Improved cascade structure adaptive wave trap group and instantaneous frequency detection method thereof

Abstract

The invention relates to the technical field of self-adaptive wave trap groups, in particular to an improved cascade structure self-adaptive wave trap group and an instantaneous frequency detection method thereof, which can overcome the respective defects of CANFB, TCANFB and PPANFB and realize more efficient multi-frequency detection. According to the invention, the method is a series-parallel adaptive wave trap group (Cascade Pre-filled and Parallel Update ANFB, CPANFB), the calculation efficiency is high, and the calculation amount is obviously reduced while the detection precision is ensured; compared with the PPANFB, the CPANFB does not need to accurately set an initial value, so that the robustness and adaptability of the system are enhanced; by optimizing the design of pre-filtering and main filtering, the CPANFB can maintain high frequency estimation precision in various signal-to-noise ratio environments.

Description

Technical Field

[0001] The invention relates to the technical field of adaptive notch filter groups, and in particular to an adaptive notch filter group with an improved cascade structure and an instantaneous frequency detection method thereof. Background Art

[0002] Adaptive Notch Filter (ANF) is widely used in instantaneous frequency detection due to its advantages such as simple structure and low computational complexity. However, when faced with signals containing multiple sinusoidal frequencies, the existing cascaded adaptive notch filter groups (ANFBs), such as series ANFB (CANFB), triangle cascade ANFB (TCANFB) and dual parallel filter ANFB (PPANFB), each have different limitations, including but not limited to inconsistent convergence speed, high computational cost and sensitivity to initial coefficients. These problems limit the widespread application of ANFB in multi-frequency adaptive detection. Each unit of CANFB is used as an update structure for estimating the frequency coefficient. The structure is as follows: Figure 1 The coefficients between units are updated iteratively independently, and its complexity and computational complexity are lower than other ANFBs. However, the convergence speed and convergence performance of each update unit cannot be kept consistent, which leads to the inability to guarantee the same reliability of each estimation result when estimating multiple frequencies. The TCANFB structure contains a pre-filter unit group and a main filter update unit. The update unit is at the end of the structure, such as Figure 2 As shown. By introducing the difference, other sinusoidal components with interference are removed in advance, while the target frequency is retained, which ensures that each update unit converges in a similar way. However, it also introduces a huge amount of calculation. The higher the target frequency, the more the calculation cost increases. PPANFB also uses the idea of ​​difference, combining the advantages of the first two CANFB and PPANFB. While ensuring that the update unit works under similar signal-to-noise ratio conditions, only a small amount of calculation of the update unit is required to achieve good steady-state performance, as shown in Figure 3 As shown. However, it has a strong dependence on the initial coefficients. The initial unit coefficients must be properly selected and distributed according to the target frequency to ensure that each filter unit has a greater probability of converging to the final target frequency. This requires a certain prior knowledge of the target signal when using this structure, otherwise it is easy to cause frequency detection failure. This defect may bring certain limitations and difficulties in practical applications.

[0003] Therefore, the above three commonly used ANFB structures all have certain defects that restrict their application in adaptive frequency detection. It is urgent to design a new ANFB framework that can have the excellent characteristics of the existing ANFB and can make up for the above shortcomings to the greatest extent. Summary of the invention

[0004] In view of this, the present invention provides an improved cascade structure adaptive notch filter group and instantaneous frequency detection method thereof, which can overcome the respective defects of CANFB, TCANFB and PPANFB and realize more efficient multi-frequency detection.

[0005] To achieve the above purpose, the technical solution of the present invention is as follows:

[0006] An improved cascade structure adaptive notch filter group includes a plurality of pre-filter units in series and a main filter update unit in parallel; wherein each pre-filter unit does not update coefficients, and its coefficients are directly called by the update unit of the main filter; the main filter group considers retaining the balanced and consistent convergence ability between the update units, and is composed of update units with adaptive update ability placed in parallel.

[0007] Among them, the pre-filter unit does not perform coefficient update, and uses the coefficient provided by the main filter update unit; the pre-filter unit processes the input signal through a cascade structure to remove interference from non-target frequencies.

[0008] Among them, the main filter update units are arranged in parallel and receive the pre-processed signals using the parallel structure. Each unit is responsible for tracking a specific frequency component and adjusting its notch frequency in real time through an adaptive algorithm.

[0009] By adjusting the coefficients of the main filter update unit, each frequency component in the input signal is tracked and estimated.

[0010] Each update unit tracks a single-frequency sinusoidal signal and estimates and updates the corresponding frequency coefficient in real time.

[0011] Among them, the corresponding frequency coefficients are updated using the PG algorithm or the NG algorithm.

[0012] The present invention also provides an improved instantaneous frequency detection method of a cascade structure adaptive notch filter group, characterized in that it comprises the following steps:

[0013] Step 1: Acquire a multi-frequency signal to be monitored; and input the signal into the improved cascade structure adaptive notch filter group as described in any one of claims 1 to 5;

[0014] Step 2: Set the number of update units according to the number of frequencies to be detected;

[0015] Step 3: Use the adaptive algorithm in the CPANFB structure to estimate the frequency coefficient of each update unit in real time and calculate the corresponding real-time angular frequency.

[0016] Beneficial effects:

[0017] 1. The improved cascade structure adaptive notch filter group (cascade pre-filtered and parallel updated ANFB, CPANFB) of the present invention has high computational efficiency, and significantly reduces the amount of computation while ensuring detection accuracy; compared with PPANFB, CPANFB does not require precise setting of initial values, which enhances the robustness and adaptability of the system; by optimizing the design of pre-filtering and main filtering, CPANFB can maintain high frequency estimation accuracy in a variety of signal-to-noise ratio environments. .

[0018] 2. In the improved cascade structure adaptive notch filter group of the present invention, the pre-filter unit does not perform coefficient update, but uses the coefficient provided by the main filter update unit to improve the signal-to-noise ratio of the input signal and maintain the frequency "mutual exclusion" capability between units.

[0019] 3. In the improved cascade structure adaptive notch filter group of the present invention, the main filter update units are arranged in parallel to ensure that all update units have balanced and consistent convergence performance, thereby achieving effective detection of multi-frequency signals.

[0020] 4. In the method of the present invention, the pre-filter unit processes the input signal through a cascade structure to remove the interference of non-target frequencies; the main filter update unit receives the pre-processed signal using a parallel structure, and each unit is responsible for tracking a specific frequency component and adjusting its notch frequency in real time through an adaptive algorithm; by adjusting the coefficients of the main filter update unit, each frequency component in the input signal can be effectively tracked and estimated. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a schematic diagram of the CANFB structure.

[0022] Figure 2 This is a schematic diagram of the TCANFB structure.

[0023] Figure 3 Schematic diagram of the PPANFB structure.

[0024] Figure 4 This is a structural diagram of the CPANFB detection framework of the present invention. DETAILED DESCRIPTION

[0025] The present invention is described in detail below with reference to the accompanying drawings and embodiments.

[0026] The present invention proposes an improved cascade structure adaptive notch filter group, which is a new series-parallel type adaptive notch filter group (Cascade Pre-filtered and Parallel Updated ANFB, CPANFB), such as Figure 4As shown, it consists of multiple pre-filter units in series and main filter update units in parallel. The use of the pre-filter group in CPANFB is expected to maximize the signal-to-noise ratio of the input signal of each update unit. Ideally, only the target frequency signal and part of the residual white noise are retained. In order to retain the ability of "mutual exclusion" of frequencies between units and avoid the initial filtering converging to the same frequency, a CANFB series structure that is not restricted by the initial value is adopted. Each pre-filter unit does not update the coefficients, and its coefficients are directly called by the update unit of the main filter. The main filter group considers retaining the balanced and consistent convergence ability between the update units, and is composed of update units with adaptive update capabilities placed in parallel. Specifically:

[0027] The pre-filter unit does not perform coefficient updates, but uses the coefficients provided by the main filter update unit, aiming to improve the signal-to-noise ratio of the input signal of each update unit while maintaining the frequency "mutual exclusion" capability between units; the pre-filter unit processes the input signal through a cascade structure, removes interference from non-target frequencies, and creates favorable conditions for subsequent frequency detection.

[0028] The main filter update units are arranged in parallel to ensure that all update units have balanced and consistent convergence performance, thereby achieving effective detection of multi-frequency signals; the main filter update units receive pre-processed signals using a parallel structure, and each unit is responsible for tracking a specific frequency component and adjusting its notch frequency in real time through an adaptive algorithm. By adjusting the coefficients of the main filter update units, each frequency component in the input signal can be effectively tracked and estimated.

[0029] In the cascade structure of the notch pre-filter bank of CPANFB, the output residual signal e of the i-th unit is c,i (n) is expressed as:

[0030]

[0031] The corresponding bandpass signal x c,i (n) is expressed as:

[0032] x c,i (n) = x(n) - e c,i (n),i=1(3)

[0033] x c,i (n) = e c,i-1 (n)-e c,i (n),i>1(4)

[0034] In the formula, x(n) represents the original input signal, x c,i (n) represents the bandpass signal output by each pre-filtering unit, c i(n) represents the update coefficient of the i-th update unit. Since the pre-filter unit is not updated, it is directly called by the coefficient in the update unit, and ρ is the notch width.

[0035] In the main filter bank with adaptive updating capability, each update unit is formed in parallel, and the input signal, output signal and gradient signal of the ith update unit of each update unit are respectively expressed as:

[0036]

[0037]

[0038] g i (n) = x p,i (n-1)-ρe i (n-1)(7)

[0039] In the formula, x p,i (n) represents the input signal of the i-th update unit, e i (n) The residual signal of the ith update unit, g i (n) is the gradient signal of each update unit, c i (n) represents the update coefficient of the i-th unit. Finally, each update unit tracks a single frequency sinusoidal signal, estimates and updates the corresponding frequency coefficient c in real time i (n) can be updated using the PG algorithm or the NG algorithm. When using the PG algorithm, it is expressed as:

[0040] c i (n+1)=c i (n)-μe p,i (n)g i (n)(8)

[0041] When using the NG algorithm, it is expressed as:

[0042]

[0043] Among them, μ is the step size parameter that controls the convergence speed; ∈ represents a small positive number, which is introduced to avoid division by zero; α∈[0,1) represents a forgetting factor that is less than but very close to 1, such as 0.98, 0.99, etc. The NG algorithm is equivalent to low-pass filtering based on the PG algorithm, which only increases the amount of calculation but improves the filtering performance, and is widely used.

[0044] In summary, the overall structure of CPANFB of the present invention is similar to that of PPANFB, they have the same number of units and a smaller amount of calculation. The series structure used by the pre-filter group in CPANFB is the same as that of CANFB, so that the pre-filter signal obtained will not be affected by the distribution of the initial value, and the constraint of the parallel structure on the empirical initial value in PPANFB is eliminated. The parallel form used in the update filter group can retain the advantage of consistent convergence performance of each unit. CPANFB retains the advantages of the existing ANFB with small amount of calculation, high estimation efficiency and high detection accuracy to the greatest extent as a whole, and optimizes its adaptive performance that is not limited by the empirical initial value, and improves and enhances the multi-frequency detection method based on ANFB from many aspects.

[0045] The present invention also provides an instantaneous multi-frequency detection method based on the improved cascade structure adaptive notch filter group of the present invention. The method of the present invention is described in detail by taking the multi-frequency mixed signal in Table 1 as an example.

[0046] Table 1 Multi-frequency mixed signal related parameters

[0047]

[0048] The specific steps of the method of the present invention are as follows:

[0049] Step 1. Signal input: obtain the multi-frequency signal x(n) to be monitored and input it into CPANFB;

[0050] Step 2. Parameter setting: Set the number of update units p according to the number of frequencies to be detected;

[0051] Step 3. Characteristic frequency extraction and calculation: According to formula (8) or (9), the instantaneous frequency coefficient c of each update unit is obtained: i (n), and the corresponding real-time angular frequency is obtained from the relationship c = -2cosω: i (n):

[0052]

[0053] If it is to be converted into frequency, it is necessary to combine the sampling frequency f of the actual signal s To calculate, the conversion formula is:

[0054]

[0055] Using the mixed signal in Table 1 under different SNR environments, multi-frequency detection is performed using CPANFB and PPANFB respectively, and the success rate, accuracy (MSE) and stability (Std) estimation performance are calculated. Table 2 shows the performance comparison of PPANFB and CPANFB with and without initial values ​​under different SNRs, which proves the superiority of the method of the present invention.

[0056] Table 2 Performance comparison of PPANFB and CPANFB with and without initial value at different SNRs

[0057]

[0058]

[0059] In summary, the above are only preferred embodiments of the present invention and are not intended to limit the protection scope of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An improved cascade structure adaptive notch filter group, characterized in that: It includes multiple pre-filtering units in series and main filter updating units in parallel; wherein each pre-filtering unit does not update coefficients, and its coefficients are directly called by the updating unit of the main filter; the main filter group considers retaining the balanced and consistent convergence ability between the updating units, and is composed of updating units with adaptive updating ability placed in parallel.

2. The improved cascade structure adaptive notch filter group as claimed in claim 1, characterized in that: The pre-filter unit does not perform coefficient update and uses the coefficients provided by the main filter update unit; the pre-filter unit processes the input signal through a cascade structure to remove interference from non-target frequencies.

3. The improved cascade structure adaptive notch filter group as claimed in claim 2, characterized in that: The main filter update units are arranged in parallel and receive the pre-processed signals using the parallel structure. Each unit is responsible for tracking a specific frequency component and adjusting its notch frequency in real time through an adaptive algorithm. By adjusting the coefficients of the main filter update unit, each frequency component in the input signal is tracked and estimated.

4. The improved cascade structure adaptive notch filter group according to any one of claims 1 to 3, characterized in that ,Each update unit tracks a sinusoidal signal of a single frequency, and estimates and updates the ,corresponding frequency coefficient in real time.

5. The improved cascade structure adaptive notch filter group as claimed in claim 4, characterized in that: Use the PG algorithm or NG algorithm to update the corresponding frequency coefficients.

6. A method for detecting instantaneous frequency of an improved cascade structure adaptive notch filter group according to any one of claims 1 to 5, characterized in that , including the following steps: Step 1: Acquire a multi-frequency signal to be monitored; and input the signal into the improved cascade structure adaptive notch filter group as described in any one of claims 1 to 5; Step 2: Set the number of update units according to the number of frequencies to be detected; Step 3: Use the adaptive algorithm in the CPANFB structure to estimate the frequency coefficient of each update unit in real time and calculate the corresponding real-time angular frequency.

Images