Adaptive filtering method, noise reduction system

Abstract

The application belongs to the technical field of signal processing, and particularly relates to a signal cascade noise reduction method, a noise reduction system and adaptive filtering, wherein the signal cascade noise reduction method comprises the following steps: step S1, pre-noise reduction; and step S2, selecting corresponding adaptive filtering according to a pre-noise reduction result to filter out noise in the signal; wherein the optimization parameters of the adaptive filtering are obtained according to the signal after pre-noise reduction, the anti-power frequency interference performance of related products and systems can be improved, and the adaptive filtering can also be used for noise reduction in a non-power frequency interference scene.

Description

[0001] The application is a divisional application of the original application with the application number 2022105478298 and the filing date of May 18, 2022; the divisional application is entitled Signal Cascade Noise Reduction Method, Noise Reduction System, and Adaptive Filtering. TECHNICAL FIELD

[0002] The application belongs to the technical field of signal processing, and particularly relates to a signal cascade noise reduction method, a noise reduction system, and adaptive filtering. BACKGROUND

[0003] Electroencephalogram (EEG) is a typical physiological weak signal. When collecting EEG data, the device is extremely susceptible to other electronic device power frequency signals and environmental noise, resulting in a decrease in the signal-to-noise ratio of the data, and in severe cases, the data may be completely invalid. Therefore, under the premise of fixed hardware conditions and environment, the advantages and disadvantages of the noise reduction method become a key factor in determining the data quality of the EEG device.

[0004] Therefore, based on the above technical problems, a new signal cascade noise reduction method, noise reduction system, and adaptive filtering need to be designed. SUMMARY

[0005] The application provides a signal cascade noise reduction method, a noise reduction system, and adaptive filtering.

[0006] In a first aspect, the application provides a signal cascade noise reduction method, comprising the following steps:

[0007] Step S1, pre-noise reduction;

[0008] Step S2, selecting a corresponding adaptive filter according to the pre-noise reduction result to filter out the noise in the signal; wherein

[0009] The optimization parameters of the adaptive filter are obtained according to the pre-noise reduction signal.

[0010] Further, the pre-noise reduction includes at least one of removing power frequency interference and removing wideband noise; wherein

[0011] The removal of the power frequency interference includes: pre-filtering fixed strong noise by using a narrow stopband frequency domain filtering method; and

[0012] The removal of the wideband noise includes: filtering high-frequency noise by using a low-pass filter;

[0013] The pre-filtering of the fixed strong noise by using the narrow stopband frequency domain filtering method includes:

[0014] A digital notch filter is set, i.e., the system function is set to zero at the frequency ;

[0015] leaving frequency the response quickly recovers to a constant, i.e. a pole, after the step;

[0016] system function is ;

[0017] wherein z represents an independent variable of the system function after Z-transformation, and j represents an imaginary unit; the digital notch filter is a 2nd order IIR filter, and coefficients of the digital notch filter are ,

[0018] , is a notch frequency, is a normalized stopband width of -3dB.

[0019] Further, the adaptive filter includes at least one of Wiener filter, RLS adaptive filter, LMS adaptive filter, and Kalman filter.

[0020] The pre-noise reduction result is whether the signal after pre-noise reduction contains stationary noise and non-stationary noise.

[0021] The adaptive filter is selected according to the pre-noise reduction result.

[0022] When the signal after pre-noise reduction contains stationary noise, the adaptive filter selects at least one of Wiener filter and RLS adaptive filter.

[0023] When the signal after pre-noise reduction contains non-stationary noise, the adaptive filter selects at least one of LMS adaptive filter and Kalman filter.

[0024] When the signal after pre-noise reduction contains both stationary noise and non-stationary noise, the adaptive filter selects at least one of LMS adaptive filter and Kalman filter.

[0025] Further, the method for determining whether the signal after pre-noise reduction contains stationary noise and non-stationary noise includes: calculating statistical characteristics of the noise; if the statistical characteristics change with time, the noise is non-stationary, otherwise, the noise is stationary.

[0026] Further, the optimization parameter of the Wiener filter includes an order.

[0027] The optimization parameter of the RLS adaptive filter includes a step and a forgetting factor.

[0028] The optimization parameter of the LMS adaptive filter includes a step.

[0029] The optimization parameter of the Kalman filter includes a covariance initial value.

[0030] When the adaptive filter selects RLS adaptive filter or LMS adaptive filter, the step is optimized according to the proportion of the sine wave in the pre-noise-reduced signal, that is

[0031] The size of the step is positively correlated with the proportion value of the sine wave;

[0032] The optimization of the step according to the proportion of the sine wave in the pre-noise-reduced signal comprises:

[0033] The range of the zero-crossing number per second of the offline signal is divided into several subintervals;

[0034] Each subinterval is optimized to obtain a corresponding set of steps, forming a step table;

[0035] The zero-crossing number per second obtained from the real-time signal is compared with the step table, and the step of the corresponding subinterval is selected.

[0036] Further, the LMS adaptive filter comprises:

[0037] The sine component in the observation signal is deleted, that is, according to and , the estimation of the sine component in the observation signal is completed, and it is subtracted from the observation signal, wherein and are the weighting coefficients to be estimated;

[0038] The signal to be filtered is determined, that is, according to the prior information , wherein f is the filtering frequency corresponding to the interference signal to be suppressed;

[0039] The filtered signal is determined, that is, set is the known measured signal, and let the sine wave approximate , and the approximation error is the filtered signal;

[0040] Iterative approximation of the weighting vector, that is, first set the initial value of the weighting vector, that is , and then iteratively approximate the weighting vector point by point towards the direction in which the cost function has the maximum random gradient with a fixed step : , wherein represents the step, , ;

[0041] The LMS adaptive filter is completed to filter out the noise in the signal, that is, to take the point-by-point approximation error as the filtering output.

[0042] Further, the RLS adaptive filter comprises:​

[0043] Remove the sinusoidal component from the observed signal, i.e., according to Sure and This involves estimating the sinusoidal component in the observed signal and subtracting it from the observed signal. and These are the weighting coefficients to be estimated;

[0044] Determine the signal to be filtered, i.e., based on prior information. , where f is the filtering frequency corresponding to suppressing interference signals;

[0045] Determine the filtered signal, i.e., set Given a known measured signal, let the sine wave... Approaching The approximation error is the filtered signal;

[0046] Iterative approximation of the weighted vector involves first setting initial values ​​for the weighted vector. Then, with fixed steps For weighted vectors By iteratively approximating the cost function point-by-point in the direction of maximum global gradient, a sinusoidal interference signal is obtained. In the formula Indicates step, and These represent the covariance matrix of the observed signal and the cross-correlation vector between the observed signal and the target signal, respectively.

[0047] The RLS adaptive filtering is used to specifically remove noise from the signal, that is, to remove the sinusoidal interference signal from the known measured signal and use it as the filtered output.

[0048] Secondly, the present invention also provides a signal denoising system for performing the above-described cascaded signal denoising method, comprising:

[0049] Processor module;

[0050] The pre-noise reduction module performs pre-noise reduction on the signal; and

[0051] The adaptive filtering module filters out noise from the pre-denoised signal.

[0052] The processor module controls the adaptive filtering module to perform corresponding adaptive filtering based on the signal pre-denoising results and obtains the optimized parameters of the adaptive filtering.

[0053] Thirdly, the present invention also provides an LMS adaptive filter for signal denoising, comprising:

[0054] Remove the sinusoidal component from the observed signal, i.e., according to determining and , the estimation of the sinusoidal component in the observation signal is completed and is subtracted from the observation signal, wherein and are the weighting coefficients to be estimated;

[0055] the signal to be filtered is determined, that is, according to prior information , wherein f is the filtering frequency corresponding to the interference signal to be suppressed;

[0056] the filtered signal is determined, that is, set is the known measured signal, and let the sinusoidal wave approximate , and the approximation error is the filtered signal;

[0057] iterative approximation of the weighting vector, that is, first set the initial value of the weighting vector, that is, , and then iteratively approximate the weighting vector point by point towards the direction in which the cost function has the maximum random gradient with a fixed step : , wherein represents the step, , ;

[0058] the LMS adaptive filtering is completed to filter out the noise in the signal, that is, the point-by-point approximation error is taken as the filtering output.

[0059] In a fourth aspect, the present application also provides an RLS adaptive filtering for signal noise reduction, comprising:

[0060] the sinusoidal component in the observation signal is deleted, that is, according to determining and , the estimation of the sinusoidal component in the observation signal is completed and is subtracted from the observation signal, wherein and are the weighting coefficients to be estimated;

[0061] the signal to be filtered is determined, that is, according to prior information , wherein f is the filtering frequency corresponding to the interference signal to be suppressed;

[0062] the filtered signal is determined, that is, set is the known measured signal, and let the sinusoidal wave approximate , and the approximation error is the filtered signal;

[0063] iterative approximation of the weighting vector, that is, first set the initial value of the weighting vector, that is, fixed step to the weighting vector Point-by-point iterative approximation is carried out towards the direction in which the cost function is maximum in the overall gradient, and the sinusoidal interference signal is obtained: , wherein represents a step, and respectively represent an observation signal covariance matrix and a cross-correlation vector of the observation signal and the target signal;

[0064] The RLS adaptive filtering is completed to filter out the noise in the signal, that is, the sinusoidal interference signal is removed from the known measured signal as a filtering output.

[0065] The beneficial effects of the present application are that, through step S1, pre-noise reduction; step S2, selecting a corresponding adaptive filter according to the pre-noise reduction result to filter out the noise in the signal; wherein the optimization parameters of the adaptive filter are obtained according to the signal after pre-noise reduction, which can improve the anti-power frequency interference performance of related products and systems, and can also be used for denoising in non-power frequency interference scenarios.

[0066] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the present application. The objects and other advantages of the present application will be realized and achieved by the structure particularly pointed out in the specification, claims and drawings.

[0067] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the following preferred embodiments are described in detail below, and the accompanying drawings are described as follows. BRIEF DESCRIPTION OF DRAWINGS

[0068] In order to more clearly illustrate the specific embodiments of the present application or the technical solutions in the prior art, the following will briefly introduce the drawings needed to be used in the specific embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0069] Figure 1 is a notch bandwidth selection schematic diagram in the prior art;

[0070] Figure 2 is a general applicability schematic diagram in the prior art;

[0071] Figure 3 is a flowchart of the cascaded noise reduction method of the signal in the present application;

[0072] Figure 4 is a schematic diagram of step-by-step filtering of power frequency interference in the present application;

[0073] Figure 5 is a schematic diagram of data polluted by power frequency in the present application;

[0074] Figure 6 is a schematic diagram of processing results of comparative scheme 1 in the present application;

[0075] Figure 7 is a schematic diagram of processing results of comparative scheme 2 in the present application;

[0076] Figure 8 is a schematic diagram of processing results of scheme in the present application;

[0077] Figure 9 is a schematic diagram of fixed filtering coefficient switching channel in the present application. DETAILED DESCRIPTION

[0078] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme of the present application will be described clearly and completely below with reference to the drawings. Obviously, the described embodiments are some embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the present application.

[0079] As shown in Figure 1 and Figure 2 , in the prior method, frequency domain filtering / notch is the most commonly used noise reduction method at present. However, it has the problem of frequency band overlap: when the frequency bands of noise and electroencephalogram overlap, useful signals will be inevitably lost while reducing noise. The following Figure 1 Taking power frequency interference as an example, there are two difficulties in designing a notch scheme: 1) notch bandwidth selection: too wide to lose useful signals; too narrow to have poor notch performance; 2) poor universality: the notch coefficient is preset and fixed, and does not adaptively adjust with the subject, channel and environment. The 50Hz notch width is 6Hz, and the 100Hz notch width is 0.4Hz: the same notch coefficient, channel 7 has good effect, and channel 1 has poor effect. In order to solve the above-mentioned problems of frequency band overlap and poor universality, time domain adaptive methods represented by Wiener filtering are used for electroencephalogram noise reduction. However, such methods require a high input signal-to-noise ratio (in the original data before noise reduction, the noise energy is significantly smaller than the target signal), which is difficult to meet in actual application.

[0080] Embodiment 1

[0081] As shown in Figure 3As shown, the embodiment 1 provides a cascaded noise reduction method of a signal, comprising the following steps: step S1, pre-noise reduction; step S2, selecting a corresponding adaptive filter according to the pre-noise reduction result to filter out the noise in the signal; wherein the optimization parameters of the adaptive filter are obtained according to the signal after pre-noise reduction; first, a fixed and narrow stop-band frequency domain filter is used to pre-filter out the fixed strong noise in the environment (such as 50Hz / 60Hz and its multiple frequency interference), and then a time domain adaptive filter is used to more specifically filter out the random noise in the environment. The narrow stop-band filter in the first step does not lose useful signals and improves the signal-to-noise ratio of the data to a certain extent, creating a favorable premise for the performance of the subsequent time domain adaptive filter. It can improve the anti-power frequency interference performance of related products and systems, and can also be used for noise reduction in non-power frequency interference scenes.

[0082] In the embodiment, after pre-noise reduction, a corresponding adaptive filter can be selected according to the pre-noise reduction result to filter out the noise in the signal, and the step of the LMS filter is optimized according to the zero-crossing times per second after pre-noise reduction and the time domain adaptive noise reduction is performed, which can improve the effect and precision of noise reduction.

[0083] In the embodiment, the pre-noise reduction includes at least one of removing power frequency interference and removing wideband noise; wherein the removing power frequency interference includes pre-filtering out fixed strong noise by using a narrow stop-band frequency domain filter; and the removing wideband noise includes filtering out high-frequency noise by using a low-pass filter; the pre-filtering out fixed strong noise by using a narrow stop-band frequency domain filter includes setting a digital notch filter, i.e. making the system function zero at the frequency , and the response quickly returns to a constant, i.e. a pole, after the frequency .

[0084] The system function is .

[0085] In the formula, z represents the independent variable of the system function after Z-transform, and j represents the imaginary unit; according to the system function, the digital notch filter is a 2-order IIR filter, and the coefficients are ,

[0086] , is the notch frequency, and the normalized stop-band width of -3dB is

[0087] The pre-noise-reduction signal should continue to select which adaptive filter, thereby reducing the difficulty of algorithm implementation, and there is no corresponding theoretical basis. In the embodiment, the following adaptive filters are listed based on specific practice. The adaptive filter includes at least one of Wiener filter, RLS adaptive filter, LMS adaptive filter and Kalman filter; the pre-noise-reduction result is whether the pre-noise-reduction signal contains stationary noise and non-stationary noise; and the adaptive filter is selected according to the pre-noise-reduction result, including: when the pre-noise-reduction signal contains stationary noise, the adaptive filter selects at least one of the Wiener filter and the RLS adaptive filter; when the pre-noise-reduction signal contains non-stationary noise, the adaptive filter selects at least one of the LMS adaptive filter and the Kalman filter; and when the pre-noise-reduction signal contains both stationary noise and non-stationary noise, the adaptive filter selects at least one of the LMS adaptive filter and the Kalman filter.

[0088] In the embodiment, the non-stationary noise signal contained in the pre-noise-reduction signal is used as an index to select an adaptive filter, and the parameters of the adaptive filter are optimized to reduce the calculation difficulty and facilitate implementation. The method for determining whether the pre-noise-reduction signal contains stationary noise and non-stationary noise includes: calculating the statistical characteristics of the noise; if the statistical characteristics change over time, the noise is non-stationary, otherwise, the noise is stationary. The reason for selecting the adaptive filter method according to the stationarity and non-stationarity of the data is that different adaptive filter methods use different past data to obtain the current point processing result. For example, the LMS method only needs to use 1 point of past data to obtain the current point processing result, while the RLS method needs to use several points of past data. When the data is non-stationary, the statistical characteristics of the past multiple points of data are unstable, which leads to a serious decline in the performance of the RLS, and the RLS is not applicable. Therefore, the former can consider both stationary and non-stationary signals, and the latter can only process stationary signals. It can be seen that selecting the adaptive filter method according to the stationarity of the pre-noise-reduction data is of great significance to the filtering effect. At the same time, compared with the Wiener filter, the RLS adaptive filter, which processes stationary noise, does not require prior knowledge of the statistical quantities of the target signal or the noise, such as the cross-correlation vector of the target signal and the actual observation signal, the autocorrelation vector of the noise, etc. These statistical quantities are difficult to obtain directly, and the difficulty of indirect estimation is also greater. Therefore, the realizability is an advantage of the RLS adaptive filter method over the Wiener filter. Compared with the Kalman filter, the LMS adaptive filter, which processes non-stationary noise, only needs to calculate the error between the estimated value and the actual observation value point by point, and combine the step parameter that can be optimized to complete the iterative update of the filter coefficient (weight vector

[0089] ​In the embodiment, the optimized parameter of LMS filtering is step, because step is the key parameter determining the convergence speed and precision of LMS method. Large step has fast convergence speed but low precision, and small step has high precision but slow convergence speed. Therefore, it is necessary to select step matching different output data. The step optimization strategy for power frequency interference is zero-crossing number per second, which can measure the proportion of strong sinusoidal wave in data. After pre-noise reduction, the value is still large, and large step can be selected to suppress power frequency interference more quickly; the value is small, and small step can be selected to filter out power frequency interference more accurately. The specific embodiment of preferred step is to divide the value range of zero-crossing number per second into several subintervals, and each subinterval obtains a group of steps based on offline data optimization. When running the noise reduction algorithm, the step corresponding to the subinterval is obtained according to the real-time calculated zero-crossing number per second. Of course, the corresponding step can also be obtained through the empirical value of zero-crossing number per second.

[0090] In the embodiment, the optimized parameter of Wiener filter includes order; the optimized parameter of RLS adaptive filter includes step and forgetting factor; the optimized parameter of LMS adaptive filter includes step; the optimized parameter of Kalman filter includes covariance initial value; when the adaptive filter selects RLS adaptive filter or LMS adaptive filter, the step is optimized according to the proportion of sinusoidal wave in pre-noise reduced signal, that is, the size of the step is positively correlated with the proportion value of the sinusoidal wave; the step optimization according to the proportion of sinusoidal wave in pre-noise reduced signal includes: dividing the value range of zero-crossing number per second of offline signal into several subintervals; obtaining a corresponding group of steps for each subinterval to form a step table; comparing the zero-crossing number per second obtained from real-time signal with the step table to select the step of the corresponding subinterval.

[0091] In the embodiment, the Wiener filter and its LMS implementation are as follows: and represent the target signal and the additive background noise respectively, and the actual observation signal is:

[0092] The basic idea of Wiener filter is to obtain N-dimensional vector , which is close to the convolution of the observation signal at the previous time, that is, , and the approximation criterion is minimum mean square error (MMSE), that is:

[0093] ;

[0094] The above objective function (cost function) is a concave function about , and the global minimum value of corresponding to the global minimum value of the objective function (cost function) can be obtained by the partial derivative method: ;​ and respectively represent the observation signal covariance matrix and the cross-correlation vector between the observation signal and the target signal. Further, when the noise and the target signal are independent of each other, ; is the noise autocorrelation vector; and are difficult to obtain, and the above solution is only optimal for stationary signals. In actual filtering, the RLS and LMS methods are usually used.

[0095] In this embodiment, the RLS implementation of the Wiener filter: with a certain fixed step 𝜇 towards the direction of the maximum overall gradient of the cost function, the weighted vector is iteratively approximated point by point:

[0096] ;

[0097] In theory, when the number of iterations tends to infinity and , the RLS method is the optimal approximation of the Wiener filter, where is the largest eigenvalue of the observation signal covariance matrix ;

[0098] The LMS implementation of the Wiener filter: with a certain fixed step 𝜇 towards the direction of the maximum random gradient of the cost function, the weighted vector is iteratively approximated point by point:

[0099]

[0100] The LMS filter is a random gradient descent method, which is simple to calculate and has strong feasibility.

[0101] In this embodiment, the LMS adaptive filter includes: deleting the sinusoidal component in the observation signal, i.e., according to to and complete the estimation of the sinusoidal component in the observation signal and subtract it from the observation signal, where and are the weighted coefficients to be estimated;

[0102] Determine the signal to be filtered, i.e., according to the prior information , where f is the filter frequency corresponding to the interference signal;

[0103] Determine the filtered signal, i.e., set For the known measured signal, the method uses reverse thinking, taking the known y(k) as the "virtual" observation signal to be filtered. The y(k) is weighted and filtered by w to approximate the sine wave in the real observation signal x(k), and then the approximated sine wave is subtracted from the observation signal x(k). At this time, the approximation error is the filtered signal, and the sine wave approximation , and the approximation error is the filtered signal;

[0104] The iterative approximation of the weighting vector is that the initial value of the weighting vector is set, that is, , and the fixed step is used to iteratively approximate the weighting vector point by point in the direction of the maximum random gradient of the cost function: , wherein represents the step, , ;

[0105] The LMS adaptive filter is completed to filter out the noise in the signal, that is, the point-by-point approximation error is taken as the filter output.

[0106] For the LMS filter of 50Hz / 100Hz power frequency interference, , the and are determined, that is, the estimation of the sine component in the observation signal is completed, and the sine component is subtracted from the observation signal. The signal to be filtered: according to the prior information , Hz; is the known actual observation signal, and the sine wave , and the approximation error is the filtered electroencephalogram signal; the first step: the initial value of the weighting vector: ; the step optimization scheme: based on the zero-crossing number per second after pre-noise reduction; the second step: iterative approximation of the weighting vector:

[0107] ,

[0108] , ;

[0109] The third step: completing noise reduction, taking the point-by-point approximation error as the filter output.

[0110] In this embodiment, the RLS adaptive filter includes: removing the sine component in the observation signal, that is, determining , the and , the estimation of the sinusoidal component in the observation signal is completed and subtracted from the observation signal, wherein and are the weighting coefficients to be estimated;

[0111] The signal to be filtered is determined, i.e. according to prior information , wherein f is the filtering frequency corresponding to the interference signal;

[0112] The filtered signal is determined, i.e. set is the known measured signal, and let the sinusoidal wave approximate , and the approximation error is the filtered signal;

[0113] The iterative approximation of the weighting vector is, i.e. first set the initial value of the weighting vector, i.e. , and then iteratively approximate the weighting vector point by point towards the direction of the maximum overall gradient of the cost function to obtain the sinusoidal interference signal: , wherein represents the step, and represent the observation signal covariance matrix and the cross-correlation vector of the observation signal and the target signal, respectively; The RLS adaptive filtering is completed to filter out the noise in the signal, i.e. the sinusoidal interference signal is removed from the known measured signal as the filtering output.

[0114]

[0115] ​In this embodiment, a specific embodiment for power frequency interference. Among them, 1) using LMS method to fit the power frequency interference in the data point by point, and then subtracting the noise reduction scheme, and 2) LMS step optimization strategy based on zero crossing number / second, are designed around the power frequency interference. When facing electromyography, electrooculography and other noises, noise reduction processing can also be carried out through "time-frequency domain cascade structure + second level coefficient optimization": 1) electromyography and electrooculography are wideband noise, and the first level of pre-noise reduction needs to design other filtering methods to replace the notch. Taking electromyography as an example, its typical frequency range is 10-500Hz, and the high frequency component is more than that of electroencephalogram. Therefore, the first level of pre-noise reduction can use low-pass filter to filter out high frequency electromyography, and the remaining low frequency electromyography overlaps with electroencephalogram in frequency band, and the second level is suppressed by time domain adaptive algorithm. If the time domain adaptive algorithm has to be set, the high frequency-low frequency energy ratio can be used to optimize the coefficient, which has the same effect as optimizing the step of LMS using zero crossing number / second. 2) According to whether the noise is stationary, the second level of time domain adaptive algorithm can be divided into two categories. One is more suitable for stationary noise, represented by Wiener filter and RLS; the other can deal with both stationary and non-stationary noise, represented by Kalman filter and LMS. The cascade noise reduction method of the signal in this embodiment can be combined with the characteristics of different types of noise to optimize the time domain adaptive algorithm in actual application. 3) Parameters to be optimized for different time domain adaptive filters: Wiener filter (order), recursive least square RLS (step, forgetting factor), least mean square LMS (step), Kalman filter (covariance initial value).

[0116] In this embodiment, as shown in Figure 4 , the cascade noise reduction method of the signal is used to realize the step-by-step filtering of power frequency interference; as shown in Figures 5-8 , comparison scheme 1: only frequency domain notch (50Hz stop band width=6Hz, 100Hz stop band width=0.4Hz); comparison scheme 2: only time domain LMS filter (50Hz step=0.08, 100Hz step=0.003); the signal cascade noise reduction method scheme used in this embodiment: pre-noise reduction notch (50Hz stop band width=1.4Hz, 100Hz stop band width=0.2Hz); time domain LMS filter (optimized according to zero crossing number / second: 50Hz step=0.009, 100Hz step=0.0015); the signal cascade noise reduction method scheme used in this embodiment can suppress 50Hz and 100Hz power frequency interference and does not lose high frequency information, and the noise reduction performance of the comparison scheme is poor and easy to lose high frequency information; as shown in Figure 9 , fixed filter coefficient, switch different channels: the scheme in this embodiment maintains good noise reduction performance, and the noise reduction performance of the comparison scheme is seriously reduced.

[0117] Embodiment 2

[0118] Based on the embodiment 1, the embodiment 2 further provides a signal denoising system for implementing the cascade denoising method of the signal, comprising:

[0119] a processor module;

[0120] a pre-denoising module for pre-denoising the signal; and

[0121] an adaptive filtering module for filtering the noise in the pre-denoised signal;

[0122] the processor module controls the adaptive filtering module to perform corresponding adaptive filtering and acquires the optimized parameters of the adaptive filtering according to the pre-denoising result of the signal.

[0123] Embodiment 3

[0124] Based on the embodiment 1, the embodiment 2 further provides an LMS adaptive filtering for implementing the signal denoising in the embodiment 1, comprising:

[0125] deleting the sinusoidal component in the observation signal, i.e. according to determining and , completing the estimation of the sinusoidal component in the observation signal and subtracting it from the observation signal, wherein and are the weighting coefficients to be estimated;

[0126] determining the signal to be filtered, i.e. according to the prior information , wherein f is the filtering frequency corresponding to the interference signal;

[0127] determining the filtered signal, i.e. setting as the known measured signal, and letting the sinusoidal wave approximate , and the approximation error is the filtered signal;

[0128] iterative approximation of the weighting vector, i.e. setting the initial value of the weighting vector, i.e. , and then performing point-by-point iterative approximation of the weighting vector toward the direction where the cost function has the maximum random gradient with a fixed step : , wherein represents the step, , ;

[0129] completing the LMS adaptive filtering to filter the noise in the signal, i.e. taking the point-by-point approximation error as the filtering output.

[0130] Embodiment 4

[0131] On the basis of embodiment 1, the embodiment 4 further provides an RLS adaptive filter for signal noise reduction in embodiment 1, comprising:

[0132] The sinusoidal component in the observation signal is deleted, that is, according to The determination And The estimation of the sinusoidal component in the observation signal is completed, and is subtracted from the observation signal, wherein And The weighted coefficient to be estimated is determined;

[0133] The signal to be filtered is determined, that is, according to the prior information Wherein f is the filtering frequency corresponding to the interference signal;

[0134] The filtered signal is determined, that is, set As the known measured signal, let the sinusoidal wave Approximate The approximation error is the filtered signal;

[0135] The iterative approximation of the weighted vector is that the initial value of the weighted vector is set, that is The weighted vector Is iteratively approximated point by point in the direction of the maximum overall gradient of the cost function with a fixed step To obtain the sinusoidal interference signal: , wherein Indicates the step, And Indicate the observation signal covariance matrix and the cross-correlation vector of the observation signal and the target signal respectively;

[0136] The RLS adaptive filter is completed, and the noise in the signal is filtered, that is, the sinusoidal interference signal is removed from the known measured signal as the filtering output.

[0137] In summary, the present application reduces the noise by step S1, and selects the corresponding adaptive filter according to the pre-noise reduction result to filter the noise in the signal in step S2, wherein the optimization parameters of the adaptive filter are obtained according to the signal after pre-noise reduction, which can improve the anti-power frequency interference performance of the related product and system, and can also be used for denoising in non-power frequency interference scene.

[0138] In several embodiments provided in the present application, it should be understood that the disclosed apparatus and method can also be implemented by other manners. The apparatus embodiments described above are merely illustrative, for example, the flowcharts and block diagrams in the drawings show the possible implementation architecture, function and operation of the apparatus, method and computer program product according to the embodiments of the present application. In this regard, each block in the flowchart or block diagram can represent a module, a program segment or a part of code, which contains one or more executable instructions for implementing the specified logic function. It should also be noted that in some alternative implementation manners, the functions noted in the blocks can also occur in different order from that noted in the drawings. For example, two consecutive blocks can actually be executed substantially in parallel, and they can also be executed in reverse order in some cases, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and the combination of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system for executing the specified function or action, or can be implemented by a combination of dedicated hardware and computer instructions.

[0139] In addition, each functional module in the embodiments of the present application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0140] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application or the part of the present application that essentially contributes to the prior art or the part of the technical solutions can be embodied in the form of a software product, which is stored in a storage medium and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in the embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0141] Based on the above ideal embodiments according to the present application, through the above description, relevant personnel can make various changes and modifications without deviating from the scope of the technical idea of the present application. The technical scope of the present application is not limited to the contents in the specification, and must be determined by the scope of the claims.

Claims

1. An adaptive filtering method applied to electroencephalogram signals, characterized in that, a corresponding adaptive filter is selected according to a pre-noise reduction result to filter out noise in the signals; wherein the optimization parameters of the adaptive filter are obtained according to the signals after pre-noise reduction; the adaptive filter includes at least one of Wiener filter, RLS adaptive filter, LMS adaptive filter and Kalman filter; the optimization parameters of the Wiener filter include order; the optimization parameters of the RLS adaptive filter include step and forgetting factor; the optimization parameters of the LMS adaptive filter include step; the optimization parameters of the Kalman filter include covariance initial value; when the adaptive filter selects RLS adaptive filter or LMS adaptive filter, the step is optimized according to the proportion of sinusoidal waves in the signals after pre-noise reduction, that is, the size of the step is positively correlated with the proportion value of the sinusoidal waves; optimizing the step according to the proportion of the sinusoidal waves in the signals after pre-noise reduction includes: dividing the value range of the zero-crossing number per second of the offline signals into several subintervals; optimizing each subinterval to obtain a corresponding set of steps to form a step table; comparing the zero-crossing number per second obtained from the real-time signals with the step table to select the step corresponding to the subinterval.

2. The adaptive filtering method according to claim 1, characterized in that, the pre-noise reduction result is whether the signals after pre-noise reduction contain stationary noise and non-stationary noise; selecting a corresponding adaptive filter according to the pre-noise reduction result includes: when the signals after pre-noise reduction contain stationary noise, the adaptive filter selects at least one of Wiener filter and RLS adaptive filter; when the signals after pre-noise reduction contain non-stationary noise, the adaptive filter selects at least one of LMS adaptive filter and Kalman filter; when the signals after pre-noise reduction contain both stationary noise and non-stationary noise, the adaptive filter selects at least one of LMS adaptive filter and Kalman filter.

3. The adaptive filtering method according to claim 2, characterized in that, the method for determining whether the signals after pre-noise reduction contain stationary noise and non-stationary noise includes: calculating the statistical characteristics of the noise; if the statistical characteristics change over time, it is non-stationary noise, otherwise it is stationary noise.

4. The adaptive filtering method according to claim 1, characterized in that, the LMS adaptive filter includes: The sinusoidal component in the observation signal is removed, i.e. is determined from The estimation of the sinusoidal component in the observation signal is completed and subtracted from the observation signal, wherein are the weighting coefficients to be estimated;​​​ determining the signal to be filtered, i.e. according to prior information where f is the filtering frequency corresponding to the interference signal to be suppressed The filtered signal is determined, i.e. set as the known measured signal, and the sine wave approximation is approximated, with an approximation error being the filtered signal; The iterative approximation of the weighting vector, i.e. setting the initial value of the weighting vector, i.e. , and then performing the fixed step to the weighting vector point-by-point in the direction of the maximum of the cost function in the random gradient: , wherein denotes the step, , ; The LMS adaptive filtering is done to filter out the noise in the signal, i.e. to approximate the error point by point as the filtered output.

5. The adaptive filtering method according to claim 1, characterized in that, the RLS adaptive filter includes: The sinusoidal component in the observation signal is deleted, i.e. is determined from The estimation of the sinusoidal component in the observation signal is completed and is subtracted from the observation signal, wherein are the weighting coefficients to be estimated;​​​ determining the signal to be filtered, i.e. according to prior information where f is the filtering frequency corresponding to the interference signal to be suppressed The filtered signal is determined, i.e. set as the known measured signal, and the sine wave approximation is approximated, with an approximation error being the filtered signal; The iterative approximation of the weight vector, i.e. setting the initial value of the weight vector, is , and then performing the iteration approximation in a fixed step to the weight vector in the direction of the maximum overall gradient of the cost function, to obtain the sinusoidal interference signal: , wherein represents the step, and respectively represent the observation signal covariance matrix and the cross-correlation vector of the observation signal and the target signal; RLS adaptive filtering is completed to filter out noise in the signals, that is, the sinusoidal interference signal is removed from the measured signals to obtain the filtering output.

6. A signal denoising system applied to electroencephalogram signals, characterized in that, including: an adaptive filter module for filtering out noise in the signals after pre-noise reduction; a processor module configured to control the adaptive filter module to perform corresponding adaptive filtering and obtain the optimization parameters of the adaptive filter according to the signal pre-noise reduction result, that is, to perform the adaptive filtering method according to any one of claims 1-5.

Images