Adaptive signal filtering method and system based on multi-band fusion

Abstract

The invention provides a self-adaptive signal filtering method and system based on multi-band fusion, and the method comprises the steps: firstly obtaining a multi-band signal set containing original signal components of different bands in a target environment, carrying out the frequency band decomposition of the multi-band signal set to obtain a plurality of sub-band signals, and then carrying out the self-adaptive filtering operation of each sub-band signal, the method comprises the following steps: dynamically adjusting a filtering coefficient according to noise characteristics of a sub-band signal, generating a filtered sub-band signal, carrying out multi-band fusion on the filtered sub-band signal to obtain a target filtering signal, and finally adjusting a segmentation strategy of band decomposition or a dynamic adjustment rule of adaptive filtering according to a quality evaluation result of the target filtering signal. And outputting the optimized target filtering signal, thereby effectively coping with a complex multi-band signal, adaptively adjusting the filtering process, and improving the quality and stability of the filtering signal.

Description

Technical Field

[0001] The present invention relates to the technical field of signal filtering, and in particular, to an adaptive signal filtering method and system based on multi-band fusion. Background Art

[0002] In today's signal processing field, signal filtering is a crucial technology, widely used in many fields such as communication, radar, audio processing, etc. Its purpose is to remove noise interference in the signal to obtain high-quality useful signals.

[0003] Most traditional signal filtering methods adopt a single fixed filtering strategy. When dealing with multi-band signals containing original signal components of different frequency bands, this single filtering strategy has obvious limitations. Since signals of different frequency bands have different characteristics, such as frequency range, noise distribution, etc., a single filtering parameter is difficult to meet the filtering requirements of signals in each frequency band simultaneously. This leads to the situation that during the filtering process, some frequency band signals may be over-filtered, while some other frequency band signals may be under-filtered, thus affecting the quality of the final filtered signal.

[0004] In addition, existing filtering technologies usually lack the adaptive ability to the real-time changes of signals. In practical applications, the signal characteristics of the target environment are often dynamically changing, and the characteristics of noise also change with time and environment. However, traditional filtering methods cannot dynamically adjust the filtering coefficients according to the real-time noise characteristics of the signal, making it difficult to maintain a stable filtering effect and difficult to provide reliable filtering results in a complex and changeable signal environment. Summary of the Invention

[0005] In view of the problems mentioned above, in combination with the first aspect of the present invention, embodiments of the present invention provide an adaptive signal filtering method based on multi-band fusion, and the method includes: Obtain a multi-band signal set of the target environment, where the multi-band signal set contains original signal components of different frequency bands; Perform frequency band decomposition processing on the multi-band signal set to obtain a plurality of sub-band signals; Perform an adaptive filtering operation on each of the sub-band signals to generate corresponding filtered sub-band signals; wherein, the adaptive filtering operation includes dynamically adjusting the filtering coefficients according to the noise characteristics of the sub-band signals; Perform multi-band fusion processing on the filtered sub-band signals to generate a target filtered signal; According to the signal quality evaluation result of the target filtered signal, adjust the segmentation strategy of the frequency band decomposition processing or the dynamic adjustment rule of the adaptive filtering operation, and output an optimized target filtered signal.

[0006] In a possible implementation of the first aspect, the process of performing frequency band decomposition on the multi-band signal set to obtain multiple sub-band signals includes: Determine the target decomposition level according to the initial frequency band division rule, and preferentially perform dynamic iterative adjustment on the number of frequency bands and the bandwidth range of the target decomposition level based on the spectral characteristics of the multi-band signal set during the signal processing. Wherein, if the frequency band energy balance characteristic after the dynamic iterative adjustment does not meet the preset threshold, further optimize the frequency band division rule according to the frequency band energy ratio; Perform overlapping segmentation processing on each original signal component in the multi-band signal set based on the target decomposition level to generate initial segmented signals; Perform a frequency domain transformation operation on the initial segmented signals to obtain frequency domain energy distribution characteristics; Map the initial segmented signals to the corresponding target frequency bands according to the matching degree between the frequency domain energy distribution characteristics and the frequency band division rule to generate the sub-band signals.

[0007] In a possible implementation of the first aspect, the process of performing overlapping segmentation processing on each original signal component in the multi-band signal set based on the target decomposition level to generate initial segmented signals includes: Determine the sampling rate of the original signal component, and calculate the number of sample points corresponding to the segmentation window according to the minimum bandwidth range in the target decomposition level and the sampling rate, so that the duration of the segmentation window is proportional to the reciprocal of the minimum bandwidth; Determine the number of overlapping sample points between adjacent segmentation windows based on the number of frequency bands in the target decomposition level, where the number of overlapping sample points has an inverse proportional relationship with the number of frequency bands; Generate an equidistant sliding window sequence according to the number of sample points of the segmentation window and the number of overlapping sample points.

[0008] In a possible implementation of the first aspect, the process of performing an adaptive filtering operation on each sub-band signal to generate a corresponding filtered sub-band signal includes: Extract the time domain waveform characteristics and frequency domain energy characteristics of the sub-band signal; Determine the signal stationarity index according to the time domain waveform characteristics, and determine the noise energy ratio according to the energy ratio of the noise-dominated frequency band in the frequency domain energy characteristics. Wherein, the noise-dominated frequency band is identified through a preset noise reference interval or a dynamic noise estimation method based on the minimum statistic method, specifically including extracting the background noise energy distribution in the signal silent segment and iteratively updating the noise energy threshold; Construct a dynamic weight factor based on the signal stationarity index and the noise energy ratio, and adjust the preset filter parameter set according to the dynamic weight factor; Call the set of filtering parameters to perform recursive filtering on the sub-band signal, and in the process of filtering, detect the pole positions of the filter in real time to constrain its stability. If the detected poles exceed the unit circle range, reset the filtering parameters to the preset safety values to generate an initial filtered signal; Perform short-time Fourier transform processing on the initial filtered signal to generate a time-frequency energy distribution matrix; According to the frequency band division rule of the current target decomposition level, extract the characteristic frequency bands related to noise in the time-frequency energy distribution matrix; Calculate the ratio of the average energy in the characteristic frequency band to the average energy of the signal dominant frequency band. If the ratio exceeds the preset noise threshold, it is determined that there is residual noise energy, and the set of filtering parameters is iteratively updated until the residual noise energy is lower than the preset threshold, and the final filtered sub-band signal is output.

[0009] In a possible implementation manner of the first aspect, the determining the signal stationarity index according to the time-domain waveform feature includes: Calculate the time-domain variance sequence of the sub-band signal and the difference mean value between adjacent sampling points; Determine the first stationarity factor according to the fluctuation amplitude of the time-domain variance sequence; Determine the second stationarity factor according to the absolute value of the difference mean value; Perform standardization conversion processing on the first stationarity factor and the second stationarity factor respectively to map them to a unified ratio interval, and dynamically allocate weight coefficients of the standardized first stationarity factor and the second stationarity factor according to the frequency band position of the sub-band signal, and then perform weighted summation to generate the signal stationarity index; Wherein, the weight coefficients of the first stationarity factor and the second stationarity factor are dynamically allocated according to the frequency band position of the sub-band signal.

[0010] In a possible implementation manner of the first aspect, the performing multi-band fusion processing on the filtered sub-band signal to generate a target filtered signal includes: Perform signal reconstruction processing on each of the filtered sub-band signals to generate a time-domain signal component of the same length as the original signal component; Dynamically determine the frequency band weight coefficient of each time-domain signal component according to the target decomposition level and the signal quality score; Based on the frequency band weight coefficient, perform time-domain alignment and weighted superposition processing on the time-domain signal components to generate an initial fusion signal; Extract the phase difference features of the adjacent frequency band signal components in the initial fusion signal, and dynamically calculate the phase correction amount of each frequency band according to the phase difference features; Based on the phase correction amounts of each frequency band, perform phase alignment on the signal components of the initial fusion signal through time-domain delay compensation or interpolation processing. For the change in signal length caused by time-domain delay compensation, use cyclic shift or edge-symmetric interpolation methods to keep the signal components the same length as the original signal, and generate a phase-aligned signal; Perform time-domain smoothing processing on the phase-aligned signal to eliminate the instantaneous mutation noise introduced by phase correction, and obtain the corrected initial fusion signal; Generate the target filtering signal according to the corrected initial fusion signal.

[0011] In a possible implementation manner of the first aspect, the adjusting the segmentation strategy of the frequency band decomposition processing or the dynamic adjustment rule of the adaptive filtering operation according to the signal quality evaluation result of the target filtering signal includes: Extract the time-domain signal-to-noise ratio feature, frequency band energy balance feature, and phase continuity feature of the target filtering signal; Calculate a first optimization index according to the time-domain signal-to-noise ratio feature, calculate a second optimization index according to the frequency band energy balance feature, and calculate a third optimization index according to the phase continuity feature; Perform range normalization processing on the first optimization index, the second optimization index, and the third optimization index respectively, calculate the normalized weighted sum based on the preset weight coefficient, and generate a comprehensive optimization score; If the comprehensive optimization score is lower than the preset optimization threshold, then perform at least one of the following adjustment operations: (1) Adjust the number of frequency bands or the bandwidth range in the frequency band division rule according to the frequency band energy balance feature; (2) Adjust the calculation rule of the dynamic weight factor in the adaptive filtering operation according to the time-domain signal-to-noise ratio feature; (3) Adjust the phase correction parameter in the multi-band fusion processing according to the phase continuity feature.

[0012] In a possible implementation manner of the first aspect, the adjusting the number of frequency bands or the bandwidth range in the frequency band division rule according to the frequency band energy balance feature includes: According to the frequency band energy balance feature of the target filtering signal, obtain the energy proportion sequence of each sub-band and its standard deviation; Judge whether the standard deviation exceeds the preset balance threshold. If it exceeds, then perform the following frequency band adjustment operations: (a) Determine the upper limit of the number of frequency bands allowed to be adjusted and the minimum bandwidth constraint value according to the number of frequency bands and the bandwidth range of each frequency band in the current frequency band division rule; (b) Screen out the first target frequency band corresponding to the highest energy proportion value and the second target frequency band corresponding to the lowest energy proportion value from the energy proportion sequence; (c) Split the first target frequency band: Based on the minimum bandwidth constraint value, calculate the maximum number of sub - frequency bands that can be split, and divide this frequency band into multiple sub - frequency bands according to the energy - ratio gradient, ensuring that the bandwidth of each sub - frequency band is not less than the minimum bandwidth constraint value; (d) Merge the second target frequency band: Identify the energy ratio of its adjacent frequency band. If the energy ratio of the adjacent frequency band is lower than the preset merging threshold, then merge the second target frequency band with the adjacent frequency band into a new frequency band, and the bandwidth of the new frequency band is the sum of the two; (e) Update the number of frequency bands and the bandwidth range of each frequency band in the frequency - band division rule, and recalculate the standard deviation of the adjusted energy - ratio sequence of each sub - band; Iteratively execute steps (a) to (e) until the standard deviation is lower than the equilibrium threshold or reaches the upper limit of the number of frequency bands, and output the adjusted frequency - band division rule.

[0013] In a possible implementation manner of the first aspect, the adjustment of the calculation rule of the dynamic weight factor in the adaptive filtering operation according to the time - domain signal - to - noise ratio feature includes: According to the time - domain signal - to - noise ratio feature of the target filtering signal, obtain the ratio of the average energy of the signal - dominant time - domain window to the average energy of the noise time - domain window; Map the ratio to a filtering intensity level, where the filtering intensity level is negatively correlated with the ratio; Based on the filtering intensity level, adjust the proportional relationship between the noise - energy weight coefficient and the signal - smoothness weight coefficient in the dynamic weight - factor calculation formula, specifically including: (f) When the filtering intensity level is high, increase the proportion of the noise - energy weight coefficient so that its proportion in the dynamic weight factor is increased to the preset upper limit value; (g) When the filtering intensity level is low, increase the proportion of the signal - smoothness weight coefficient so that its proportion in the dynamic weight factor is increased to the preset lower limit value; Substitute the adjusted weight - coefficient ratio into the dynamic weight - factor calculation formula to generate a new dynamic weight factor; Update the filtering - parameter set of the adaptive filtering operation according to the new dynamic weight factor and apply it to the recursive filtering process of the next - round sub - band signal.

[0014] For example, in a possible implementation manner of the first aspect, the adjustment of the phase - correction parameter in the multi - band fusion processing according to the phase - continuity feature includes: According to the phase - continuity feature of the target filtering signal, obtain the mean value and variance of the phase difference of the adjacent - band signal components in the overlapping time - domain window; Determine the overall phase offset based on the mean of the phase differences, and determine the local phase fluctuation intensity based on the variance; Perform the following phase correction parameter adjustment operations: (h) If the mean of the phase differences exceeds the first correction threshold, calculate the time-domain delay compensation amount for each frequency band according to the overall phase offset. For the non-integer sample point shift requirement, use polynomial interpolation to generate a continuous time-domain signal, and achieve phase alignment through resampling; (i) If the variance exceeds the second correction threshold, insert interpolation points within the overlapping time-domain window, and generate a smooth transition curve based on the interpolated phase sequence to eliminate local phase mutations; Update the time-domain delay amount or the interpolation point density in the phase correction parameters, and recalculate the mean and variance of the phase differences; Iteratively adjust the parameters until both the mean and variance of the phase differences are lower than the corresponding thresholds, and output the optimized set of phase correction parameters.

[0015] On the other hand, an embodiment of the present invention further provides an adaptive signal filtering system based on multi-band fusion, including a processor and a machine-readable storage medium. The machine-readable storage medium is connected to the processor. The machine-readable storage medium is used to store programs, instructions, or codes, and the processor is used to execute the programs, instructions, or codes in the machine-readable storage medium to implement the above method.

[0016] Based on the above aspects, the embodiments of the present invention can effectively handle complex multi-band signals containing original signal components of different frequency bands. By using frequency band decomposition, the multi-band signal is subdivided into multiple sub-band signals, enabling subsequent filtering operations to be carried out according to the characteristics of each sub-band, avoiding the limitations of a single filtering strategy in processing signals of different frequency bands. The adaptive filtering operation dynamically adjusts the filtering coefficients according to the noise characteristics of the sub-band signals, and can more accurately remove the noise within each sub-band, improving the pertinence and effectiveness of filtering. The multi-band fusion processing integrates the filtered sub-band signals into the target filtered signal, ensuring the integrity and coherence of the signal. And dynamically adjusting the segmentation strategy of the frequency band decomposition processing or the dynamic adjustment rule of the adaptive filtering operation according to the signal quality evaluation result of the target filtered signal makes the entire filtering process have self-adaptability and self-optimization capabilities, and can optimize the filtering effect in real time according to the actual signal situation, significantly improving the quality and stability of the filtered signal. Description of the Drawings

[0017] Figure 1 It is a schematic flowchart of the execution process of the adaptive signal filtering method based on multi-band fusion provided by the embodiment of the present invention.

[0018] Figure 2It is a schematic diagram of exemplary hardware and software components of an adaptive signal filtering system based on multi - band fusion provided by an embodiment of the present invention. Detailed implementation manners

[0019] The present invention will be specifically described below in conjunction with the accompanying drawings of the specification. Figure 1 It is a schematic flowchart of an adaptive signal filtering method based on multi - band fusion provided by an embodiment of the present invention. The adaptive signal filtering method based on multi - band fusion will be introduced in detail below.

[0020] Step S110: Obtain a multi - band signal set of the target environment, where the multi - band signal set includes original signal components of different frequency bands.

[0021] In the field of signal processing, the target environment can be a communication network environment. For a communication network environment, a broadband antenna is often used as a signal acquisition device. A broadband antenna has a relatively wide frequency response range and can receive wireless signals of different frequency bands. Its working principle is based on electromagnetic induction. When an external electromagnetic wave signal acts on the antenna, an induced current will be generated inside the antenna. By detecting and processing the induced current, it is converted into an electrical signal for output. Among them, when the signal acquisition device collects signals, it needs to follow certain sampling rules. Sampling is the process of converting a continuous analog signal into a discrete digital signal. The core of the sampling rule is the selection of the sampling frequency. According to the Nyquist sampling theorem, the sampling frequency must be at least twice the highest frequency of the signal to ensure the complete acquisition of signal information. Assuming that the highest frequency of the signal in the target environment is fmax, then the sampling frequency fs should satisfy fs≥2fmax. In practical applications, to improve the accuracy and reliability of sampling, a sampling frequency higher than the theoretical value is usually selected.

[0022] During the sampling process, it is also necessary to consider the sampling time interval and the sampling duration. The sampling time interval Δt is equal to the reciprocal of the sampling frequency, that is, Δt = 1 / fs. The sampling duration T is determined according to actual requirements and the characteristics of the signal. If the signal is continuously changing, it is necessary to collect for a long enough time to capture the complete characteristics of the signal; if the signal is a short - term burst signal, the sampling duration can be relatively short.

[0023] The collected multi - band signal set is composed of original signal components of multiple different frequency bands. These original signal components may come from different signal sources, and each signal source generates signals with unique frequency ranges and characteristics.

[0024] Step S120: Perform frequency - band decomposition processing on the multi - band signal set to obtain multiple sub - band signals.

[0025] Step S121: Determine the target decomposition level according to the initial frequency band division rule, and during the signal processing process, preferentially perform dynamic iterative adjustment on the number of frequency bands and the bandwidth range of the target decomposition level based on the spectral characteristics of the multi-band signal set. Among them, if the frequency band energy balance characteristic after the dynamic iterative adjustment does not meet the preset threshold, further optimize the frequency band division rule according to the frequency band energy ratio.

[0026] The initial frequency band division rule is the basis for frequency band decomposition processing and is formulated based on the preliminary understanding of the target environmental signal and past experience. For example, for a communication network environment, a rough frequency band division scheme can be preset according to the frequency band ranges of different communication protocols. Suppose the initial frequency band division rule divides the entire frequency range into n frequency bands, and each frequency band has a specific bandwidth range.

[0027] The target decomposition level is determined based on the initial frequency band division rule, which stipulates the degree of detail for decomposing the multi-band signal set. The higher the target decomposition level, the more detailed the decomposition of the signal, the larger the number of frequency bands, and the narrower the bandwidth of each frequency band. After determining the target decomposition level, it is necessary to perform spectral analysis on the multi-band signal set to obtain its spectral characteristics.

[0028] Spectral analysis is the process of converting a time-domain signal into a frequency-domain signal. Through spectral analysis, the distribution of the signal at different frequencies can be clearly understood. Common spectral analysis methods include Fourier transform, fast Fourier transform, etc. Taking the fast Fourier transform (FFT) as an example, it is an efficient Fourier transform algorithm that can quickly convert a time-domain signal into a frequency-domain signal. Suppose the collected multi-band signal set is x(t), and its frequency-domain representation X(f) can be obtained through the FFT algorithm. The spectral characteristics include information such as the frequency components of the signal, the frequency range, the density of the frequency distribution, and the energy distribution of each frequency component.

[0029] Based on the spectral characteristics, perform dynamic iterative adjustment on the number of frequency bands and the bandwidth range of the target decomposition level. Specifically, if the spectral analysis result shows that the signal energy in some frequency bands is concentrated and the bandwidth of this frequency band is relatively wide, it may be necessary to further subdivide this frequency band to better process the signal; if the signal energy in some frequency bands is weak and the bandwidth is narrow, it may be possible to merge it with adjacent frequency bands to simplify the processing process. After each adjustment, it is necessary to calculate the frequency band energy balance characteristic.

[0030] The energy balance feature of frequency bands is used to measure the degree of uniformity of the energy distribution among different frequency bands. One way to calculate the energy balance feature of frequency bands is to first calculate the energy of each frequency band. Suppose the signal of the i-th frequency band is xi(t), and its energy Ei can be obtained by integrating the square of the signal of this frequency band over a period of time, that is, Ei = ∫[xi(t)]²dt. Then, calculate the average value Eavg and the standard deviation σ of the energies of all frequency bands. The energy balance feature of frequency bands can be represented by the ratio σ / Eavg of the standard deviation to the average value. If this ratio is small, it indicates that the energy distribution is relatively uniform; if the ratio is large, it indicates that the energy distribution is non-uniform.

[0031] The preset threshold is a standard value set according to actual requirements and experience. If the energy balance feature of frequency bands after dynamic iterative adjustment does not meet the preset threshold, that is, the situation of non-uniform energy distribution is relatively serious, it is necessary to further optimize the frequency band division rule according to the energy proportion of frequency bands. The energy proportion of a frequency band refers to the proportion of the energy of each frequency band in the total energy. Suppose the total energy is Etotal and the energy of the i-th frequency band is Ei, then the energy proportion Pi of this frequency band is Pi = Ei / Etotal.

[0032] For the frequency bands with a high energy proportion, they can be further subdivided to improve the processing accuracy of the signals in this frequency band; for the frequency bands with a low energy proportion, they can be merged with adjacent frequency bands to improve the uniformity of the energy distribution. In the process of optimizing the frequency band division rule, it is necessary to continuously repeat the steps of dynamic iterative adjustment and calculation of the energy balance feature of frequency bands until the energy balance feature of frequency bands meets the preset threshold.

[0033] Step S122: Based on the target decomposition level, perform overlapping segmentation processing on each original signal component in the multi-band signal set to generate initial segmented signals.

[0034] Step S1221: Determine the sampling rate of the original signal component, and calculate the number of sample points corresponding to the segmentation window according to the minimum bandwidth range in the target decomposition level and the sampling rate, so that the duration of the segmentation window is proportional to the reciprocal of the minimum bandwidth.

[0035] The sampling rate of the original signal component is an important parameter determined in the signal acquisition stage. It is related to the fineness of signal discretization and must satisfy the Nyquist sampling theorem to ensure the complete acquisition of signal information. When performing overlapping segmentation processing, it is necessary to calculate the number of sample points corresponding to the segmentation window according to the minimum bandwidth range in the target decomposition level and the sampling rate.

[0036] The minimum bandwidth range reflects the minimum granularity of the target decomposition level for signal subdivision processing. According to the basic principle of signal processing, the duration of the segmentation window is proportional to the reciprocal of the minimum bandwidth, that is, the segmentation window duration T and the minimum bandwidth range Δfmin satisfy the relationship T = 1 / Δfmin. This is because in the frequency domain, a narrower bandwidth corresponds to a longer time window. To accurately analyze the signal in a frequency band with a smaller bandwidth, a longer segmentation window corresponding to it is required.

[0037] When specifically calculating the number of sample points, the time length T is converted into the number of sample points N in combination with the sampling rate fs. Since the sampling rate fs represents the number of samples collected per second, the number of sample points N = fs×T = fs / Δfmin. The number of sample points calculated in this way can ensure that the duration of the segmentation window is proportional to the reciprocal of the minimum bandwidth, so that the segmentation window can effectively cover the signal in frequency bands with different bandwidths.

[0038] Step S1222: Determine the number of overlapping sample points between adjacent segmentation windows based on the number of frequency bands at the target decomposition level, where the number of overlapping sample points is in an inverse proportional relationship with the number of frequency bands.

[0039] After determining the number of sample points corresponding to the segmentation window, it is necessary to determine the number of overlapping sample points between adjacent segmentation windows. There is an overlap between adjacent segmentation windows to avoid information loss or discontinuity during the signal segmentation process. The determination of the number of overlapping sample points is related to the number of frequency bands at the target decomposition level, and they are in an inverse proportional relationship.

[0040] When the number of frequency bands is large, it means that the signal is more finely divided, the bandwidth of each frequency band is relatively narrow, and the signal change may be more complex. At this time, if the number of overlapping sample points between adjacent segmentation windows is too large, it will increase the calculation amount and processing complexity, and may also introduce unnecessary redundant information. Therefore, when the number of frequency bands is large, appropriately reducing the number of overlapping sample points can improve the processing efficiency.

[0041] On the contrary, when the number of frequency bands is small, the bandwidth of each frequency band is relatively wide, and the signal change is relatively gentle. To ensure that the signal characteristics can be accurately captured during the segmentation process and avoid information loss, it is necessary to increase the number of overlapping sample points between adjacent segmentation windows. This can make there be more information overlap between adjacent segmentation windows, so as to better connect and transition different segmented signals.

[0042] When specifically determining the number of overlapping sample points, to avoid negative values and out-of-range situations, let the number of frequency bands at the target decomposition level be \(n\), the number of sample points in the segmentation window be \(N\), and the number of overlapping sample points be \(M\). First, set a proportional coefficient function \(\alpha(n)\) related to the number of frequency bands. This function satisfies \(\alpha(n)=\max(0,\min(1,a - b\times n))\), where \(a\) and \(b\) are constants, and \(a>0\), \(b>0\). Then calculate the number of overlapping sample points \(M = \alpha(n)\times N\). In this way, the number of overlapping sample points can be dynamically adjusted according to the number of frequency bands at the target decomposition level, while ensuring that the number of overlapping sample points is within a reasonable range to achieve the best segmentation processing effect.

[0043] Step S1223: Generate an equally spaced sliding window sequence based on the number of sample points in the segmentation window and the number of overlapping sample points.

[0044] After determining the number of sample points \(N\) in the segmentation window and the number of overlapping sample points \(M\) between adjacent segmentation windows, an equally spaced sliding window sequence can be generated. The sliding window sequence is a tool for segmenting signals. It can slide on the signal sequence at a set interval and intercept a segment of the signal as a segmentation window each time.

[0045] The specific process of generating the sliding window sequence is as follows: First, take the starting point of the signal sequence as the starting position of the first sliding window, and intercept a segment of the signal according to the number of sample points \(N\) in the segmentation window to form the first segmentation window. Then, determine the starting position of the next sliding window according to the number of overlapping sample points \(M\). The number of sample points that the starting position of the next sliding window moves backward relative to the starting position of the current sliding window is \(N - M\). Repeat this process until the sliding window covers the entire signal sequence, thus generating a series of equally spaced sliding windows.

[0046] The signal segment intercepted by each sliding window is the initial segmented signal. These initial segmented signals will serve as the basis for subsequent processing. By further analyzing and processing each initial segmented signal, the characteristics of the signal can be better extracted and noise can be removed.

[0047] Step S123: Perform a frequency domain transformation operation on the initial segmented signal to obtain the frequency domain energy distribution characteristics.

[0048] Performing a frequency domain transformation operation on the initial segmented signal is to convert the time domain signal into a frequency domain signal, so as to obtain the energy distribution characteristics of the signal in the frequency domain. Commonly used methods for the frequency domain transformation operation include Fourier transform and fast Fourier transform (FFT).

[0049] Taking the Fast Fourier Transform as an example, it is an efficient Fourier transform algorithm that can quickly convert a time-domain signal into a frequency-domain signal. Suppose the initial segmented signal is x[n], where n = 0, 1, …, N - 1, and N is the number of sample points in the segmentation window. Through the FFT algorithm, its frequency-domain representation X[k] can be obtained, where k = 0, 1, …, N - 1.

[0050] The frequency-domain energy distribution characteristic reflects the energy distribution of the signal at different frequencies. For the frequency-domain signal X[k], the energy of each frequency component can be obtained by calculating the square of the amplitude of that frequency component. That is, the energy Ek of the k-th frequency component is Ek = |X[k]|². Arranging the energies of all frequency components in frequency order gives the frequency-domain energy distribution characteristic.

[0051] The frequency-domain energy distribution characteristic can intuitively show the energy concentration of the signal at different frequencies. For example, if the energy is high in a certain frequency range, it means that the signal components in that frequency range are strong; if the energy is low in a certain frequency range, it means that the signal components in that frequency range are weak. By analyzing the frequency-domain energy distribution characteristic, the frequency characteristics of the signal can be better understood.

[0052] Step S124: Map the initial segmented signal to the corresponding target frequency band according to the matching degree between the frequency-domain energy distribution characteristic and the frequency band division rule, and generate the sub-band signal.

[0053] After obtaining the frequency-domain energy distribution characteristic and determining the frequency band division rule, it is necessary to map the initial segmented signal to the corresponding target frequency band according to their matching degree, so as to generate the sub-band signal.

[0054] The frequency band division rule stipulates the frequency ranges of different frequency bands. For example, suppose the frequency band division rule divides the entire frequency range into m frequency bands, and the frequency range of the j-th frequency band is [fj1, fj2], where j = 1, 2, …, m.

[0055] For each frequency component in the frequency-domain energy distribution characteristic, determine the frequency band to which it belongs. If the frequency fk of the k-th frequency component satisfies fj1 ≤ fk ≤ fj2, then map the time-domain signal component corresponding to that frequency component to the j-th target frequency band.

[0056] The specific mapping process is as follows: First, according to the relationship between the frequency-domain signal X[k] and the time-domain signal x[n], convert the frequency-domain signal back to the time-domain signal. This can be achieved through the Inverse Fast Fourier Transform (IFFT). Then, extract the time-domain signal components belonging to the same target frequency band and form the sub-band signal of that target frequency band.

[0057] For example, for the j-th target frequency band, extract the time-domain signal components corresponding to all frequency components that satisfy fj1 ≤ fk ≤ fj2, and obtain the sub-band signal yj[n] of this frequency band after IFFT transformation. In this way, the initial segmented signal is mapped to the corresponding target frequency band, generating multiple sub-band signals. These sub-band signals will be used as the input for subsequent adaptive filtering operations.

[0058] Step S130: Perform an adaptive filtering operation on each of the sub-band signals to generate a corresponding filtered sub-band signal; wherein, the adaptive filtering operation includes dynamically adjusting the filtering coefficients according to the noise characteristics of the sub-band signals.

[0059] Step S131: Extract the time-domain waveform features and frequency-domain energy features of the sub-band signals.

[0060] The sub-band signals contain rich information. By extracting their time-domain waveform features and frequency-domain energy features, the characteristics of the sub-band signals can be better understood, providing a basis for the adaptive filtering operation.

[0061] The time-domain waveform features reflect the changes of the sub-band signals in the time domain. Common time-domain waveform features include the amplitude, period, rising edge time, falling edge time, etc. of the signal. There are various methods for extracting time-domain waveform features. For example, the amplitude characteristics of the signal can be described by calculating statistical quantities such as the maximum value, minimum value, and average value of the signal; the period of the signal can be determined by detecting the zero-crossing points of the signal.

[0062] The frequency-domain energy features reflect the energy distribution of the sub-band signals in the frequency domain. As mentioned above, the frequency-domain signal can be obtained by performing a frequency-domain transformation (such as FFT) on the time-domain signal, and then calculating the energy of each frequency component to obtain the frequency-domain energy features. The frequency-domain energy features can help understand the intensity of different frequency components in the signal, and judge whether there is noise in the signal and the frequency range of the noise.

[0063] Step S132: Determine the signal stationarity index according to the time-domain waveform features, and determine the noise energy ratio according to the energy ratio of the noise-dominated frequency band in the frequency-domain energy features, where the noise-dominated frequency band is identified by a preset noise reference interval or a dynamic noise estimation method based on the minimum statistic method, specifically including extracting the background noise energy distribution in the signal silent segment and iteratively updating the noise energy threshold.

[0064] The signal stationarity index is used to measure the stability of the sub-band signals in time. One method for determining the signal stationarity index according to the time-domain waveform features is to comprehensively consider the time-domain variance sequence of the signal and the differential mean between adjacent sampling points.

[0065] First, calculate the time-domain variance sequence of the sub-band signal. The time-domain variance reflects the degree of fluctuation of the signal over a period of time. Let the sub-band signal be x[n], where n = 0, 1, …, N - 1. The signal is divided into multiple time periods, and each time period contains L sample points. For the i-th time period, its variance σi² can be calculated by the following formula: σi² = (1 / L) × ∑[x[n] - μi]², where n ranges from iL to (i + 1)L - 1, and μi is the average value of the signal within this time period.

[0066] Then, calculate the differential mean between adjacent sampling points. The differential mean reflects the degree of change of the signal between adjacent sampling points. Let the differential sequence be d[n] = x[n + 1] - x[n], where n = 0, 1, …, N - 2. Then the differential mean md can be calculated by the following formula: md = (1 / (N - 1)) × ∑d[n].

[0067] Determine the first stationary factor according to the fluctuation amplitude of the time-domain variance sequence. The fluctuation amplitude can be represented by calculating the standard deviation of the time-domain variance sequence. Let the standard deviation of the time-domain variance sequence be σv. Then the first stationary factor s1 can be expressed as s1 = 1 / (1 + σv). The larger the value of the first stationary factor, the smaller the fluctuation of the signal and the better the stationarity.

[0068] Determine the second stationary factor according to the absolute value of the differential mean. The larger the absolute value of the differential mean, the greater the change of the signal between adjacent sampling points and the worse the stationarity. Therefore, the second stationary factor s2 can be expressed as s2 = 1 / (1 + |md|).

[0069] Perform standardized transformation on the first stationary factor and the second stationary factor respectively to map them to a unified ratio interval. The standardized transformation can adopt the method of linear transformation. For example, map the first stationary factor and the second stationary factor to the interval [0, 1]. Let the standardized first stationary factor be s1', and the standardized second stationary factor be s2'.

[0070] Dynamically allocate the weight coefficients of the first stationary factor and the second stationary factor after standardized conversion according to the frequency band position of the sub-band signal. Signals in different frequency bands may have different characteristics, so it is necessary to dynamically adjust the weight coefficients according to the frequency band position. Suppose the sub-band signal is in the low-frequency band. Signals in the low-frequency band are usually relatively stable, and the weight of the first stationary factor may be relatively large. If it is in the high-frequency band, the signal may change more violently, and the weight of the second stationary factor may be larger. Let the weight coefficient of the first stationary factor be w1, the weight coefficient of the second stationary factor be w2, and w1 + w2 = 1. A weight allocation function can be established according to the frequency band position. For example, the weight coefficients can be determined according to the center frequency f0 of the frequency band. Suppose the frequency range of the low-frequency band is from f_low to f_mid, and the frequency range of the high-frequency band is from f_mid to f_high. When f0 is in the low-frequency band, w1 can be set to a relatively large value, such as w1 = 0.7 and w2 = 0.3. When f0 is in the high-frequency band, w1 can be set to a relatively small value, such as w1 = 0.3 and w2 = 0.7. Then, the standardized first stationary factor and the second stationary factor are weighted and summed according to the weight coefficients to generate the signal stationarity index S, that is, S = w1×s1' + w2×s2'.

[0071] Next, determine the proportion of noise energy. There are two ways to identify the noise-dominated frequency band. One is through a preset noise reference interval. The preset noise reference interval is determined based on past experience or prior knowledge of the target environmental noise characteristics. The frequency components within this reference interval are considered to be the noise-dominated frequency bands. The other way is a dynamic noise estimation method based on the minimum statistic method. This method first extracts the background noise energy distribution in the signal silent segment. The signal silent segment refers to the time period when there are no effective signal components in the signal and only background noise exists. In the signal silent segment, the signal is transformed into the frequency domain to obtain the frequency domain energy distribution of the background noise. Then, the noise energy threshold is iteratively updated. The initial noise energy threshold can be determined according to the statistical characteristics of the background noise energy distribution. For example, the average value of the background noise energy is taken as the initial threshold. During the subsequent signal processing, continuously monitor the frequency domain energy distribution of the signal. When the energy of a certain frequency component is lower than the current noise energy threshold, this frequency component is considered a noise component. As the signal changes, continuously update the noise energy threshold to adapt to different noise environments.

[0072] After determining the noise-dominated frequency band, calculate the proportion of the energy of the noise-dominated frequency band. First, calculate the total energy E_noise of all frequency components within the noise-dominated frequency band, and then calculate the total energy E_total of the entire sub-band signal. The noise energy proportion P_noise = E_noise / E_total.

[0073] Step S133: Construct a dynamic weight factor based on the signal stationarity index and the noise energy ratio, and adjust the preset filter parameter set according to the dynamic weight factor.

[0074] The dynamic weight factor is used to comprehensively consider the stationarity of the signal and the noise situation to more flexibly adjust the filter parameters. Let the dynamic weight factor be W, which is a function of the signal stationarity index S and the noise energy ratio P_noise. A simple functional relationship can be constructed, for example, W = a×S + b×(1 - P_noise), where a and b are weight coefficients and a + b = 1. The values of a and b can be determined according to the specific application scenario and the emphasis on signal stationarity and noise suppression. If more emphasis is placed on signal stationarity, the value of a can be relatively large; if more emphasis is placed on noise suppression, the value of b can be relatively large.

[0075] The preset filter parameter set is a set of filter parameters preset before performing the adaptive filtering operation, such as the coefficients of the filter, etc. Adjust the preset filter parameter set according to the dynamic weight factor W. Assume the preset filter parameter set is F_set, and the adjusted filter parameter set F_set' can be calculated in the following way: for each parameter f_i in the filter parameter set, f_i' = f_i×W, where f_i' is the adjusted parameter. In this way, the filter parameters are dynamically adjusted according to the actual situation of the signal, enabling the filter to better adapt to different signal characteristics and noise environments.

[0076] Step S134: Call the filter parameter set to perform recursive filtering on the sub-band signal, and detect the pole positions of the filter in real time during the filtering process to constrain its stability. If it is detected that the poles exceed the unit circle range, reset the filter parameters to the preset safety values to generate an initial filtered signal.

[0077] Recursive filtering is a commonly used filtering method that uses the current output of the filter, the past output, and the current input to calculate the current output. Let the sub-band signal be x[n], and the filter parameter set be F_set'. The output y[n] of the recursive filtering can be calculated through the following recursive relationship: y[n] = ∑(a_i×x[n - i]) + ∑(b_j×y[n - j]), where a_i and b_j are the coefficients in the filter parameter set.

[0078] During the filtering process, in order to ensure the stability of the filter, it is necessary to monitor the pole positions of the filter. For the problem that it is difficult to analytically solve the pole positions of high-order filters in real time, an approximate method can be used to judge the stability of the filter. When designing the filter, the structure and parameters of the filter can be analyzed in advance to determine the approximate range of the filter poles. During the recursive filtering process, the stability of the filter can be indirectly judged by monitoring the amplitude and change trend of the filter output.

[0079] For example, set an amplitude threshold A. If the amplitude of the filter output exceeds this threshold, or the amplitude of the output shows a trend of rapid growth, it is considered that the filter may be unstable. When it is judged that the filter may be unstable, reset the filtering parameters to the preset safe values. The preset safe values are a set of pre-set filtering parameters that can ensure the stability of the filter. After resetting the filtering parameters to the preset safe values, perform the recursive filtering process again until the filter output is stable, generating the initial filtered signal y_initial[n].

[0080] Step S135: Perform short-time Fourier transform processing on the initial filtered signal to generate a time-frequency energy distribution matrix.

[0081] Short-time Fourier transform (STFT) is a method for converting a time-domain signal into a time-frequency domain signal, which can show the energy distribution of the signal at different times and frequencies. When performing short-time Fourier transform processing on the initial filtered signal y_initial[n], first select a suitable window function, such as a Hann window, a Hamming window, etc. The role of the window function is to window the signal in the time domain, making the signal approximately stationary within a local range.

[0082] Let the window function be w[n] with a length of L. Multiply the initial filtered signal y_initial[n] by the window function w[n] to obtain the windowed signal y_windowed[n]. Then perform Fourier transform on the windowed signal to obtain the frequency-domain representation at each time point. Let the time point be t, and perform Fourier transform on y_windowed[n+t] to obtain Y(t, f), where f represents the frequency.

[0083] Combining the frequency-domain representations at different time points, the time-frequency energy distribution matrix M is obtained. Each element M(t, f) of the matrix M represents the signal energy at time t and frequency f. Through the time-frequency energy distribution matrix, the energy change of the signal at different times and frequencies can be intuitively observed.

[0084] Step S136: According to the frequency band division rule of the current target decomposition level, extract the characteristic frequency bands related to noise in the time-frequency energy distribution matrix.

[0085] The frequency band division rule at the current target decomposition level stipulates the frequency ranges of different frequency bands. According to this rule, the characteristic frequency bands related to noise are extracted from the time-frequency energy distribution matrix M. The noise-dominated frequency bands have been identified previously through a preset noise reference interval or a dynamic noise estimation method based on the minimum statistic method. In the time-frequency energy distribution matrix, find the frequency ranges corresponding to the noise-dominated frequency bands.

[0086] Assume that the frequency range of the noise-dominated frequency band is from f_noise1 to f_noise2. Extract the elements with frequency ranges between f_noise1 and f_noise2 from the time-frequency energy distribution matrix M to form the characteristic frequency band matrix M_noise related to noise. The characteristic frequency band matrix M_noise shows the energy distribution of noise at different times and frequencies.

[0087] Step S137: Calculate the ratio of the average energy within the characteristic frequency band to the average energy of the signal-dominated frequency band. If the ratio exceeds the preset noise threshold, it is determined that there is residual noise energy, and the set of filtering parameters is iteratively updated until the residual noise energy is lower than the preset threshold, and the final filtered sub-band signal is output.

[0088] First, calculate the average energy E_noise_avg within the characteristic frequency band. Sum all the elements in the characteristic frequency band matrix M_noise, and then divide by the number of elements in the matrix to obtain the average energy E_noise_avg = ∑M_noise(t, f) / N_noise, where N_noise is the number of elements in the characteristic frequency band matrix M_noise.

[0089] The signal-dominated frequency band refers to the frequency band with stronger energy and containing the main information in the signal. According to the frequency band division rule, determine the frequency range of the signal-dominated frequency band, extract the corresponding elements of the signal-dominated frequency band from the time-frequency energy distribution matrix M to form the signal-dominated frequency band matrix M_signal. Calculate the average energy E_signal_avg = ∑M_signal(t, f) / N_signal within the signal-dominated frequency band, where N_signal is the number of elements in the signal-dominated frequency band matrix M_signal.

[0090] Calculate the ratio R = E_noise_avg / E_signal_avg of the average energy within the characteristic frequency band to the average energy of the signal-dominated frequency band. The preset noise threshold is a pre-set value used to determine whether there is residual noise energy. If the ratio R exceeds the preset noise threshold, it indicates that there is residual noise energy.

[0091] When it is determined that there is residual noise energy, the set of filtering parameters is iteratively updated. The dynamic weight factor W can be adjusted according to the magnitude of the ratio R, and then the set of filtering parameters F_set' is adjusted. For example, if the ratio R is large, it indicates that the noise energy is relatively high, and it is necessary to increase the part related to noise suppression in the dynamic weight factor, that is, increase the value of b and decrease the value of a. Then, the set of filtering parameters is recalculated according to the adjusted dynamic weight factor, and the steps of recursive filtering, short-time Fourier transform, extracting the characteristic frequency band, and calculating the ratio are performed again until the ratio R is lower than the preset noise threshold. At this time, the final filtered sub-band signal y_final[n] is output.

[0092] Step S140: Perform multi-band fusion processing on the filtered sub-band signal to generate a target filtered signal.

[0093] Step S141: Perform signal reconstruction processing on each of the filtered sub-band signals to generate a time-domain signal component of the same length as the original signal component.

[0094] The filtered sub-band signal is a signal after adaptive filtering processing, usually obtained by processing in the frequency domain. In order to fuse these filtered sub-band signals, they need to be converted back to the time domain and a time-domain signal component of the same length as the original signal component is generated.

[0095] Perform signal reconstruction processing on each filtered sub-band signal y_final[n] using the inverse short-time Fourier transform (ISTFT). The inverse short-time Fourier transform is the inverse process of the short-time Fourier transform, which converts the time-frequency domain signal back to the time-domain signal. When performing the inverse short-time Fourier transform, the same window function w[n] as the short-time Fourier transform needs to be used.

[0096] Let the time-frequency representation of the filtered sub-band signal be Y_final(t, f), and convert it back to the time-domain signal y_reconstructed[n] through the inverse short-time Fourier transform. During the reconstruction process, attention needs to be paid to the length of the signal to ensure that the generated time-domain signal component is of the same length as the original signal component. The matching of the signal length can be achieved by performing appropriate truncation or zero-padding operations on the reconstructed signal.

[0097] Step S142: Dynamically determine the band weight coefficient of each time-domain signal component according to the target decomposition level and the signal quality score.

[0098] The target decomposition level specifies the number of frequency bands into which the signal is decomposed and the bandwidth range of each frequency band. Different frequency bands may have different importance in the signal, so it is necessary to dynamically determine the band weight coefficient of each time-domain signal component according to the target decomposition level and the signal quality score.

[0099] The signal quality score is an index for evaluating the quality of each time-domain signal component. The signal quality score can be calculated based on indicators such as the signal-to-noise ratio and distortion degree of the signal. For example, for a certain time-domain signal component, its signal-to-noise ratio SNR can be calculated. The higher the signal-to-noise ratio, the better the signal quality.

[0100] Suppose the target decomposition level divides the signal into m frequency bands, and the time-domain signal component of the j-th frequency band is y_reconstructed_j[n], and its signal quality score is Q_j. According to the target decomposition level and the signal quality score, the frequency band weight coefficient w_j of each time-domain signal component is determined. A functional relationship can be established, for example, w_j = k×Q_j / ∑Q_i, where k is a normalization coefficient to ensure that the sum of all frequency band weight coefficients is 1, and i ranges from 1 to m. In this way, the better the signal quality of the time-domain signal component, the larger its frequency band weight coefficient, and the greater the proportion it occupies in the subsequent fusion process.

[0101] Step S143: Perform time-domain alignment and weighted superposition processing on the time-domain signal components based on the frequency band weight coefficients to generate an initial fusion signal.

[0102] Before performing the weighted superposition processing, it is necessary to perform time-domain alignment on the time-domain signal components. Since signals in different frequency bands may have different time delays during processing, it is necessary to align them in the time domain to ensure the correct fusion of the signals.

[0103] The method of time-domain alignment can determine the time delay of the signal by detecting the characteristic points of the signal, such as zero-crossing points, peak points, etc. Then, translation operations are performed on the time-domain signal components with different time delays to align their characteristic points.

[0104] Suppose the time-domain signal component of the j-th frequency band after time-domain alignment is y_aligned_j[n], and the frequency band weight coefficient is w_j. All time-domain signal components are weighted and superposed according to the frequency band weight coefficients to generate an initial fusion signal y_initial_fusion[n]=∑(w_j×y_aligned_j[n]), where j ranges from 1 to m.

[0105] Step S144: Extract the phase difference characteristics of adjacent frequency band signal components in the initial fusion signal, and dynamically calculate the phase correction amount of each frequency band according to the phase difference characteristics.

[0106] There may be a phase difference between adjacent frequency band signal components, and this phase difference will affect the fusion effect of the signal, resulting in signal distortion. Therefore, it is necessary to extract the phase difference characteristics of adjacent frequency band signal components in the initial fusion signal, and dynamically calculate the phase correction amount of each frequency band according to these characteristics.

[0107] Perform a frequency-domain transformation on the initial fusion signal y_initial_fusion[n] to obtain its frequency-domain representation Y_initial_fusion(f). According to the frequency band division rule, divide the frequency-domain representation into signal components of different frequency bands. For two adjacent frequency bands j and j+1, extract their phase information φ_j(f) and φ_j+1(f).

[0108] Calculate the phase difference of the signal components in adjacent frequency bands Δφ(f)=φ_j+1(f)-φ_j(f). Perform statistical analysis on the phase difference Δφ(f) to obtain the mean value Δφ_avg and variance Δφ_var of the phase difference.

[0109] Dynamically calculate the phase correction amount for each frequency band according to the phase difference characteristics. If the mean value of the phase difference Δφ_avg is large, it indicates that there is an overall phase shift, and an overall phase correction needs to be performed on the frequency band; if the variance of the phase difference Δφ_var is large, it indicates that there is local phase fluctuation, and a local phase correction needs to be performed on the frequency band.

[0110] Let the phase correction amount of the j-th frequency band be Δφ_correction_j(f), and the phase correction amount can be determined according to the mean value and variance of the phase difference. For example, for the overall phase shift, Δφ_correction_j(f)=-Δφ_avg can be set; for local phase fluctuation, the phase correction amount can be determined by interpolation or filtering according to the specific situation of the phase difference.

[0111] Step S145: Based on the phase correction amounts of each frequency band, perform phase alignment on the signal components of the initial fusion signal through time-domain delay compensation or interpolation processing. For the change in signal length caused by time-domain delay compensation, use the cyclic shift or edge symmetric interpolation method to keep the signal components the same length as the original signal, and generate a phase-aligned signal.

[0112] According to the phase correction amounts Δφ_correction_j(f) of each frequency band, perform phase alignment on the signal components of the initial fusion signal. Phase alignment can be achieved through time-domain delay compensation or interpolation processing.

[0113] Time-domain delay compensation is to calculate the corresponding time delay according to the phase correction amount, and then perform a delay operation on the signal component. Let the phase correction amount be Δφ_correction_j(f), and the corresponding time delay be Δt_j=Δφ_correction_j(f) / (2πf). Perform a delay operation on the signal component y_aligned_j[n] of the j-th frequency band to obtain the delayed signal component y_delayed_j[n].

[0114] Interpolation processing is to insert new sampling points into the signal to adjust the phase of the signal. Methods such as linear interpolation and polynomial interpolation can be used. For example, for two adjacent sampling points y_aligned_j[n] and y_aligned_j[n + 1], a new sampling point y_interpolated[n + 0.5] is inserted between them, and it is calculated through the linear interpolation formula y_interpolated[n + 0.5] = 0.5×(y_aligned_j[n] + y_aligned_j[n + 1]).

[0115] When performing time-domain delay compensation, it may cause a change in the signal length. To keep the signal components the same length as the original signal, methods such as circular shift or edge-symmetric interpolation can be used. Circular shift is to move a part of the signal to the other end of the signal, keeping the signal length unchanged; edge-symmetric interpolation is to insert symmetric sampling points at the edges of the signal to fill the gaps caused by the delay.

[0116] After the phase alignment process, a phase-aligned signal y_phase_aligned[n] is generated.

[0117] Step S146: Perform time-domain smoothing processing on the phase-aligned signal to eliminate the instantaneous mutation noise introduced by phase correction, and obtain the corrected initial fusion signal.

[0118] Instantaneous mutation noise may be introduced during the phase correction process, and these noises will affect the signal quality. To eliminate these noises, it is necessary to perform time-domain smoothing processing on the phase-aligned signal.

[0119] Time-domain smoothing processing can use methods such as moving average filtering and median filtering. Moving average filtering is to calculate the average value of adjacent sampling points for each sampling point of the signal as the smoothed value of this sampling point. Let the phase-aligned signal be y_phase_aligned[n], and the window length of moving average filtering be L, then the smoothed signal y_smoothed[n] = (1 / L)×∑y_phase_aligned[n + i], where i ranges from -(L - 1) / 2 to (L - 1) / 2.

[0120] Median filtering is to take the median value of adjacent sampling points for each sampling point of the signal as the smoothed value of this sampling point. Through time-domain smoothing processing, the instantaneous mutation noise introduced by phase correction is eliminated, and the corrected initial fusion signal y_corrected_fusion[n] is obtained.

[0121] Step S147: Generate the target filtering signal according to the corrected initial fusion signal.

[0122] The corrected initial fusion signal y_corrected_fusion[n] has undergone processes such as signal reconstruction, weighted superposition, phase alignment, and time-domain smoothing, removing noise and interference and ensuring the phase consistency of the signal. The corrected initial fusion signal is output as the target filtered signal, which has good quality, contains useful information in the original signal, and reduces the influence of noise and interference at the same time.

[0123] Step S150: According to the signal quality evaluation result of the target filtered signal, adjust the segmentation strategy of the frequency band decomposition process or the dynamic adjustment rule of the adaptive filtering operation, and output the optimized target filtered signal.

[0124] Step S151: Extract the time-domain signal-to-noise ratio feature, frequency-band energy balance feature, and phase continuity feature of the target filtered signal.

[0125] The time-domain signal-to-noise ratio feature reflects the ratio relationship between the useful signal and noise in the target filtered signal. The method for calculating the time-domain signal-to-noise ratio is to first divide the target filtered signal into a signal segment and a noise segment. The signal segment is the time period containing the useful signal, and the noise segment is the time period containing only noise. The signal segment and the noise segment can be divided by detecting features such as the amplitude and frequency of the signal.

[0126] Let the signal in the signal segment be s[n], and the signal in the noise segment be n[n]. The energy of the signal segment can be obtained by summing the squares of the signals in the signal segment, that is, the signal energy E_s = sum of the squares of s[n] within the signal segment; the energy of the noise segment is obtained by summing the squares of the signals in the noise segment, that is, the noise energy E_n = sum of the squares of n[n] within the noise segment. The time-domain signal-to-noise ratio SNR = the ratio of the signal energy E_s to the noise energy E_n.

[0127] The frequency-band energy balance feature is used to measure the degree of uniform energy distribution of the target filtered signal in each frequency band. First, the target filtered signal is decomposed into frequency bands to obtain the signal components in each frequency band. Let the target filtered signal be decomposed into m frequency bands, and the signal component in the j-th frequency band be s_j[n]. The energy of this frequency band E_j = sum of the squares of s_j[n] within the corresponding frequency band. Calculate the average value E_avg of the energies of all frequency bands = sum of the energies E_j of all frequency bands divided by the number of frequency bands m. Then calculate the deviation of the energy of each frequency band from the average value, and the sum of the squares of the deviations is divided by the number of frequency bands m to obtain the variance σ². The frequency-band energy balance feature can be represented by the variance σ². The smaller the variance, the more balanced the energy distribution.

[0128] The phase continuity feature reflects the degree of continuity of the phase change of the target filtered signal between different frequency bands. The frequency domain transformation is performed on the target filtered signal to obtain the frequency domain representation, and the frequency domain representation is divided into signal components of different frequency bands according to the frequency band division rule. For two adjacent frequency bands j and j+1, their phase information φ_j(f) and φ_j+1(f) are extracted. The phase difference Δφ(f)=φ_j+1(f)-φ_j(f) of the adjacent frequency band signal components is calculated. Statistical analysis is performed on the phase difference Δφ(f) to obtain the mean value Δφ_avg and variance Δφ_var of the phase difference. The mean value of the phase difference reflects the overall phase offset situation, and the variance reflects the local phase fluctuation situation.

[0129] Step S152: Calculate the first optimization index according to the time domain signal-to-noise ratio feature, calculate the second optimization index according to the frequency band energy balance feature, and calculate the third optimization index according to the phase continuity feature.

[0130] The first optimization index is related to the time domain signal-to-noise ratio feature of the target filtered signal. The higher the time domain signal-to-noise ratio, the more useful information relative to noise in the signal, and the better the signal quality. A certain transformation can be performed on the time domain signal-to-noise ratio SNR to obtain the first optimization index I_1. For example, a linear transformation I_1=k_1×SNR is adopted, where k_1 is a proportionality coefficient used to adjust the scale of the first optimization index.

[0131] The second optimization index is based on the frequency band energy balance feature. The frequency band energy balance feature is represented by the variance σ². The smaller the variance, the more balanced the energy distribution and the better the signal quality. The second optimization index I_2 can be obtained by transforming the variance σ². For example, an inverse proportional transformation I_2=k_2 / (1+σ²) is adopted, where k_2 is a proportionality coefficient, so that the smaller the variance, the larger the value of the second optimization index.

[0132] The third optimization index is calculated according to the phase continuity feature. The phase continuity feature includes the mean value Δφ_avg and variance Δφ_var of the phase difference. These two values can be combined to obtain the third optimization index I_3. For example, a weighted summation method I_3=k_31×(1 / (1+|Δφ_avg|))+k_32×(1 / (1+Δφ_var)) is adopted, where k_31 and k_32 are weight coefficients and k_31+k_32=1. By adjusting the weight coefficients, the emphasis on the overall phase offset and local phase fluctuation can be adjusted according to actual needs.

[0133] Step S153: Perform range normalization processing on the first optimization index, the second optimization index, and the third optimization index respectively, calculate the normalized weighted sum based on the preset weight coefficient, and generate a comprehensive optimization score.

[0134] Range normalization is to map the values of each optimization index to the interval [0, 1]. For the first optimization index I_1, let its maximum value be I_1_max and its minimum value be I_1_min. The normalized first optimization index I_1_norm = (I_1 - I_1_min) / (I_1_max - I_1_min).

[0135] For the second optimization index I_2, let its maximum value be I_2_max and its minimum value be I_2_min. The normalized second optimization index I_2_norm = (I_2 - I_2_min) / (I_2_max - I_2_min).

[0136] For the third optimization index I_3, let its maximum value be I_3_max and its minimum value be I_3_min. The normalized third optimization index I_3_norm = (I_3 - I_3_min) / (I_3_max - I_3_min).

[0137] The preset weight coefficients are used to determine the importance of each normalized optimization index in the comprehensive optimization score. Let the weight coefficient of the first optimization index be w_1, the weight coefficient of the second optimization index be w_2, and the weight coefficient of the third optimization index be w_3, and w_1 + w_2 + w_3 = 1. The comprehensive optimization score Score = w_1 × I_1_norm + w_2 × I_2_norm + w_3 × I_3_norm.

[0138] Step S154: If the comprehensive optimization score is lower than the preset optimization threshold, perform at least one of the following adjustment operations.

[0139] The preset optimization threshold is a pre-set value used to judge whether the quality of the target filtered signal meets the requirements. If the comprehensive optimization score Score is lower than the preset optimization threshold, it means that there is still room for improvement in the quality of the target filtered signal, and adjustment operations need to be performed.

[0140] (1) Adjust the number of frequency bands or the bandwidth range in the frequency band division rule according to the frequency band energy balance characteristic.

[0141] Step S1541: According to the frequency band energy balance characteristic of the target filtered signal, obtain the energy proportion sequence and its standard deviation of each sub-band.

[0142] The energy E_j of each frequency band of the target filtered signal has been calculated previously. The total energy E_total = sum of all frequency band energies E_j. The energy proportion P_j of each sub-band = E_j / E_total, so the energy proportion sequence {P_1, P_2,..., P_m} of each sub-band is obtained.

[0143] Calculate the standard deviation σ_P of the energy proportion sequence. First, calculate the average value P_avg of the energy proportion sequence, which is the sum of all energy proportions P_j divided by the number of frequency bands m. Then, calculate the deviation of each energy proportion from the average value. The sum of the squares of the deviations is divided by the number of frequency bands m to obtain the variance σ_P², and the standard deviation σ_P is the square root of the variance σ_P².

[0144] Step S1542: Determine whether the standard deviation exceeds a preset equilibrium threshold. If it exceeds, perform the following frequency band adjustment operations.

[0145] The preset equilibrium threshold is a preset value used to determine whether the frequency band energy distribution is balanced. If the standard deviation σ_P exceeds the preset equilibrium threshold, it indicates that the frequency band energy distribution is uneven and frequency band adjustment operations need to be performed.

[0146] (a)According to the number of frequency bands and the bandwidth range of each frequency band in the current frequency band division rule, determine the upper limit of the number of frequency bands allowed to be adjusted and the minimum bandwidth constraint value.

[0147] The current frequency band division rule stipulates the number of frequency bands m and the bandwidth range of each frequency band. The upper limit N_max of the number of frequency bands allowed to be adjusted is a value preset according to the actual situation and processing capacity, and the number of frequency bands cannot be increased without limit. The minimum bandwidth constraint value B_min is also preset to ensure that the bandwidth of each frequency band cannot be too small, otherwise it will increase the processing complexity and calculation amount.

[0148] (b)Screen out the first target frequency band corresponding to the highest energy proportion value and the second target frequency band corresponding to the lowest energy proportion value from the energy proportion sequence.

[0149] In the energy proportion sequence {P_1, P_2,..., P_m}, by comparing the values of each energy proportion, find the maximum value P_max and the minimum value P_min. The frequency band corresponding to the maximum value P_max is the first target frequency band, and the frequency band corresponding to the minimum value P_min is the second target frequency band.

[0150] (c)Perform frequency band splitting on the first target frequency band: Based on the minimum bandwidth constraint value, calculate the maximum number of sub-frequency bands that can be split, and divide the frequency band into multiple sub-frequency bands according to the energy distribution, ensuring that the bandwidth of each sub-frequency band is not less than the minimum bandwidth constraint value.

[0151] Let the bandwidth of the first target frequency band be \(B_1\). According to the minimum bandwidth constraint value \(B_{min}\), the maximum number of sub - frequency bands \(n_{max}\) that can be split is \(n_{max}=\lfloor B_1 / B_{min}\rfloor\). Further analyze the signals within the first target frequency band, divide the frequency range within this band into several small frequency intervals, and calculate the energy of each small frequency interval. According to the energy distribution, divide the regions with large energy differences into different sub - frequency bands. During the division process, ensure that the bandwidth of each sub - frequency band is not less than the minimum bandwidth constraint value \(B_{min}\).

[0152] (d)Perform frequency band merging on the second target frequency band: Identify the energy proportion of its adjacent frequency bands. If the energy proportion of the adjacent frequency band is lower than the preset merging threshold, and the bandwidth of the new frequency band after merging does not violate the minimum bandwidth constraint and the upper limit requirement of the maximum number of frequency bands, then merge the second target frequency band with the adjacent frequency band into a new frequency band, and the bandwidth of the new frequency band is the sum of the two.

[0153] The preset merging threshold is a value preset according to historical data and experience, and is used to judge whether to perform frequency band merging. For the second target frequency band, identify its adjacent frequency bands. Let the energy proportions of the adjacent frequency bands be \(P_{adj1}\) and \(P_{adj2}\) respectively. If \(P_{adj1}\) or \(P_{adj2}\) is lower than the preset merging threshold, and the bandwidth of the new frequency band after merging is not less than the minimum bandwidth constraint value \(B_{min}\), and at the same time does not cause the number of frequency bands to exceed the upper limit \(N_{max}\) of the allowable adjustment of the number of frequency bands, then merge the second target frequency band with the corresponding adjacent frequency band into a new frequency band. The bandwidth of the new frequency band is the sum of the bandwidth of the second target frequency band and the bandwidth of the adjacent frequency band.

[0154] (e)Update the number of frequency bands and the bandwidth range of each frequency band in the frequency band division rule, and recalculate the standard deviation of the adjusted energy proportion sequence of each sub - frequency band.

[0155] After the frequency band splitting and merging operations, update the number of frequency bands and the bandwidth range of each frequency band in the frequency band division rule. Decompose the target filtered signal into frequency bands again, calculate the energy \(E_j'\) of each frequency band, and the total energy \(E_{total}'=\sum_{j}E_j'\) (the sum of the energies \(E_j'\) of all frequency bands). The energy proportion \(P_j'\) of each sub - frequency band is \(P_j' = E_j' / E_{total}'\), obtaining the adjusted energy proportion sequence \(\{P_1', P_2',\cdots, P_m'\}\) of each sub - frequency band. According to the method of calculating the standard deviation above, recalculate the standard deviation \(\sigma_{P}'\) of the adjusted energy proportion sequence.

[0156] Iteratively execute steps (a) to (e) until the standard deviation is lower than the equilibrium threshold or reaches the upper limit of the number of frequency bands, and output the adjusted frequency band division rule.

[0157] Continuously repeat the above frequency band adjustment operation, and calculate the standard deviation of the energy ratio sequence after each adjustment until the standard deviation σ_P' is lower than the equilibrium threshold or the number of frequency bands reaches the upper limit N_max of the allowable number of adjusted frequency bands. At this time, output the adjusted frequency band division rule, and use the adjusted frequency band division rule for frequency band decomposition processing subsequently.

[0158] (2) Adjust the calculation rule of the dynamic weight factor in the adaptive filtering operation according to the time-domain signal-to-noise ratio characteristics.

[0159] Step S1543: According to the time-domain signal-to-noise ratio characteristics of the target filtered signal, obtain the ratio of the average energy of the signal-dominated time-domain window to the average energy of the noise time-domain window.

[0160] When calculating the time-domain signal-to-noise ratio, the signal segment and the noise segment have been divided. The signal-dominated time-domain window is the signal segment, and the noise time-domain window is the noise segment. The average energy E_s_avg of the signal-dominated time-domain window = the energy E_s of the signal segment divided by the length of the signal segment; the average energy E_n_avg of the noise time-domain window = the energy E_n of the noise segment divided by the length of the noise segment. The ratio R_energy = E_s_avg / E_n_avg.

[0161] Step S1544: Map the ratio to a filtering intensity level, where the filtering intensity level is negatively correlated with the ratio.

[0162] Set multiple filtering intensity levels, for example, divided into three levels: low, medium, and high. Map the ratio R_energy to the corresponding filtering intensity level according to its magnitude. Since the filtering intensity level is negatively correlated with the ratio, that is, the larger the ratio, the more useful information in the signal relative to the noise, and the relatively lower the filtering intensity; the smaller the ratio, the more noise, and a higher filtering intensity is required. The mapping can be performed by setting different threshold intervals. For example, when R_energy is greater than the threshold T_1, it is mapped to the low filtering intensity level; when R_energy is between the thresholds T_1 and T_2, it is mapped to the medium filtering intensity level; when R_energy is less than the threshold T_2, it is mapped to the high filtering intensity level.

[0163] Step S1545: Based on the filtering intensity level, adjust the proportional relationship between the noise energy weight coefficient and the signal smoothness weight coefficient in the dynamic weight factor calculation formula.

[0164] The dynamic weight factor calculation formula is W = a×S + b×(1 - P_noise), where a is the signal smoothness weight coefficient, b is the noise energy weight coefficient, and a + b = 1.

[0165] (f) When the filtering intensity level is high, increase the proportion of the noise energy weight coefficient so that its proportion in the dynamic weight factor is increased to the preset upper limit value.

[0166] The preset upper limit value is a value set in advance to limit the maximum value of the noise energy weight coefficient. When the filtering intensity level is high, it indicates that there is relatively more noise and stronger noise suppression is required. Increase the noise energy weight coefficient b to the preset upper limit value b_max. Correspondingly, the signal smoothness weight coefficient a = 1 - b_max.

[0167] (g) When the filtering intensity level is low, increase the proportion of the signal smoothness weight coefficient so that its proportion in the dynamic weight factor is increased to the preset lower limit value.

[0168] The preset lower limit value is a value set in advance to limit the minimum value of the signal smoothness weight coefficient. When the filtering intensity level is low, it indicates that there is relatively more useful information in the signal and excessive filtering is not required. Increase the signal smoothness weight coefficient a to the preset lower limit value a_min. Correspondingly, the noise energy weight coefficient b = 1 - a_min.

[0169] Step S1546: Substitute the adjusted weight coefficient ratio into the dynamic weight factor calculation formula to generate a new dynamic weight factor.

[0170] Substitute the adjusted signal smoothness weight coefficient a and noise energy weight coefficient b into the dynamic weight factor calculation formula W = a×S + b×(1 - P_noise) to generate a new dynamic weight factor W'.

[0171] Step S1547: Update the filter parameter set of the adaptive filtering operation according to the new dynamic weight factor and apply it to the recursive filtering process of the next sub-band signal.

[0172] According to the new dynamic weight factor W', update the filter parameter set F_set'' of the adaptive filtering operation according to the method of adjusting the filter parameter set before. When performing recursive filtering on the new sub-band signal subsequently, use the updated filter parameter set F_set''.

[0173] (3) Adjust the phase correction parameter in the multi-band fusion processing according to the phase continuity feature.

[0174] Step S1548: According to the phase continuity feature of the target filtered signal, obtain the mean and variance of the phase difference of adjacent band signal components within the overlapping time domain window.

[0175] When calculating the phase continuity feature above, the phase difference Δφ(f) of the adjacent frequency band signal components has been obtained. For the phase difference of the adjacent frequency band signal components in the overlapping time domain window, extract the phase difference data in the window and calculate its mean Δφ_avg_overlap and variance Δφ_var_overlap.

[0176] Step S1549: Determine the overall phase offset based on the phase difference mean, and determine the local phase fluctuation intensity based on the variance.

[0177] In order to solve the phase winding problem, the circular statistical method is used to calculate the mean and variance of the phase difference. For the phase difference Δφ(f) of the signal components in adjacent frequency bands, it is first converted to the main value interval (-π to π).

[0178] For the determination of the overall phase offset, the weighted average method is used to calculate the average phase difference. Assume that the phase difference sequence is {Δφ_1, Δφ_2, …, Δφ_N}, first convert each phase difference Δφ_i to the main value interval, and then calculate the average phase difference Δφ_avg. The calculation process is: first convert all phase differences into complex form, that is, z_i=exp(j×Δφ_i), calculate the average value of the complex number Z_avg=(∑z_i) / N, and finally calculate the phase of Z_avg to obtain the average phase difference Δφ_avg=angle(Z_avg), which is the overall phase offset.

[0179] For the determination of the intensity of local phase fluctuations, the variance is also calculated based on the phase difference after conversion to the main value interval. First, the deviation of each phase difference from the average phase difference is calculated, and the deviation is also converted to the main value interval. Then, the sum of the squares of these deviations is calculated and divided by the number of samples to obtain the variance Δφ_var. The intensity of local phase fluctuations can be represented by this variance Δφ_var. The larger the variance, the more severe the local phase fluctuation.

[0180] Perform the following phase correction parameter adjustment operations.

[0181] (h) If the phase difference mean exceeds the first correction threshold, the time domain delay compensation amount of each frequency band is calculated according to the overall phase offset. For non-integer sample point shift requirements, a polynomial interpolation method is used to generate a continuous time domain signal, and phase alignment is achieved through resampling.

[0182] The first correction threshold is a pre-set value used to determine whether there is a large overall phase shift. If the mean phase difference Δφ_avg_overlap exceeds the first correction threshold, it indicates that there is an obvious overall phase shift. Calculate the time-domain delay compensation amount Δt for each frequency band according to the overall phase shift amount Δφ_avg_overlap. For the non-integer sample point shift requirement, use polynomial interpolation to insert new sampling points into the signal to generate a continuous time-domain signal. Then, through resampling operation, shift the signal according to the time-domain delay compensation amount to achieve phase alignment.

[0183] (i)If the variance exceeds the second correction threshold, interpolation points are inserted within the overlapping time-domain window, and a smooth transition curve is generated based on the interpolated phase sequence to eliminate local phase mutations.

[0184] The second correction threshold is a pre-set value used to determine whether there is a large local phase fluctuation. If the variance Δφ_var_overlap exceeds the second correction threshold, it indicates that there are obvious local phase mutations. Insert interpolation points within the overlapping time-domain window, and generate new phase values through an interpolation algorithm. Based on the interpolated phase sequence, use methods such as curve fitting to generate a smooth transition curve to eliminate local phase mutations.

[0185] Then, update the time-domain delay amount or interpolation point density in the phase correction parameters, and recalculate the mean phase difference and variance.

[0186] According to the above phase correction parameter adjustment operation, update the time-domain delay amount or interpolation point density in the phase correction parameters. Process the target filtered signal again, and calculate the mean phase difference and variance of the signal components in adjacent frequency bands within the overlapping time-domain window.

[0187] Iteratively adjust the parameters until both the mean phase difference and variance are lower than the corresponding thresholds, and output the optimized set of phase correction parameters.

[0188] Continuously repeat the above steps of phase correction parameter adjustment operation and calculation of mean phase difference and variance until both the mean phase difference and variance are lower than the corresponding thresholds (the first correction threshold and the second correction threshold). At this time, output the optimized set of phase correction parameters, and use the optimized set of phase correction parameters for phase correction in subsequent multi-band fusion processing.

[0189] After the above possible adjustment operations, perform frequency band decomposition processing using the adjusted frequency band division rule, perform adaptive filtering operation using the updated set of filtering parameters, and perform multi-band fusion processing using the optimized set of phase correction parameters to finally obtain the optimized target filtered signal. This optimized target filtered signal has been improved in terms of time-domain signal-to-noise ratio, frequency band energy balance, and phase continuity, and has better signal quality.

[0190] Figure 2 FIG. shows a schematic diagram of exemplary hardware and software components of an adaptive signal filtering system 100 based on multi-band fusion that can implement the idea of the present application provided by some embodiments of the present application. For example, the processor 120 can be used on the adaptive signal filtering system 100 based on multi-band fusion and is used to execute the functions in the present application.

[0191] The adaptive signal filtering system 100 based on multi-band fusion can be a general-purpose server or a special-purpose server, both of which can be used to implement the multi-band fusion-based adaptive signal filtering method of the present application. Although only one server is shown in the present application, for convenience, the functions described in the present application can be implemented in a distributed manner on multiple similar platforms to balance the processing load.

[0192] For example, the adaptive signal filtering system 100 based on multi-band fusion can include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and different forms of storage media 140, such as disks, ROM, or RAM, or any combination thereof. Exemplarily, the adaptive signal filtering system 100 based on multi-band fusion can also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The method of the present application can be implemented according to these program instructions. The adaptive signal filtering system 100 based on multi-band fusion also includes an I / O interface 150 between the computer and other input / output devices.

[0193] For ease of illustration, only one processor is described in the adaptive signal filtering system 100 based on multi-band fusion. However, it should be noted that the adaptive signal filtering system 100 based on multi-band fusion in the present application can also include multiple processors. Therefore, the steps executed by one processor described in the present application can also be jointly executed or separately executed by multiple processors. For example, if the processor of the adaptive signal filtering system 100 based on multi-band fusion executes steps A and B, it should be understood that steps A and B can also be jointly executed by two different processors or separately executed in one processor. For example, the first processor executes step A, the second processor executes step B, or the first processor and the second processor jointly execute steps A and B.

[0194] In addition, an embodiment of the present invention also provides a readable storage medium, in which computer-executable instructions are preset. When the processor executes the computer-executable instructions, the above multi-band fusion-based adaptive signal filtering method is implemented.

[0195] It should be noted that, in order to simplify the description of the disclosure of the present invention and thus assist in the understanding of one or more embodiments of the invention, in the foregoing description of the embodiments of the present invention, sometimes multiple features are incorporated into one embodiment, drawing or description thereof.

Claims

1. An adaptive signal filtering method based on multi - band fusion, characterized in that, The method includes: Obtaining a multi-band signal set of a target environment, where the multi-band signal set includes original signal components of different frequency bands; Performing frequency band decomposition processing on the multi-band signal set to obtain a plurality of sub-band signals; Performing an adaptive filtering operation on each of the sub-band signals to generate a corresponding filtered sub-band signal; wherein, the adaptive filtering operation includes dynamically adjusting filtering coefficients according to the noise characteristics of the sub-band signal; Performing multi-band fusion processing on the filtered sub-band signals to generate a target filtered signal; Adjusting the segmentation strategy of the frequency band decomposition processing or the dynamic adjustment rule of the adaptive filtering operation according to the signal quality evaluation result of the target filtered signal, and outputting an optimized target filtered signal.

2. The adaptive signal filtering method based on multi - band fusion according to claim 1, characterized in that, The performing frequency band decomposition processing on the multi-band signal set to obtain a plurality of sub-band signals includes: Determining a target decomposition level according to an initial frequency band division rule, and dynamically iteratively adjusting the number of frequency bands and the bandwidth range of the target decomposition level based on the spectral characteristics of the multi-band signal set during the signal processing. Wherein, if the frequency band energy balance characteristic after the dynamic iterative adjustment does not meet a preset threshold, the frequency band division rule is further optimized according to the frequency band energy ratio; Performing overlapping segmentation processing on each original signal component in the multi-band signal set based on the target decomposition level to generate an initial segmented signal; Performing a frequency domain transformation operation on the initial segmented signal to obtain a frequency domain energy distribution characteristic; Mapping the initial segmented signal to a corresponding target frequency band according to the matching degree between the frequency domain energy distribution characteristic and the frequency band division rule to generate the sub-band signal.

3. The adaptive signal filtering method based on multi - band fusion according to claim 2, characterized in that, The performing overlapping segmentation processing on each original signal component in the multi-band signal set based on the target decomposition level to generate an initial segmented signal includes: Determining the sampling rate of the original signal component, and calculating the number of sample points corresponding to the segmentation window according to the minimum bandwidth range in the target decomposition level and the sampling rate, so that the duration of the segmentation window is proportional to the reciprocal of the minimum bandwidth; Determining the number of overlapping sample points between adjacent segmentation windows based on the number of frequency bands of the target decomposition level, where the number of overlapping sample points is in an inverse proportional relationship with the number of frequency bands; Generating an equally spaced sliding window sequence according to the number of sample points of the segmentation window and the number of overlapping sample points.

4. The adaptive signal filtering method based on multi - band fusion according to claim 1, characterized in that, The performing an adaptive filtering operation on each of the sub-band signals to generate a corresponding filtered sub-band signal includes: Extracting the time domain waveform feature and the frequency domain energy feature of the sub-band signal; Determining a signal stationarity index according to the time domain waveform feature, and determining a noise energy ratio according to the energy ratio of the noise-dominated frequency band in the frequency domain energy feature, where the noise-dominated frequency band is identified by a preset noise reference interval or a dynamic noise estimation method based on the minimum statistic method, specifically including extracting the background noise energy distribution in the signal silent segment and iteratively updating the noise energy threshold; Constructing a dynamic weight factor based on the signal stationarity index and the noise energy ratio, and adjusting a preset set of filtering parameters according to the dynamic weight factor; Call the set of filtering parameters to perform recursive filtering on the sub-band signal, and in the process of filtering, detect the pole positions of the filter in real time to constrain its stability. If the detected poles exceed the unit circle range, reset the filtering parameters to the preset safety value to generate an initial filtered signal; Perform short-time Fourier transform on the initial filtered signal to generate a time-frequency energy distribution matrix; Extract the feature frequency bands related to noise in the time-frequency energy distribution matrix according to the frequency band division rules of the current target decomposition level; Calculate the ratio of the average energy in the feature frequency band to the average energy of the signal dominant frequency band. If the ratio exceeds the preset noise threshold, it is determined that there is residual noise energy, and the set of filtering parameters is iteratively updated until the residual noise energy is lower than the preset threshold, and the final filtered sub-band signal is output.

5. The adaptive signal filtering method based on multi - band fusion according to claim 4, characterized in that, The determination of the signal stationarity index according to the time-domain waveform characteristics includes: Calculate the time-domain variance sequence of the sub-band signal and the differential mean between adjacent sampling points; Determine the first stationarity factor according to the fluctuation amplitude of the time-domain variance sequence; Determine the second stationarity factor according to the absolute value of the differential mean; Perform standardization conversion on the first stationarity factor and the second stationarity factor respectively to map them to a unified ratio interval, and dynamically allocate the weight coefficients of the first stationarity factor and the second stationarity factor after standardization conversion according to the frequency band position of the sub-band signal, and then perform weighted summation to generate the signal stationarity index; Among them, the weight coefficients of the first stationarity factor and the second stationarity factor are dynamically allocated according to the frequency band position of the sub-band signal.

6. The adaptive signal filtering method based on multi - band fusion according to claim 2, characterized in that, The multi-band fusion processing of the filtered sub-band signal to generate a target filtered signal includes: Perform signal reconstruction on each filtered sub-band signal to generate a time-domain signal component of the same length as the original signal component; Dynamically determine the frequency band weight coefficient of each time-domain signal component according to the target decomposition level and the signal quality score; Perform time-domain alignment and weighted superposition processing on the time-domain signal components based on the frequency band weight coefficients to generate an initial fusion signal; Extract the phase difference characteristics of adjacent frequency band signal components in the initial fusion signal, and dynamically calculate the phase correction amount of each frequency band according to the phase difference characteristics; Based on the phase correction amounts of each frequency band, perform phase alignment on the signal components of the initial fusion signal through time-domain delay compensation or interpolation processing. For the change in signal length caused by time-domain delay compensation, use the cyclic shift or edge symmetric interpolation method to keep the signal components the same length as the original signal to generate a phase-aligned signal; Perform time-domain smoothing processing on the phase-aligned signal to eliminate the instantaneous mutation noise introduced by phase correction, and obtain the corrected initial fusion signal; Generate the target filtered signal according to the corrected initial fusion signal.

7. The adaptive signal filtering method based on multi - band fusion according to claim 2, wherein, The adjustment of the segmentation strategy of the frequency band decomposition process or the dynamic adjustment rule of the adaptive filtering operation according to the signal quality evaluation result of the target filtered signal includes: Extract the time-domain signal-to-noise ratio characteristics, frequency band energy balance characteristics, and phase continuity characteristics of the target filtered signal; Calculate a first optimization index according to the time-domain signal-to-noise ratio characteristic, calculate a second optimization index according to the frequency-band energy balance characteristic, and calculate a third optimization index according to the phase continuity characteristic; Perform range normalization processing on the first optimization index, the second optimization index, and the third optimization index respectively, calculate the normalized weighted sum based on a preset weight coefficient, and generate a comprehensive optimization score; If the comprehensive optimization score is lower than a preset optimization threshold, perform at least one of the following adjustment operations: (1)Adjust the number of frequency bands or the bandwidth range in the frequency-band division rule according to the frequency-band energy balance characteristic; (2)Adjust the calculation rule of the dynamic weight factor in the adaptive filtering operation according to the time-domain signal-to-noise ratio characteristic; (3)Adjust the phase correction parameter in the multi-band fusion processing according to the phase continuity characteristic.

8. The adaptive signal filtering method based on multi - band fusion according to claim 7, wherein, The adjusting the number of frequency bands or the bandwidth range in the frequency-band division rule according to the frequency-band energy balance characteristic includes: Obtain the energy proportion sequence and its standard deviation of each sub-band according to the frequency-band energy balance characteristic of the target filtered signal; Judge whether the standard deviation exceeds a preset balance threshold. If it exceeds, perform the following frequency-band adjustment operations: (a)Determine the upper limit of the number of frequency bands allowed to be adjusted and the minimum bandwidth constraint value according to the number of frequency bands and the bandwidth range of each frequency band in the current frequency-band division rule; (b)Select the first target frequency band corresponding to the highest energy proportion value and the second target frequency band corresponding to the lowest energy proportion value from the energy proportion sequence; (c)Perform frequency-band splitting on the first target frequency band: Based on the minimum bandwidth constraint value, calculate the maximum number of sub-bands that can be split, and divide the frequency band into multiple sub-bands according to the energy proportion gradient to ensure that the bandwidth of each sub-band is not less than the minimum bandwidth constraint value; (d)Perform frequency-band merging on the second target frequency band: Identify the energy proportion of its adjacent frequency band. If the energy proportion of the adjacent frequency band is lower than the preset merging threshold, merge the second target frequency band with the adjacent frequency band into a new frequency band, and the bandwidth of the new frequency band is the sum of the two; (e)Update the number of frequency bands and the bandwidth range of each frequency band in the frequency-band division rule, and recalculate the standard deviation of the energy proportion sequence of each adjusted sub-band; Iteratively execute steps (a) to (e) until the standard deviation is lower than the balance threshold or reaches the upper limit of the number of frequency bands, and output the adjusted frequency-band division rule.

9. The adaptive signal filtering method based on multi - band fusion according to claim 7, wherein, The adjusting the calculation rule of the dynamic weight factor in the adaptive filtering operation according to the time-domain signal-to-noise ratio characteristic includes: Obtain the ratio of the average energy of the signal-dominant time-domain window to the average energy of the noise time-domain window according to the time-domain signal-to-noise ratio characteristic of the target filtered signal; Map the ratio to a filtering intensity level, where the filtering intensity level is negatively correlated with the ratio; Adjust the proportional relationship between the noise energy weight coefficient and the signal smoothness weight coefficient in the dynamic weight factor calculation formula based on the filtering intensity level, specifically including: (f)When the filtering intensity level is high, increase the proportion of the noise energy weight coefficient so that its proportion in the dynamic weight factor is increased to a preset upper limit value; (g) When the filtering intensity level is low, increase the proportion of the signal smoothness weight coefficient so that its proportion in the dynamic weight factor is increased to the preset lower limit value; Substitute the adjusted weight coefficient ratio into the dynamic weight factor calculation formula to generate a new dynamic weight factor; Update the filter parameter set of the adaptive filtering operation according to the new dynamic weight factor and apply it to the recursive filtering process of the next round of sub-band signals.

10. An adaptive signal filtering system based on multi - band fusion, wherein, It includes a processor and a memory. The memory is connected to the processor. The memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the multi-band fusion-based adaptive signal filtering method described in any one of claims 1-9 above.

Images