A signal noise reduction method and device based on improved AKF algorithm

Abstract

The present invention discloses a signal denoising method and device based on an improved AKF algorithm, the method comprising: based on the existing AKF denoising method, introducing an improved AKF algorithm to improve the denoising accuracy. The present invention proposes two improved optimization methods, considering that the measurement value error and the true value error are the influencing factors affecting the denoising result in the iterative process of the AKF algorithm, setting a threshold for the error value, and optimizing and updating the denoising result that exceeds the error threshold value. At the same time, considering the influencing factors of the error, a related optimization relationship is established, and the denoising accuracy is effectively improved. The present invention adopts the denoising method of the improved AKF algorithm to solve the problem of low denoising accuracy of the existing AKF algorithm.

Description

Technical Field

[0001] The invention belongs to the technical field of signal noise reduction, and relates to a noise reduction method and a device based on an improved Kalman filter AKF algorithm. Background Art

[0002] During the signal acquisition and transmission process, due to the interference of the external environment and the influence of the instrument itself, it is inevitable that there will be noise mixed in it. Noise is an important factor affecting the detection and recognition performance of the target signal. Especially in the analysis of some high-precision data, even weak noise will have a huge impact on the analysis results. Therefore, in the signal analysis process, the first thing to do is to denoise the signal. The commonly used signal noise reduction method uses the filtering method, which is to extract useful signals from the received signal containing interference. Filtering is divided into classical filtering and modern filtering. Classical filtering is an engineering concept proposed based on Fourier analysis and transformation. According to classical filtering theory, any signal that meets certain conditions can be regarded as a superposition of infinite sine waves. Classical filtering allows signal components within a certain frequency range to pass normally, while preventing other frequency components from passing. Modern filtering uses analog electronic circuits to filter analog signals. Its basic principle is to use the frequency characteristics of the circuit to select the frequency components in the signal.

[0003] Widely used signal filters include time domain filtering, frequency domain filtering, linear filters, nonlinear filters, active filters and passive filters. Existing filters still have the following two problems for signal noise reduction:

[0004] (1) Since the received signal changes dynamically, the traditional noise reduction method can only be implemented under the condition that the useful signal and the noise have different frequency bands. The adaptive Kalman filter has linear filtering and prediction theory. Its characteristic is that it processes the noisy input and observation signals based on the linear state space representation to obtain the system state or real signal.

[0005] (2) The adaptive Kalman filter predicts the output estimate of the current moment with the output result of the previous moment through the prediction and update stages, and uses the measured value to estimate the expected output result of the current moment, and expresses the AKF denoising output result through the transformation relationship. Since the AKF algorithm uses the signal iteration relationship, and the error generated by each signal denoising is transferable, this error transfer affects the accuracy of the AKF denoising output result. It is necessary to optimize the error of the AKF denoising output result to make the denoising output result more accurate. Summary of the invention

[0006] Objective: In order to overcome the deficiencies in the prior art, the present invention provides a signal denoising method and device based on an improved AKF algorithm, which uses the improved AKF algorithm to perform denoising on the received signal, overcomes the problem of low denoising accuracy of the existing AKF algorithm, and improves the denoising accuracy by proposing an optimization scheme.

[0007] Technical solution: To solve the above technical problems, the technical solution adopted by the present invention is:

[0008] In a first aspect, a signal noise reduction method is provided, comprising:

[0009] (1) Acquire received signal data to be subjected to noise reduction processing, wherein the signal data is a signal measurement value matrix Z;

[0010] (2) The signal measurement value matrix Z is subjected to denoising using the adaptive Kalman filter algorithm to obtain the denoised X value and the measurement value error ε t and / or true value error δ t ;

[0011] (3) Using an improved Kalman filter algorithm based on measurement error correction or true value error correction to perform noise reduction processing, the noise reduction result after correction processing is obtained, including:

[0012] (3a) In response to |ε t |≥gate1 t , an improved Kalman filter algorithm based on measurement error correction is used for noise reduction:

[0013] Using the measurement error value Indicates the expected output of the Kalman filter AKF at time t Deviation measurement value Z t By setting the first correction threshold by gate1 t , the measurement error value ε of each iteration t With gate1 t In comparison, if |ε t |≥gate1 t , then correct The corrected Used for the next iteration to finally obtain the denoised output result

[0014] or, 3b) in response to |δ t |≥gate2 t , the improved Kalman filter algorithm based on the correction of the true value error is used for noise reduction:

[0015] The correction is based on the true value error, and the true value error value Represents the true value of the signal X t And the predicted true value The error between the actual value and the error between the predicted and measured values ​​is composed of the predicted and measured noises. By setting the second correction threshold, gate2 t , the true value error value δ of each iteration t With gate2 t In comparison, if |δ t |≥gate2 t , then correct The corrected calculation is obtained by Obtained Update Output as the denoising output result.

[0016] In some embodiments, the noise reduction process using an adaptive Kalman filter algorithm includes:

[0017] Step L1: The true value X of the signal received at time t t and the measured value Z t satisfy:

[0018] X t =FX t-1 +N t Formula 1

[0019] Z t =HX t +V t Formula 2

[0020] Among them, X t-1 is the true value vector at time t-1; F is the state transfer matrix, which expresses the true value of the received signal at time t-1 as the true value at time t; H is the measurement matrix, which maps the true value to the observation space, that is, the true value of the signal that cannot be measured is expressed as a measurable quantity; the vector N t The mean is zero and the covariance matrix is ​​Q t Gaussian white noise; vector V t The mean is zero and the covariance matrix is ​​R t Gaussian white noise;

[0021] Step L2: In the prediction stage, AKF estimates the posterior true value at time t-1 (Output result of AKF at time t-1) calculation is the prior estimate of the true value at time t, according to P t-1 and Q t Calculate the prior error covariance matrix P at time t t|t-1 ; is the posterior error covariance matrix at time t-1, and cov represents the covariance calculation;

[0022]

[0023] P t|t-1 =FP t-1 F T +Q t Formula 4

[0024] Among them, F T represents the transposed matrix of the state transfer matrix F;

[0025] Step L3: In the update phase, first calculate the expected measurement estimate at time t (The expected output of the AKF algorithm at time t), and then calculate the Kalman gain matrix K at time t t , finally according to K t , and Z t Obtain the posterior true value estimate at time t (The output result after denoising by the AKF algorithm at time t), update the posterior error covariance matrix P at time t t :

[0026]

[0027] K t =P t|t-1 H T (HP t|t-1 H T +R t ) -1 Formula 6

[0028]

[0029] P t =P t|t-1 -K t HP t|t-1 Formula 8

[0030] In the noise covariance estimation stage, update Q t and R t , used for the next iteration; the update method is as follows:

[0031]

[0032]

[0033]

[0034] Among them, ε t Represents Z t and The error between the measured value and the expected output value is measured; H T represents the transposed matrix of the measurement matrix H; b is the forgetting factor.

[0035] Furthermore, F and H are represented by M-order identity matrices.

[0036] Furthermore, P t-1 The initial value P0 is represented by the M-order unit matrix; The initial value of Take the first measurement.

[0037] Furthermore, the range of b is 0.95 to 0.995.

[0038] In some embodiments, the noise reduction process is performed based on the improved Kalman filter algorithm corrected by the measurement error value, including:

[0039] Set the threshold value gate1 at time t t as follows:

[0040]

[0041] gate1 t =μr1 t Formula 13

[0042] Where R t is the measurement error, R t (1,1) is R t The value in the first row and first column, r1 t is the measurement error R t The square root of , μ is the coefficient;

[0043] Depend on Reflection: At time t, the expected output of AKF Deviation measurement value Z t Degree:

[0044] When |ε t |<gate1 t , the output value at time t-1 Used as the output result at time t;

[0045] When |ε t |≥gate1 t , correct get

[0046]

[0047] c t =μr1 t / Rt (1,1) Formula 15

[0048] where c t is the correction factor;

[0049] Obtained Used for the next iteration to finally obtain the denoised output result

[0050] Furthermore, the coefficient μ ranges from 0.5 to 0.9.

[0051] In some embodiments, the improved Kalman filter algorithm based on the true value error value correction performs noise reduction processing, including:

[0052] Set the threshold value gate2 at time t t as follows:

[0053]

[0054] gate2 t =θr2 t Formula 17

[0055] Where Q t-1 The system noise at the previous moment, R t (1,1) is the measurement error R t The value in the first row and first column, r2 t is the square root of the sum of system noise and measurement error; K t is the Kalman gain matrix, θ is the coefficient;

[0056] The error value from the true value Represents the true value of the signal X t And the predicted true value The error between

[0057] Will With gate2 t Compare, if |δ t |<gate2 t Indicates that the error is within the threshold value range and the noise reduction output result More accurate, that is, the output after the adaptive Kalman filter algorithm That is the true value of the signal after noise reduction; if |δ t |≥gate2 t ,but Inaccurate, based on The corrected Corrected It is expressed as follows:

[0058]

[0059] Obtained Update Output as the denoising output result.

[0060] Furthermore, the coefficient θ ranges from 0.7 to 1.

[0061] In a second aspect, the present invention provides a signal noise reduction device based on an improved AKF algorithm, comprising a processor and a storage medium;

[0062] The storage medium is used to store instructions;

[0063] The processor is used to operate according to the instructions to execute the steps of the method according to the first aspect.

[0064] Beneficial effects: The signal denoising method and device based on the improved AKF algorithm provided by the present invention have the following advantages:

[0065] (1) The present invention proposes an improved AKF denoising method, analyzes the influencing factors from the perspective of the error of the denoising result, proposes an AKF denoising method based on the measurement result error improvement and an AKF denoising method based on the true value error improvement, analyzes from multiple angles, sets a threshold value for the error, retains the results below the threshold value, and further optimizes and updates the results above the threshold value, thereby further improving the denoising accuracy.

[0066] (2) The present invention proposes an improvement method for noise reduction results with large errors. The method considers the influencing factors of the errors and establishes a correlation equation, that is, a correction equation is added to the original noise reduction output result, so that the improved noise reduction output effect is better. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 is a complete algorithm process flow chart of an embodiment of the present invention;

[0068] Figure 2 AKF algorithm flow chart of the embodiment;

[0069] Figure 3 It is a flow chart of the AKF denoising algorithm based on the improvement of the measurement value error in the embodiment;

[0070] Figure 4 It is a flow chart of the AKF denoising algorithm improved based on the true value error of the embodiment. DETAILED DESCRIPTION

[0071] The present invention will be further described below in conjunction with the accompanying drawings and embodiments. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0072] In the description of the present invention, "several" means more than one, "many" means more than two, "greater than", "less than", "exceed", etc. are understood to exclude the number itself, and "above", "below", "within", etc. are understood to include the number itself. If there is a description of "first" or "second", it is only used for the purpose of distinguishing the technical features, and cannot be understood as indicating or implying the relative importance or implicitly indicating the number of the indicated technical features or implicitly indicating the order of the indicated technical features.

[0073] In the description of the present invention, the description with reference to the terms "one embodiment", "some embodiments", "illustrative embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0074] Example 1

[0075] A signal noise reduction method based on an improved AKF algorithm, comprising:

[0076] (1) Acquire received signal data to be subjected to noise reduction processing, wherein the signal data is a signal measurement value matrix Z;

[0077] (2) The signal measurement value matrix Z is subjected to denoising using the adaptive Kalman filter algorithm to obtain the denoised X value and the measurement value error ε t and / or true value error δ t ;

[0078] (3) Using an improved Kalman filter algorithm based on measurement error correction or true value error correction to perform noise reduction processing, the noise reduction result after correction processing is obtained, including:

[0079] (3a) In response to |ε t |≥gate1 t , an improved Kalman filter algorithm based on measurement error correction is used for noise reduction:

[0080] Using the measurement error value Indicates the expected output of the Kalman filter AKF at time t Deviation measurement value Z t By setting the first correction threshold by gate1 t , the measurement error value ε of each iteration t With gate1 t In comparison, if |ε t |≥gate1t , then correct The corrected Used for the next iteration to finally obtain the denoised output result

[0081] or, 3b) in response to |δ t |≥gate2 t , the improved Kalman filter algorithm based on the correction of the true value error is used for noise reduction:

[0082] The correction is based on the true value error, and the true value error value Represents the true value of the signal X t And the predicted true value The error between the actual value and the error between the predicted and measured values ​​is composed of the predicted and measured noises. By setting the second correction threshold, gate2 t , the true value error value δ of each iteration t With gate2 t In comparison, if |δ t |≥gate2 t , then correct The corrected calculation is obtained by Obtained Update Output as the denoising output result.

[0083] In some embodiments, Figure 1 As shown, a signal noise reduction method based on an improved AKF algorithm comprises:

[0084] Step 1: Build a received signal database;

[0085] Step 2: Measure the signal value at the test point multiple times to obtain the measurement value matrix Z to be processed by noise reduction;

[0086] Step 3: Use the adaptive Kalman filter algorithm to obtain the x value after noise reduction;

[0087] Step 4: Considering that the measurement value is the influencing factor of the current noise reduction output result, an improved Kalman filter noise reduction method based on measurement error value correction is adopted. Indicates the expected output of AKF at time t Deviation measurement value Z t By setting the correction threshold, gate1 t , the error value ε of each iteration t In comparison, if |ε t |≥gate1 t , then correct according to Recalculate the expected value at time t Thus, the The accuracy can achieve the purpose of real-time correction.

[0088] Step 5: Consider the difference between the noise reduction output result and the real signal as an indicator of noise reduction accuracy, and use the real value error to make corrections, Represents the true value of the signal X t To predict the true value The error between the actual value and the measured value is composed of the prediction noise and the measurement noise. The correction threshold is set by gate2 t , the error value ε of each iteration t In comparison, if |δ t |≥gate2 t , then correct The corrected calculation is obtained by Thereby improving the The accuracy can achieve the purpose of real-time correction.

[0089] like Figure 2 The specific steps of the AKF algorithm shown are as follows:

[0090] The AKF denoising method uses the denoising output result obtained at the previous moment to adjust the filter parameters at the current moment, thereby achieving denoising of the measured signal. The AKF denoising method is based on two basic conditions: 1. The true value at time t can be expressed by the true value at time t-1, and the measured value at time t can be expressed by the true value. 2. The expression and measurement process can be described by a linear algorithm.

[0091] Step L1: The true value X of the signal received at time t t and the measured value Z t satisfy:

[0092] X t =FX t-1 +N t Formula 1

[0093] Z t =HX t +V t Formula 2

[0094] X t-1 is the true value vector at time t-1. F is the state transfer matrix, whose function is to express the true value of the received signal at time t-1 as the true value at time t; H is the measurement matrix, which mainly maps the true value to the observation space, that is, to express the true value of the signal that cannot be measured as a measurable quantity. The only difference between the measured value and the true value is the noise. At different times, the only difference between different measured values ​​is the random value of the noise, so F and H can be represented by the M-order unit matrix. Vector Nt The mean is zero and the covariance matrix is ​​Q t Gaussian white noise; vector V t The mean is zero and the covariance matrix is ​​R t Gaussian white noise.

[0095] Step L2: In the prediction stage, AKF estimates the posterior true value at time t-1 (Output result of AKF at time t-1) calculation is the prior estimate of the true value at time t, and according to P t-1 and Q t Calculate the prior error covariance matrix P at time t t|t-1 . is the posterior error covariance matrix at time t-1, and cov represents the covariance calculation. t-1 Reflects the accuracy of the output result at time t-1:

[0096]

[0097] P t|t-1 =FP t-1 F T +Q t Formula 4

[0098] Since the true value of the received signal cannot be measured at any time, P t-1 The initial value P0 of is difficult to determine. Generally speaking, when P0 is not a zero matrix, it does not affect the convergence of AKF, so the present invention uses an M-order unit matrix to represent it.

[0099] The initial value of Take the first measurement.

[0100] Step L3: In the update phase, first calculate the expected measurement estimate at time t (The expected output of the AKF algorithm at time t), and then calculate the Kalman gain matrix K at time t t , finally according to K t , and Z t Obtain the posterior true value estimate at time t (The output result after denoising by the AKF algorithm at time t), update the posterior error covariance matrix P at time t t :

[0101]

[0102] K t =P t|t-1 H T (HPt|t-1 H T +R t ) -1 Formula 6

[0103]

[0104] P t =P t|t-1 -K t HP t|t-1 Formula 8

[0105] In the noise covariance estimation stage, update Q t and R t , used for the next iteration. The update method is as follows:

[0106]

[0107]

[0108]

[0109] Among them, ε t Represents Z t and The error between the measured value and the expected output value. T represents the transposed matrix of the state transfer matrix F; H T represents the transposed matrix of the measurement matrix H;

[0110] b is the forgetting factor. The function of b is to reduce the influence of the noise reduction output result (old data) at time t-1 on the data used for iteration at time t (new data). As the number of iterations increases, the error covariance matrix will become smaller and even become a zero matrix. t|t-1 →0, K t →0, then K t It is unable to correct the noise reduction, that is, the old data makes the AKF algorithm saturated, which is called the "data saturation" phenomenon. This phenomenon will cause the new data to fail to play the noise reduction function in the iteration process, and the noise reduction input and the new data are closely related, and the old data will bury the new data. In order to remove the influence of the old data on noise reduction, the forgetting factor b is introduced into the AKF algorithm to speed up its convergence. The range of b is 0.95 to 0.995.

[0111] like Figure 3 As shown, the specific steps of the embodiment of the AKF noise reduction method based on the measurement value error improvement are as follows:

[0112] The random signal generated by the system is used as the real signal value X t , in this embodiment, the system noise Nt is a Gaussian white noise with a mean of zero and a covariance matrix of the unit matrix. The signal generated by the model is used as the measurement value Z t In this embodiment, the measurement noise is a Gaussian white noise with a mean of zero and a covariance matrix of a unit matrix, and time t is 100. After the AKF denoising method, the measurement error ε is obtained. t .

[0113] Depend on Reflects: At time t, the expected output of AKF Deviation measurement value Z t When ε t When ε is small, the output value at time t-1 can be used as the output result at time t. t When it is large, the model parameters are inaccurate or the measured value Z t When there is a large fluctuation, then at time t, the error between the expected output and the measured value is large. Based on the formula The output result obtained by the AKF algorithm is not accurate. In order to improve the accuracy of the output result of the AKF algorithm at time t, With Z t Adjustments are made to the errors between them, that is, corrections are made using the measurement errors.

[0114] The present invention is based on t Set judgment criteria and propose correction criteria to reduce the expected output at time t With the measured value Z t The error between the expected output and the measured value makes the output of the AKF algorithm closer to the true value, that is, the correction initial condition is based on the error between the expected output and the measured value being greater than a certain correction threshold.

[0115]

[0116] gate1 t =μr1 t Formula 13

[0117] Where R t To measure noise, R t (1,1) is R t The value in the first row and first column in this embodiment is R t (1,1) = 1. The correction threshold at time t is determined by gate1 t The measurement noise R tThe coefficient μ ranges from 0.5 to 0.9. The purpose is to balance the influence of the measured value at time t and the expected output on the output result of the AKF algorithm, while increasing the trust in the measured r value. Because the AKF model is an approximation of the actual scene, the parameters cannot be truly accurate, and its output result is based on the measured value and then indirectly obtained through a series of iterative processes. Therefore, the trust in the measured value should be increased so that the output result after denoising by the AKF algorithm is closer to the true value. In this embodiment, μ=0.8.

[0118] Compare the error value with the threshold value. With gate1 t Compare, if |ε t |<gate1 t Indicates that the error is within the threshold value range and the noise reduction output result More accurate, that is, the output after the adaptive Kalman filter algorithm This is the true value of the signal after noise reduction. t |≥gate1 t ,Depend on Indicates expected output With the measured value Z t The deviation is greater than the threshold value. express Not accurate, based on Right now If the deviation is large, Corrected to get the corrected

[0119] Corrected It is expressed as follows:

[0120]

[0121] c t =μr1 t / R t (1,1) Formula 15

[0122] where c t is the correction factor;

[0123] Obtained Used for the next iteration to finally obtain the denoised output result

[0124] Depend on The true value error of the denoised output result of this embodiment is calculated, and the error value is less than 0.02. The error between the denoised output result and the true value before the improvement is less than 0.4.

[0125] like Figure 4As shown, the specific steps of the embodiment of the AKF denoising method based on the true value error are as follows:

[0126] The random signal generated by the system is used as the real signal value X t , in this embodiment, the system noise N t is a Gaussian white noise with a mean of zero and a covariance matrix of the unit matrix. The signal generated by the model is used as the measurement value Z t In this embodiment, the measurement noise is a Gaussian white noise with a mean of zero and a covariance matrix of a unit matrix, and time t is 100. After the AKF denoising method, the measurement error ε is obtained. t .

[0127] make Represents the true value of the signal X t To predict the true value The error between It can be seen that the predicted true value is determined by the prior true estimate and measurement error The prior true estimate From the real value of the previous moment Obtained through the state transfer matrix, so the posterior true value estimate at time t The noise is related to the system noise Q at the previous moment t-1 and the prediction model noise R t Related to, that is, δ t With Q t-1 , R t related.

[0128] If δ t is larger, indicating that at time t The error is large, so Correction Thus, the Therefore, the present invention is based on δ t Set the judgment criteria and propose correction criteria to reduce the true value X of the signal at time t t To predict the true value The error between the expected output and the actual value makes the output of the AKF algorithm denoising closer to the true value, that is, the correction initial condition is based on the error between the expected output and the actual value being greater than a certain correction threshold.

[0129]

[0130] gate2 t =θr2 t Formula 17

[0131] Where R t(1,1) is the measurement error R t The value in the first row and first column, r2 t is the square root of the sum of system noise and measurement error; K t is the Kalman gain matrix. The correction threshold at time t is given by gate2 t The coefficient θ ranges from 0.7 to 1, and its purpose is to balance the error between the true value at time t and the expected output. In this embodiment, θ=0.9.

[0132] Compare the true value error value with the threshold value, that is, With gate2 t Compare, if |δ t |<gate2 t Indicates that the error is within the threshold value range and the noise reduction output result More accurate, that is, the output after the adaptive Kalman filter algorithm This is the true value of the signal after noise reduction. t |≥gate2 t ,but Inaccurate, based on The corrected Corrected It is expressed as follows:

[0133]

[0134] Obtained Update Output as the denoising output result.

[0135] Depend on The true value error of the denoised output result of this embodiment is calculated, and the error value is less than 0.02. The error between the denoised output result and the true value before the improvement is less than 0.4.

[0136] Example 2

[0137] In a second aspect, this embodiment provides a signal noise reduction device based on an improved AKF algorithm, including a processor and a storage medium;

[0138] The storage medium is used to store instructions;

[0139] The processor is used to operate according to the instructions to execute the steps of the method according to embodiment 1.

[0140] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0141] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0142] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0143] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0144] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A signal noise reduction method, characterized in that: include: (1) Acquire received signal data to be subjected to noise reduction processing, wherein the signal data is a signal measurement value matrix Z; (2) The signal measurement value matrix Z is subjected to denoising using the adaptive Kalman filter algorithm to obtain the denoised X value and the measurement value error ε t and / or true value error δ t ; (3) Using an improved Kalman filter algorithm based on measurement error correction or true value error correction to perform noise reduction processing, the noise reduction result after correction processing is obtained, including: (3a) In response to |ε t |≥gate1 t , an improved Kalman filter algorithm based on measurement error correction is used for noise reduction: Using the measurement error value Indicates the expected output of the Kalman filter AKF at time t Deviation measurement value Z t By setting the first correction threshold by gate1 t , the measurement error value ε of each iteration t With gate1 t In comparison, if |ε t |≥gate1 t , then correct The corrected Used for the next iteration to finally obtain the denoised output result 3b) In response to |δ t |≥gate2 t , the improved Kalman filter algorithm based on the correction of the true value error is used for noise reduction: The correction is based on the true value error, and the true value error value Represents the true value of the signal X t And the predicted true value The error between the actual value and the error between the predicted and measured values ​​is composed of the predicted and measured noises. By setting the second correction threshold, gate2 t , the true value error value δ of each iteration t With gate2 t In comparison, if |δ t |≥gate2 t , then correct The corrected calculation is obtained by Obtained Update Output as denoising output result; The improved Kalman filter algorithm based on the correction of the true value error value performs noise reduction processing, including: Set the second correction threshold by gate2 t as follows: gate2 t =θr2 t Where Q t-1 is the system noise at the previous moment, R t (1,1) is the measurement error R t The value in the first row and first column, r2 t is the square root of the sum of system noise and measurement error; K t is the Kalman gain matrix, θ is the coefficient; The error value from the true value Represents the true value of the signal X t And the predicted true value The error between Will With gate2 t Compare, if |δ t | <gate2 t Indicates that the error is within the threshold value range and the noise reduction output result More accurate, that is, the output after the adaptive Kalman filter algorithm That is the true value of the signal after noise reduction; if |δ t |≥gate2 t ,but Inaccurate, based on The corrected Corrected It is expressed as follows: Obtained Update Output as the denoising output result.

2. The signal noise reduction method according to claim 1, characterized in that: The noise reduction process using the adaptive Kalman filter algorithm includes: Step L1: The true value X of the signal received at time t t and the measured value Z t satisfy: X t =FX t-1 +N t Z t =HX t +V t Among them, X t-1 is the true value vector at time t-1; F is the state transfer matrix, which expresses the true value of the received signal at time t-1 as the true value at time t; H is the measurement matrix, which maps the true value to the observation space, that is, the true value of the signal that cannot be measured is expressed as a measurable quantity; the vector N t The mean is zero and the covariance matrix is ​​Q t Gaussian white noise; vector V t The mean is zero and the covariance matrix is ​​R t Gaussian white noise; Step L2: In the prediction stage, AKF estimates the posterior true value at time t-1 calculate That is, the output result of AKF at time t-1, is the prior estimate of the true value at time t, according to P t-1 and Q t Calculate the prior error covariance matrix P at time t t|t-1 ; is the posterior error covariance matrix at time t-1, and cov represents the covariance calculation; P t|t-1 =FP t-1 F T +Q t Among them, F T represents the transposed matrix of the state transfer matrix F; Step L3: In the update phase, first calculate the expected measurement estimate at time t It is the expected output result of the AKF algorithm at time t, and then calculates the Kalman gain matrix K at time t t , finally according to K t , and Z t Obtain the posterior true value estimate at time t It is the output result of AKF algorithm after noise reduction at time t, and updates the posterior error covariance matrix P at time t t : K t =P t|t-1 H T (HP t|t-1 H T +R t ) -1 P t =P t|t-1 -K t HP t|t-1 In the noise covariance estimation stage, update Q t and R t , used for the next iteration; the update method is as follows: Among them, ε t Represents Z t and The error between the measured value and the expected output value is measured; H T represents the transposed matrix of the measurement matrix H; b is the forgetting factor.

3. The signal noise reduction method according to claim 2, characterized in that: F and H are represented by M-order identity matrices.

4. The signal noise reduction method according to claim 2, characterized in that: P t-1 The initial value P0 is represented by the M-order unit matrix; The initial value of Take the first measurement.

5. The signal noise reduction method according to claim 2, characterized in that: The range of b is 0.95 to 0.

995.

6. The signal noise reduction method according to claim 1, characterized in that: The improved Kalman filter algorithm based on the measurement error correction is used for noise reduction, including: Set the first correction threshold by gate1 t as follows: gate1 t =μr1 t Where R t is the measurement error, R t (1,1) is R t The value in the first row and first column, r1 t is the measurement error R t The square root of , μ is the coefficient; Depend on Reflection: At time t, the expected output of AKF Deviation measurement value Z t Degree: When |ε t | <gate1 t , the output value at time t-1 Used as the output result at time t; When |ε t |≥gate1 t , correct get c t =μr1 t / R t (1,1) where c t is the correction factor; Obtained Used for the next iteration to finally obtain the denoised output result 7. The signal noise reduction method according to claim 6, characterized in that: The coefficient μ ranges from 0.5 to 0.

9.

8. The signal noise reduction method according to claim 1, characterized in that: The coefficient θ ranges from 0.7 to 1.

9. A signal noise reduction device, characterized in that: include: including processor and storage medium; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of the method according to any one of claims 1 to 8.

Images