Position intersection and time delay difference fusion positioning method based on inverse variance adaptive weighting

Abstract

The application discloses a position intersection and time delay difference fusion positioning method based on inverse variance adaptive weighting, which is used for position intersection positioning of a sound source signal to obtain a position intersection positioning estimation result; the sound source signal is positioned by using a time delay difference to obtain a time delay difference positioning estimation result; a position intersection positioning error variance and a time delay difference positioning error variance are estimated in real time; inverse variances of the position intersection positioning error variance and the time delay difference positioning error variance are determined to obtain corresponding adaptive fusion weights, an adaptive weighted fusion model is constructed, the adaptive fusion weights are updated in real time, and a final fusion positioning result is obtained. The TDOA positioning process is constrained by fully utilizing angle information, which is helpful to reduce the influence of sound velocity uncertainty and synchronization error on the positioning result. Meanwhile, time information provided by the TDOA can eliminate the geometric degradation problem in the position intersection to a certain extent, and the adaptability and stability in a complex environment are improved.

Description

Technical Field

[0001] This invention belongs to the field of signal processing and sound source localization technology. The invention relates to a localization method based on inverse variance adaptive weighting and fusion of azimuth intersection and time delay difference. Background Technology

[0002] In modern sonar, radar, navigation, and wireless communication systems, target localization technology is a core component for target identification, tracking, and navigation control. Passive sound source localization, due to its high concealment and low energy consumption, is widely used in military reconnaissance, marine monitoring, intelligent transportation, and wireless positioning. Traditional passive localization methods mainly include bearing-only localization and time difference of arrival (TDOA) based methods. Both methods have their advantages but also significant limitations, especially in complex environments where their performance is significantly affected by factors such as changes in sound speed, synchronization errors, and sensor placement.

[0003] Pure azimuth intersection positioning measures the incident azimuth angle of a target using multiple spatially distributed sensor arrays and determines the target's position through geometric intersection. This method has advantages such as simplicity of implementation and low time synchronization requirements, making it particularly suitable for scenarios where the target is far away and the speed of sound is difficult to measure. However, its positioning accuracy is highly dependent on the geometric relationship between the sensors, sensitive to angle measurement errors, and when multiple azimuth lines are nearly collinear, geometric degradation occurs, leading to a significant amplification of the positioning error.

[0004] The TDOA method infers the target position based on the time difference between the sound source signals received by multiple sensors, possessing high theoretical positioning accuracy and suitable for real-time tracking of dynamic targets. However, this method relies on high-precision time synchronization between sensors and requires accurate knowledge of the speed of sound or electromagnetic waves in the propagation medium. In practical applications, the speed of sound is affected by factors such as ambient temperature, humidity, air pressure, and wind speed, exhibiting significant spatiotemporal nonuniformity, often making it difficult to obtain accurate prior information. Furthermore, multipath propagation, non-line-of-sight effects, and signal attenuation can also significantly interfere with time delay estimation. Summary of the Invention

[0005] The purpose of this invention is to provide a azimuth intersection and time delay difference (TDOA) fusion positioning method based on inverse variance adaptive weighting. This method fully utilizes angle information to constrain the TDOA positioning process, helping to reduce the impact of sound speed uncertainty and synchronization error on the positioning results. Simultaneously, the time information provided by TDOA can also, to some extent, eliminate the geometric degradation problem in azimuth intersection, improving adaptability and stability in complex environments.

[0006] The technical solution to achieve the purpose of this invention is as follows: A localization method based on inverse variance adaptive weighting and fusion of azimuth intersection and time delay difference includes the following steps: A cross-shaped sensor array is constructed, the time delay between acoustic sensors is extracted, the azimuth angle of the sound source is calculated based on the time delay and the geometric relationship between the sensor array, and azimuth intersection localization is performed to obtain the azimuth intersection localization estimation result. To locate the sound source signal by time delay difference, the time delay difference between acoustic sensors is calculated. The difference between the observed time delay difference and the calculated time delay difference is used as the residual. The objective function composed of all residuals is optimized using nonlinear least squares to obtain the time delay difference localization estimation result. Real-time estimation of the variance of the azimuth intersection positioning error and the variance of the time delay difference positioning error; The corresponding adaptive fusion weights are determined based on the inverse variances of the azimuth intersection positioning error variance and the time delay difference positioning error variance. The azimuth intersection positioning estimation result and the time delay difference positioning estimation result are then weighted and fused to construct an adaptive weighted fusion model. The adaptive fusion weights are then updated in real time to obtain the final fused positioning result.

[0007] In the preferred technical solution, the azimuth intersection positioning estimation results include: A four-element cross array sensor array is constructed, with four acoustic sensors symmetrically arranged along two mutually perpendicular axes. Acquire the sound source signals received by each acoustic sensor, and perform a Fourier transform on the sound source signals to obtain frequency domain signals; The cross-power spectrum between sensor pairs was calculated using the phase transform generalized cross-correlation algorithm. , in, This is the frequency domain signal of sensor 1. This is the complex conjugate of the frequency domain signal of sensor 3. Indicates the modulus; Fourier inversion is performed on the cross-power spectrum to obtain the cross-correlation function, and the time delay estimate between sensors 1 and 3 is determined based on the peak position of the cross-correlation function. ; Obtaining the azimuth angle of a signal using geometric relationships and sound wave transmission characteristics:

[0008] in, For the speed of sound propagation, The distance between the two sensors; A set of linear equations constraining the location of sound sources is constructed based on the arrival angles of multiple sound sources, and the location of the sound sources is solved by the least squares method to obtain the azimuth intersection positioning estimation result.

[0009] In the preferred technical solution, the time delay difference positioning estimation results include: Using a reference sensor as a benchmark, the TDOA values ​​of other sensors relative to the reference sensor are obtained as follows: For the first i The sensors have the following distance relationships:

[0010] in, Represents the norm, Location of the sound source For the first i The spatial coordinates of the sensors, according to the definition of TDOA, are:

[0011] in, For the speed of sound propagation, The distance between the first sensor and the sound source; The actual observed TDOA is , and The difference is the residual, and the residual for each sensor is:

[0012] in, The spatial coordinates of the first sensor; The objective function composed of all residuals is:

[0013] Where N is the number of sensors; Using a nonlinear least squares optimization algorithm, the estimated sound source location is iteratively updated to achieve the objective function. To reach the minimum:

[0014] Output the final optimal solution. Estimated location of the sound source .

[0015] In the preferred technical solution, the real-time estimation of the variance of the azimuth intersection positioning error and the variance of the positioning error based on the time delay difference include: Estimating the variance of angle measurements :

[0016] Where SNR is the signal-to-noise ratio. These are system calibration constants; Estimated delay and estimated variance :

[0017] in, For signal duration, The effective bandwidth of the signal; The sound velocity profile is measured in real time using a temperature-depth-salt probe integrated in a sensor array, and its short-time statistical variance is calculated to estimate the sound velocity variance. Alternatively, the variance of the speed of sound can be estimated based on prior environmental knowledge. ; Variance of azimuth intersection positioning error:

[0018] in, For sensor spacing, A constant factor related to the geometric layout; Time delay difference positioning error variance:

[0019] in, For reference delay, This is a coefficient related to the sensor's geometric dilution accuracy factor.

[0020] In the preferred technical solutions, the final fusion positioning results include: The corresponding adaptive fusion weights are dynamically determined based on the inverse variance relationship between the real-time updated azimuth intersection positioning error variance and the time delay difference positioning error variance. The adaptive fusion weights corresponding to the azimuth intersection positioning are as follows:

[0021] The adaptive fusion weights corresponding to the time delay difference positioning are:

[0022] in, The variance of the orientation intersection positioning error. The time delay difference represents the positioning error variance; Based on the adaptive fusion weights, the azimuth intersection localization estimation result and the time delay difference localization estimation result are weighted and fused to obtain the fused sound source location as follows:

[0023] in, The location of the sound source is estimated for azimuth intersection positioning. The location of the sound source is estimated based on the time delay difference. The adaptive fusion weights and fusion localization results are dynamically updated based on continuously acquired sound source observation data.

[0024] This invention also discloses a azimuth intersection and time delay difference fusion positioning system based on inverse variance adaptive weighting, used to implement the aforementioned azimuth intersection and time delay difference fusion positioning method based on inverse variance adaptive weighting, comprising: The azimuth intersection localization estimation module constructs a cross-shaped sensor array, extracts the time delay between acoustic sensors, calculates the azimuth angle of the sound source based on the time delay and the geometric relationship between the sensor array, performs azimuth intersection localization, and obtains the azimuth intersection localization estimation result. The time delay difference localization estimation module locates the sound source signal by time delay difference, calculates the time delay difference between acoustic sensors, uses the difference between the observed time delay difference and the calculated time delay difference as the residual, and uses nonlinear least squares optimization on the objective function composed of all residuals to obtain the time delay difference localization estimation result. The positioning error variance calculation module estimates the variance of the azimuth intersection positioning error and the variance of the time delay difference positioning error in real time. The fusion positioning module determines the corresponding adaptive fusion weights based on the inverse variances of the azimuth intersection positioning error variance and the time delay difference positioning error variance. It then performs weighted fusion on the azimuth intersection positioning estimation result and the time delay difference positioning estimation result to construct an adaptive weighted fusion model. The adaptive fusion weights are then updated in real time to obtain the final fusion positioning result.

[0025] The present invention also discloses a computer storage medium storing a computer program, wherein when the computer executes the computer program, it implements the orientation intersection and time delay difference fusion positioning method based on inverse variance adaptive weighting as described above.

[0026] The present invention also discloses an electronic device, including a memory and a processor. The memory stores a computer program, and the processor runs the computer program stored in the memory. When the computer program is executed, it implements the orientation intersection and time delay difference fusion positioning method based on inverse variance adaptive weighting as described above.

[0027] Compared with the prior art, the significant advantages of this invention are: This invention achieves higher positioning accuracy and stronger robustness than existing pure orientation intersection and time delay difference methods by integrating the results of pure orientation intersection and time delay difference-based positioning methods. It is not only adaptable to complex environments, but also has good versatility and high engineering application value.

[0028] Compared to fixed-weighted average fusion or weighting based on empirical upper limits, this adaptive weighting method dynamically adjusts the fusion weights according to real-time environmental changes: when the estimation accuracy of a method is high (small variance), its weight automatically increases; conversely, when the accuracy of a method deteriorates due to noise or geometric degradation, its weight automatically decreases. This maintains optimal fusion performance across all scenarios, significantly improving the system's robustness and adaptability. Attached Figure Description

[0029] Figure 1 This is a flowchart of the orientation intersection and time delay difference fusion positioning method based on inverse variance adaptive weighting in this embodiment; Figure 2 This is a structural diagram of the cross array sensor in this embodiment; Figure 3 This is a comparison chart of the errors of the three positioning methods in this embodiment; Figure 4 This is a schematic diagram of the orientation intersection and time delay difference fusion positioning system based on inverse variance adaptive weighting in this embodiment. Detailed Implementation

[0030] The principle of this invention is to overcome the shortcomings of existing single positioning methods by fusing the azimuth intersection and TDOA positioning results, using angle information to constrain the TDOA positioning process, and using time information to eliminate the geometric degradation problem in the azimuth intersection, thereby ultimately improving positioning accuracy and system adaptability.

[0031] Example 1: like Figure 1 As shown, a azimuth intersection and time delay difference fusion positioning method based on inverse variance adaptive weighting includes the following steps: A cross-shaped sensor array is constructed, the time delay between acoustic sensors is extracted, the azimuth angle of the sound source is calculated based on the time delay and the geometric relationship between the sensor array, and azimuth intersection localization is performed to obtain the azimuth intersection localization estimation result. To locate the sound source signal by time delay difference, the time delay difference between acoustic sensors is calculated. The difference between the observed time delay difference and the calculated time delay difference is used as the residual. The objective function composed of all residuals is optimized using nonlinear least squares to obtain the time delay difference localization estimation result. Real-time estimation of the variance of the azimuth intersection positioning error and the variance of the time delay difference positioning error; The corresponding adaptive fusion weights are determined based on the inverse variances of the azimuth intersection positioning error variance and the time delay difference positioning error variance. The azimuth intersection positioning estimation result and the time delay difference positioning estimation result are then weighted and fused to construct an adaptive weighted fusion model. The adaptive fusion weights are then updated in real time to obtain the final fused positioning result.

[0032] The azimuth convergence localization here can be called azimuth convergence localization based on array angle of arrival estimation, or simply azimuth convergence localization. The estimation result of azimuth convergence localization is: The corresponding error variance is The estimation result based on time delay difference positioning is: The corresponding error variance is The adaptive weight is then defined as:

[0033] The location of the merged sound source is:

[0034] Among them, error variance and It is not a fixed constant, but is obtained by real-time dynamic estimation based on the current signal quality and environmental parameters.

[0035] Specifically: The variance estimate of the error in azimuth intersection positioning is: ,in The variance of the angle measurement is obtained by mapping the Cramér-Rao lower bound or the measured signal-to-noise ratio in array signal processing. For sensor spacing, For a geometric layout constant factor (for a cross matrix), Take 1.5 to 2.0).

[0036] The error variance estimate for time delay difference-based positioning is: ,in To estimate the variance of the time delay, For the variance of sound speed, For the speed of sound propagation, The reference delay (which can be the average delay of each sensor) is used as a reference. This is a coefficient related to the sensor's geometric dilution accuracy factor (GDOP).

[0037] Compared to fixed-weighted average fusion or weighting based on empirical upper limits, this adaptive weighting method dynamically adjusts the fusion weights according to real-time environmental changes: when the estimation accuracy of a method is high (small variance), its weight automatically increases; conversely, when the accuracy of a method deteriorates due to noise or geometric degradation, its weight automatically decreases. This maintains optimal fusion performance across all scenarios, significantly improving the system's robustness and adaptability.

[0038] The following is a specific example illustrating the process of the orientation intersection and time delay difference fusion positioning method based on inverse variance adaptive weighting, including the following steps: Step S1 involves using a four-element cross array sensor array to locate the sound source signal by azimuth intersection.

[0039] First, a typical cross-shaped sensor structure is built, such as... Figure 2 As shown. The cross-shaped sensor array consists of two mutually perpendicular supports of equal length. An acoustic sensor, numbered 1 to 4, is mounted at each end of each support. The origin of the coordinate system is set at the intersection point, and the sensors are distributed along the horizontal and vertical axes, forming a stable and symmetrical spatial structure. Regarding the signal model, assuming a single sound source in the far field emits a signal, the signals received by the four sensors can be represented as:

[0040] in, Let m be the signal attenuation factor at sensor m. The time-domain signal received by sensor m, where n represents time. Original sound source signal The signal propagated to sensor m, This represents the propagation time from the sound source to the sensor m. The noise is Gaussian white noise, and it is assumed that the noise of each sensor is uncorrelated.

[0041] The time delay difference is extracted using the Phase Transform Generalized Cross-Correlation (PHAT-GCC) method. First, a Fourier transform is performed on the received signal, resulting in the signal's frequency domain model as follows:

[0042] in, Represents angular frequency. Represents the imaginary unit. This represents the modeling of the original sound source signal in the frequency domain. This represents the modeling of Gaussian white noise in the frequency domain.

[0043] The frequency domain phase-transform cross power spectrum of one pair of sensors (e.g., 1 and 3) is calculated as follows:

[0044] in, Indicates complex conjugation. This represents the magnitude. This processing essentially preserves the signal phase difference while discarding the amplitude component, exhibiting significant robustness. It is particularly suitable for extracting phase information in high-noise or non-ideal attenuation environments and is a concrete manifestation of the PHAT (Phase Transform) weighting strategy in the generalized cross-correlation (GCC) method.

[0045] Phase-transformed cross power spectrum Performing an inverse Fourier transform, the cross-correlation function between the two signals is obtained as follows:

[0046] in, For reference delay.

[0047] Furthermore, the time delay between the two sensors is estimated by finding the location of the maximum peak in the cross-correlation function:

[0048] The azimuth angle of the signal can be obtained by calculating the time delay based on geometric relationships and the characteristics of sound wave propagation. :

[0049] in, For the speed of sound propagation, Let be the distance between the two sensors. From this, a linear equation can be established from the sound source to the sensor. Let the location of a sound source on the plane be . , No. The location of each sensor is The equation can be obtained as follows:

[0050] Then we can conclude Equations of a straight line:

[0051] The equation can be converted into matrix form and solved using the least squares method:

[0052] The solution obtained is the location of the sound source. .

[0053] Step S2 involves using a four-element cross array sensor array to locate the sound source signal based on time delay difference (TDOA).

[0054] Assume there are N cross-shaped sensors in the sensor array, and the spatial coordinates of the i-th sensor are known. The location of the sound source is Let the speed of sound (the speed of sound propagation) be... Taking the first sensor as a reference, the TDOA of the other sensors are respectively... For the first sensor in the sensor array i The sensors have the following distance relationships:

[0055] in, Let represent the norm. According to the definition of TDOA, we have:

[0056] The actual observed TDOA is Theoretical TDOA and The difference between the two is the residual. The residual for each sensor is defined as:

[0057] The objective function (sum of squared residuals) composed of all residuals is:

[0058] Using a nonlinear least squares optimization algorithm, the location of the sound source is first determined. Set an initial estimate (such as the geometric center of the sensor array), and then iteratively update the sound source location estimate to make the objective function... To reach the minimum:

[0059] Output the final optimal solution. Estimated location of the sound source .

[0060] Step S3 is an adaptive weighted fusion based on inverse variance.

[0061] S3.1 Real-time estimation of angle measurement variance

[0062] After estimating the angle of arrival using the PHAT-GCC method for the four received signals, the angle measurement variance is estimated based on the peak sharpness of the cross-correlation function and the signal-to-noise ratio (SNR). This embodiment uses an empirical mapping function:

[0063] in These are the system calibration constants, determined through offline experiments.

[0064] S3.2 Real-time estimation of delay and variance

[0065] Estimation is performed using the second derivative near the main peak of the generalized cross-correlation function or based on the Cramér-Rao lower bound. This embodiment uses the following approximation formula:

[0066] in, Signal duration (in seconds). The effective bandwidth of the signal (unit: Hz). The ratio of received signal power to noise power is calculated in real time.

[0067] S3.3 Estimating the variance of sound speed

[0068] The speed of sound is affected by temperature, salinity, and pressure. The sound speed profile can be measured in real time using a temperature-depth-salinity probe integrated into the sensor array, and its short-time statistical variance can be calculated. If the system is not equipped with a probe, a typical fluctuation range of 1% to 5% of the speed of sound propagation is set based on prior environmental knowledge. In this embodiment... .

[0069] S3.4 Calculate the variance of the positioning error

[0070] Variance of azimuth intersection positioning error:

[0071] For the cross formation, The distance between two relative sensors (e.g., sensors 1 and 3). Take 1.8 (empirical value).

[0072] Based on the time delay difference, the variance of the positioning error is:

[0073] in, Based on the geometric dilution factor of precision (GDOP) of the sensor array, the cross array in this embodiment... Take 2.0; Take the average absolute value of the time delay for each sensor (unit: seconds).

[0074] S3.5 Calculate and fuse adaptive weights

[0075] The above estimates are obtained and Substitute into the weight formula:

[0076] Then calculate the fused localization result:

[0077] S3.6 Dynamic Updates

[0078] Each time a new signal observation (or each frame of data) is acquired, steps S3.1 to S3.5 are repeated to achieve real-time adaptive updates of the weights. In continuous tracking mode, the fusion result of the previous frame can also be used as the initial estimate of the current frame to further improve stability.

[0079] The positioning method of the present invention is compared experimentally with the pure orientation intersection positioning method and the time delay difference-based positioning method.

[0080] To further verify the advantages of the azimuth intersection and time delay difference fusion positioning of the present invention in practical applications, this experiment, under the premise that other laboratory factors remain unchanged, selected signal-to-noise ratios of 10dB, 20dB, and 30dB, and respectively conducted error mean analysis on pure azimuth intersection positioning, time delay difference-based positioning, and the positioning of the present invention.

[0081] Figure 3 The diagram compares the positioning errors of the three positioning methods. As the signal-to-noise ratio (SNR) increases, the positioning errors of all three methods decrease significantly. The proposed fusion positioning method exhibits the best positioning accuracy under all SNR conditions. Specifically, when the SNR is 10 dB, 20 dB, and 30 dB, the error of the fusion positioning method is lower than the other two methods, and the advantage of the fusion positioning method becomes more pronounced as the SNR increases.

[0082] The main reason for the above results is that single positioning methods (such as pure direction intersection positioning or time delay difference-based positioning) are susceptible to environmental noise and system errors. The fusion algorithm proposed in this paper combines two types of positioning information, exhibiting significant advantages in suppressing noise interference and improving positioning robustness. Experimental results further verify the effectiveness and practical value of the fusion positioning method.

[0083] Experimental results fully verify the superiority of the azimuth intersection and time delay difference fusion localization method of this invention in complex environments. This method not only significantly improves the accuracy of azimuth estimation, but also significantly enhances the system's adaptability under different environmental conditions, providing reliable technical support for practical sound source localization applications.

[0084] Example 2

[0085] Figure 4 The figure shown is a azimuth intersection and time delay difference fusion positioning system based on inverse variance adaptive weighting proposed in this embodiment, which includes the following: The azimuth intersection localization estimation module constructs a cross-shaped sensor array, extracts the time delay between acoustic sensors, calculates the azimuth angle of the sound source based on the time delay and the geometric relationship between the sensor array, performs azimuth intersection localization, and obtains the azimuth intersection localization estimation result. The time delay difference localization estimation module locates the sound source signal by time delay difference, calculates the time delay difference between acoustic sensors, uses the difference between the observed time delay difference and the calculated time delay difference as the residual, and uses nonlinear least squares optimization on the objective function composed of all residuals to obtain the time delay difference localization estimation result. The positioning error variance calculation module estimates the variance of the azimuth intersection positioning error and the variance of the time delay difference positioning error in real time. The fusion positioning module determines the corresponding adaptive fusion weights based on the inverse variances of the azimuth intersection positioning error variance and the time delay difference positioning error variance. It then performs weighted fusion on the azimuth intersection positioning estimation result and the time delay difference positioning estimation result to construct an adaptive weighted fusion model. The adaptive fusion weights are then updated in real time to obtain the final fusion positioning result.

[0086] The specific implementation method is the one described above, and will not be repeated here.

[0087] In another embodiment, a computer storage medium stores a computer program thereon, and when the computer executes the computer program, it implements the orientation intersection and time delay difference fusion positioning method based on inverse variance adaptive weighting as described in Embodiment 1 above.

[0088] The specific implementation method is the one described above, and will not be repeated here.

[0089] In another embodiment, an electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor runs the computer program stored in the memory. When the computer program is executed, it implements the orientation intersection and time delay difference fusion positioning method based on inverse variance adaptive weighting as described in Embodiment 1 above.

[0090] The specific implementation method is the one described above, and will not be repeated here.

[0091] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A method for position fusion based on inverse variance adaptive weighting of bearing intersection and time delay difference, characterized in that, Includes the following steps: A cross-shaped sensor array is constructed, the time delay between acoustic sensors is extracted, the azimuth angle of the sound source is calculated based on the time delay and the geometric relationship between the sensor array, and azimuth intersection localization is performed to obtain the azimuth intersection localization estimation result. To locate the sound source signal by time delay difference, the time delay difference between acoustic sensors is calculated. The difference between the observed time delay difference and the calculated time delay difference is used as the residual. The objective function composed of all residuals is optimized using nonlinear least squares to obtain the time delay difference localization estimation result. Real-time estimation of the variance of the azimuth intersection positioning error and the variance of the time delay difference positioning error, including: Estimating the variance of angle measurements : Where SNR is the signal-to-noise ratio. These are system calibration constants; Estimated delay and estimated variance : in, For signal duration, The effective bandwidth of the signal; The sound velocity profile is measured in real time using a temperature-depth-salt probe integrated in a sensor array, and its short-time statistical variance is calculated to estimate the sound velocity variance. Alternatively, the variance of the speed of sound can be estimated based on prior environmental knowledge. ; Variance of azimuth intersection positioning error: in, For sensor spacing, A constant factor related to the geometric layout; Based on the time delay difference, the variance of the positioning error is: in, For reference delay, The coefficient is related to the sensor's geometric dilution accuracy factor; The corresponding adaptive fusion weights are determined based on the inverse variances of the azimuth intersection positioning error variance and the time delay difference positioning error variance. The azimuth intersection positioning estimation result and the time delay difference positioning estimation result are then weighted and fused to construct an adaptive weighted fusion model. The adaptive fusion weights are updated in real time to obtain the final fused positioning result, which includes: The corresponding adaptive fusion weights are dynamically determined based on the inverse variance relationship between the real-time updated azimuth intersection positioning error variance and the time delay difference positioning error variance. The adaptive fusion weights corresponding to the azimuth intersection positioning are as follows: The adaptive fusion weights corresponding to the time delay difference positioning are: in, The variance of the azimuth intersection positioning error. The time delay difference represents the positioning error variance; Based on the adaptive fusion weights, the azimuth intersection localization estimation result and the time delay difference localization estimation result are weighted and fused to obtain the fused sound source location as follows: in, The location of the sound source is estimated for azimuth intersection positioning. The location of the sound source is estimated based on the time delay difference. The adaptive fusion weights and fusion localization results are dynamically updated based on continuously acquired sound source observation data.

2. The orientation intersection and time delay difference fusion positioning method based on inverse variance adaptive weighting according to claim 1, characterized in that, The obtained azimuth intersection positioning estimation results include: A four-element cross array sensor array is constructed, with four acoustic sensors symmetrically arranged along two mutually perpendicular axes. Acquire the sound source signals received by each acoustic sensor, and perform a Fourier transform on the sound source signals to obtain frequency domain signals; The cross-power spectrum between sensor pairs is calculated using a phase-transform generalized cross-correlation algorithm. , in, This is the frequency domain signal of sensor 1. This is the complex conjugate of the frequency domain signal of sensor 3. Indicates the modulus; Perform Fourier inversion on the cross-power spectrum to obtain the cross-correlation function, and determine the time delay estimate between sensors 1 and 3 based on the peak position of the cross-correlation function. ; Obtaining the azimuth angle of a signal using geometric relationships and sound wave transmission characteristics: in, For the speed of sound propagation, The distance between the two sensors; A set of linear equations constraining the location of sound sources is constructed based on the arrival angles of multiple sound sources, and the location of the sound sources is solved by the least squares method to obtain the azimuth intersection positioning estimation result.

3. The orientation intersection and time delay difference fusion positioning method based on inverse variance adaptive weighting according to claim 1, characterized in that, The time delay difference positioning estimation results obtained include: Using a reference sensor as a benchmark, the TDOA values ​​of other sensors relative to the reference sensor are obtained as follows: For the first i The sensors have the following distance relationships: in, Represents the norm, Location of the sound source For the first i The spatial coordinates of the sensors, according to the definition of TDOA, are: in, For the speed of sound propagation, The distance between the first sensor and the sound source; The actual observed TDOA is , and The difference is the residual, and the residual for each sensor is: in, The spatial coordinates of the first sensor; The objective function composed of all residuals is: Where N is the number of sensors; Using a nonlinear least squares optimization algorithm, the estimated sound source location is iteratively updated to achieve the objective function. To reach the minimum: Output the final optimal solution. Estimated location of the sound source .

4. A azimuth intersection and time delay difference fusion positioning system based on inverse variance adaptive weighting, characterized in that, To implement the orientation intersection and time delay difference fusion positioning method based on inverse variance adaptive weighting as described in any one of claims 1-3, the method includes: The azimuth intersection positioning estimation module constructs a cross array sensor array, extracts the time delay between acoustic sensors, calculates the azimuth angle of the sound source based on the time delay and the geometric relationship between the sensor array, performs azimuth intersection positioning, and obtains the azimuth intersection positioning estimation result. The time delay difference localization estimation module locates the sound source signal by time delay difference, calculates the time delay difference between acoustic sensors, uses the difference between the observed time delay difference and the calculated time delay difference as the residual, and uses nonlinear least squares optimization on the objective function composed of all residuals to obtain the time delay difference localization estimation result. The positioning error variance calculation module estimates the variance of the azimuth intersection positioning error and the variance of the time delay difference positioning error in real time. The fusion positioning module determines the corresponding adaptive fusion weights based on the inverse variances of the azimuth intersection positioning error variance and the time delay difference positioning error variance. It then performs weighted fusion on the azimuth intersection positioning estimation results and the time delay difference positioning estimation results to construct an adaptive weighted fusion model. The adaptive fusion weights are then updated in real time to obtain the final fusion positioning result.

5. A computer storage medium having a computer program stored thereon, characterized in that, When the computer executes the computer program, it implements the orientation intersection and time delay difference fusion positioning method based on inverse variance adaptive weighting as described in any one of claims 1-3.

6. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor runs the computer program stored in the memory. When the computer program is executed, it implements the orientation intersection and time delay difference fusion positioning method based on inverse variance adaptive weighting as described in any one of claims 1-3.

Images