Method and system for adaptive suppression of wideband noise

Abstract

The application provides a method for wideband noise adaptive suppression, comprising the following steps: S1, collecting an original signal; S2, dividing the original signal into two categories by taking the multiple of the standard deviation of the noise in the original signal as a threshold, and quantizing and transforming the original signal according to the categories to obtain a quantized transformed value; S3, calculating a transformed residual of the quantized transformation according to the quantized transformation, and adaptively filtering the obtained transformed residual value to obtain an optimal estimation value of the quantization error; and S4, integrating the optimal estimation value of the quantization error and the quantized transformed value to obtain a useful signal after noise reduction. The application can be applied to the wideband noise adaptive suppression of various signal forms containing static, steady and transient useful signals, and solves the problem that the existing adaptive filtering cannot effectively filter out the wideband noise and even introduces new noise when the characteristics of the useful signal change.

Description

Technical Field

[0001] The present invention relates to the field of information and communication technologies, and in particular to a method and system for adaptively suppressing broadband noise. Background Art

[0002] The output signal generated by any information acquisition system inevitably contains noise. Noise originates from various sources both within and outside the system. Certain noise sources produce noise with specific characteristics and properties, such as drift noise, vibration noise, and glitch noise. However, the noise generated by all or part of these sources, when combined, exhibits the characteristics of white noise or quasi-white noise: randomness and a wide bandwidth, occupying nearly the entire signal bandwidth or even wider. This broadband noise typically exhibits stationary characteristics; for example, its mean and variance remain constant as long as the system's operating conditions and environmental conditions remain unchanged. Broadband random noise is an inevitable component of the output signal generated by any information acquisition system. Eliminating or suppressing broadband noise can improve the signal-to-noise ratio of the information acquisition output and significantly enhance the quality of the information acquired. This is often essential for subsequent information applications such as measurement and control.

[0003] According to signal estimation theory, to obtain the best estimate of a useful signal from a signal superimposed with noise (hereinafter referred to as the "original signal"), the original signal must be subjected to matched filtering, also known as optimal filtering. Since the useful signal's autocorrelation function and other characteristics are often unknown, adaptive filtering is generally used in practice to implement matched filtering and obtain the best estimate of the useful signal. Adaptive filtering adjusts filter parameters according to a set rule based on an objective function, so that the adaptive filter converges to the matched filter. If the signal changes, causing the objective function parameters to change, the adaptive filter parameters are readjusted to converge to the new matched filter parameters. If the adaptive filter does not converge to the optimal filter during the adjustment period, it will not function as a matched filter, and the parameter adjustments may also introduce new interference into the filtered output. In the information acquisition output, the useful signal is unknown. Even if the noise is stationary, changes in the useful signal often cause changes in the mean and variance of the entire original signal, resulting in corresponding changes in the objective function parameters of the adaptive filter. Therefore, adaptive filtering is only suitable for removing noise from the original signal when the mean and variance of the useful signal and noise are constant. When the useful signal changes, not only will it fail to filter out the noise, but the adjustment process will also introduce new noise.

[0004] A Chinese patent document with publication number CN115842568A discloses a method for communication interference removal, comprising: receiving a signal to be processed, preprocessing the signal to be processed to obtain a digital signal corresponding to the signal to be processed; performing spectrum analysis on the digital signal to determine the signal bandwidth between the maximum and minimum signal frequencies after the spectrum analysis; performing broadband noise interference judgment and in-band noise interference judgment based on the determined signal bandwidth to perform interference removal processing; and outputting the signal after interference removal processing. This patent is not applicable to situations where the spectrum information of the useful signal is unknown, which is a typical situation in fields such as sensing and measurement. In addition, the patent performs interference identification and judgment on the digital signal after spectrum analysis, which is a complex and inefficient process.

[0005] Therefore, a method and system for adaptively suppressing broadband noise is needed. Summary of the Invention

[0006] In view of the defects in the prior art, the present invention aims to provide a method and system for adaptively suppressing broadband noise.

[0007] A method for adaptively suppressing broadband noise provided by the present invention includes:

[0008] Step S1: collecting original signals;

[0009] Step S2: Using multiples of the noise standard deviation in the original signal as a threshold, the original signal is divided into two categories, and the original signal is quantized and transformed according to the category to obtain a quantized transformation value;

[0010] Step S3: Calculating the transformation residual of the quantization transformation according to the quantization transformation, and performing adaptive filtering on the obtained transformation residual value to obtain the optimal estimation value of the quantization error;

[0011] Step S4: Integrate the optimal estimate of the quantization error and the quantization transformation value to obtain a useful signal after denoising.

[0012] Preferably, the step S1 includes:

[0013] The original signal x(n) collected is:

[0014] x(n)=s(n)+w(n)

[0015] s(n) is the useful signal, w(n) is the broadband random noise, where the standard deviation of the noise is σ.

[0016] Preferably, the step S2 includes:

[0017] Step S2.1: Calculate the difference value Δ(n) of the original signal x(n):

[0018] Δ(n) = x(n) - x(n-1)

[0019] Step S2.2: divide the original signal into two categories with the threshold of standard deviation; specifically, divide the original signal into two categories with the threshold of 6 times of the standard deviation 6σ, and compare the absolute value |Δ(n)| of the difference value with the threshold 6σ, when |Δ(n)|≥6σ, the original signal is the first category, and when |Δ(n)|<6σ, the original signal is the second category.

[0020] Step S2.3: quantize the original signal with the resolution R as the quantization scale to obtain the quantized transform value x Q (n).

[0021] Specifically:

[0022]

[0023] wherein, the resolution R is a set value.

[0024] Preferably, the step S3 comprises:

[0025] Step S3.1: calculate the transform residual Q(n) of the quantized transform of the original signal x(n):

[0026] Q(n) = x(n) - x Q (n)

[0027] Step S3.2: the quantization error ε s (n) is the error between the quantized transform of the original signal x(n) and the useful signal s(n), and is

[0028] ε s (n) = s(n) - x Q (n) = Q(n) - w(n);

[0029] Step S3.3: adaptively filter the transform residual Q(n) to obtain the optimal estimation value of the quantization error;

[0030] Specifically, take Q(n) as the reference signal and Q(n-M) as the input signal of the adaptive filter, and obtain the optimal estimation value of ε s (n) through the adaptive filter.

[0031] M is a preset integer parameter, which is a value greater than the autocorrelation length of the noise w(n) and less than the autocorrelation length of the useful signal s(n).

[0032] Preferably, the step S4 comprises:

[0033] According to the transform residual Q(n) and the optimal estimation value of the quantization error ε The denoised useful signal is obtained:

[0034]

[0035] According to the application, a system for wideband noise adaptive suppression is provided, comprising:

[0036] Module M1: collecting original signals;

[0037] Module M2: taking the multiple of the standard deviation of the noise in the original signal as the threshold value, dividing the original signal into two categories, and quantizing and transforming the original signal according to the categories to obtain the quantized transformed value;

[0038] Module M3: calculating the transformed residual of the quantized transformation according to the quantized transformation, and adaptively filtering the obtained transformed residual value to obtain the optimal estimation value of the quantization error;

[0039] Module M4: integrating the optimal estimation value of the quantization error and the quantized transformed value to obtain the denoised useful signal.

[0040] Preferably, the module M1 comprises:

[0041] The collected original signal x(n) is:

[0042] x(n) = s(n) + w(n)

[0043] s(n) is the useful signal, and w(n) is the wideband random noise, wherein the standard deviation of the noise is σ.

[0044] Preferably, the module M2 comprises:

[0045] Module M2.1: calculating the difference value Δ(n) of the original signal x(n):

[0046] Δ(n) = x(n) - x(n-1)

[0047] Module M2.2: taking the multiple of the standard deviation as the threshold value to divide the original signal into two categories; specifically, taking the 6 times of the variance as the threshold value 6σ, comparing the absolute value |Δ(n)| of the difference value with the threshold value 6σ, when |Δ(n)|≥6σ, the original signal is of the first category, and when |Δ(n)|<6σ, the original signal is of the second category;

[0048] Module M2.3: for different original signal categories, different quantizations are performed on the original signal with the resolution R as the quantization scale to obtain the quantized transformed value x Q (n),

[0049] Specifically:

[0050]

[0051] Wherein, the resolution R is a set value.

[0052] Preferably, the module M3 comprises:

[0053] Module M3.1: calculating the transform residual Q(n) of the quantization transform of the original signal x(n):

[0054] Q(n) = x(n) - x Q (n)

[0055] Module M3.2: quantization error ε s (n) is the error between the quantization transform of the original signal x(n) and the useful signal s(n), and is

[0056] ε s (n) = Q(n) - w(n);

[0057] Module M3.3: adaptively filtering the transform residual Q(n) to obtain the optimal estimation value of the quantization error;

[0058] Specifically, taking Q(n) as the reference signal and Q(n-M) as the input signal of the adaptive filter, the optimal estimation value of ε s (n) is obtained through the adaptive filter.

[0059] M is a preset integer parameter, which is a value greater than the autocorrelation length of the noise w(n) and less than the autocorrelation length of the useful signal s(n).

[0060] Preferably, the module M4 comprises:

[0061] According to the transform residual Q(n) and the optimal estimation value of the quantization error the denoised useful signal is obtained:

[0062]

[0063] Compared with the prior art, the present application has the following beneficial effects:

[0064] 1. The present application can significantly suppress wideband random noise and improve the quality of signals by the combination method of signal difference, quantization transform and adaptive filtering, and can adaptively suppress wideband noise in signals containing useful signals of various signal forms such as static, steady and transient, and solve the problem that the existing adaptive filtering cannot effectively filter out wideband noise or even introduce new noise when the characteristics of the useful signal change.

[0065] 2. The present application can also estimate the variance of the noise in the signal and the resolution of the useful signal, and can still suppress the noise in the case of unknown noise or resolution. BRIEF DESCRIPTION OF DRAWINGS

[0066] Other features, objects, and advantages of the application will become more apparent from the following detailed description of non-limiting embodiments thereof, when read in conjunction with the accompanying drawings:

[0067] Figure 1 Flow chart of the method for wideband noise adaptive suppression of the application;

[0068] Figure 2 Signal noise suppression processing flow of the first embodiment of the application;

[0069] Figure 3 Block diagram of the adaptive filter of the second embodiment of the application. DETAILED DESCRIPTION

[0070] The application will be described in detail below with specific embodiments. The following embodiments will help those skilled in the art to further understand the application, but do not limit the application in any form. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the application. These are within the scope of protection of the application.

[0071] The application provides a general wideband noise adaptive suppression method, which is suitable for wideband noise adaptive suppression of various signal forms containing static, steady, and transient useful signals, and solves the problem that existing adaptive filtering cannot effectively filter out wideband noise or even introduces new noise when the characteristics of useful signals change. The specific technical idea and implementation steps of the application are: checking the difference of the signal, quantizing and transforming according to the difference, adaptively filtering the residual of the quantizing and transforming, obtaining the optimal estimation of the useful signal and the quantizing and transforming error, and the sum of the optimal estimation and the quantizing and transforming is the signal with suppressed noise.

[0072] The application provides a wideband noise adaptive suppression method, which comprises:

[0073] Step S1: collecting an original signal;

[0074] Step S2: taking a multiple of the standard deviation of the noise in the original signal as a threshold, dividing the original signal into two categories, and quantizing and transforming the original signal according to the categories to obtain quantizing and transforming values;

[0075] Step S3: calculating the transformation residual of the quantizing and transforming according to the quantizing and transforming, adaptively filtering the obtained transformation residual value, and obtaining the optimal estimation value of the quantizing error;

[0076] Step S4: integrating the optimal estimation value of the quantizing error and the quantizing and transforming value to obtain the useful signal after noise reduction.

[0077] Specifically, the step S1 comprises:

[0078] The original signal x(n) collected is:

[0079] x(n) = s(n) + w(n)

[0080] s(n) is a useful signal, and w(n) is a wideband random noise, wherein the standard deviation of the noise is σ.

[0081] Specifically, the step S2 comprises:

[0082] Step S2.1: Calculate the difference value Δ(n) of the original signal x(n):

[0083] Δ(n) = x(n) - x(n-1)

[0084] Step S2.2: Divide the original signal into two categories with the threshold value being the multiple of the standard deviation; specifically, take 6σ as the threshold value, which is 6 times the variance, so that the absolute value |Δ(n)| of the difference value is compared with the threshold value 6σ, when |Δ(n)|≥6σ, the original signal is the first category, and when |Δ(n)|<6σ, the original signal is the second category;

[0085] Step S2.3: For different original signal categories, the original signal is quantized with the resolution R as the quantization scale to obtain the quantized transform value x Q (n),

[0086] Specifically:

[0087]

[0088] Wherein, the resolution R is a set value.

[0089] Specifically, the step S3 comprises:

[0090] Step S3.1: Calculate the transform residual Q(n) of the quantized transform of the original signal x(n):

[0091] Q(n) = x(n) - x Q (n)

[0092] Step S3.2: The quantization error ε s (n) is the error between the quantized transform of the original signal x(n) and the useful signal s(n), which is

[0093] ε s (n) = s(n) - x Q (n) = Q(n) - w(n);

[0094] Step S3.3: Perform adaptive filtering on the transform residual Q(n) to obtain the optimal estimation value of the quantization error;

[0095] Specifically, taking Q(n) as the reference signal and Q(n-M) as the input signal of the adaptive filter, the optimal estimation value of ε(n) is obtained through the adaptive filter s (n) is obtained

[0096] M is a preset integer parameter, and is a value greater than the autocorrelation length of the noise w(n) and less than the autocorrelation length of the useful signal s(n).

[0097] Specifically, the step S4 comprises:

[0098] According to the transformed residual Q(n) and the optimal estimation value of the quantization error The de-noised useful signal is obtained:

[0099]

[0100] Embodiment one

[0101] Suppose that the original signal obtained by information acquisition is x(n), x(n) = s(n) + w(n), s(n) is a useful signal, and w(n) is a wideband random noise. It is assumed that the mean of the noise is zero, the variance is σ 2 (the standard deviation σ), and the noise is considered to have the characteristics of "white noise": that is, the noise is irrelevant to the useful signal s(n), and because the frequency band of the noise is wide, the autocorrelation length of the noise is shorter than that of the useful signal. The resolution of the information acquisition system is an important parameter for system design and performance description, and is denoted as R, which is of the same dimension as the signal x(n). R is determined by the standard deviation σ of the noise, and R = mσ is usually taken, where 3 ≥ m ≥ 1. Its physical meaning is that the change of the useful signal s(n) should not be less than R to be able to detect the change of s(n) from x(n). Based on this, a wideband noise suppression method based on signal difference adaptive filtering is proposed: 1) the state of the useful signal in x(n) is determined according to the signal difference, that is, change or no change, 2) x(n) is quantized and transformed differently according to the two states, with the resolution R as the quantization level, 3) the residual of the quantized and transformed x(n) is adaptively filtered to obtain the optimal estimation of the useful signal quantized and transformed residual, and 4) the sum of the quantized and transformed x(n) and the optimal estimation of the quantized and transformed residual is the optimal estimation of x(n) without wideband noise. The specific analysis is as follows.

[0102] When |Δ(n)| ≥ 6σ, it indicates that the useful signal has a detectable change, that is, the useful signal has a change greater than the resolution R. In order to eliminate the change of the statistical characteristics of the original signal caused by the change of the useful signal, the current value of the original signal is quantized with the resolution R as the quantization scale to obtain the quantized and transformed value x Q(n) ; when | Δ(n) | < 6σ, it means that there is no detectable change in the useful signal, then quantize the original signal at the previous time with resolution R as the quantization scale to get the quantized transform value x Q (n) ;

[0103] Let the difference of signal x(n) be Δ(n), Δ(n) = x(n) - x(n-1), the quantized transform of x(n) be x Q (n), define

[0104]

[0105] Take the residual of the quantized transform of signal x(n) as

[0106] Q(n) = x(n) - x Q (n) = ε(n) (2)

[0107] Obviously when | Δ(n) | ≥ 6σ, Q(n) is the quantization error ε(n) of x(n) quantized with R as the quantization scale. Using the rounding method to quantize, the quantization noise ε(n) is distributed in with zero mean, since the mean of w(n) is also zero, so the mean of the residual of the quantized transform Q(n) is always zero. That is

[0108]

[0109] When | Δ(n) | < 6σ, the residual of the quantized transform

[0110]

[0111] So when | Δ(n) | < 6σ,

[0112]

[0113] Let w(n) obey Gaussian distribution, then its odd moments are zero, so

[0114] E[w(n-1)ε(n-1)] = σ 2 (6)

[0115] So when | Δ(n) | < 6σ, there is still

[0116]

[0117] So the residual of the quantized transform Q(n) is stationary, its mean and variance do not change with the change of the sensing quantity, that is, the useful signal s(n).

[0118] On the other hand, the residual of the quantized transform

[0119]

[0120] Let ε s (n) = Q(n) - w(n), where

[0121]

[0122] Obviously, ε s (n) is the error between the quantized signal Q(n) and the useful signal s(n). Then

[0123]

[0124] Therefore, if ε s (n) can be estimated from Q(n), the useful signal s(n) can be estimated. According to the signal estimation theory, the optimal estimation of ε s (n) can be obtained by matching filtering Q(n). Therefore, an adaptive filter can be used with Q(n) as the reference signal and Q(n-M) as the input signal of the adaptive filter. When the adaptive filter converges, the output of the filter is the optimal estimation of ε s (n). The parameter M should be selected to be larger than the autocorrelation length of the noise w(n) and smaller than the autocorrelation length of the useful signal s(n). Since the quantization residual Q(n) is stationary, once the adaptive filter converges to a steady state, it will not need to be adjusted to reach a new steady state due to the changes in the useful signal s(n). This fundamentally solves the problem of using adaptive filtering to effectively remove wideband noise in non-stationary useful signals. Because the useful signal is non-stationary, its changes can cause the adaptive filter to adjust, and the output of the filter during adjustment is definitely not the optimal filtering result. Thus, the filter output during adjustment not only cannot remove the noise in the signal, but also can increase new interference. According to the above analysis, no matter how the useful signal changes, it will not cause the mean and variance of the quantization residual Q(n) to change, so the changes in the useful signal will not cause the adaptive filter to adjust.

[0125] According to the above principle analysis, the implementation steps of the present invention are determined: 1) Calculate the difference of the original signal; 2) Perform quantization transformation on the original signal based on the comparison result of the absolute value of the difference and the resolution of the useful signal - a) When the absolute value of the difference is greater than or equal to 6 times the standard deviation of the noise, the quantization transformation of the original signal is the quantization result of the resolution as the quantization scale; b) When the absolute value of the difference is less than 6 times the standard deviation of the noise, the quantization transformation of the original signal is the quantization result of the signal at the previous moment; 3) Perform adaptive filtering on the signal and the residue (difference) of its quantization transformation - the time delay of the residue serves as the input of the adaptive filter, and the residue itself serves as the reference signal; 4) Obtain the adaptive filtering output, which is the optimal estimate of the error between the useful signal and the quantization transformation in the original signal; 5) The sum of the adaptive filter output and the quantization transformation is the optimal estimate of the original signal, that is, the output that suppresses broadband random noise. See Figure 1 .

[0126] From the above analysis, we can also see that the variance of the residual of the quantization transformation of the original signal is determined by the noise variance and the quantization level, and the quantization level is the resolution of the useful signal. In fact, the residual (error) of signal quantization has this characteristic: the variance of the quantization error is determined by the noise variance and the quantization level. Therefore, the present invention can also provide the noise variance and resolution estimation of the original signal. In other words, when the present invention is used to filter the broadband noise of the original signal, it is not necessary to predict the variance and resolution of the noise. The noise variance can be obtained by analyzing the variance of the signal quantization residual. The specific implementation steps are: take the useful signal resolution as the quantization level, quantize the original signal, calculate the variance of the quantization error, and subtract the quantization variance from the variance, that is, This is the variance of the noise. On the other hand, in a practical information acquisition system, the resolution of the useful signal is often determined by the standard deviation of the noise, typically taking 1-3 times the standard deviation of the noise as the resolution of the useful signal. Therefore, by establishing a relationship between resolution and noise standard deviation, the actual resolution can also be calculated from the noise variance.

[0127] Example 2

[0128] Known noise variance σ 2 The block diagram of the adaptive filter is as follows: Figure 3 Where Q(n) is the quantization transformation residual of the original signal x(n),

[0129] Q(n)=x(n)-x Q (n) (15)

[0130] Quantization transformation x Q (n) is defined as

[0131]

[0132] The order of the adaptive filter is L, and the filter vector H(n) is composed of filter coefficients at time n:

[0133] H(n) = [h(0, n), h(l, n),... h(L, n)] T (17)

[0134] The corresponding input signal vector is

[0135] Q(n) = [Q(n), Q(n-1),... Q(n-L)] T (18)

[0136] In this embodiment, an LMS adaptive filter is used. According to the LMS adjustment rule, the adaptive filter update is

[0137] H(n+1) = H(n) + 2μe(n)Q(n-M) (19)

[0138] μ is the iteration step, which can be taken as σ 2 is the noise variance, R is the useful signal resolution, and 0 < k < 1. The order L of the adaptive filter is selected according to the processing speed, signal frequency resolution, and other parameters. The higher the order, the higher the frequency resolution, and the slower the processing speed. Generally, it is selected between 5 and 100. The delay time M is any value greater than the autocorrelation length of the noise w(n) and less than the autocorrelation length of the useful signal s(n). M can also be selected only according to the characteristics of the noise. For example, for white noise, M = 1 can be taken.

[0139] According to the flowchart of Figure 2 , the adaptive filtering of the original signal x(n) is realized, and the output y(n) that suppresses the noise is obtained, which is also the optimal estimation of the useful signal s(n). In the processing flow, the starting point A of the adaptive filtering can be determined artificially, and A ≥ L + M needs to be ensured.

[0140] Example Three

[0141] The variance of the noise in the original signal x(n) to be processed is unknown, and x(n) containing only noise and not containing useful signal needs to be collected first to calculate its mean and variance, and obtain the variance σ 2 Then, the original signal is processed to suppress noise according to Example One.

[0142] Example Four

[0143] The variance of the noise in the original signal x(n) to be processed is unknown, and the resolution R of the useful signal is known. First, x(n) (which can contain useful signal) is quantized, and the quantization scale is the resolution of the useful signal. The variance σ Q 2 of the quantization error is calculated, and

[0144]

[0145] Calculate the noise variance σ 2 If the result is negative or 0, it means that the actual noise level of the original signal has exceeded the noise intensity that satisfies the useful signal resolution R. R is not the true resolution of the useful signal, and the noise variance calculation is invalid. It is necessary to re-analyze the resolution R of the useful signal and calculate the noise variance again using formula (20) until the correct result is obtained. If the result is positive, the correct variance σ is obtained. 2 , according to embodiment one, noise suppression processing is performed on the original signal.

[0146] Example 5

[0147] The variance of the noise and the resolution of the useful signal in the original signal x(n) being processed are unknown. Quantize x(n) (which may contain the useful signal) with the quantization scale being the resolution of the useful signal, and calculate the variance of the quantization error σ Q 2 The relationship between resolution and noise variance (standard deviation) is set as R = mσ, where 3 ≥ m ≥ 1. The noise variance σ is calculated according to formula (20): 2 and the useful signal resolution R. According to the first embodiment, the original signal is subjected to noise suppression processing.

[0148] The present invention also provides a system for broadband noise adaptive suppression, which can be implemented by executing the process steps of the broadband noise adaptive suppression method. That is, those skilled in the art can understand the broadband noise adaptive suppression method as a preferred implementation of the broadband noise adaptive suppression system.

[0149] According to the present invention, a system for adaptively suppressing broadband noise includes: a module M1 for collecting an original signal; a module M2 for dividing the original signal into two categories using a multiple of a noise standard deviation in the original signal as a threshold, and subjecting the original signal to quantization transformation according to the category to obtain a quantization transformation value; a module M3 for calculating a transformation residual of the quantization transformation based on the quantization transformation, and performing adaptive filtering on the obtained transformation residual value to obtain an optimal estimate of the quantization error; and a module M4 for integrating the optimal estimate of the quantization error with the quantization transformation value to obtain a useful signal after denoising.

[0150] Specifically, the module M1 includes: the collected original signal x(n) is:

[0151] x(n)=s(n)+w(n)

[0152] s(n) is the useful signal, w(n) is the broadband random noise, where the standard deviation of the noise is σ.

[0153] Specifically, the module M2 comprises: module M2.1: calculating the difference value Δ(n) of the original signal x(n):

[0154] Δ(n) = x(n) - x(n-1)

[0155] Module M2.2: dividing the original signal into two categories with the threshold value being the multiple of the standard deviation; specifically, taking 6σ as the threshold value, which is 6 times of the variance, and comparing the absolute value |Δ(n)| of the difference value with the threshold value 6σ, when |Δ(n)|≥6σ, the original signal is of the first category, and when |Δ(n)|<6σ, the original signal is of the second category; module M2.3: for different categories of the original signal, quantizing the original signal with the resolution R as the quantization scale to obtain the quantized transformed value x Q (n), specifically:

[0156]

[0157] Wherein, the resolution R is a set value.

[0158] Specifically, the module M3 comprises: module M3.1: calculating the transformed residual Q(n) of the quantized transformed original signal x(n):

[0159] Q(n) = x(n) - x Q (n)

[0160] Module M3.2: quantization error ε s (n) is the error between the quantized transformed original signal x(n) and the useful signal s(n), which is

[0161] ε s (n) = s(n) - x Q (n) = Q(n) - w(n);

[0162] Module M3.3: adaptively filtering the transformed residual Q(n) to obtain the optimal estimation value of the quantization error; specifically, taking Q(n) as the reference signal and Q(n-M) as the input signal of the adaptive filter, the optimal estimation value of ε s (n) is obtained through the adaptive filter. M is a preset integer parameter, which is a value greater than the autocorrelation length of the noise w(n) and less than the autocorrelation length of the useful signal s(n).

[0163] Specifically, the module M4 comprises: obtaining the denoised useful signal according to the transformed residual Q(n) and the optimal estimation value of the quantization error ε

[0164] ​Those skilled in the art will appreciate that, in addition to implementing the system and its various devices, modules, and units provided by the present invention in purely computer-readable program code, it is entirely possible to implement the same functions of the system and its various devices, modules, and units provided by the present invention in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system and its various devices, modules, and units provided by the present invention can be considered a hardware component, and the devices, modules, and units included therein for implementing various functions can also be considered as structures within the hardware component; the devices, modules, and units for implementing various functions can also be considered as both software modules implementing the method and structures within the hardware component.

[0165] The above describes specific embodiments of the present invention. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art may make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. The embodiments of this application and the features in the embodiments may be combined with each other in any manner unless there is a conflict.

Claims

1. A method for adaptively suppressing broadband noise, characterized in that: include: Step S1: collecting original signals; Step S2: Using multiples of the noise standard deviation in the original signal as a threshold, the original signal is divided into two categories, and the original signal is quantized and transformed according to the category to obtain a quantized transformation value; Step S3: Calculating the transformation residual of the quantization transformation according to the quantization transformation, and performing adaptive filtering on the obtained transformation residual value to obtain the optimal estimation value of the quantization error; Step S4: Integrate the optimal estimate of the quantization error and the quantization transformation value to obtain a useful signal after denoising; The step S2 comprises: Step S2.1: Calculate the difference value Δ(n) of the original signal x(n): Δ(n)=x(n)-x(n-1) Step S2.2: Use multiples of the standard deviation as the threshold to classify the original signal into two categories. Specifically, use 6 times the variance as the threshold 6σ, and compare the absolute value of the difference |Δ(n)| with the threshold 6σ. When |Δ(n)| ≥ 6σ, the original signal belongs to the first category, and when |Δ(n)| < 6σ, the original signal belongs to the second category. Step S2.3: For different original signal categories, the original signal is quantized with resolution R as the quantization scale to obtain the quantization transformation value x Q (n), Specifically: Among them, the resolution R is the set value; The step S3 comprises: Step S3.1: Calculate the transformation residual Q(n) of the quantized transformation of the original signal x(n): Q(n)=x(n)-x Q (n) Step S3.2: Quantization error ε s (n) is the error between the quantized transform of the original signal x(n) and the useful signal s(n), which is ε s (n) = s(n) - x Q (n) = Q(n) - w(n); Step S3.3: Adaptively filter the transform residual Q(n) to obtain the optimal estimate of the quantization error; Specifically, Q(n) is used as the reference signal, Q(nM) is used as the input signal of the adaptive filter, and ε is obtained through the adaptive filter. s The best estimate of (n) M is a preset integer parameter, which is a value greater than the autocorrelation length of the noise w(n) and smaller than the autocorrelation length of the useful signal s(n).

2. The method for adaptively suppressing broadband noise according to claim 1, wherein: The step S1 comprises: The original signal x(n) collected is: x(n)=s(n)+w(n) s(n) is the useful signal, w(n) is the broadband random noise, where the standard deviation of the noise is σ.

3. The method for adaptively suppressing broadband noise according to claim 1, wherein: The step S4 comprises: According to the optimal estimate of the transform residual Q(n) and the quantization error Get the useful signal after denoising:

4. A system for adaptive broadband noise suppression, characterized in that: include: Module M1: collects original signals; Module M2: Using the multiple of the noise standard deviation in the original signal as the threshold, the original signal is divided into two categories, and the original signal is quantized according to the category to obtain the quantized transformation value; Module M3: Calculate the transformation residual of the quantization transformation according to the quantization transformation, perform adaptive filtering on the obtained transformation residual value, and obtain the optimal estimate of the quantization error; Module M4: Integrate the optimal estimate of the quantization error and the quantization transformation value to obtain the useful signal after denoising; The module M2 includes: Module M2.1: Calculate the difference value Δ(n) of the original signal x(n): Δ(n)=x(n)-x(n-1) Module M2.2: Use multiples of the standard deviation as the threshold to classify the original signal into two categories. Specifically, use 6 times the variance as the threshold 6σ, and compare the absolute value of the difference |Δ(n)| with the threshold 6σ. When |Δ(n)| ≥ 6σ, the original signal belongs to the first category, and when |Δ(n)| < 6σ, the original signal belongs to the second category. Module M2.3: For different original signal categories, the original signal is quantized differently with resolution R as the quantization scale to obtain the quantization transformation value x Q (n), Specifically: Among them, the resolution R is the set value; The module M3 includes: Module M3.1: Calculate the transformation residual Q(n) of the quantized transformation of the original signal x(n): Q(n)=x(n)-x Q (n) Module M3.2: Quantization Error ε s (n) is the error between the quantized transform of the original signal x(n) and the useful signal s(n), which is ε s (n) = s(n) - x Q (n) = Q(n) - w(n); Module M3.3: Adaptively filter the transform residual Q(n) to obtain the optimal estimate of the quantization error; Specifically, Q(n) is used as the reference signal, Q(nM) is used as the input signal of the adaptive filter, and ε is obtained through the adaptive filter. s The best estimate of (n) M is a preset integer parameter, which is a value greater than the autocorrelation length of the noise w(n) and smaller than the autocorrelation length of the useful signal s(n).

5. The system for adaptive broadband noise suppression according to claim 4, characterized in that: The module M1 includes: The original signal x(n) collected is: x(n)=s(n)+w(n) s(n) is the useful signal, w(n) is the broadband random noise, where the standard deviation of the noise is σ.

6. The broadband noise adaptive suppression system according to claim 4, characterized in that: The module M4 includes: According to the optimal estimate of the transform residual Q(n) and the quantization error Get the useful signal after denoising:

Images