Acoustic feature enhancement method and device based on blind source separation and two-stage filtering, and medium
By employing FastICA blind source separation and two-stage filtering, the problem of amplitude and phase uncertainty in acoustic signal separation was solved, achieving effective separation of target sound sources and noise suppression, and improving the reliability of fault identification.
Patent Information
- Application Number
- CN202511473359.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-15
- Publication Date
- 2026-01-27
AI Technical Summary
The existing FastICA algorithm suffers from amplitude and phase uncertainties when separating acoustic signals, which makes it impossible to effectively separate the target source signal and affects the effectiveness of fault identification.
The FastICA blind source separation method combined with two-stage filtering is adopted. By correcting the amplitude and phase of the sub-signal, the effective source signal is selected by the energy ratio, and the target source signal is obtained by suppressing noise interference through adaptive two-stage filtering.
It effectively separates the target sound source, significantly suppresses noise interference, and improves the reliability of acoustic feature extraction and fault identification.
Smart Images
Figure CN121415798A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of fault diagnosis technology, and in particular to an acoustic feature enhancement method, device and medium based on blind source separation and two-stage filtering. Background Technology
[0002] Acoustic signal separation is an active and widely applied research field. In fault prediction and health management, microphone arrays are used to monitor the operating noise of machines (such as engines, bearings, and gearboxes), separating abnormal noises from specific components to achieve early fault diagnosis and warning. In defense and military industries, processing sonar signals can separate and identify the noise of ships and submarines from marine environmental noise for target detection, tracking, and identification. Inside moving vehicles, driver commands, passenger conversations, entertainment system sounds, and external environmental noise are separated to provide clean signals for cabin monitoring and voice control. Utilizing spatial information (Direction of Arrival (DOA) and Time Delay Asynchrony (TDOA)) provided by microphone arrays has become mainstream. Future technologies will focus more on jointly optimizing separation and sound source localization to achieve a unified understanding of both "hearing" and "seeing" sound sources.
[0003] FastICA is the most commonly used method for signal separation. Based on the theoretical framework of Independent Component Analysis (ICA), the separation process reconstructs the source signal by constructing an unmixing matrix. The optimization criterion is to maximize the non-Gaussianity of the output signal (such as negative entropy or kurtosis). It maintains high separation accuracy and robustness even in noisy environments. However, it requires the source signal to be independent and non-Gaussian. It fails to separate Gaussian distributed signals (such as white noise). Its technical limitations include: order uncertainty: the order of the separated independent components (ICs) is random and has no correspondence with the original order of the source signal; amplitude uncertainty: the amplitude (scaling ratio) of the separated signal is also uncertain. It may reconstruct a scaled version of the source signal, and the sign (positive or negative) and energy (amplitude) information will be lost. Therefore, the separated components need to be manually identified and interpreted based on domain knowledge. Summary of the Invention
[0004] The technical problem to be solved by the present invention is to provide an acoustic feature enhancement method, device and medium based on blind source separation and two-stage filtering, which can effectively separate the target source signal, reduce background noise, significantly enhance acoustic features, and thus improve the effectiveness of acoustic fault identification.
[0005] The technical solution adopted by this invention to solve its technical problem is: to provide an acoustic feature enhancement method based on blind source separation and two-stage filtering, comprising the following steps: The acquired acoustic signals were subjected to FastICA blind source separation to obtain several separated sub-signals; The amplitude and phase of each obtained sub-signal are corrected; The effective source signals are obtained by filtering the corrected sub-signals using energy ratios. The effective source signal is filtered using an adaptive two-stage filtering method to obtain the target source signal.
[0006] The acquisition of acoustic signals is subjected to FastICA blind source separation to obtain several separated sub-signals, specifically including: Calculate the mean of the acoustic signal and perform mean-removal processing on the acoustic signal; The mean-removed acoustic signal is whitened to obtain a whitened acoustic signal matrix; The independent components of the whitened acoustic signal matrix are extracted using the principle of maximizing non-Gaussianity, resulting in several separate sub-signals.
[0007] The method of extracting independent components of the whitened acoustic signal matrix using the principle of maximizing non-Gaussianity yields several separated sub-signals, specifically including: Choose a random unit vector as the weight vector. The initial value; use For the weight vector Update until convergence; where, For the new weight vector, This represents the data vector in the whitened acoustic signal matrix. Expressing expectations, It is a nonlinear function used to capture non-Gaussianity. for The derivative; The new weight vector is orthogonalized. Weight vector after orthogonalization The whitened acoustic signal matrix is projected onto the independent components to obtain the separated sub-signals.
[0008] The correction of the amplitude and phase of each obtained sub-signal specifically includes: The phase of a sub-signal is reconstructed by utilizing the spectral correlation between adjacent frequency points in the sub-signal; The minimum distortion criterion is used to determine the amplitude correction weight, and the amplitude of the sub-signal is corrected by the amplitude correction weight.
[0009] The step of filtering the corrected sub-signals using energy ratios to obtain effective source signals specifically includes: Perform Fourier transform on each corrected sub-signal and the acquired acoustic signal to obtain the spectra of several sub-signals and the spectrum of the acoustic signal. Calculate the ratio of the energy of the target characteristic frequency points and harmonic points in the spectrum of each sub-signal to the energy of the acoustic signal spectrum; The sub-signal corresponding to the maximum energy ratio is selected as the effective source signal.
[0010] The step of filtering the effective source signal using an adaptive two-stage filtering method to obtain the target source signal specifically includes: The center frequency and the first bandwidth are determined, and the effective source signal is filtered using an adaptive bandpass filter to obtain an intermediate signal; wherein, the center frequency is selected from the main characteristic frequency point in the spectrum of the effective source signal; The bandwidth is set to the second bandwidth, and an adaptive bandpass filter is used to filter the intermediate signal to obtain the source signal; wherein, the second bandwidth is less than the first bandwidth; Repeat the above steps n times, with each time the center frequency is different, and superimpose the n source signals to obtain the target source signal.
[0011] The adaptive bandpass filter includes: A forward filtering unit is used to apply a Butterworth bandpass filter to the effective source signal in chronological order to obtain a first filtered signal; The inverse filtering unit is used to reverse the timing of the first filtered signal and then generate a second filtered signal through the Butterworth bandpass filter. The result inversion unit is used to invert the timing of the second filtered signal to obtain an intermediate signal.
[0012] The transfer function of the Butterworth bandpass filter is expressed as: ,in, Let be the transfer function of the Butterworth bandpass filter. For passband gain, The center angular frequency, For quality factor, This represents the complex frequency vector of the effective source signal.
[0013] The technical solution adopted by the present invention to solve its technical problem is: to provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, it implements the steps of the above-mentioned acoustic feature enhancement method based on blind source separation and two-stage filtering.
[0014] The technical solution adopted by the present invention to solve its technical problem is: to provide a computer-readable storage medium on which a computer program is stored, wherein when the computer program is executed by a processor, the steps of the above-mentioned acoustic feature enhancement method based on blind source separation and two-stage filtering are implemented.
[0015] Beneficial effects Due to the adoption of the above technical solutions, this invention has the following advantages and positive effects compared with the prior art: This invention performs amplitude and phase correction on the sub-signals of FastICA blind source separation, making up for the uncertainty of amplitude and phase, and obtaining useful acoustic information with reasonable amplitude and coherent phase; it performs two-stage filtering on the screened useful signals, and by changing the center frequency bandwidth and combining it with improved adaptive filtering, it effectively separates the target sound source, significantly suppresses noise interference, and improves the reliability of acoustic feature extraction and fault identification. Attached Figure Description
[0016] Figure 1 This is a flowchart of the acoustic feature enhancement method based on blind source separation and two-stage filtering according to the first embodiment of the present invention; Figure 2 This is a flowchart of FastICA blind source separation in the first embodiment of the present invention; Figure 3 These are the time-domain diagram and spectrum diagram of the acoustic signal in this embodiment of the invention; Figure 4 This is a time-domain diagram of the separated sub-signals in an embodiment of the present invention; Figure 5 This is the power spectrum of the separated sub-signals in this embodiment of the invention; Figure 6 These are the time-domain diagram and spectrum diagram of the effective source signal in the first embodiment of the present invention; Figure 7 This is a schematic diagram of an adaptive bandpass filter in the first embodiment of the present invention; Figure 8 These are the time-domain diagram and spectrum diagram of the gearbox planetary separation signal in an embodiment of the present invention. Detailed Implementation
[0017] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined by the appended claims.
[0018] The first embodiment of the present invention relates to an acoustic feature enhancement method based on blind source separation and two-stage filtering, such as... Figure 1 As shown, it includes the following steps: Step 1: Perform FastICA blind source separation on the acoustic signal acquired by the microphone under steady-state conditions to decompose the frequency domain separation matrix. and the initially estimated sub-signal Among them, FastICA blind source separation technology is as follows: Figure 2 As shown, the specific steps are as follows: Step S11, calculate the acoustic signal The mean of the denominator is then removed, i.e.: ; in, For the first The group's acoustic signals, For the first The first group of acoustic signals The value of each sampling point, The number of sampling points, The first one after mean removal processing The group's acoustic signals.
[0019] Step S12: The mean-removed acoustic signal is whitened to remove correlation between signals, resulting in the whitened acoustic signal matrix: in, It is a matrix composed of eigenvectors column by column. It is a diagonal matrix composed of eigenvalues. It is a whitened acoustic signal matrix.
[0020] Step S13: Extract the whitened acoustic signal matrix using the principle of maximizing non-Gaussianity. The independent components are used to obtain several separated sub-signals. The specific steps are as follows: First, initialization is performed, which involves selecting a random unit vector as the weight vector. The initial value.
[0021] Next, iterative updates are performed at fixed points, that is: for the weight vector Apply the following rules to update until convergence.
[0022] ; in, For the new weight vector, This represents the data vector in the whitened acoustic signal matrix. Expressing expectations, It is a nonlinear function used to capture non-Gaussianity. It is its derivative. After the update, the new weight vector can also be applied. Normalization is performed.
[0023] Then orthogonalization is performed, which is done by applying the new weight vector. Orthogonalization is performed to ensure that they are independent of each other.
[0024] Then, the independent components are calculated, i.e., the obtained weight vector is used. The whitened acoustic signal matrix Projected onto the independent components, it is represented as: .
[0025] If multiple independent components need to be extracted, the steps of orthogonalization and calculation of independent components need to be repeated, and each time the extraction is performed, it must be ensured that the new weight vector is orthogonal (or independent) to the previously extracted independent components.
[0026] As an example, this embodiment uses acoustic signals collected by a gearbox test bench. The rotational speed is set to 2700 rpm, with no load, a sampling frequency of 51200 Hz, and a sampling time of 1 second. The time-domain plot and spectrum of the original acoustic signal are shown below. Figure 3 As shown, the time-domain plots and power spectrum plots of the 16 sub-signals obtained through step 1 are as follows: Figure 4 and Figure 5 As shown in the figure, FastICA can separate the meshing frequency of gears with strong signals and their harmonics and sidebands. However, the disadvantage of FastICA is that the amplitude and order of the separated source signals are uncertain, and it cannot directly determine the source signal of the corresponding gear, shaft or bearing.
[0027] Step 2: Correct the amplitude and phase of each obtained sub-signal. This step is detailed below: Step S21: The phase of the sub-signal is reconstructed by utilizing the spectral correlation between adjacent frequency points in the sub-signal to ensure the channel consistency of each sub-signal at each frequency point. Step S22: The minimum distortion criterion is used to determine the amplitude correction weight, and the amplitude of the sub-signal is corrected by the amplitude correction weight, thereby completing the estimation of the sub-signal.
[0028] As an example, in this embodiment, after correcting the amplitude and phase of the sub-signal, the following can be obtained: Figure 6 The time-domain plot and spectrum plot shown are obtained by comparing with... Figure 3 , Figure 4 and Figure 5 The comparison shows that the amplitude of the time-domain waveform and the amplitude of the spectrum are significantly corrected.
[0029] Step 3: Filter the corrected sub-signals using energy ratios to obtain valid source signals. In this step, firstly, perform a Fourier transform on each corrected sub-signal and the acquired acoustic signal to obtain several sub-signal spectra and the acoustic signal spectrum. Then, based on the characteristics of the target sound source, obtain the characteristic frequencies and their harmonics of the target sound source, and calculate the ratio of the sum of the energy at the target characteristic frequency points and harmonics points in each sub-signal spectrum to the energy of the acoustic signal spectrum. Finally, select the sub-signal corresponding to the maximum energy ratio as the valid source signal. Wherein, the first... The energy ratio of each sub-signal The calculation method is expressed as follows: ; in, The spectral energy of the acquired acoustic signal. The sum of the energy of the target characteristic frequency points and harmonics points in the sub-signal spectrum.
[0030] As an example, the selected valid source signal in this embodiment is as follows: Figure 6 As shown.
[0031] Step 4: The effective source signal is filtered using an adaptive two-stage filtering method to obtain the target source signal. The specific steps for performing this step are as follows: Step S41: Determine the center frequency and bandwidth, and use an adaptive bandpass filter to filter the effective source signal to obtain an intermediate signal. The center frequency is selected from the main characteristic frequency point in the spectrum of the effective source signal, and the bandwidth is set to a wide bandwidth.
[0032] Step S42: Set the bandwidth to a narrow bandwidth and use an adaptive bandpass filter to filter the intermediate signal to obtain the source signal; Step S43: Repeat the above steps n times, with each time the center frequency is different. Superimpose the obtained n source signals to obtain the target source signal. The target source signal effectively enhances the acoustic characteristics of the source signal and suppresses background noise interference.
[0033] The adaptive bandpass filter used in this step eliminates the phase distortion introduced by traditional filtering through bidirectional filtering, such as... Figure 7 As shown, it specifically includes: A forward filtering unit is used to apply a Butterworth bandpass filter to the effective source signal in time sequence to obtain a first filtered signal; wherein the transfer function of the Butterworth bandpass filter is expressed as: ; in, Let be the transfer function of the Butterworth bandpass filter. For passband gain, The center angular frequency, For quality factor, and These are the upper and lower cutoff frequencies. This represents the complex frequency vector of the effective source signal. (Through...) The transfer function of a Butterworth bandpass filter can be converted into a digital filter. Directly representing a high-order digital filter in transfer function form can lead to numerical instability. Therefore, the transfer function of a high-order digital filter can be decomposed into the product of multiple second-order sections. The entire filter, through the cascading of second-order sections, makes the numerical calculation more stable. Each second-order section has a specific structure... Represented as: ; in, , , These represent the coefficients of the numerator polynomial of the second-order section transfer function, respectively. , and These represent the coefficients of the denominator polynomial of the second-order section transfer function, respectively. The Z-domain representation of the effective source signal.
[0034] The inverse filtering unit is used to reverse the timing of the first filtered signal and then generate a second filtered signal through the Butterworth bandpass filter. The result inversion unit is used to invert the timing of the second filtered signal to obtain an intermediate signal.
[0035] In this embodiment, the phase responses of the forward and reverse filters cancel each other out, resulting in a total phase delay of 0.
[0036] In this embodiment, the selected center frequency is the theoretical value of the gearbox planetary meshing frequency of 154Hz, its second harmonic 308Hz, and its third harmonic 462Hz. The first filter bandwidth is set to 100Hz, and the second filter bandwidth is set to 10Hz. The input signal is filtered in two stages using these three center frequencies as the center. The filtering results are as follows: Figure 8 As shown. Figure 8 The spectrum clearly shows that the gearbox planetary meshing frequency and its third harmonic signal are effectively separated, and the frequency domain characteristics of the target sound source are enhanced.
[0037] It is easy to see that this invention performs amplitude and phase correction on the sub-signals of FastICA blind source separation, making up for the uncertainty of amplitude and phase, and obtaining useful acoustic information with reasonable amplitude and coherent phase. Two-stage filtering is performed on the screened useful signals. By changing the center frequency bandwidth and combining it with improved adaptive filtering, the target sound source is effectively separated, and noise interference is significantly suppressed, improving the reliability of acoustic feature extraction and fault identification.
[0038] A second embodiment of the present invention relates to an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the acoustic feature enhancement method based on blind source separation and two-stage filtering of the first embodiment.
[0039] The third embodiment of the present invention relates to a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the acoustic feature enhancement method based on blind source separation and two-stage filtering of the first embodiment.
[0040] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.
[0041] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0042] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction methods implemented in a process. Figure 1 One or more processes and / or boxes Figure 1The function specified in one or more boxes.
[0043] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0044] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.
Claims
1. An acoustic feature enhancement method based on blind source separation and two-stage filtering, characterized in that, Includes the following steps: The acquired acoustic signals were subjected to FastICA blind source separation to obtain several separated sub-signals; The amplitude and phase of each obtained sub-signal are corrected; The effective source signals are obtained by filtering the corrected sub-signals using energy ratios. The effective source signal is filtered using an adaptive two-stage filtering method to obtain the target source signal.
2. The acoustic feature enhancement method based on blind source separation and two-stage filtering according to claim 1, characterized in that, The acquisition of acoustic signals is subjected to FastICA blind source separation to obtain several separated sub-signals, specifically including: Calculate the mean of the acoustic signal and perform mean-removal processing on the acoustic signal; The mean-removed acoustic signal is whitened to obtain a whitened acoustic signal matrix; The independent components of the whitened acoustic signal matrix are extracted using the principle of maximizing non-Gaussianity, resulting in several separate sub-signals.
3. The acoustic feature enhancement method based on blind source separation and two-stage filtering according to claim 1, characterized in that, The method of extracting independent components of the whitened acoustic signal matrix using the principle of maximizing non-Gaussianity yields several separated sub-signals, specifically including: Choose a random unit vector as the weight vector. The initial value; use For the weight vector Update until convergence; where, For the new weight vector, This represents the data vector in the whitened acoustic signal matrix. Expressing expectations, It is a nonlinear function used to capture non-Gaussianity. for The derivative; The new weight vector is orthogonalized. Weight vector after orthogonalization The whitened acoustic signal matrix is projected onto the independent components to obtain the separated sub-signals.
4. The acoustic feature enhancement method based on blind source separation and two-stage filtering according to claim 1, characterized in that, The correction of the amplitude and phase of each obtained sub-signal specifically includes: The phase of a sub-signal is reconstructed by utilizing the spectral correlation between adjacent frequency points in the sub-signal; The minimum distortion criterion is used to determine the amplitude correction weight, and the amplitude of the sub-signal is corrected by the amplitude correction weight.
5. The acoustic feature enhancement method based on blind source separation and two-stage filtering according to claim 1, characterized in that, The step of filtering the corrected sub-signals using energy ratios to obtain effective source signals specifically includes: Perform Fourier transform on each corrected sub-signal and the acquired acoustic signal to obtain the spectra of several sub-signals and the spectrum of the acoustic signal. Calculate the ratio of the energy of the target characteristic frequency points and harmonic points in the spectrum of each sub-signal to the energy of the acoustic signal spectrum; The sub-signal corresponding to the maximum energy ratio is selected as the effective source signal.
6. The acoustic feature enhancement method based on blind source separation and two-stage filtering according to claim 1, characterized in that, The step of filtering the effective source signal using an adaptive two-stage filtering method to obtain the target source signal specifically includes: The center frequency and the first bandwidth are determined, and the effective source signal is filtered using an adaptive bandpass filter to obtain an intermediate signal; wherein, the center frequency is selected from the main characteristic frequency point in the spectrum of the effective source signal; The bandwidth is set to the second bandwidth, and an adaptive bandpass filter is used to filter the intermediate signal to obtain the source signal; wherein, the second bandwidth is less than the first bandwidth; Repeat the above steps n times, with each time the center frequency is different, and superimpose the n source signals to obtain the target source signal.
7. The acoustic feature enhancement method based on blind source separation and two-stage filtering according to claim 6, characterized in that, The adaptive bandpass filter includes: A forward filtering unit is used to apply a Butterworth bandpass filter to the effective source signal in chronological order to obtain a first filtered signal; The inverse filtering unit is used to reverse the timing of the first filtered signal and then generate a second filtered signal through the Butterworth bandpass filter. The result inversion unit is used to invert the timing of the second filtered signal to obtain an intermediate signal.
8. The acoustic feature enhancement method based on blind source separation and two-stage filtering according to claim 6, characterized in that, The transfer function of the Butterworth bandpass filter is expressed as: ,in, Let be the transfer function of the Butterworth bandpass filter. For passband gain, The center angular frequency, For quality factor, This represents the complex frequency vector of the effective source signal.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the acoustic feature enhancement method based on blind source separation and two-stage filtering as described in any one of claims 1-8.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the acoustic feature enhancement method based on blind source separation and two-stage filtering as described in any one of claims 1-8.