A method and system for measuring ultra-low frequency audio in performance settings

By collecting and processing audio data in different regions, and combining noise suppression and mode decomposition, the instability problem of ultra-low frequency audio measurement in performance settings has been solved, and the accurate reconstruction of the low-frequency sound field in the audience area and the enhancement of sound effects have been achieved.

CN121677913BActive Publication Date: 2026-11-10HANGZHOU ELITE DIGITAL TECH CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202511948351.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2025-12-23
Publication Date
2026-11-10
Estimated Expiration
2045-12-23

AI Technical Summary

Technical Problem

In performance settings, the measurement and analysis of ultra-low frequency audio is difficult to achieve using conventional acoustic methods, resulting in unstable measurement results that cannot accurately reflect the characteristics of the low-frequency sound field and affect the optimization of performance acoustics.

Method used

Audio data is collected in different areas, abnormal data is deleted and target frequency bands are filtered. Through noise suppression, singular value decomposition and mode decomposition, the low-frequency sound field of the audience area is reconstructed. Natural low-frequency signals are used as excitation sources to avoid interference from dedicated equipment.

Benefits of technology

It achieves precise reconstruction of the low-frequency sound field in the audience area, enhances the impact and uniformity of ultra-low frequency sound effects, adapts to the needs of smooth performance, and ensures the accuracy of measurement and scene adaptability.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121677913B_ABST
    Figure CN121677913B_ABST
Patent Text Reader

Abstract

This invention relates to the field of acoustic measurement and performance sound technology, and particularly to a method and system for measuring ultra-low frequency audio in performance settings. The method includes acquiring first audio data from the stage perimeter and second audio data from the audience area; extracting target frequency bands and suppressing noise to obtain first and second standard signals; extracting sound source constraints from the first standard signal; constructing a sound field data matrix using the second standard signal; decomposing and approximating the matrix to obtain a low-rank approximation matrix; combining the constraint conditions with modal decomposition to obtain a modal distribution; and superimposing these modal distributions to obtain the low-frequency sound field distribution in the audience area and calculating ultra-low frequency audio data. This invention improves the signal-to-noise ratio, accurately reconstructs the low-frequency sound field in the audience area, adapts to the requirements of performance smoothness, and the output data can support acoustic optimization, effectively enhancing the impact and uniformity of live ultra-low frequency sound effects.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of acoustic measurement and performance sound technology, and in particular to an ultra-low frequency audio measurement method and system for performance settings. Background Technology

[0002] In music festivals, stadiums, and large-scale concerts, large-scale ultra-low frequency (URF) sound systems (subwoofer arrays) are commonly used as core equipment to enhance the low-frequency impact and fullness of the sound effects, forming a crucial component in determining the performance's acoustic experience. However, the physical characteristics of URF audio (long wavelength, unique propagation characteristics) and the complex environment of performance venues (numerous interference, restricted processes) make it difficult to measure and analyze its sound field characteristics using conventional acoustic measurement methods. Current technologies for URF audio measurement and analysis are mostly based on high-frequency measurement designs, which are not suitable for the physical characteristics and specific scenarios of URF audio. Due to the long wavelength and significant standing wave effect of URF waves, the received signal strength varies greatly depending on the receiver's location. Conventional point measurement methods cannot accurately reflect the low-frequency sound field characteristics. To ensure smooth performance, low-frequency signals can only be collected using natural low-frequency signals from the music program as the excitation source. Furthermore, wind noise and stage machinery vibrations overlap with the URF target signal, ultimately leading to unstable measurement results and making it difficult to measure URF sound field data for performance acoustic optimization.

[0003] Therefore, it is necessary to propose an ultra-low frequency audio measurement method and system for performance occasions to solve the above problems. Summary of the Invention

[0004] The purpose of this invention is to provide an ultra-low frequency audio measurement method and system for performance occasions, in order to solve the problems that conventional point measurement methods cannot accurately reflect the characteristics of low frequency sound fields due to the long wavelength of ultra-low frequency waves, and that in order to ensure smooth performance, low frequency signals can only be collected using natural low frequency signals in music programs as excitation sources, while wind noise and stage machinery vibrations overlap with ultra-low frequency target signals, resulting in unstable measurement results and making it difficult to measure ultra-low frequency sound fields to achieve acoustic optimization of performances.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A method and system for measuring ultra-low frequency audio in performance settings, the method comprising: First audio data and second audio data of the performance venue are collected respectively. The first audio data is used to indicate the audio signal of the measuring point around the stage, and the second audio data is used to indicate the audio signal of the measuring point in the audience area. Based on the first audio data and the second audio data, audio data corresponding to the target frequency band is extracted to obtain the first target signal and the second target signal. Noise suppression processing is performed on the first target signal and the second target signal to obtain the first standard signal and the second standard signal. The target frequency band is used to indicate the frequency band of the ultra-low frequency audio part in the audio data. The sound source characteristics are extracted from the first standard signal to obtain the sound source constraint conditions; The sound field data matrix is ​​constructed by using the signal set of the second standard signal as input. The sound field data matrix is ​​then decomposed and approximated in low rank to obtain a low rank approximation matrix. Based on the low-rank approximation matrix and the sound source constraint, the second standard signal is decomposed into signal modes by mode decomposition, and the signal modes are processed by the low-rank approximation matrix and the sound source constraint to obtain the mode distribution; The low-frequency sound field distribution is obtained by processing the modal distribution, and the ultra-low frequency audio data is obtained by analyzing and calculating the low-frequency sound field distribution.

[0006] Preferably, the step of extracting audio data corresponding to the target frequency band based on the first audio data and the second audio data to obtain the first target signal and the second target signal includes: The first audio data and the second audio data are traversed by time sequence signal respectively. After identifying and deleting abnormal data, the first audio data and the second audio data are uniformly converted into a time sequence format including time frames and sound pressure levels. Abnormal data includes signal disconnection data, zero value data and interference data. The first audio data and the second audio data are respectively filtered by bandpass filters to obtain the first frequency band signal and the second frequency band signal; Short-time Fourier transforms are performed on the first and second frequency band signals respectively to generate time-frequency diagrams. The time-frequency components of the target frequency band are filtered out by the frequency axis in the time-frequency diagram. The time-frequency diagram is a three-dimensional time-frequency diagram that includes time, frequency and sound pressure level. The time-frequency components include the frequency axis, the time axis and the original audio data. Perform an inverse short-time Fourier transform on the time-frequency components to obtain the first target signal and the second target signal.

[0007] Preferably, the target frequency band is 20Hz-100Hz.

[0008] Preferably, the step of performing noise suppression processing on the first target signal and the second target signal to obtain the first standard signal and the second standard signal includes: Combining the time-frequency characteristics of the time-frequency graph and the signal source scenario, the specific frequency mechanical vibration noise of the first target signal is identified, and the center frequency and bandwidth are extracted. The wind noise of the second target signal is identified, and the triggering time and duration of the wind noise are extracted. The time-frequency characteristics include the mechanical vibration noise characteristics on the frequency axis of the time-frequency graph and the noise variation law over time on the time axis. The first target signal is noise-suppressed by narrowband notch filtering and spectrum compensation, and the second target signal is noise-suppressed by adaptive filtering and wavelet threshold denoising. The output is a first standard signal and a second standard signal in time frame and sound pressure level time sequence format.

[0009] Preferably, the step of extracting sound source characteristics from the first standard signal to obtain sound source constraint conditions includes: By using short-time Fourier transform, the proportion of sound pressure level at each frequency point within the target frequency band is statistically analyzed to obtain the frequency distribution at the sound source end. Record the initial phase at each frequency point; The source constraint conditions are formed by the frequency distribution and initial phase at the source end.

[0010] Preferably, the step of constructing a sound field data matrix by using the signal set of the second standard signal as input, and decomposing and approximating the sound field data matrix to obtain a low-rank approximation matrix includes: Using the signal set of the second standard signal as input, a matrix is ​​constructed to obtain a two-dimensional sound field data matrix. The row dimension of the sound field data matrix is ​​the time frame, the column dimension is the audience area measurement point, and the matrix elements represent the sound pressure level of the corresponding time frame and audience area measurement point. Singular value decomposition is performed on the sound field data matrix to obtain a left singular vector matrix, a singular value diagonal matrix, and a right singular vector matrix. Each column of the left singular vector matrix is ​​an eigenvector of the time dimension, each row of the right singular vector matrix is ​​an eigenvector of the spatial dimension, and the diagonal elements of the singular value diagonal matrix are non-zero, while the other elements are zero. The diagonal elements of the singular value diagonal matrix are sorted in descending order, and the three largest singular values ​​are selected to form a maximum singular value sequence. The left and right singular vectors in the corresponding left and right singular vector matrices are selected according to the largest singular values, and the eigenvectors are formed by the left and right singular vectors. The magnitude of the singular values ​​of the diagonal elements of the singular value diagonal matrix is ​​used to indicate the degree of contribution of the corresponding eigenvectors to the sound field information. The magnitude of the singular values ​​is positively correlated with the degree of contribution to the sound field information. A low-rank approximate matrix is ​​obtained by constructing a matrix using the maximum singular value and the corresponding eigenvector.

[0011] Preferably, after selecting three maximum singular values ​​to form a maximum singular value sequence, the corresponding feature vectors are grouped in descending order according to the size of the singular values ​​to obtain the first feature vector, the second feature vector, and the third feature vector in sequence.

[0012] Preferably, the step of using mode decomposition to split the second standard signal into signal modes based on the low-rank approximation matrix and sound source constraints, and processing the signal modes using the low-rank approximation matrix and sound source constraints to obtain the mode distribution includes: The right singular vector of the first eigenvector is mapped to the direct sound mode, the right singular vector of the second eigenvector is mapped to the reflected sound mode, and the right singular vector of the third eigenvector is mapped to the standing wave mode. The signal modes include the direct sound mode, the reflected sound mode, and the standing wave mode. Based on the characteristic vectors of each group, and combined with the sound source constraints, the direct sound mode, reflected sound mode, and standing wave mode are matched and calculated respectively to obtain the measurement point mode distribution of the direct sound mode, reflected sound mode, and standing wave mode at the measurement point in the audience area; Based on multiple measurement points in the audience area, the mode distribution of each measurement point is extended to the entire audience area through linear interpolation, generating the direct sound, reflected sound, and standing wave mode distribution of the low-frequency sound field in the audience area.

[0013] Preferably, the low-frequency sound field distribution of the audience area is obtained by linearly superimposing the distribution results of the direct sound mode, the reflected sound mode and the standing wave mode, and the ultra-low frequency audio data of the audience area is calculated using the low-frequency sound field distribution results.

[0014] Preferably, an ultra-low frequency audio measurement system for performance settings, the system comprising: The acquisition module is used to acquire first audio data and second audio data in the performance setting; The preprocessing module is used to perform abnormal data deletion on the raw audio data acquired by the acquisition module, and to filter the target time-frequency components through the generated time-frequency diagram to obtain the first target signal and the second target signal; A noise suppression module is used to identify the mechanical vibration noise of the first target signal and the wind noise of the second target signal, and output a first standard signal and a second standard signal. The extraction module is used to extract the sound source characteristics of the first standard signal to obtain the sound source constraint conditions; The processing module is used to construct a two-dimensional sound field data matrix by taking the second standard signal as input, perform singular value decomposition on the matrix and select the maximum singular value and the corresponding eigenvector to construct a low-rank approximate matrix; The decomposition module is used to map the eigenvectors of the low-rank approximation matrix to the modes, perform matching calculations on each mode in combination with the sound source constraints, obtain the mode distribution of the measurement point, and then extend it to the entire audience area through linear interpolation to generate the mode distribution. The analysis and calculation module is used to linearly superimpose the distributions of direct sound, reflected sound, and standing wave modes to obtain the low-frequency sound field distribution in the audience area, and then analyze and calculate the sound field distribution to obtain ultra-low frequency audio data.

[0015] The technical effects and advantages of the present invention in the above technical solution are as follows: This invention collects audio data in different regions, removes abnormal data and filters target frequency bands, and performs targeted noise reduction on the audio data, effectively improving the signal-to-noise ratio of ultra-low frequency signals and ensuring the purity of the input data. It obtains sound source constraints from the stage area audio data, simplifies the sound field data of the audience area audio data using singular value decomposition and low-rank approximation while retaining core features, and extends to the entire audience area through modal decomposition and linear interpolation. This overcomes the measurement limitations caused by the long wavelength and significant standing wave effect of ultra-low frequencies, achieving accurate reconstruction of the low-frequency sound field in the audience area and avoiding the limitations of conventional point measurement methods. Furthermore, the solution uses natural low-frequency signals from the performance as the excitation source, eliminating the need for dedicated excitation equipment and adapting to the requirements of performance smoothness. The final output ultra-low frequency audio data can directly support performance acoustic optimization, significantly improving the impact and uniformity of the live ultra-low frequency sound effects, balancing measurement accuracy, scene adaptability, and practical application value. Attached Figure Description

[0016] Figure 1 This is a flowchart of an ultra-low frequency audio measurement method for performance occasions according to the present invention.

[0017] Figure 2 This is a structural diagram of an ultra-low frequency audio measurement system for performance occasions according to the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1: The present invention provides as follows Figure 1 The method shown is an ultra-low frequency audio measurement method for performance occasions, the method includes: Step 101: Collect the first audio data and the second audio data of the performance venue respectively. The first audio data is used to indicate the audio signal of the measuring points around the stage, and the second audio data is used to indicate the audio signal of the measuring points in the audience area. A small number of measurement points (e.g., 4-8) are set up in the audience area and around the stage. Ultra-low frequency signals are collected through array deployment and sparse sampling strategy. This adapts to the actual scene constraints of the performance venue. A small number of measurement points can avoid dense deployment occupying too much space, while reducing costs and operational complexity, reducing the number of dedicated ultra-low frequency sensors to be purchased, and simplifying the installation and calibration process of measurement points. In performance settings, the area around the stage is the signal source of the ultra-low frequency sound system (subwoofer array). The first audio data can directly reflect the original characteristics of the sound source, including output spectrum, initial phase, etc., providing a benchmark for audio decomposition. The audience area is the final coverage and experience area of ​​the ultra-low frequency sound field. The ultra-low frequency signal contained in the second audio data is the core audio data for subsequent construction of the sound field data matrix and realization of sound field reconstruction.

[0020] Step 102: Extract audio data corresponding to the target frequency band based on the first audio data and the second audio data respectively to obtain the first target signal and the second target signal; The above method is specifically implemented through the following steps: performing time-series signal traversal on the first audio data and the second audio data respectively, identifying and deleting abnormal data, and then uniformly converting the first audio data and the second audio data into a time-series format including time frames and sound pressure levels. The abnormal data includes signal disconnection data, zero-value data, and interference data. By performing time-series signal traversal on the first audio data and the second audio data respectively, it is possible to accurately identify and delete signal disconnection data that may be caused by sensor failure, zero value data caused by poor wiring contact, and interference data caused by audience screams and equipment collisions, so as to avoid such abnormal data from being mixed into subsequent processing. Because the time frame-sound pressure level time sequence format of the first audio data and the second audio data keeps the dimensions and units of the two types of data consistent, the processing flow is not interrupted due to differences in data format.

[0021] The first audio data and the second audio data are respectively filtered by bandpass filters to obtain the first frequency band signal and the second frequency band signal; By performing filtering through a bandpass filter, the target frequency band of 20Hz-100Hz is accurately preserved, and interference outside the frequency band is eliminated, namely infrasound below 20Hz and mid-to-high frequency signals above 100Hz. Preferably, the filter parameter configuration selects and sets the filter order to 4-6. For the first audio data, the <20Hz infrasound unique to the stage periphery is filtered out, and the 20Hz-100Hz ultra-low frequency components are retained to obtain the first frequency band signal; for the second audio data, the >100Hz mid-high frequency interference commonly found in the audience area is filtered out, and the 20Hz-100Hz ultra-low frequency components are retained to obtain the second frequency band signal.

[0022] Short-time Fourier transforms are performed on the first and second frequency band signals respectively to generate time-frequency diagrams. The time-frequency components of the target frequency band are then selected by filtering the frequency axis in the time-frequency diagram. The time-frequency diagram is a three-dimensional time-frequency diagram that includes time, frequency, and sound pressure level. The time-frequency components include the frequency axis, the time axis, and the original audio data. By using the frequency axis of the time-frequency diagram, all time-frequency components in the 20Hz-100Hz range can be filtered out, removing band edge interference and residual signals that were not completely filtered out by the bandpass filter, while retaining the dynamic characteristics of the two types of frequency band signals, providing accurate time-frequency data support for the first target signal and the second target signal.

[0023] Perform an inverse short-time Fourier transform on the time-frequency components to obtain the first target signal and the second target signal.

[0024] The first target signal and the second target signal are subjected to noise suppression processing to obtain the first standard signal and the second standard signal, wherein the target frequency band is used to indicate the frequency band of the ultra-low frequency audio part in the audio data; The above method is specifically implemented through the following steps: combining the time-frequency characteristics of the time-frequency graph and the signal source scenario, identifying the specific frequency mechanical vibration noise of the first target signal and extracting the center frequency and bandwidth; identifying the wind noise of the second target signal and extracting the wind noise trigger time and duration; wherein, the time-frequency characteristics include the mechanical vibration noise characteristics on the frequency axis of the time-frequency graph and the noise change law over time on the time axis. The three-dimensional time-frequency graphs (including time axis, frequency axis, and sound pressure level axis) corresponding to the first and second target signals generated in the previous steps are called. For the time-frequency graph of the first target signal, by traversing the frequency axis from 20Hz to 100Hz, a fixed frequency component with a sound pressure level that is consistently higher than the surrounding frequencies and has no obvious fluctuations on the time axis is identified, which corresponds to mechanical vibration noise. The center frequency of this noise is located using the peak sound pressure level of the time-frequency graph. Then, based on the center frequency, the frequency range corresponding to when the sound pressure level drops to half of the peak value is calculated to determine the noise bandwidth. For the time-frequency graph of the second target signal, by traversing the time axis, a fluctuating component with irregular sudden rises and falls in sound pressure level and covering multiple ultra-low frequency bands on the frequency axis is identified, which corresponds to wind noise. The moment when the sound pressure level first rises is recorded as the wind noise trigger moment, and the moment when the sound pressure level falls back to the normal level is recorded as the wind noise end moment. The difference between the two is the duration of the wind noise. Thus, combined with the frequency axis, the influence range of wind noise on the ultra-low frequency target band is confirmed. By using contextual features and accurately locating key noise parameters (such as center frequency, bandwidth, trigger time, and duration) in the time-frequency graph, we can avoid the problem of accidentally filtering out ultra-low frequency target signals or not completely filtering out interference during subsequent noise suppression, thus ensuring the accuracy of the entire ultra-low frequency audio measurement process.

[0025] The first target signal is noise-suppressed by narrowband notch filtering and spectrum compensation, and the second target signal is noise-suppressed by adaptive filtering and wavelet threshold denoising. The first and second standard signals in time frame and sound pressure level time sequence format are output. Based on the scene-specific interference differences between the first target signal and the second target signal in step 102, noise suppression is used to remove interference signals while preserving the integrity and authenticity of the ultra-low frequency target signal to the maximum extent, and finally outputting a signal with high signal-to-noise ratio and standardized format. For the first target signal, narrowband notch filtering can selectively filter out the fixed-frequency interference based on the center frequency and bandwidth of the mechanical vibration noise extracted in the above steps, while spectrum compensation can restore the target frequency band signal strength and details that may be lost during the filtering process. For the second target signal, adaptive filtering can dynamically track the irregular fluctuations of wind noise and adjust the filtering parameters in real time to continuously suppress wind noise, while wavelet threshold denoising can further eliminate random interference components in wind noise, ensuring that the second standard signal accurately reflects the propagation and distribution characteristics of the ultra-low frequency signal in the audience area.

[0026] Step 103: Extract the sound source characteristics from the first standard signal to obtain the sound source constraint conditions; Specifically, the above method is implemented through the following steps: by using short-time Fourier transform, the sound pressure level ratio of each frequency point in the target frequency band is statistically analyzed to obtain the frequency distribution at the sound source end; Record the initial phase at each frequency point; The sound source constraint conditions are constituted by the frequency distribution and initial phase at the sound source end; The propagation path of ultra-low frequency (ULF) signals in the audience area is affected by the propagation path. The propagation forms of ULF signals include direct transmission of audio signals, reflection of audio signals when they encounter walls, and standing waves formed by the superposition of direct transmission and reflection. The frequency distribution and phase characteristics may deviate from the original characteristics of the sound source. Therefore, the frequency distribution at the sound source end in the sound source constraint condition can be used as the spectral matching benchmark for the direct sound mode, and the initial phase of each frequency point can be used as the phase verification basis for the direct sound mode. Finally, the distinction and matching calculation of each signal mode can be achieved through this constraint condition, providing accurate sound field data support for the acoustic optimization of the performance.

[0027] Step 104: Use the signal set of the second standard signal as input to construct the sound field data matrix, and decompose and perform low-rank approximation on the sound field data matrix to obtain the low-rank approximation matrix. Specifically, the above method is implemented through the following steps: a matrix is ​​constructed using the signal set of the second standard signal as input to obtain a two-dimensional sound field data matrix, wherein the row dimension of the sound field data matrix is ​​the time frame, the column dimension is the audience area measurement point, and the matrix elements represent the sound pressure level of the corresponding time frame and the audience area measurement point; Singular value decomposition is performed on the sound field data matrix to obtain a left singular vector matrix, a singular value diagonal matrix, and a right singular vector matrix. Each column of the left singular vector matrix is ​​an eigenvector of the time dimension, each row of the right singular vector matrix is ​​an eigenvector of the spatial dimension, and the diagonal elements of the singular value diagonal matrix are non-zero, while the other elements are zero. The diagonal elements of the singular value diagonal matrix are sorted in descending order, and the three largest singular values ​​are selected to form a maximum singular value sequence. The left and right singular vectors in the corresponding left and right singular vector matrices are selected according to the largest singular values, and the eigenvectors are formed by the left and right singular vectors. The magnitude of the singular values ​​of the diagonal elements of the singular value diagonal matrix is ​​used to indicate the degree of contribution of the corresponding eigenvectors to the sound field information. The magnitude of the singular values ​​is positively correlated with the degree of contribution to the sound field information. A low-rank approximate matrix is ​​obtained by constructing a matrix using the maximum singular value and the eigenvector corresponding to the maximum singular value. In the embodiments of the present invention, based on the second standard signal, a two-dimensional sound field data matrix is ​​constructed using the time frame-sound pressure level timing data output from the signal set of the input second standard signal, ensuring that the time frame in the row dimension is completely synchronized with the timing of the second standard signal, and the audience area measurement points in the column dimension correspond to the actual audience area. Singular value decomposition (SVD) is performed on the matrix. The resulting left singular vector matrix is ​​used to obtain the dynamic law of the ultra-low frequency sound field over time, and the right singular vector matrix is ​​used to reflect the spatial distribution pattern of the sound field in the audience area. Only the main diagonal elements of the singular value diagonal matrix are non-zero, and the magnitude of the singular value directly corresponds to the contribution of the eigenvector to the sound field information. The larger the value, the more core information such as direct sound and reflected sound is contained. Then, the three largest singular values ​​are selected to form a sequence. The eigenvectors corresponding to these three largest singular values ​​already contain most of the key information of the ultra-low frequency sound field. By extracting the left and right singular vectors corresponding to the singular values ​​and combining them into eigenvectors, the core sound field information is focused. Finally, a low-rank approximate matrix is ​​constructed through the largest singular values ​​and the corresponding eigenvectors. This not only greatly simplifies the dimension of the original matrix and reduces the computational complexity, but also preserves the core spatiotemporal characteristics of the sound field and avoids redundant data interference.

[0028] Furthermore, after selecting three maximum singular values ​​to form a maximum singular value sequence, the corresponding eigenvectors are grouped in descending order according to the size of the singular values, to obtain the first eigenvector, the second eigenvector, and the third eigenvector in sequence. Since the magnitude of singular values ​​is positively correlated with the contribution of eigenvectors to the sound field information, three maximum singular values ​​and their corresponding eigenvectors are selected. The eigenvectors are grouped from largest to smallest singular values ​​to obtain the first eigenvector, the second eigenvector, and the third eigenvector. This realizes the association between the core sound field components and the eigenvectors. Specifically, the first eigenvector with the highest contribution is used to reflect the direct sound, the second eigenvector with the second highest contribution is used to reflect the reflected sound, and the third eigenvector with the lowest contribution is used to reflect the standing wave. By clearly defining the functional positioning of the feature vectors through grouping, a clear input basis can be provided for subsequent modal matching calculations. At the same time, the information dimension of the low-rank approximation matrix is ​​further simplified, ensuring that the spatiotemporal characteristics of the ultra-low frequency sound field are preserved and redundant information interference is avoided during the subsequent modal decomposition and full-field sound field reconstruction process. Ultimately, this guarantees the accuracy of the low-frequency sound field distribution in the audience area and the calculation of ultra-low frequency audio data.

[0029] Step 105: Based on the low-rank approximation matrix and the sound source constraint, the second standard signal is split into signal modes by mode decomposition, and the signal modes are processed by the low-rank approximation matrix and the sound source constraint to obtain the mode distribution; Specifically, the above method is implemented through the following steps: the right singular vector of the first feature vector is mapped to the direct sound mode, the right singular vector of the second feature vector is mapped to the reflected sound mode, and the right singular vector of the third feature vector is mapped to the standing wave mode. The signal modes include the direct sound mode, the reflected sound mode, and the standing wave mode. Based on the characteristic vectors of each group, and combined with the sound source constraints, the direct sound mode, reflected sound mode, and standing wave mode are matched and calculated respectively to obtain the measurement point mode distribution of the direct sound mode, reflected sound mode, and standing wave mode at the measurement point in the audience area; Based on multiple measurement points in the audience area, the mode distribution of each measurement point is extended to the entire audience area through linear interpolation, generating the direct sound, reflected sound, and standing wave mode distribution of the low-frequency sound field in the audience area; In this embodiment of the invention, the modes are directionally correlated with the feature vectors. The first feature vector right singular vector, which has the highest contribution and corresponds to the main component of the ultra-low frequency sound field, is correlated with the direct sound mode that propagates directly from the stage sound source to the audience area. The second feature vector right singular vector, which has the second highest contribution and corresponds to the minor component, is correlated with the reflected sound mode that propagates after being reflected from the site boundary (such as indoor walls or outdoor ground). The third feature vector right singular vector, which has the lowest contribution and corresponds to the local wave component, is correlated with the standing wave mode formed by the superposition of direct sound and reflected sound. Perform modal matching calculation, call the sound source constraint conditions extracted from the first standard signal in the above steps, calculate the cosine similarity between the frequency distribution of sound pressure level at each measurement point reflected by the right singular vector of the first feature vector and the frequency distribution at the sound source end for the direct sound mode, and at the same time check the difference between the signal at each measurement point and the initial phase of the sound source to ensure that it meets the characteristics of the direct sound spectrum being consistent with the sound source and the phase having no abrupt change, and obtain the direct sound mode distribution at each measurement point; For the reflected sound mode, the time difference between the signal at each measurement point and the signal of the direct sound mode, as reflected by the right singular vector of the second eigenvector, is calculated, and the sound pressure level is less than the sound pressure level of the direct sound. After verification, the distribution of the reflected sound mode at each measurement point is obtained. For standing wave modes, the sound pressure level fluctuation period of each measuring point is calculated as reflected by the right singular vector of the third eigenvector. After locating the antinodes and nodes, the standing wave mode distribution of each measuring point is obtained, and finally a complete mode distribution of the measuring points in the audience area is formed. The entire area is expanded. Based on the modal distribution of the 4-8 sparse measurement points already deployed in the audience area, a grid is constructed with the audience area planar coordinates, the X-axis as the distance from the stage and the Y-axis as the horizontal width. The modal values ​​of the blank areas in the grid are calculated by linear interpolation algorithm.

[0030] Furthermore, by linearly superimposing the distribution results of direct sound mode, reflected sound mode and standing wave mode, the low-frequency sound field distribution of the audience area is obtained, and the ultra-low frequency audio data of the audience area is calculated using the low-frequency sound field distribution results.

[0031] Step 106: Process the modal distribution to obtain the low-frequency sound field distribution, and analyze and calculate the low-frequency sound field distribution to obtain ultra-low frequency audio data.

[0032] Example 2: The present invention provides as follows Figure 2The system shown is an ultra-low frequency audio measurement system for performance occasions. The system includes: The acquisition module is used to acquire the first and second audio data during the performance. The preprocessing module is used to perform abnormal data deletion on the raw audio data acquired by the acquisition module, and to filter the target time-frequency components through the generated time-frequency diagram to obtain the first target signal and the second target signal. The noise suppression module is used to identify the mechanical vibration noise of the first target signal and the wind noise of the second target signal, and output the first standard signal and the second standard signal. The extraction module is used to extract the sound source characteristics from the first standard signal to obtain the sound source constraint conditions; The processing module is used to construct a two-dimensional sound field data matrix using the second standard signal as input, perform singular value decomposition on the matrix and select the maximum singular value and the corresponding eigenvector to construct a low-rank approximate matrix. The decomposition module is used to map the eigenvectors of the low-rank approximation matrix to the modes, and to perform matching calculations on each mode in combination with the sound source constraints to obtain the modal distribution of the measurement point. Then, it is extended to the entire audience area through linear interpolation to generate the modal distribution. The analysis and calculation module is used to linearly superimpose the distributions of direct sound, reflected sound, and standing wave modes to obtain the low-frequency sound field distribution in the audience area, and then analyze and calculate the sound field distribution to obtain ultra-low frequency audio data.

[0033] The ultra-low frequency audio measurement system for performance occasions provided in this embodiment is deployed in a hardware architecture with a processor as the computing core and a memory as the carrier for data and program storage. When the computer program pre-stored in the memory runs on the processor, it can realize the ultra-low frequency audio measurement of the present invention for performance occasions. At the same time, the memory can synchronously store intermediate data and final results of audio measurement. The processor can transmit the measurement results to the staff control terminal in real time according to the requirements of the ultra-low frequency audio measurement system. If remote work is required, data synchronization with remote terminals can also be achieved through the system interface, supporting remote viewing of audio measurement data, assisting multiple people in collaborative debugging of measurement parameters, and improving the practicality and intelligence level of the audio measurement system.

[0034] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A method for measuring ultra-low frequency audio in performance settings, characterized in that, The method includes: First audio data and second audio data of the performance venue are collected respectively. The first audio data is used to indicate the audio signal of the measuring point around the stage, and the second audio data is used to indicate the audio signal of the measuring point in the audience area. Based on the first audio data and the second audio data, audio data corresponding to the target frequency band is extracted to obtain the first target signal and the second target signal. Noise suppression processing is performed on the first target signal and the second target signal to obtain the first standard signal and the second standard signal. The target frequency band is used to indicate the frequency band of the ultra-low frequency audio part in the audio data. The sound source characteristics are extracted from the first standard signal to obtain the sound source constraint conditions; The sound field data matrix is ​​constructed by using the signal set of the second standard signal as input. The sound field data matrix is ​​then subjected to singular value decomposition, and the maximum singular value and the corresponding eigenvector are selected to obtain a low-rank approximate matrix. Based on the low-rank approximation matrix and sound source constraints, the second standard signal is decomposed into direct sound, reflected sound and standing wave signal modes using mode decomposition. The signal modes are then processed using the low-rank approximation matrix and sound source constraints to obtain the mode distribution at the measurement point. Linear interpolation is then used to extend the mode distribution at the measurement point to the entire audience area to generate the mode distribution of the entire area. Based on the modal distribution of the signal modes across the entire region, the low-frequency sound field distribution in the audience area is obtained by linear superposition, and ultra-low frequency audio data is obtained by analyzing and calculating the low-frequency sound field distribution.

2. The ultra-low frequency audio measurement method for performance occasions according to claim 1, characterized in that, The step of extracting audio data corresponding to the target frequency band based on the first audio data and the second audio data to obtain the first target signal and the second target signal includes: The first audio data and the second audio data are traversed by time sequence signal respectively. After identifying and deleting abnormal data, the first audio data and the second audio data are uniformly converted into a time sequence format including time frames and sound pressure levels. Abnormal data includes signal disconnection data, zero value data and interference data. The first audio data and the second audio data are respectively filtered by bandpass filters to obtain the first frequency band signal and the second frequency band signal; Short-time Fourier transforms are performed on the first and second frequency band signals respectively to generate time-frequency diagrams. The time-frequency components of the target frequency band are filtered out by the frequency axis in the time-frequency diagram. The time-frequency diagram is a three-dimensional time-frequency diagram that includes time, frequency and sound pressure level. The time-frequency components include the frequency axis, the time axis and the original audio data. Perform an inverse short-time Fourier transform on the time-frequency components to obtain the first target signal and the second target signal.

3. The ultra-low frequency audio measurement method for performance occasions according to claim 2, characterized in that: The target frequency band is 20Hz-100Hz.

4. The ultra-low frequency audio measurement method for performance occasions according to claim 1, characterized in that, The step of performing noise suppression processing on the first target signal and the second target signal to obtain the first standard signal and the second standard signal includes: Combining the time-frequency characteristics of the time-frequency graph and the signal source scenario, the specific frequency mechanical vibration noise of the first target signal is identified, and the center frequency and bandwidth are extracted. The wind noise of the second target signal is identified, and the triggering time and duration of the wind noise are extracted. The time-frequency characteristics include the mechanical vibration noise characteristics on the frequency axis of the time-frequency graph and the noise variation law over time on the time axis. The first target signal is noise-suppressed by narrowband notch filtering and spectrum compensation, and the second target signal is noise-suppressed by adaptive filtering and wavelet threshold denoising. The output is a first standard signal and a second standard signal in time frame and sound pressure level time sequence format.

5. The ultra-low frequency audio measurement method for performance occasions according to claim 1, characterized in that, The step of extracting sound source characteristics from the first standard signal to obtain sound source constraint conditions includes: By using short-time Fourier transform, the proportion of sound pressure level at each frequency point within the target frequency band is statistically analyzed to obtain the frequency distribution at the sound source end. Record the initial phase at each frequency point; The source constraint conditions are formed by the frequency distribution and initial phase at the source end.

6. The ultra-low frequency audio measurement method for performance occasions according to claim 1, characterized in that: The steps of constructing a sound field data matrix by using the signal set of the second standard signal as input, performing singular value decomposition on the sound field data matrix, selecting the maximum singular value and the corresponding eigenvector, and obtaining a low-rank approximate matrix include: Using the signal set of the second standard signal as input, a matrix is ​​constructed to obtain a two-dimensional sound field data matrix. The row dimension of the sound field data matrix is ​​the time frame, the column dimension is the audience area measurement point, and the matrix elements represent the sound pressure level of the corresponding time frame and audience area measurement point. Singular value decomposition is performed on the sound field data matrix to obtain a left singular vector matrix, a singular value diagonal matrix, and a right singular vector matrix. Each column of the left singular vector matrix is ​​an eigenvector of the time dimension, each row of the right singular vector matrix is ​​an eigenvector of the spatial dimension, and the diagonal elements of the singular value diagonal matrix are non-zero, while the other elements are zero. The diagonal elements of the singular value diagonal matrix are sorted in descending order, and the three largest singular values ​​are selected to form a maximum singular value sequence. The left and right singular vectors in the corresponding left and right singular vector matrices are selected according to the largest singular values, and the eigenvectors are formed by the left and right singular vectors. The magnitude of the singular values ​​of the diagonal elements of the singular value diagonal matrix is ​​used to indicate the degree of contribution of the corresponding eigenvectors to the sound field information. The magnitude of the singular values ​​is positively correlated with the degree of contribution to the sound field information. A low-rank approximate matrix is ​​obtained by constructing a matrix using the maximum singular value and the corresponding eigenvector.

7. The ultra-low frequency audio measurement method for performance occasions according to claim 6, characterized in that: After selecting three maximum singular values ​​to form a maximum singular value sequence, the corresponding feature vectors are grouped in descending order of the singular values ​​to obtain the first feature vector, the second feature vector, and the third feature vector.

8. The ultra-low frequency audio measurement method for performance occasions according to claim 6, characterized in that, The steps of using mode decomposition to split the second standard signal into signal modes based on a low-rank approximation matrix and sound source constraints, and then processing the signal modes using the low-rank approximation matrix and sound source constraints to obtain the mode distribution, include: The right singular vector of the first eigenvector is mapped to the direct sound mode, the right singular vector of the second eigenvector is mapped to the reflected sound mode, and the right singular vector of the third eigenvector is mapped to the standing wave mode. The signal modes include the direct sound mode, the reflected sound mode, and the standing wave mode. Based on the characteristic vectors of each group, and combined with the sound source constraints, the direct sound mode, reflected sound mode, and standing wave mode are matched and calculated respectively to obtain the measurement point mode distribution of the direct sound mode, reflected sound mode, and standing wave mode at the measurement point in the audience area; Based on multiple measurement points in the audience area, the mode distribution of each measurement point is extended to the entire audience area through linear interpolation, generating the direct sound, reflected sound, and standing wave mode distribution of the low-frequency sound field in the audience area.

9. The ultra-low frequency audio measurement method for performance occasions according to claim 8, characterized in that: The low-frequency sound field distribution of the audience area is obtained by linearly superimposing the distribution results of the direct sound mode, reflected sound mode and standing wave mode, and the ultra-low frequency audio data of the audience area is calculated using the low-frequency sound field distribution results.

10. An ultra-low frequency audio measurement system for performance settings, characterized in that, The system includes: The acquisition module is used to acquire first audio data and second audio data during the performance. The first audio data is used to indicate the audio signals of measuring points around the stage, and the second audio data is used to indicate the audio signals of measuring points in the audience area. The preprocessing module is used to perform abnormal data deletion on the first audio data and the second audio data acquired by the acquisition module, and to filter the time-frequency components of the target frequency band through the generated time-frequency diagram to obtain the first target signal and the second target signal. A noise suppression module is used to perform noise suppression processing on a first target signal and a second target signal, and output a first standard signal and a second standard signal. The extraction module is used to extract the sound source characteristics of the first standard signal to obtain the sound source constraint conditions; The processing module is used to construct a two-dimensional sound field data matrix by taking the second standard signal as input, perform singular value decomposition on the matrix and select the maximum singular value and the corresponding eigenvector to construct a low-rank approximate matrix; The decomposition module is used to decompose the second standard signal into signal modes of direct sound, reflected sound and standing wave based on the low-rank approximation matrix and sound source constraints. The signal modes are then processed to obtain the mode distribution of the measurement point. The mode distribution of the measurement point is then extended to the entire audience area through linear interpolation to generate the mode distribution of the entire area. The analysis and calculation module is used to linearly superimpose the modal distribution of the signal modes in the entire area to obtain the low-frequency sound field distribution in the audience area, and then analyze and calculate the sound field distribution to obtain ultra-low frequency audio data.

Citation Information

Patent Citations

  • Reactor abnormal sound identification method and device

    CN114993454A

  • Automatic noise monitoring system and method based on multi-sensor data fusion

    CN120493030A