Tri-axial acoustic sensor directional method and system based on multi-band progressive correlation

Abstract

This invention discloses a direction-finding method and system for a triaxial acoustic sensor based on multi-band progressive correlation, belonging to the field of acoustic detection and signal processing technology. The specific steps of the direction-finding method are as follows: Ⅰ: Acquire acoustic signals from different orthogonal directions in the spatial sound field and divide all acquired acoustic signals into multiple sub-bands; Ⅱ: Preprocess each sub-band, perform direction scanning and confidence calculation, and obtain the optimal estimated direction corresponding to each sub-band. This invention can gradually focus on the main energy frequency band of the signal, avoid the influence of noise and interference in the broadband, effectively eliminate abnormal frequency bands, improve the accuracy of sound source direction estimation, and make the estimation results gradually converge as the frequency band processing progresses, achieving higher direction-finding accuracy, and without the need for multi-station array deployment.

Description

Technical Field

[0001] This invention relates to the field of acoustic detection and signal processing technology, and in particular to a method and system for orientation of a triaxial acoustic sensor based on multi-band asymptotic correlation. Background Technology

[0002] Sound source localization technology is widely used in industrial noise monitoring, intelligent voice interaction, security monitoring and other fields. Traditional sound source localization methods mostly use microphone array beamforming technology, but there are problems such as large array size and complicated deployment process. Moreover, due to the limitation of array aperture, the directivity performance in the low frequency band is not good. On the other hand, the localization method based on single-point triaxial acoustic sensors (such as particle velocity sensors) has become a research hotspot due to its advantages of small size and high portability. These sensors can directly obtain the direction of sound energy flow by measuring the three orthogonal components of particle velocity in the sound field, thereby estimating the location of the sound source.

[0003] However, traditional orientation methods based on single-point triaxial sensors have the following shortcomings: Existing technologies typically employ single-band processing, directly performing vector analysis on full-band signals or a single band. Although they can utilize the covariance information of triaxial signals, they cannot effectively distinguish the frequency components of signals from noise and interference. When the sound source signal contains multiple frequency components, or when there are multipath reflections or broadband noise in the environment, the processing results of a single band are easily contaminated, leading to a significant decrease in orientation accuracy. To address this, we propose a orientation method and system based on a multi-band asymptotic correlation triaxial acoustic sensor. Summary of the Invention

[0004] The purpose of this invention is to address the deficiencies in the prior art by proposing a triaxial acoustic sensor orientation method and system based on multi-band progressive correlation.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: A triaxial acoustic sensor orientation method based on multi-band asymptotic correlation is described in the following steps: Ⅰ: Collect acoustic signals from different orthogonal directions in the spatial sound field, and divide all collected acoustic signals into multiple sub-frequency bands; II: Preprocess each sub-band, perform direction scanning and confidence calculation, and obtain the best estimated direction corresponding to each sub-band; III: Compare the best estimated direction of each sub-band with the cumulative estimated direction of the processed frequency bands and remove abnormal sub-bands that fail the test; IV: Based on the comparative test results of all sub-bands, generate the final source direction estimate and confidence level.

[0006] As a further aspect of the present invention, the specific steps of dividing all the acquired acoustic signals into multiple sub-frequency bands in step I are as follows: S1.1: The particle velocity or sound pressure gradient in the three orthogonal directions of X, Y and Z in the spatial sound field is sensed by a triaxial acoustic sensor, and the measured physical quantity is converted into three independent analog voltage signals. Then, the three analog voltage signals are converted into corresponding digital signals, and finally three-channel time series data corresponding to the X, Y and Z directions are formed. The triaxial acoustic sensor used can be a particle velocity sensor, a sound pressure gradient sensor or an array composed of three orthogonally placed sound pressure sensors. S1.2: Based on the preset specific division method, the overall frequency band of all acquired three-channel time series data is divided in order from narrow to wide, and each frequency band is divided into multiple sub-frequency bands. Then, the specific parameters of each sub-frequency band are obtained, the lower limit frequency and upper limit frequency of each sub-frequency band are determined, and a unique frequency band range is determined. After that, all the divided sub-frequency bands are sorted in order from narrow to wide to generate an ordered sub-frequency band list. The preset specific division method includes octave, fractional octave, and adaptive methods.

[0007] As a further aspect of the present invention, the specific steps for obtaining the optimal estimation direction corresponding to each sub-frequency band in step II are as follows: S2.1: Starting with the narrowest sub-band, process each sub-band sequentially. First, use a Butterworth filter or frequency domain filtering method to bandpass filter the three-channel time series data of each sub-band, extract the signal components within each sub-band, suppress out-of-band noise and interference, and obtain the filtered X, Y, and Z channel signal data. Each channel has a length of... The vectors are denoted as follows: , as well as ; S2.2: Perform zero-mean processing on the filtered three-channel data, and then combine the zero-mean data into a single data set. The matrix is ​​then used to calculate a 3×3 symmetric matrix, which is the covariance matrix. The specific calculation formula for zero-mean processing is as follows: In the formula, These represent the mean values ​​of each channel; where, the... The specific representation of the matrix is ​​as follows ; The specific formula for calculating the covariance matrix is ​​as follows: Expanded to: In the formula, It is the variance of the X channel; It is the covariance of X and Y; S2.3: Decompose the eigenvalues ​​of the covariance matrix and obtain the eigenvectors of each eigenvalue. Then, arrange the eigenvalues ​​in descending order. The corresponding feature vectors are v1, v2, and v3, and then the largest eigenvalue is selected. The corresponding feature vector v1 is used as the main direction of the current sub-band. Spatial grid scanning is performed within the preset azimuth range of 0°~360° and elevation range of -90°~90°, and the scanning direction is calculated. The unit direction vector is then used to calculate the directional consistency factor of the scanning direction using the formula, where the specific formula for calculating the unit direction vector is as follows: ; The specific formula for calculating the directional consistency factor is as follows: ; S2.4: Calculate the polarization linearity parameter according to the formula, and then calculate the polarization linearity parameter for each scanning direction. The confidence level is then determined, and the direction corresponding to the maximum confidence level among all scanning directions is selected. The direction of the best estimation of the sound source in the current sub-band is denoted as , and its corresponding confidence level is denoted as . The specific formula for calculating the polarization linearity parameter is as follows: ; The specific formula for calculating the confidence level is as follows: .

[0008] As a further aspect of the present invention, the specific steps for eliminating abnormal sub-frequency bands that failed the test in step III are as follows: S3.1: Obtain the currently processed sub-band and determine if it is the first sub-band in ascending order of width. If it is the first sub-band, no consistency check is required, and its best estimated direction can be directly determined. As the initial cumulative estimation direction The direction is converted into a unit direction vector as the initial accumulated direction vector. If it is a subsequent sub-band, the best estimated direction of the current sub-band is used. Convert to a unit direction vector, where the specific conversion formula for the unit direction vector is as follows: ; S3.2: Based on the current direction unit vector With the existing cumulative direction unit vector Calculate the angle difference between the two directions, and then compare the calculated angle difference with a preset threshold. If the angle difference is less than the preset threshold, the current sub-band is determined to pass the consistency test; if the angle difference is greater than or equal to the preset threshold, the current sub-band is determined to be an abnormal band and the band is directly removed. S3.3: The vector averaging method is used, and the sub-bands that pass the consistency test are weighted in combination with the confidence level. Then, the new cumulative direction vector is calculated, and the corresponding new cumulative azimuth angle is calculated by back-calculating the three components of the new cumulative direction vector. and new cumulative elevation angle The cumulative confidence score is then updated according to the formula, with the specific steps for creating the new cumulative direction vector as follows: In the formula, This represents the cumulative confidence level, initially set to 0. The specific formula for calculating the updated cumulative confidence level is as follows: .

[0009] As a further aspect of the present invention, the specific steps for generating the final sound source direction estimate and confidence level in step IV are as follows: S4.1: Count the number of processed and unprocessed sub-bands to confirm whether all sub-bands have been processed. Retrieve the latest cumulative estimated direction and cumulative confidence level, and set the convergence verification result to "not satisfied". Then, record the cumulative estimated direction of each update after the processing of the most recent preset number of sub-bands that have passed the consistency check and completed the cumulative update. Calculate the angle difference between two adjacent cumulative estimated directions. If all angle differences are less than the preset change threshold, the convergence condition is deemed to be met. Finally, count the total number of sub-bands that have completed per-band processing up to the present. If this value reaches the preset maximum processing number, the convergence condition is deemed to be met. S4.2: If all sub-bands have been processed, or the cumulative estimation results meet any convergence condition, the final result output process is triggered. From the current cumulative estimation data, the latest cumulative azimuth is extracted as the final azimuth, and the latest cumulative elevation is extracted as the final elevation. At the same time, the corresponding cumulative confidence is obtained. Then, according to the preset output format, the final sound source direction estimate is output. and confidence level ; S4.3: If neither of the two conditions is met, no output is triggered, and the sub-band processing index is shifted one position to the right. The next unprocessed sub-band is selected in order from narrow to wide. At the same time, the process returns to the frequency band processing stage and repeats the entire process of frequency band processing, consistency check and cumulative update until the final result output process is triggered.

[0010] A triaxial acoustic sensor orientation system based on multi-band progressive correlation includes a triaxial acoustic sensor module, a data acquisition module, a data preprocessing module, a progressive frequency band division module, a frequency band-by-frequency processing module, a consistency verification module, an accumulation fusion module, and an output display module. The triaxial acoustic sensor module is used to sense the particle velocity or sound pressure gradient in three orthogonal directions in the spatial sound field and convert it into three independent analog voltage signals. The data acquisition module is used to synchronously sample and convert three independent analog voltage signals into analog-to-digital data to form a three-channel digital time-series data stream. The data preprocessing module is used to preprocess the three-channel digital time series data stream to generate an optimized three-channel time series digital signal; The progressive frequency band division module is used to divide the three-channel time series digital signal according to a preset method, and then generate a sub-frequency band list based on the division results; The frequency band processing module is used to process each sub-frequency band sequentially, extract the signal components within the frequency band, and estimate the optimal estimation direction corresponding to the frequency band. The consistency check module is used to compare the best estimated direction of the current sub-frequency band with the direction accumulated from the previously processed frequency bands, and to eliminate abnormal sub-frequency bands. The accumulation fusion module is used to fuse the sub-frequency band results that have passed the consistency test and update the accumulation direction and confidence. The output display module outputs the final cumulative direction and cumulative confidence level through a human-machine interface or data interface.

[0011] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention uses a triaxial acoustic sensor to collect particle velocity or sound pressure gradients in the X, Y, and Z orthogonal directions of a spatial sound field. These are then converted into three analog voltage signals and subjected to analog-to-digital conversion to form three-channel time-series data. This data is then divided into multiple sub-bands according to a preset method from narrowest to widest. The upper and lower frequency limits of each sub-band are defined, and a sub-band list is generated by sorting the sub-bands by width. Processing begins sequentially from the narrowest sub-band. First, bandpass filtering extracts the signal components. The filtered data is then subjected to zero-mean processing and a matrix is ​​constructed. The covariance matrix is ​​calculated, and eigenvalue decomposition is performed. The eigenvalues ​​are then sorted in descending order, and the eigenvector corresponding to the largest eigenvalue is selected as the main signal direction. Scanning is performed within a preset azimuth and elevation range, calculating the unit vector, direction consistency factor, and polarization linearity parameter for each direction to obtain the confidence level. The direction corresponding to the maximum value is then selected as the best estimated direction for the current sub-band, and the confidence level is recorded. Finally, it is determined whether this is the first sub-band. If it is the first sub-band... If a sub-band is selected, it is directly set as the initial accumulation direction. If it is a subsequent sub-band, the angle difference between it and the accumulation direction is calculated. If the angle difference is less than a preset threshold, it passes the consistency check; otherwise, it is discarded. For the sub-bands that pass the check, a new accumulation direction vector is calculated by weighting the confidence level. The new accumulation azimuth and elevation angles are then deduced and the accumulation confidence level is updated. Finally, the sub-band processing status is statistically analyzed to verify the convergence condition that the change in the estimation results of multiple consecutive sub-bands is less than the threshold or the number of processed sub-bands reaches the maximum value. When all sub-bands have been processed or the convergence condition is met, the final azimuth, elevation angle, and confidence level are extracted and output in a preset format. If the condition is not met, the next sub-band is processed, and the process is repeated until the output is triggered. This method can gradually focus on the main energy frequency band of the signal, avoid the influence of noise and interference in the wideband, effectively remove abnormal frequency bands, improve the accuracy of sound source direction estimation, and make the estimation results gradually converge as the frequency band processing progresses, achieving higher directional accuracy without the need for multi-station array deployment. Attached Figure Description

[0012] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0013] Figure 1 This is a flowchart of the triaxial acoustic sensor orientation method based on multi-band progressive correlation proposed in this invention; Figure 2 This is a flowchart illustrating the processing flow of the triaxial acoustic sensor orientation method based on multi-band progressive correlation proposed in this invention. Figure 3 This is a system block diagram of the triaxial acoustic sensor orientation system based on multi-band progressive correlation proposed in this invention; Figure 4 This is a flowchart illustrating the operation of the triaxial acoustic sensor orientation system based on multi-band progressive correlation proposed in this invention. Detailed Implementation

[0014] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0015] Example 1, referring to Figure 1 and Figure 2 This embodiment discloses a triaxial acoustic sensor orientation method based on multi-band asymptotic correlation. The specific steps of the orientation method are as follows: Acquire acoustic signals from different orthogonal directions in the spatial sound field, and divide all acquired acoustic signals into multiple sub-frequency bands.

[0016] Specifically, a triaxial acoustic sensor senses the particle velocity or sound pressure gradient in the three orthogonal directions (X, Y, and Z) of the spatial sound field, converting the measured physical quantities into three independent analog voltage signals. These analog voltage signals are then converted into corresponding digital signals, ultimately forming three-channel time-series data corresponding to the X, Y, and Z directions. Based on a preset specific division method, the overall frequency band of all acquired three-channel time-series data is divided in ascending order of narrowness, with each frequency band further divided into multiple sub-bands. The specific parameters of each sub-band are then obtained, defining the lower and upper frequency limits of each sub-band to determine a unique frequency range. Finally, all the divided sub-bands are sorted in ascending order of bandwidth to generate an ordered sub-band list.

[0017] It should be further noted that the triaxial acoustic sensor used can be a particle velocity sensor, a sound pressure gradient sensor, or an array of three orthogonally placed sound pressure sensors; the preset specific division methods include octave band, fractional octave band, and adaptive mode.

[0018] Each sub-band is preprocessed, and direction scanning and confidence calculation are performed to obtain the best estimated direction for each sub-band.

[0019] Specifically, starting with the narrowest sub-band, each sub-band is processed sequentially. First, a Butterworth filter or frequency domain filtering method is used to bandpass filter the three-channel time series data of each sub-band, extracting the signal components within each sub-band and suppressing out-of-band noise and interference to obtain the filtered X, Y, and Z channel signal data. Zero-mean processing is then applied to the filtered three-channel data, and finally, the zero-mean data is combined... The matrix is ​​then used to calculate a 3×3 symmetric matrix, which is the covariance matrix. The eigenvalues ​​of the covariance matrix are decomposed, and the eigenvectors of each eigenvalue are obtained. The eigenvalues ​​are then arranged in descending order. The corresponding feature vectors are v1, v2, and v3, and then the largest eigenvalue is selected. The corresponding feature vector v1 is used as the main direction of the current sub-band. Spatial grid scanning is performed within the preset azimuth range of 0°~360° and elevation range of -90°~90°, and the scanning direction is calculated. The unit direction vector is then used to calculate the direction consistency factor of the scanning direction using the formula, and the polarization linearity parameter is calculated using the formula. Finally, the calculation is performed for each scanning direction. The confidence level is then determined, and the direction corresponding to the maximum confidence level among all scanning directions is selected. The direction of the best estimation of the sound source in the current sub-band is denoted as , and its corresponding confidence level is denoted as . .

[0020] It should be further noted that each channel is [length missing]. The vectors are denoted as follows: , as well as .

[0021] It should be further explained that the specific calculation formula for zero-mean processing is as follows: In the formula, These represent the mean values ​​of each channel; where, the... The specific representation of the matrix is ​​as follows ; The specific formula for calculating the covariance matrix is ​​as follows: Expanded to: In the formula, It is the variance of the X channel; It is the covariance of X and Y; The specific formula for calculating the unit direction vector is as follows: ; The specific formula for calculating the directional consistency factor is as follows: ; The specific formula for calculating the polarization linearity parameter is as follows: ; The specific formula for calculating the confidence level is as follows: .

[0022] The optimal estimated direction of each sub-band is compared with the cumulative estimated direction of the processed bands for verification, and abnormal sub-bands that fail the verification are eliminated.

[0023] Specifically, the currently processed sub-band is obtained, and it is determined whether it is the first sub-band in ascending order of narrowness. If it is the first sub-band, there is no need to perform a consistency check, and its best estimated direction is directly obtained. As the initial cumulative estimation direction The direction is converted into a unit direction vector as the initial accumulated direction vector. If it is a subsequent sub-band, the best estimated direction of the current sub-band is used. Convert to a unit direction vector, based on the current direction unit vector. With the existing cumulative direction unit vector The angle difference between the two directions is calculated and then compared with a preset threshold. If the angle difference is less than the preset threshold, the current sub-band is deemed to have passed the consistency test; if the angle difference is greater than or equal to the preset threshold, the current sub-band is deemed an abnormal band and is directly removed. The vector averaging method is used, and the sub-bands that have passed the consistency test are weighted in conjunction with the confidence level. A new cumulative direction vector is then calculated, and the corresponding new cumulative azimuth angle is calculated by back-calculating the three components of the new cumulative direction vector. and new cumulative elevation angle And update the cumulative confidence level according to the formula.

[0024] It should be further explained that the specific conversion formula for the unit direction vector is as follows: ; The specific steps for creating the new cumulative direction vector are as follows: In the formula, This represents the cumulative confidence level, initially set to 0. The specific formula for calculating the updated cumulative confidence level is as follows: .

[0025] Based on the comparative test results of all sub-bands, the final source direction estimate and confidence level are generated.

[0026] Specifically, the process involves counting the number of processed and unprocessed sub-bands to confirm whether all sub-bands have been processed. The latest cumulative estimated direction and cumulative confidence level are retrieved, and the convergence verification result is set to "not satisfied." Then, the cumulative estimated direction of each update is recorded after processing the most recently preset number of sub-bands that have passed the consistency check and completed cumulative updates. The angle difference between two adjacent cumulative estimated directions is then calculated. If all angle differences are less than a preset change threshold, the convergence condition is considered met. Finally, the total number of sub-bands that have completed band-by-band processing is counted. If this number reaches the preset maximum processing number, the convergence condition is considered met. If all sub-bands have been processed, or the cumulative estimation result meets any convergence condition, the final result output process is triggered. From the current cumulative estimation data, the latest cumulative azimuth is extracted as the final azimuth, and the latest cumulative elevation is extracted as the final elevation. The corresponding cumulative confidence level is also obtained. Finally, the final sound source direction estimate is output according to a preset output format. and confidence level If neither of the two conditions is met, no output is triggered, and the sub-band processing index is shifted one position to the right. The next unprocessed sub-band is selected in order from narrow to wide. At the same time, the process returns to the per-band processing stage and repeats the entire process of per-band processing, consistency check and cumulative update until the final result output process is triggered.

[0027] Example 2, refer to Figure 3 and Figure 4 This embodiment discloses a triaxial acoustic sensor orientation system based on multi-band progressive correlation, including a triaxial acoustic sensor module, a data acquisition module, a data preprocessing module, a progressive frequency band division module, a frequency band processing module, a consistency verification module, an accumulation fusion module, and an output display module; The triaxial acoustic sensor module is used to sense the particle velocity or sound pressure gradient in three orthogonal directions in a spatial sound field and convert it into three independent analog voltage signals.

[0028] The data acquisition module is used to synchronously sample and convert three independent analog voltage signals into analog-to-digital signals, forming a three-channel digital time series data stream; the data preprocessing module is used to preprocess the three-channel digital time series data stream to generate an optimized three-channel time series digital signal.

[0029] The progressive frequency band division module is used to divide the three-channel time-series digital signal according to a preset method, and then generate a sub-frequency band list based on the division results; the frequency band processing module is used to process each sub-frequency band in turn, extract the signal components in the frequency band, and estimate the best estimation direction corresponding to the frequency band.

[0030] The consistency check module is used to compare the best estimated direction of the current sub-band with the direction accumulated from the previously processed frequency bands, and to eliminate abnormal sub-bands.

[0031] The cumulative fusion module is used to fuse the sub-band results that have passed the consistency test and update the cumulative direction and confidence level.

[0032] The output display module outputs the final cumulative direction and cumulative confidence level through a human-machine interface or data interface.

Claims

1. A triaxial acoustic sensor orientation method based on multi-band asymptotic correlation, characterized in that, The specific steps of this targeting method are as follows: Ⅰ: Collect acoustic signals from different orthogonal directions in the spatial sound field, and divide all collected acoustic signals into multiple sub-frequency bands; II: Preprocess each sub-band, perform direction scanning and confidence calculation, and obtain the best estimated direction corresponding to each sub-band; III: Compare the best estimated direction of each sub-band with the cumulative estimated direction of the processed frequency bands and remove abnormal sub-bands that fail the test; IV: Based on the comparative test results of all sub-bands, generate the final source direction estimate and confidence level.

2. The triaxial acoustic sensor orientation method based on multi-band asymptotic correlation according to claim 1, characterized in that, The specific steps for dividing all the acquired acoustic signals into multiple sub-frequency bands as described in step I are as follows: S1.1: The particle velocity or sound pressure gradient in the three orthogonal directions of X, Y and Z in the spatial sound field is sensed by a triaxial acoustic sensor, and the measured physical quantity is converted into three independent analog voltage signals. Then, the three analog voltage signals are converted into corresponding digital signals, and finally three-channel time series data corresponding to the X, Y and Z directions are formed. The triaxial acoustic sensor used can be a particle velocity sensor, a sound pressure gradient sensor or an array composed of three orthogonally placed sound pressure sensors. S1.2: Based on the preset specific division method, the overall frequency band of all acquired three-channel time series data is divided in order from narrow to wide, and each frequency band is divided into multiple sub-frequency bands. Then, the specific parameters of each sub-frequency band are obtained, the lower limit frequency and upper limit frequency of each sub-frequency band are determined, and a unique frequency band range is determined. After that, all the divided sub-frequency bands are sorted in order from narrow to wide to generate an ordered sub-frequency band list. The preset specific division method includes octave, fractional octave, and adaptive methods.

3. The triaxial acoustic sensor orientation method based on multi-band asymptotic correlation according to claim 2, characterized in that, The specific steps for obtaining the optimal estimation direction for each sub-band in step II are as follows: S2.1: Starting with the narrowest sub-band, process each sub-band sequentially. First, use a Butterworth filter or frequency domain filtering method to bandpass filter the three-channel time series data of each sub-band, extract the signal components within each sub-band, suppress out-of-band noise and interference, and obtain the filtered X, Y, and Z channel signal data. Each channel has a length of... The vectors are denoted as follows: , as well as ; S2.2: Perform zero-mean processing on the filtered three-channel data, and then combine the zero-mean data into a single data set. The matrix is ​​then used to calculate a 3×3 symmetric matrix, which is the covariance matrix. The specific calculation formula for zero-mean processing is as follows: In the formula, These represent the mean values ​​of each channel; where, the... The specific representation of the matrix is ​​as follows ; The specific formula for calculating the covariance matrix is ​​as follows: Expanded to: In the formula, It is the variance of the X channel; It is the covariance of X and Y; S2.3: Decompose the eigenvalues ​​of the covariance matrix and obtain the eigenvectors of each eigenvalue. Then, arrange the eigenvalues ​​in descending order. The corresponding feature vectors are v1, v2, and v3, and then the largest eigenvalue is selected. The corresponding feature vector v1 is used as the main direction of the current sub-band. Spatial grid scanning is performed within the preset azimuth range of 0°~360° and elevation range of -90°~90°, and the scanning direction is calculated. The unit direction vector is then used to calculate the directional consistency factor of the scanning direction using the formula, where the specific formula for calculating the unit direction vector is as follows: ； The specific formula for calculating the directional consistency factor is as follows: ； S2.4: Calculate the polarization linearity parameter according to the formula, and then calculate the polarization linearity parameter for each scanning direction. The confidence level is then determined, and the direction corresponding to the maximum confidence level among all scanning directions is selected. The direction of the best estimation of the sound source in the current sub-band is denoted as , and its corresponding confidence level is denoted as . The specific formula for calculating the polarization linearity parameter is as follows: ； The specific formula for calculating the confidence level is as follows: 。 4. The triaxial acoustic sensor orientation method based on multi-band progressive correlation according to claim 3, characterized in that, The specific steps for eliminating the abnormal sub-frequency bands that failed the test in step III are as follows: S3.1: Obtain the currently processed sub-band and determine if it is the first sub-band in ascending order of width. If it is the first sub-band, no consistency check is required, and its best estimated direction can be directly obtained. As the initial cumulative estimation direction The direction is converted into a unit direction vector as the initial accumulated direction vector. If it is a subsequent sub-band, the best estimated direction of the current sub-band is used. Convert to a unit direction vector, where the specific conversion formula for the unit direction vector is as follows: ； S3.2: Based on the current direction unit vector With the existing cumulative direction unit vector Calculate the angle difference between the two directions, and then compare the calculated angle difference with a preset threshold. If the angle difference is less than the preset threshold, the current sub-band is determined to pass the consistency test; if the angle difference is greater than or equal to the preset threshold, the current sub-band is determined to be an abnormal band and the band is directly removed. S3.3: The vector averaging method is used, and the sub-bands that pass the consistency test are weighted in combination with the confidence level. Then, the new cumulative direction vector is calculated, and the corresponding new cumulative azimuth angle is calculated by back-calculating the three components of the new cumulative direction vector. and new cumulative elevation angle The cumulative confidence score is then updated according to the formula, with the specific steps for creating the new cumulative direction vector as follows: In the formula, This represents the cumulative confidence level, initially set to 0. The specific formula for calculating the updated cumulative confidence level is as follows: 。 5. The triaxial acoustic sensor orientation method based on multi-band asymptotic correlation according to claim 4, characterized in that, The specific steps for generating the final sound source direction estimate and confidence level in step IV are as follows: S4.1: Count the number of processed and unprocessed sub-bands to confirm whether all sub-bands have been processed. Retrieve the latest cumulative estimated direction and cumulative confidence level, and set the convergence verification result to "not satisfied". Then, record the cumulative estimated direction of each update after the processing of the most recent preset number of sub-bands that have passed the consistency check and completed the cumulative update. Calculate the angle difference between two adjacent cumulative estimated directions. If all angle differences are less than the preset change threshold, the convergence condition is deemed to be met. Finally, count the total number of sub-bands that have completed per-band processing up to the present. If this value reaches the preset maximum processing number, the convergence condition is deemed to be met. S4.2: If all sub-bands have been processed, or the cumulative estimation results meet any convergence condition, the final result output process is triggered. From the current cumulative estimation data, the latest cumulative azimuth is extracted as the final azimuth, and the latest cumulative elevation is extracted as the final elevation. At the same time, the corresponding cumulative confidence is obtained. Then, according to the preset output format, the final sound source direction estimate is output. and confidence level ; S4.3: If neither of the two conditions is met, no output is triggered, and the sub-band processing index is shifted one position to the right. The next unprocessed sub-band is selected in order from narrow to wide. At the same time, the process returns to the frequency band processing stage and repeats the entire process of frequency band processing, consistency check and cumulative update until the final result output process is triggered.

6. A triaxial acoustic sensor orientation system based on multi-band progressive correlation, used to implement the triaxial acoustic sensor orientation method based on multi-band progressive correlation as described in any one of claims 1-5, characterized in that, It includes a triaxial acoustic sensor module, a data acquisition module, a data preprocessing module, a progressive frequency band division module, a frequency band processing module, a consistency verification module, a cumulative fusion module, and an output display module; The triaxial acoustic sensor module is used to sense the particle velocity or sound pressure gradient in three orthogonal directions in the spatial sound field and convert it into three independent analog voltage signals. The data acquisition module is used to synchronously sample and convert three independent analog voltage signals into analog-to-digital data to form a three-channel digital time-series data stream. The data preprocessing module is used to preprocess the three-channel digital time series data stream to generate an optimized three-channel time series digital signal; The progressive frequency band division module is used to divide the three-channel time series digital signal according to a preset method, and then generate a sub-frequency band list based on the division results; The frequency band processing module is used to process each sub-frequency band sequentially, extract the signal components within the frequency band, and estimate the optimal estimation direction corresponding to the frequency band. The consistency check module is used to compare the best estimated direction of the current sub-frequency band with the direction accumulated from the previously processed frequency bands, and to eliminate abnormal sub-frequency bands. The accumulation fusion module is used to fuse the sub-frequency band results that have passed the consistency test and update the accumulation direction and confidence. The output display module outputs the final cumulative direction and cumulative confidence level through a human-machine interface or data interface.

Images