure 4 : Influence of the step parameter in the HT-VSS-LMS algorithm on the step on the step;

[0082] Figure 5 : Influence of the step parameter in the HT-VSS-LMS algorithm on the step on the step;

[0083] Figure 6 : Simultaneous and co-frequency self-interference cancellation performance curve based on HT-VSS-LMS. Specific implementation manners

[0084] The embodiments of the present invention will be described in detail below. The embodiments are exemplary and are intended to explain the present invention, but should not be construed as limiting the present invention.

[0085] Embodiment 1:

[0086] In simultaneous and co-frequency communication, the digital-domain self-interference cancellation scheme based on the LMS algorithm has been relatively common. As Figure 1 shown, is the transmitted signal of simultaneous and co-frequency communication and is also the input signal of the adaptive filter in the self-interference cancellation system, which is used for self-interference channel estimation. The adaptive filter processes and outputs . The actual received signal of simultaneous and co-frequency communication is the desired signal of the adaptive filter, which contains the communication signal to be received and the self-interference signal. is the error signal between the output signal of the adaptive filter and the actual received signal , which is used to adjust the adaptive filter and is also the received signal after self-interference cancellation.

[0087] The principle of the simultaneous and co-frequency self-interference cancellation scheme based on the LMS algorithm is as follows: The adaptive filter performs iterative update calculations on the weight coefficient and the error . After iteration, the adaptive filter outputs the reconstructed self-interference signal . Then, the reconstructed self-interference signal is subtracted from the actual received signal . The obtained error signal is the received signal after self-interference cancellation.

[0088] The traditional LMS algorithm uses a fixed step-size parameter to iteratively update the weight coefficients, making it difficult to effectively address the challenges brought by the time-varying characteristics of the self-interference channel. There are also solutions that adopt variable step-size designs, but it is difficult for the variable step-size adjustment method to simultaneously improve the two important performances of the convergence speed and the steady-state error. Therefore, in this embodiment, a self-interference cancellation method based on the variable step-size LMS algorithm with hyperbolic tangent function is proposed. A non-linear correlation function between the error signal and the iteration step-size is constructed using the hyperbolic tangent function to achieve dynamic parameter adjustment: when the error is large at the initial stage of system convergence, a larger step-size value is adopted to improve the convergence speed. When the error is small in the steady-state working range, a smaller step-size value is adopted to reduce the steady-state error. The strategy of dynamically adjusting the step-size in this embodiment overcomes the contradiction between the convergence speed and the steady-state accuracy caused by the fixed step-size of the traditional LMS algorithm. Simulation verification shows that the present invention has good control over the step-size with different step-size control parameters, and improves the performance of the algorithm when applied to self-interference cancellation.

[0089] The specific steps in this embodiment are as follows:

[0090] Step 1: Obtain the transmitted signal of the simultaneous and co-frequency communication and the received signal , because the present invention is based on the digital-domain simultaneous and co-frequency self-interference cancellation model, so represents the th point;

[0091] Step 2: Calculate the self-interference signal

[0092]

[0093] according to the formula , where is the weight coefficient, and the superscript T represents the transpose;

[0094] Step 3: Calculate the error signal

[0095]

[0096] according to the formula , and the obtained

[0097] is the received signal after self-interference cancellation;

[0098]

[0099] Step 4: Update and iterate the weight coefficients according to the weight coefficient update formula until the iteration termination condition is reached; where is the weight coefficient at the th point,

[0100]

[0101] Based on this, parameters are introduced , , jointly control the change curve of the step size, adjust the value of the step size according to the error value, better control the change of the step size value, and the obtained step size formula is

[0102]

[0103] The improved hyperbolic tangent function has the characteristic of being symmetric about relative to a certain value. When the error signal has a large value (i.e., at the initial stage of the algorithm), the corresponding step size is also large, and at this time, the convergence speed of the algorithm can be accelerated. When the value of the error signal tends to (i.e., the algorithm converges and stably approaches the communication signal to be actually received), the corresponding step size value is small, and at this time, the steady-state error of the algorithm can be reduced. The improved hyperbolic tangent function conforms to the step size selection principle of the optimized LMS algorithm, can overcome the defect that the two important performances of the existing LMS algorithm, namely the convergence speed and the steady-state error, cannot be improved simultaneously, and at the same time has a high convergence speed and a small steady-state error.

[0104] Example 2:

[0105] In this embodiment, in order to reduce the influence of random variables, the error signal in the step size formula is replaced by its correlation value , so as to adjust the step size formula to:

[0106]

[0107] where is the error signal at the th point.

[0108] Example 3:

[0109] In this embodiment, in order to further strengthen the relationship between the step size and the input signal and improve the tracking effect of the algorithm, a step size feedback factor is added to the step size formula:

[0110]

[0111] where is a set coefficient, is the step size feedback factor at the th point. In this embodiment, the initial value of the step size feedback factor is taken as 1, and the set coefficient Comprehensively consider according to the step-size curve.

[0112] Thus, the step-size formula is adjusted to:

[0113] .

[0114] An important difference between the HT-VSS-LMS algorithm and the LMS algorithm in this embodiment is the selection of the step size. In this embodiment, finally, through , , and control the step size, and adjust the value of the step size according to the error value. The relationship between the step size and the error signal is as shown in Figures 2 to 5 .

[0115] The step-size parameter has a great influence on the value of the step size . As increases, the value of the step size also increases; it can be seen from Figures 3 to 5 that the step-size parameters , , affect the shape of the step-size curve. When is smaller, is larger, and is smaller, the corresponding step-size value is smaller under the same error signal value. Combining the analysis of the influence of Figures 2 to 5 different parameters on the step-size value, in actual engineering applications, take the step size when the algorithm is close to the convergence state, so as to comprehensively consider the influence of each parameter on the step size and obtain better performance.

[0116] Example 4:

[0117] In this embodiment, in order to reduce the influence of the sudden increase of the received signal on the algorithm, a normalization algorithm is used to limit the step size , to avoid a large value of the step size, and thus the weight coefficient update formula is adjusted to

[0118]

[0119] where is a set constant, and its value is 1 in this embodiment.

[0120] Example 5:

[0121] This embodiment proposes a self-interference cancellation system based on a hyperbolic tangent function variable step-size LMS algorithm, where the adaptive filter weight coefficient is , The update formula is

[0122]

[0123] where is a set constant, and its value is 1 in this embodiment; is the transmitted signal of co - time and co - frequency communication, is the received signal of co - time and co - frequency communication is the error signal between and the self - interference signal output by the adaptive filter, that is, the received signal after self - interference cancellation, is the step size, which is obtained according to the formula

[0124]

[0125] where , , are all control parameters for controlling the change of the step size, is the step - size feedback factor:

[0126]

[0127] where is the set coefficient, and the set coefficient is considered comprehensively according to the step - size curve.

[0128] Embodiment 6:

[0129] This embodiment proposes a self - adaptive update strategy for step - size control parameters, aiming to solve the problem that in the common variable - step - size self - interference cancellation algorithm, the step - size parameter needs to be manually adjusted, resulting in an unreliable algorithm or a too high complexity in self - interference cancellation.

[0130] Design the step - size control parameters , , self - adaptive update strategy:

[0131] Adopt the exponential moving average method to achieve that when the error is large increases to accelerate convergence; when the error tends to be stable, decreases to enhance the steady - state performance:

[0132]

[0133] where, is the smoothing coefficient, ρ ∈ [ 0 , 1 ] , is the scaling factor;

[0134] Use the error variance to adaptively adjust the curvature parameter of the function :

[0135]

[0136] wherein, is the mapping ratio, , is the sliding window length, L ∈ [ 10 , 30 ] , is the average error within the sliding window range, is the mapping ratio control parameter, is the set minimum value to prevent , and the value is ;

[0137] Optimizing the response speed to accelerate the exponential contraction effect and improve the dynamic adaptability, is set as:

[0138]

[0139] After substituting the above step size control parameter to update the adaptive update strategy, the step size formula expression can be obtained as:

[0140] .

[0141] By updating and iterating the weight coefficients of the adaptive filter, simultaneous co-frequency self-interference cancellation is achieved. The following is the simulation analysis:

[0142] The self-interference signal and communication signal in the simulation are generated by BPSK modulation. The signal-to-interference ratio (SIR) is set to -50~20 dB. Under different SIRs, by comparing the residual self-interference signal power and the received self-interference signal power, the full-duplex self-interference cancellation performance of the algorithm is analyzed. For the convenience of comparison, the communication signal power is respectively compared with the power values of the residual self-interference signal and the received self-interference signal. Define RES-SIR to represent the ratio of the communication signal power to the residual self-interference signal power. The self-interference cancellation performance curve based on HT-VSS-LMS is as Figure 6 shown.

[0143] According to Figure 6, it can be seen from the analysis in the direction of increasing the power of the self-interference signal from small to large. When the SIR is greater than 0 dB, the RES_SIR of the residual self-interference signal remains basically unchanged and is near 0 dB as the SIR value decreases. At this time, the received communication signal power is much greater than the received self-interference signal, and the self-interference suppression effect of the algorithm is not obvious. When the SIR is between -40 dB and 0 dB, as the SIR value decreases, the difference between the RES-SIR value of the residual self-interference signal and the SIR also increases, indicating that as the intensity of the self-interference signal increases, the self-interference cancellation performance of the HT-VSS-LMS algorithm improves. When the INR is less than -40 dB, a linear relationship is shown between the RES-SIR value of the residual self-interference signal and the SIR value. Because in the process of self-interference cancellation, the process of simulating the self-interference signal is a convergent process, and this convergent process is also reflected in the residual self-interference signal, so there is a linear relationship between the two. The difference between the two is about 33 dB, indicating that under the current parameter settings, the HT-VSS-LMS algorithm can achieve a self-interference cancellation effect of about 33 dB.

[0144] Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention without departing from the principles and purposes of the present invention.

Claims

1. A self-interference cancellation method based on the hyperbolic tangent function variable step size LMS algorithm, characterized in that: Including the following steps: Step 1: Obtain the transmitted signal for co-simultaneous and co-frequency communication and the received signal , denote the th point; Step 2: According to the formula Calculate the self-interference signal , where is the weight coefficient; Step 3: According to the formula The error signal is calculated which is the received signal after self-interference cancellation; Step 4: According to the weight coefficient update formula Update and iterate the weight coefficients until the iteration termination condition is reached; where is the weight coefficient of the -th point, is the step size, and according to the formula obtained, where , , are all control parameters for controlling the step size change.

2. The self-interference cancellation method based on the hyperbolic tangent function variable step size LMS algorithm according to claim 1, wherein: Adjust the step size formula to: wherein is the error signal of the nth point 3. The self-interference cancellation method based on the hyperbolic tangent function variable step size LMS algorithm according to claim 1, characterized in that: Adjust the step size formula to: Among them is the step feedback factor 4. The self-interference cancellation method based on the hyperbolic tangent function variable step size LMS algorithm according to claim 3, wherein: Design step size control parameter , , Adaptive update strategy: Adopt the exponential moving average method to perform adaptive update: Among them, is the smoothing coefficient, , is the scaling factor; Adaptive adjustment is performed using the error variance for : Among them, is the mapping ratio, , is the sliding window length, is the average error within the sliding window range, is the mapping ratio control parameter, is the set minimum value; The adaptive adjustment formula is as follows: After substituting the above step size control parameter adaptive update strategy, the step size formula expression obtained is: 。 5. The self-interference cancellation method based on the hyperbolic tangent function variable step-size LMS algorithm according to claim 3 or 4, characterized in that: The step size feedback factor is based on the formula Calculated, where is a set coefficient,[ is the step feedback factor of the 6. The self-interference cancellation method based on the hyperbolic tangent function variable step size LMS algorithm according to claim 1, characterized in that: Adjust the weight coefficient update formula to wherein is a set constant.

7. A self-interference cancellation system based on a hyperbolic tangent function variable step size LMS algorithm, characterized in that: The adaptive filter weight coefficients of the self-interference cancellation system are , and the update formula is wherein is the transmission signal for simultaneous co-frequency communication, is the received signal for simultaneous co-frequency communication and the self-interference signal output by the adaptive filter is the error signal therebetween, that is, the received signal after self-interference cancellation, is the weight coefficient of the th point, is the step size, according to the formula obtained, where , , are all control parameters for controlling the step size change.

8. The self-interference cancellation system based on the hyperbolic tangent function variable step-size LMS algorithm according to claim 7, wherein: Design step size control parameter 、 、 Adaptive update strategy: Adopt the exponential moving average method to perform adaptive update: Among them, is the smoothing coefficient, , is the scaling factor; Adaptive adjustment is performed using the error variance for : Among them, is the mapping ratio, , is the sliding window length, is the average error within the sliding window range, is the mapping ratio control parameter, is the set minimum value; The adaptive adjustment formula is as follows: After substituting the above step size control parameter adaptive update strategy, the step size formula expression obtained is: 。 9. The self-interference cancellation system based on the hyperbolic tangent function variable step size LMS algorithm according to claim 8, characterized in that: The step size feedback factor is based on the formula Calculated, where is a set coefficient, is the step feedback factor of the -th point.

10. The self-interference cancellation system based on the hyperbolic tangent function variable step size LMS algorithm according to claim 7, characterized in that: Adjust the weight coefficient update formula to wherein is a set constant.

Images