Adaptive filtering noise reduction system based on variable step size minimum mean square error algorithm

Abstract

The invention discloses a self-adaptive filtering noise reduction system based on a variable step size minimum mean square error algorithm, which comprises a signal main link and a closed-loop control module, and is characterized in that the signal main link is sequentially integrated with a low-noise amplifier, an analog-to-digital converter and a self-adaptive filter and is responsible for accurate amplification, efficient analog-to-digital conversion and flexible filtering processing of signals; the closed-loop control system captures and analyzes noise spectrum characteristics in real time through close cooperation of a noise detection unit, a spectrum analysis unit and a parameter adjustment unit, and then dynamically optimizes filtering parameters to achieve the optimal noise reduction effect. A variable step size minimum mean square error algorithm is adopted in a parameter adjusting unit, dynamic adjustment of step size is ingeniously achieved by means of a unique error signal real-time amplitude feedback mechanism in combination with nonlinear characteristics of a hyperbolic tangent function, and therefore ideal self-adaptive balance is achieved between fast convergence and low steady-state errors.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of communication, and particularly relates to a design of an adaptive filtering and noise reduction system based on a variable step size least mean square error algorithm. BACKGROUND

[0002] In a wireless communication system, the quality of the front-end signal processing of the receiver directly affects the reliability and sensitivity of the communication link. With the rapid development of 5G, Internet of Things and high-frequency communication technologies, the receiver needs to process multi-band, time-varying noise interference, including white noise, narrowband interference, impulse noise and the like, in a complex electromagnetic environment. The traditional fixed parameter filter is difficult to effectively suppress non-stationary noise due to the lack of dynamic adaptability, resulting in signal distortion, increased error rate and decreased overall system performance.

[0003] Currently, adaptive filtering technology is widely used for dynamic noise suppression, but its core defect is that the traditional least mean square (LMS) algorithm uses a fixed step size parameter, which is difficult to balance the convergence speed and steady-state error. In addition, existing adaptive filtering schemes rely on single noise feature adjustment parameters, which have insufficient adaptability to complex frequency spectrum distribution scenarios, resulting in limited noise suppression effect.

[0004] In recent years, researchers have proposed improved algorithms based on LMS to address the above problems, but there are still problems such as low dynamic adjustment efficiency, imperfect closed-loop control and high hardware implementation complexity. Therefore, there is an urgent need for an adaptive filtering and noise reduction circuit that takes into account convergence speed, steady-state accuracy and hardware efficiency to meet the stringent requirements of modern wireless communication systems for high sensitivity, low error rate and complex noise suppression capability. SUMMARY

[0005] The purpose of the present application is to solve the problem of low noise reduction performance of existing adaptive filtering and noise reduction technology based on LMS algorithm, and to propose an adaptive filtering and noise reduction system based on a variable step size least mean square error algorithm.

[0006] The technical solution of the present application is: an adaptive filtering and noise reduction system based on a variable step size least mean square error algorithm, comprising a signal main link and a closed-loop control module.

[0007] The signal main link comprises a low noise amplifier, an analog-to-digital converter and an adaptive filter connected in sequence, the low noise amplifier is used for amplifying the input signal to obtain an amplified signal, the analog-to-digital converter is used for analog-to-digital conversion of the amplified signal to obtain a digital signal, and the adaptive filter is used for filtering the digital signal to obtain an output signal.

[0008] The closed-loop control module includes a noise detection unit, a spectrum analysis unit and a parameter adjustment unit; the input end of the noise detection unit is connected to the output end of the analog-to-digital converter, and is used to extract the noise component from the digital signal and calculate the root mean square value of the noise component; the input end of the spectrum analysis unit is connected to the output end of the noise detection unit, and is used to analyze the noise power spectrum density according to the root mean square value of the noise component using fast Fourier transform, and adjust the smoothing parameter and scaling parameter of the variable step size minimum mean square error algorithm according to the noise power spectrum density; the input end of the parameter adjustment unit is connected to the output end of the spectrum analysis unit, and its output end is connected to the adaptive filter, and is used to dynamically adjust the tap coefficient of the adaptive filter using the variable step size minimum mean square error algorithm to form a closed-loop control of the signal main link.

[0009] Furthermore, the noise detection unit extracts the noise component from the digital signal using the following formula: in n is the index representing the discrete time series, represents the noise component, represents the ideal reference signal, Represents the output vector generated by taking the inner product of the input signal vector and the tap coefficients of the adaptive filter.

[0010] Furthermore, the formula for calculating the root mean square value of the noise component in the noise detection unit is: in represents the RMS value of the noise component, N Indicates the number of sampling points for signal analysis, Represents the noise component.

[0011] Furthermore, the spectrum analysis unit responds to the root mean square value of the noise component , the noise power spectral density is analyzed using fast Fourier transform.

[0012] Furthermore, the formula for analytical noise power spectral density is: in Indicates the frequency k The noise power density at represents a single time domain sampling point, n is the index representing the discrete time series, N Indicates the number of sampling points for signal analysis, j Represents an imaginary unit.

[0013] Furthermore, the method for adjusting the smoothing parameter and scaling parameter of the variable step size minimum mean square error algorithm according to the noise power spectrum density in the noise detection unit is specifically as follows: If the variance of the noise power spectral density at each frequency point is σ ²≤0.1, the noise is determined to be flat noise, and the smoothing parameter of the variable step size minimum mean square error algorithm is set to Increase from the initial value of 0.5 to 0.8 to accelerate convergence; if the variance of the noise power spectrum density at each frequency point σ If ²>0.1, the noise is determined to be non-flat noise, and the scaling parameter of the variable step size minimum mean square error algorithm is set to Reduced from the initial value of 0.05 to 0.02 to improve steady-state accuracy.

[0014] Furthermore, the variable step size minimum mean square error algorithm in the parameter adjustment unit includes the following steps: S1. Initialize the parameters of the variable step size minimum mean square error algorithm.

[0015] S2, according to the noise component Calculate the filter step size parameter .

[0016] S3, according to the filter step size parameter The tap coefficients of the adaptive filter are dynamically adjusted.

[0017] Furthermore, step S1 is specifically as follows: Initialize the maximum value of the filter step size parameter μ max =0.1, the minimum value of the filter step parameter μ min =0.001, smoothing parameter The initial value is 0.5, and the scaling parameter The initial value of is 0.05, and the initial value of the tap coefficient vector of the adaptive filter is 0.

[0018] Furthermore, the filter step size parameter in step S2 is The calculation formula is: in represents the smoothing parameter, represents the scaling parameter, represents the hyperbolic tangent function.

[0019] Furthermore, the formula for dynamically adjusting the tap coefficients of the adaptive filter in step S3 is: in represents the filter tap coefficient vector before adjustment, denotes an adjusted filter tap coefficient vector, denotes an input signal vector.

[0020] The present application has the following advantages: (1) The adaptive filter noise reduction system in the present application is based on dynamic step adjustment of variable step size least mean square error algorithm, which significantly improves the convergence speed, while realizing higher signal-to-noise ratio gain, effective narrowband interference suppression and extremely low bit error rate, has high efficient noise reduction performance, significantly enhances the communication reliability in multi-band, time-varying noise environment, and provides an efficient and reliable solution for the front-end signal preprocessing of high-sensitivity receivers.

[0021] (2) The present application realizes fast closed-loop response of the adaptive filter noise reduction system through fast Fourier transform processing and dynamic adaptive filter parameter updating, improves the real-time performance of the system, and reduces the energy efficiency of the system.

[0022] (3) The present application realizes efficient resource scheduling and low delay operation through real-time closed-loop control strategy, and at the same time adopts parallel processing technology and pipeline structure, which greatly shortens the operation time of the key modules. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 Fig. 1 is a block diagram of an adaptive filter noise reduction system based on variable step size least mean square error algorithm provided by an embodiment of the present application.

[0024] Figure 2 Fig. 3 is a comparison diagram of bit error rate suppression performance of different algorithms provided by an embodiment of the present application.

[0025] Figure 3 Fig. 4 is a comparison diagram of convergence performance of different algorithms provided by an embodiment of the present application. DETAILED DESCRIPTION

[0026] Exemplary embodiments of the present application will now be described in detail with reference to the accompanying drawings. It should be understood that the embodiments shown and described herein are merely exemplary and are intended to illustrate the principles and spirit of the present application, but are not intended to limit the scope of the present application.

[0027] An adaptive filter noise reduction system based on variable step size least mean square error (VSLMS) algorithm is provided by an embodiment of the present application, as shown in Figure 1 Fig. 1, which comprises a signal main link and a closed-loop control module.

[0028] As shown in Figure 1 Fig. 1, the signal main link comprises a low noise amplifier (LNA), an analog-to-digital converter (ADC) and an adaptive filter connected in sequence.

[0029] The low noise amplifier is used for amplifying the input signal to obtain an amplified signal.

[0030] The analog-to-digital converter is used for analog-to-digital conversion of the amplified signal to obtain a digital signal.

[0031] The adaptive filter is used for filtering the digital signal to obtain an output signal.

[0032] As shown in Figure 1 , the closed-loop control module includes a noise detection unit, a spectrum analysis unit and a parameter adjustment unit.

[0033] The input end of the noise detection unit is connected with the output end of the analog-to-digital converter, and is used for extracting a noise component from the digital signal and calculating a root mean square value of the noise component.

[0034] In the embodiment of the application, the formula for the noise detection unit to extract the noise component from the digital signal is: wherein n is an index representing a discrete time sequence, represents the noise component, represents an ideal reference signal, represents an output vector generated by performing inner product of an input signal vector and tap coefficients of the adaptive filter.

[0035] The formula for the noise detection unit to calculate the root mean square value of the noise component is: wherein represents the root mean square value of the noise component, N represents a sampling point number of signal analysis, represents the noise component.

[0036] The input end of the spectrum analysis unit is connected with the output end of the noise detection unit, and is used for analyzing a noise power spectral density by using fast Fourier transform (FFT) according to the root mean square value of the noise component, and adjusting a smoothing parameter and a scaling parameter of a variable step size least mean square error algorithm according to the noise power spectral density.

[0037] In the embodiment of the application, in response to the root mean square value of the noise component , the noise power spectral density is analyzed by using fast Fourier transform, and the specific formula is: wherein denotes the noise power density at frequency k , denotes a single time-domain sample, i.e. the n th sample value (scalar) of the digital signal sequence after analog-to-digital conversion, n is an index denoting the discrete-time sequence, N denotes the number of samples of the signal analysis, j denotes the imaginary unit.

[0038] In the embodiment of the present application, if the variance σ of the noise power spectral density at each frequency point is ≤0.1, because the power values at each frequency point are close to uniform distribution, it is indicated that the noise power spectrum fluctuation is small, the noise is determined to be flat noise (i.e. white noise, the power spectral density at all frequency points is close to a constant value, which shows a flat spectrum curve), and the smoothing parameter σ of the variable step-size least mean square error algorithm is raised from the initial value 0.5 to 0.8 to accelerate convergence; if the variance of the noise power spectral density at each frequency point is >0.1, because there are significant peaks , σ denotes the maximum power value in the noise power spectral density, and denotes the average power value of the noise power spectral density), or multiple peaks and valleys, the noise is determined to be non-flat noise, the non-flat noise power spectral density will have obvious peaks at some frequency points, at this time the variance 2 is large, and the spectrum curve is not flat, the scaling parameter of the variable step-size least mean square error algorithm is reduced from the initial value 0.05 to 0.02 to improve the steady-state accuracy. The updated smoothing parameter and scaling parameter are written into the FPGA register through the SPI interface to ensure that the next clock cycle takes effect.

[0039] The input end of the parameter adjustment unit is connected with the output end of the spectrum analysis unit, and the output end thereof is connected with the adaptive filter, which is used to dynamically adjust the tap coefficient of the adaptive filter by using the variable step-size least mean square error algorithm, so as to form a closed-loop control on the signal main link.

[0040] The variable step-size least mean square error algorithm includes the following steps S1-S3 in the embodiment of the present application:

[0041] In the embodiment of the present application, the maximum value of the filter step parameter is initialized to be μ max =0.1, and the minimum value of the filter step parameter isμ min =0.001, smoothing parameter =0.5, scaling parameter =0.05, initial value of tap coefficient vector of adaptive filter

[0042] S2, according to noise component Calculate filter step parameter : wherein denotes a smoothing parameter, denotes a scaling parameter, denotes a hyperbolic tangent function, and the expression is: The hyperbolic tangent function maps the error amplitude to the interval [0, 1) non-linearly, and after scaling by the scaling parameter , the scaled .

[0043] S3, according to filter step parameter Dynamically adjust the tap coefficient of the adaptive filter, and the adjustment formula is: wherein denotes the filter tap coefficient vector before adjustment, denotes the filter tap coefficient vector after adjustment, denotes an input signal vector. In the embodiment of the present application, the additional limiter ensures , to avoid divergence caused by step out of bounds.

[0044] In the embodiment of the present application, the hardware optimization design uses parallelization and pipeline technology to improve efficiency. The FFT processor is optimized based on the base-2 algorithm, and the operation time of 1024 points is compressed to 5 μs . The coefficient update of the adaptive filter adopts a four-stage pipeline structure, and the single update delay is ≤10 ns. In the FPGA, the priority scheduling threads are divided: the response time of the parameter adjustment thread is ≤50 ns, the spectrum analysis period is 1 ms, and the noise detection is a background task. When the frequency jumps, an interrupt is triggered, and the parameters are forced to reload.

[0045] The performance verification of the VSLMS algorithm provided by the embodiment of the present application can be completed through simulation and actual measurement.

[0046] In the hardware measurement, when the 2.4GHz Wi-Fi signal is interfered by -70dBm narrowband interference, the interference suppression depth reaches 27dB, as shown in Figure 2 , the bit error rate of the VSLMS algorithm provided by the embodiment of the present application is reduced from 10-3 decreased to 1.3 x 10 -7 .

[0047] In MATLAB simulation, when 5G NR signal is superimposed with 0dB Gaussian white noise, the VSLMS algorithm provided by the embodiment of the present application converges within 200 iterations, as shown in Figure 3 the table, compared with the traditional LMS algorithm, the steady-state error is decreased to 10 -4 , and the signal-to-noise ratio is increased by 12dB.

[0048] Those skilled in the art will realize that the embodiments described herein are for the purpose of helping the reader understand the principles of the present application, and should be understood as not limiting the scope of protection of the present application to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations according to the technical inspiration disclosed in the present application without departing from the essence of the present application, and these modifications and combinations are still within the scope of protection of the present application.

Claims

1. An adaptive filtering noise reduction system based on a variable step-size minimum mean square error algorithm, characterized in that: Including signal main link and closed-loop control module; The signal main link includes a low-noise amplifier, an analog-to-digital converter, and an adaptive filter connected in sequence, wherein the low-noise amplifier is used to amplify the input signal to obtain an amplified signal; the analog-to-digital converter is used to perform analog-to-digital conversion on the amplified signal to obtain a digital signal; and the adaptive filter is used to filter the digital signal to obtain an output signal. The closed-loop control module includes a noise detection unit, a spectrum analysis unit and a parameter adjustment unit; The input end of the noise detection unit is connected to the output end of the analog-to-digital converter, and is used to extract the noise component from the digital signal and calculate the root mean square value of the noise component; The input end of the spectrum analysis unit is connected to the output end of the noise detection unit, and is used to analyze the noise power spectrum density using fast Fourier transform according to the root mean square value of the noise component, and adjust the smoothing parameter and scaling parameter of the variable step size minimum mean square error algorithm according to the noise power spectrum density; The input end of the parameter adjustment unit is connected to the output end of the spectrum analysis unit, and its output end is connected to the adaptive filter, which is used to dynamically adjust the tap coefficients of the adaptive filter using a variable step size minimum mean square error algorithm to form a closed-loop control of the signal main link.

2. The adaptive filtering noise reduction system based on the variable step size minimum mean square error algorithm according to claim 1, characterized in that: The noise detection unit extracts the noise component from the digital signal using the following formula: in n is the index representing the discrete time series, represents the noise component, represents the ideal reference signal, Represents the output vector generated by taking the inner product of the input signal vector and the tap coefficients of the adaptive filter.

3. The adaptive filtering noise reduction system based on the variable step size minimum mean square error algorithm according to claim 1, characterized in that: The formula for calculating the root mean square value of the noise component in the noise detection unit is: in represents the RMS value of the noise component, N Indicates the number of sampling points for signal analysis, Represents the noise component.

4. The adaptive filtering noise reduction system based on the variable step size minimum mean square error algorithm according to claim 1, characterized in that: The spectrum analysis unit responds to the root mean square value of the noise component , the noise power spectral density is analyzed using fast Fourier transform.

5. The adaptive filtering noise reduction system based on the variable step size minimum mean square error algorithm according to claim 4, characterized in that: The formula for the analytical noise power spectral density is: in Indicates the frequency k The noise power density at represents a single time domain sampling point, n is the index representing the discrete time series, N Indicates the number of sampling points for signal analysis, j Represents an imaginary unit.

6. The adaptive filtering noise reduction system based on the variable step size minimum mean square error algorithm according to claim 1, characterized in that: The method for adjusting the smoothing parameter and scaling parameter of the variable step size minimum mean square error algorithm according to the noise power spectrum density in the noise detection unit is specifically as follows: If the variance of the noise power spectral density at each frequency point is σ ²≤0.1, the noise is determined to be flat noise, and the smoothing parameter of the variable step size minimum mean square error algorithm is set to Increase from the initial value of 0.5 to 0.8 to speed up convergence; If the variance of the noise power spectral density at each frequency point is σ If ²>0.1, the noise is determined to be non-flat noise, and the scaling parameter of the variable step size minimum mean square error algorithm is set to Reduced from the initial value of 0.05 to 0.02 to improve steady-state accuracy.

7. The adaptive filtering noise reduction system based on the variable step size minimum mean square error algorithm according to claim 1, characterized in that: The variable step size minimum mean square error algorithm in the parameter adjustment unit comprises the following steps: S1, initialize the parameters of the variable step size minimum mean square error algorithm; S2, according to the noise component Calculate the filter step size parameter ; S3, according to the filter step size parameter The tap coefficients of the adaptive filter are dynamically adjusted.

8. The adaptive filtering noise reduction system based on the variable step size minimum mean square error algorithm according to claim 7, characterized in that: The step S1 is specifically as follows: Initialize the maximum value of the filter step size parameter μ max =0.1, the minimum value of the filter step parameter μ min =0.001, smoothing parameter The initial value is 0.5, and the scaling parameter The initial value of is 0.05, and the initial value of the tap coefficient vector of the adaptive filter is 0.

9. The adaptive filtering noise reduction system based on the variable step size minimum mean square error algorithm according to claim 7, characterized in that: The filter step size parameter in step S2 The calculation formula is: in represents the smoothing parameter, represents the scaling parameter, represents the hyperbolic tangent function.

10. The adaptive filtering noise reduction system based on the variable step size minimum mean square error algorithm according to claim 7, characterized in that: The formula for dynamically adjusting the tap coefficients of the adaptive filter in step S3 is: in represents the filter tap coefficient vector before adjustment, represents the adjusted filter tap coefficient vector, Represents the input signal vector.

Images