A high-resolution imaging method, apparatus, and storage medium for a planar microphone array in the differential domain.

Abstract

This application proposes a high-resolution imaging method, apparatus, and storage medium for a planar microphone array in the differential domain. The method is applied to typical planar sparse array structures such as two-dimensional nested arrays. By vectorizing the covariance matrix of the received signal, the differential comatrix of the original array is obtained. Acoustic imaging is then performed based on this differential comatrix, resulting in a uniform rectangular array with higher degrees of freedom in the differential domain. This allows for high-resolution acoustic imaging with a small number of microphones. Numerical simulation experiments show that, with the same number of microphones, the acoustic imaging localization algorithm based on the differential domain of the nested planar array achieves higher resolution than existing acoustic imaging algorithms that directly locate typical planar arrays.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of acoustic imaging positioning, and particularly relates to a high-resolution imaging method of a planar microphone array. BACKGROUND

[0002] Sound source positioning refers to determining the position of a sound source by using sound signals received by a microphone. At present, it is widely applied in the fields of intelligent transportation, abnormal sound of mechanical equipment, robots, etc. Acoustic imaging technology based on a microphone array is one of the main methods of sound source positioning. It refers to converting the spatial distribution of sound signals into image signals by using a signal processing algorithm according to the multi-channel sound pressure data collected by the microphone array, so as to visualize the spatial distribution of the sound source and quickly locate the position of the sound source. The position of the sound source can be directly obtained by using acoustic imaging for sound source positioning, and therefore it is widely applied.

[0003] Microphone array pattern has been a research hotspot in the field of acoustic imaging. The design of microphone array pattern is mainly to arrange microphones at different positions to improve the imaging resolution, etc. The initial research is almost regular geometric array such as cross array, rectangular array and circular array, etc. However, when using regular geometric array for acoustic imaging, spatial aliasing phenomenon will occur, which will cause grating lobe to interfere with the identification of main lobe. In order to solve the problem of grating lobe, Dougherty proposed irregular single-arm and multi-arm spiral array. In the design of single-arm spiral array, Archimedes spiral array uses spiral formula to adjust the design of spiral radius and number of turns. Snellen et al. analyzed the variation of over-the-top noise level of an airplane by using a 32-element single-arm spiral array. Underbrink multi-arm spiral array is improved on the basis of the exponential multi-arm spiral array proposed by Dougherty. The aperture area of the array is equally divided by each microphone, and there is almost no grating lobe problem in imaging, and the number of microphones can be reduced to a certain extent. Amaral et al. further improved the Underbrink multi-arm spiral array, and different numbers of microphones can be distributed on each ring. The array is more flexible in design, and the dynamic range is greatly improved. Liu Zhe et al. proposed an optimized 64-element spiral array structure, which realizes the positioning of automobile engine noise. Xiong et al. designed a two-dimensional Fibonacci array, which has a lower maximum side lobe level in a certain frequency range, and applied it to the fault diagnosis of track acoustic bearing. Zhang Menghao et al. proposed an optimized wheel-shaped array, and the maximum side lobe level of the imaging is reduced by 3dB compared with the general wheel-shaped array. Although the above arrays can solve the problem of grating lobe, to improve the resolution of imaging, in addition to using more complex algorithms, the main method is to increase the number of microphones or increase the array aperture, such as Siren et al. designed a 118-microphone microphone array to locate the loose transformer core; Kummritz et al. used a Fibonacci array and a star array to measure the noise distribution and source of a train, each array consisted of 120 microphones; Cao et al. used a microphone array with an aperture of 4m to conduct acoustic imaging positioning research on the noise of an 800KV substation.

[0004] In recent years, the proposal of sparse array based on difference domain, such as one-dimensional nested array and one-dimensional coprime array, makes a N-element sparse array achieve O(N 2 ) degrees of freedom, thereby greatly improving the number of target detection. One-dimensional sparse array is rapidly extended to two-dimensional sparse array, such as two-dimensional nested array and hourglass array, which greatly improves the number of target direction of arrival estimation. However, the existing research on two-dimensional sparse array is mainly applied to direction of arrival estimation, and there is little application in the field of acoustic imaging. SUMMARY

[0005] In order to solve the problem that a large number of microphones are needed to obtain higher resolution in the existing acoustic imaging positioning system, the application provides a high-resolution imaging method, device and storage medium of a differential domain planar microphone array, which reduces the number of array elements used in the existing acoustic imaging positioning system and obtains high-resolution acoustic imaging with as few microphones as possible.

[0006] The general idea of the application is to vectorize the covariance matrix of the received signal to obtain a differential common array of the original array, and then perform acoustic imaging based on the differential common array. The two-dimensional nested array and other typical planar sparse array structures can be used to obtain a uniform rectangular array with higher degrees of freedom in the differential domain, so that high-resolution acoustic imaging can be obtained with a small number of microphones.

[0007] The technical solution of the application is as follows:

[0008] In the first aspect, the application provides a high-resolution imaging method of a differential domain planar microphone array, which comprises the following steps:

[0009] Step 1: A typical planar sparse array such as a planar nested array, an hourglass array, etc. is used to collect acoustic signals with a plurality of microphones to obtain array received signals, and the covariance matrix of the array received signals is calculated.

[0010] Step 2: The covariance matrix obtained in step 1 is vectorized to obtain a single snapshot received signal vector of a differential common array, and the differential common array contains overlapping partial array element positions.

[0011] Step 3: The received signals corresponding to the array elements with the same position in the differential common array obtained in step 2 are averaged, and then the single snapshot received signal vector of the differential common array is rearranged in the order of increasing coordinate position, so as to obtain a single snapshot received signal vector of a differential common array without repeated array elements.

[0012] Step 4: The single snapshot received signal of the differential common array without repeated array elements is used as the input signal of beamforming, and the conventional beamforming algorithm is used to complete power scanning in all directions to perform acoustic imaging and positioning of the target sound source and obtain an acoustic imaging map.

[0013] Further, according to an embodiment of the application, the specific processing method of calculating the covariance matrix of the array received signals in step 1 is as follows:

[0014] Assuming that the original array received signal vector is x(t), according to the assumption that the signal and the noise are not correlated, the covariance matrix of the array received signals can be represented as

[0015]

[0016] wherein Let be the covariance matrix of the signal. Let be the power of the i-th signal. Let A be the power of the noise, and A be the array manifold matrix. This is the array guide vector.

[0017] Furthermore, according to one embodiment of this application, step 2 specifically includes:

[0018] Vectorizing the covariance matrix of the received signal yields a new single-shot observation model for the virtual array:

[0019]

[0020] in e i It is a column vector where all positions except the i-th position are zero; z, A * ⊙A、p、 These correspond to the equivalent received signal, array manifold, signal source, and noise signal of the virtual array, respectively.

[0021] The virtual array steering vector obtained from the virtual array observation model is: Therefore, the positions of the virtual array elements are obtained by... The set is given, where Let L be the position vectors of the m-th and n-th array elements, respectively; M is the number of array elements in the original array; the resulting set L is... d That is, the differential co-array of the original array; the observation vector of the differential co-array corresponding to Equation (2) contains elements with repeated array element positions.

[0022] Furthermore, according to one embodiment of this application, step 3 specifically includes:

[0023] The received signals corresponding to the same array element positions in equation (2) are averaged and arranged in ascending order of differential array element positions to remove redundancy and obtain the equivalent received signals. for

[0024]

[0025] in e is a vector whose middle elements are 1 and the rest are 0;

[0026] The differential co-matrix steering vector, after rearranging the positions of the differential array elements, is denoted as...

[0027]

[0028] Where Q is the number of differential co-array elements after removing redundancy. This indicates the position of the q-th difference co-array element.

[0029] Furthermore, according to one embodiment of this application, step 4 specifically includes:

[0030] Let w = [w1w2w] Q ] T Let w be the weight vector, where w q Let Q be the weighting coefficient corresponding to the q-th differential co-array element, where q = 1, 2, ..., Q. The beamforming output expression based on the differential co-array is:

[0031]

[0032] The expression for its weighting coefficient is as follows: The equivalent received signal of the q-th differential co-array element; time delay The azimuth and elevation angles of the scan are respectively, and their ranges are as follows: i and j are the index values ​​of the number of grid points divided by the azimuth and elevation angles, respectively;

[0033] The output power of the differential co-array beamforming is

[0034]

[0035] Where R is the covariance matrix of the differential co-array output signal.

[0036] Finally, the acoustic image of the sound source can be obtained from the output power of the virtual array beamforming in each direction.

[0037] In a second aspect of this application, a high-resolution imaging apparatus for a differential domain planar microphone array is provided, which is used to perform the method described in the first aspect above, comprising:

[0038] The covariance matrix calculation unit is used to collect sound signals by using multiple microphones to form a typical planar sparse array, obtain the array received signal, and calculate the covariance matrix of the array received signal.

[0039] A vectorization unit is used to vectorize the covariance matrix to obtain a single-shot received signal vector of the differential comatrix, wherein some array elements of the differential comatrix overlap.

[0040] The mean processing and vector rearrangement unit is used to perform mean processing on the received signals corresponding to the array elements with the same position in the differential co-array, and then rearrange the single-shot received signal vector of the differential co-array in ascending order of coordinate position, so as to obtain a single-shot received signal vector of the differential co-array without repeating array elements.

[0041] The acoustic imaging acquisition unit is used to take the single-shot received signal of the differential co-array without repeating array elements as the input signal for beamforming, and use the beamforming algorithm to complete the power scan in all directions to perform acoustic imaging and localize the target sound source to obtain an acoustic imaging map.

[0042] In a third aspect, this application also provides a computer-readable storage medium storing computer-executable instructions for causing a computer to perform the steps of the high-resolution imaging method of a planar microphone array in the differential domain as described in the first aspect of this application.

[0043] The advantages and effects of this application are as follows:

[0044] This application proposes a high-resolution imaging method for planar microphone arrays in the differential domain. It obtains the differential array elements corresponding to the nested array in the differential domain of the planar nested array. Noise signals collected by the microphone array at different positions are processed to obtain virtual signals corresponding to the positions of the differential array elements. These virtual signals are then used as input signals for beamforming to perform acoustic imaging and localization of noise. Experimental simulations show that, compared with existing acoustic imaging based on physical array elements, the differential co-array beamforming algorithm based on a two-dimensional planar nested array in the differential domain has better localization effects for both single and multiple noise signals. Compared with acoustic imaging based on the same physical array elements, the method proposed in this application has a narrower main lobe, higher resolution, and can distinguish two nearby sound sources. Even when using a nested array with far fewer array elements than other physical arrays, the method proposed in this application still has a lower main lobe width and can still achieve localization of nearby sound sources, exhibiting strong noise resistance and high spatial resolution. Attached Figure Description

[0045] Figure 1 It is an acoustic imaging signal model;

[0046] Figure 2 This is a schematic diagram of a two-dimensional nested array and its differential array;

[0047] Figure 3 This is a schematic diagram of the formation of differential co-array beams from a physical array;

[0048] Figure 4 These are array element position diagrams for different arrays, where (a) is an Archimedes spiral array, (b) is an annular array, (c) is an involute spiral array, (d) is an 81-element nested array, and (e) is a 36-element nested array.

[0049] Figure 5These are comparison diagrams of the spatial spectrum of a single sound source, where (a) is the spatial spectrum of a single sound source of an Archimedes spiral array, (b) is the spatial spectrum of a single sound source of an annural array, (c) is the spatial spectrum of a single sound source of an involute spiral array, (d) is the spatial spectrum of a single sound source of an 81-element nested array virtual array, and (e) is the spatial spectrum of a single sound source of an 36-element nested array virtual array.

[0050] Figure 6 These are comparison diagrams of the spatial spectra of two sound sources, where (a) is the spatial spectrum of the Archimedes spiral array, (b) is the spatial spectrum of the annural array, (c) is the spatial spectrum of the involute spiral array, (d) is the spatial spectrum of the 81-element nested array virtual array, and (e) is the spatial spectrum of the 36-element nested array virtual array.

[0051] Figure 7 These are comparison images of two sound sources, where (a) is a dual-source sound imaging of an Archimedes spiral array, (b) is a dual-source sound imaging of an annular array, (c) is a dual-source sound imaging of an involute spiral array, (d) is a dual-source sound imaging of an 81-element nested array virtual array, and (e) is a dual-source sound imaging of a 36-element nested array virtual array. Detailed Implementation

[0052] The following detailed description, in conjunction with the accompanying drawings, illustrates the embodiments of this application from both a principle and implementation perspective.

[0053] Symbol Explanation: In this application, thin lowercase letters, bold lowercase letters, and bold uppercase letters are used to represent scalars, vectors, and matrices, respectively. Symbol [·] T , [·] * and[·] H represents the transpose, conjugate, and conjugate transpose of a matrix or vector, respectively. I represents the identity matrix, E[·] is the expectation operation, and ⊙ is the Khatri-Rao product symbol. This is the Kronecker product notation. Given a vector a, diag(a) denotes constructing a diagonal matrix with vector a as its diagonal. For matrix A, vec(A) denotes the vectorization operation on matrix A.

[0054] The research basis of this application includes the following acoustic imaging signal model:

[0055] like Figure 1 As shown, assume that D uncorrelated far-field narrowband signals are incident on a planar array with M elements, and the incident azimuth and elevation angles of the i-th signal are θ and θ, respectively. i , The position of the m-th element is Assuming that the signals are uncorrelated with each other and with noise, and that the noise is uncorrelated additive white Gaussian noise, then the received signal on the m-th array element can be expressed as:

[0056]

[0057] Where n m (t) represents the observation noise of the m-th array element at time t, s i (t) represents the value of the i-th signal source at time t, and f is the center frequency of the sound source signal. π is the imaginary unit, and π is the value of pi.

[0058]

[0059] τ mi The time delay of the i-th signal relative to the reference element (the element at the origin) when it reaches the m-th array element is represented by λ, where λ is the wavelength of the sound source signal, and x is the wavelength of the sound source signal. m y m Let be the coordinates of the m-th array element.

[0060] make

[0061] x(t) = [x1(t), x2(t), ..., x M (t)] T ,

[0062] s(t)=[s1(t),s2(t),,s D (t)] T ,

[0063] n(t) = [n1(t), n2(t), ..., n M (t)] T

[0064] Let represent the received signal vector, the signal source vector, and the received noise signal vector, respectively. Then equation (5) can be written in vector form as follows:

[0065] x(t)=As(t)+n(t), (7)

[0066] Where A is the array manifold matrix of the planar array, it can be represented as

[0067]

[0068] Where the guiding vector for

[0069]

[0070] Acoustic imaging involves gridding the sound source surface. Based on the spatial sound signal x(t) received by the microphone array, the power output value is calculated for each discrete grid point. The power value of each grid point is used as the pixel value of the image, thus obtaining the corresponding image. Therefore, bright spots or bright areas in the image obtained by acoustic imaging correspond to the locations of the maximum output power, which ultimately correspond to the estimated location of the sound source.

[0071] Based on this array observation signal model, this application proposes the following high-resolution imaging method for a planar microphone array in the differential domain.

[0072] While traditional beamforming algorithms are computationally inexpensive, robust, and effective, their resolution is limited by the Rayleigh limit, resulting in relatively low resolution. To improve the ability to resolve sound sources, this application employs a novel beamforming algorithm for planar sparse arrays in the differential domain. Applying this algorithm to nested planar arrays, hourglass arrays, and other arrays can increase the effective aperture of the array and improve its spatial resolution.

[0073] The method described in this application first vectorizes the covariance matrix of the received array signal to obtain a differential co-array. Then, redundant array elements of the differential co-array are removed and sorted to obtain a differential co-array without repeating elements. Finally, based on the differential co-array without repeating elements, a conventional beamforming algorithm is used to complete power scanning in all directions to obtain an acoustic image. The specific steps are as follows:

[0074] Step 1: Use multiple microphones to form a typical planar sparse array, such as a nested planar array or an hourglass array, to collect sound signals, obtain the array received signal, and calculate the covariance matrix of the array received signal.

[0075] The specific method is as follows:

[0076] First, based on the assumptions about signal and noise, the covariance matrix of the array-received signal is:

[0077]

[0078] in Let be the covariance matrix of the signal. Let be the power of the i-th signal. This represents the power of the noise.

[0079] Step 2: Vectorize the covariance matrix obtained in Step 1 to obtain the received signal vector of a single snapshot of the differential co-array. Some elements of this differential co-array overlap. Specifically:

[0080] Vectorization of the received signal covariance matrix yields

[0081]

[0082] in e i It is a column vector where all positions except the i-th position are zero.

[0083] It can be seen that equation (2) and equation (7) have similar forms, and equation (2) can be considered as a new virtual array observation model. In equation (2), z and A... * ⊙A、p、 These correspond to the equivalent received signal, array manifold, signal source, and noise signal of the virtual array, respectively.

[0084] The virtual array steering vector obtained from equation (2) above is According to the steering vector of the virtual array, the positions of the new virtual array elements are determined by... The set is given, where Let L be the position vectors of the m-th and n-th array elements, respectively. The resulting set L... d It is exactly the differential array of the original array.

[0085] For ease of understanding, Figure 2 The diagram shows a two-dimensional nested array and its differential array. The two-dimensional nested array is a two-layer nested array, with two elements in the first layer and two elements in the second layer, for a total of 16 elements. The minimum spacing d between the elements is half a wavelength. The differential array of the obtained two-dimensional nested array is a uniform square array with a total of 121 elements and a spacing d between the elements. According to existing technology, without limiting the array type, the maximum degree of freedom that 16 elements can obtain should be 144. It can be seen that the actual number of differential elements obtained by the two-dimensional nested array is less than the maximum number of degrees of freedom, that is, there are multiple overlapping positions. Therefore, multiple received signals will appear at the same position in the differential array obtained by equation (2). In actual operation, the average value of the multiple received signals at the same element position needs to be calculated before beamforming acoustic imaging.

[0086] Step 3: Calculate the average value of the received signals corresponding to the array elements with the same position in the differential co-array obtained in Step 2. Then, rearrange the received signal vector of the differential co-array single snapshot in ascending order of coordinate position to obtain an equivalent received signal vector of the differential co-array single snapshot without repeating array elements.

[0087] Specifically as follows:

[0088] The equivalent received signal after removing redundancy is denoted as . It can be represented as

[0089]

[0090] in e is a vector whose middle element is 1 and the rest are 0. The virtual array steering vector after rearranging the elements according to their positions is denoted as...

[0091]

[0092] Where Q is the number of virtual difference array elements after removing redundancy. This indicates the position of the q-th difference array element.

[0093] Step 4: Use the single-snapshot received signal of the differential co-array without repeating array elements as the input signal for beamforming, and use the beamforming algorithm to complete the power scan in all directions to perform acoustic imaging and localize the target sound source, thereby obtaining an acoustic image map.

[0094] Let w = [w1 w2 … w] Q ] T ,w q Let be the weight coefficient corresponding to the q-th difference array element, where q = 1, 2, ..., Q, and Q is the number of non-overlapping difference array elements. Figure 3 As shown, the output expression for beamforming based on the differential array is:

[0095]

[0096] The expression for its weighting coefficient is as follows: This represents the equivalent received signal of the q-th differential array element. (Time delay) The azimuth and elevation angles of the scan are respectively, and their ranges are as follows: i and j are the indices of the number of grid points divided by the azimuth and elevation angles, respectively.

[0097] The output power of the virtual array beamforming can be obtained as follows:

[0098]

[0099] Where R is the covariance matrix of the differential co-array output signal. The acoustic image of the sound source can be obtained from the output power of the differential co-array beamforming in each direction.

[0100] As described above in terms of principles and implementation methods, the high-resolution imaging method for planar microphone arrays in the differential domain proposed in this application, taking a two-dimensional nested planar array as an example, obtains the differential array elements corresponding to the nested array in its differential domain. Noise signals collected by the microphone array at different positions are processed to obtain virtual array received signals corresponding to the positions of the differential array elements. These signals are then used as input signals for beamforming to perform acoustic imaging localization of the noise. This method can solve the problem that current acoustic imaging localization systems require a large number of microphones to achieve higher resolution.

[0101] Simulation results show that the acoustic imaging algorithm based on a planar differential microphone array can distinguish between two closely spaced sound source signals, and also has a higher resolution than the acoustic imaging algorithm based on physical array elements. The simulation analysis process is as follows:

[0102] 1. Simulation Result Analysis

[0103] To verify the performance of the proposed acoustic imaging algorithm, a two-dimensional planar nested array is used as an example, and the method described in this application is compared with an Archimedean spiral array, an involute spiral linear array, and an annural array. Except for the annural array, which has 90 array elements, the other arrays have 81 array elements. To further verify the performance of the two-dimensional nested array, its array element count is set to 81 and 36, respectively. All comparison arrays use conventional beamforming algorithms for power estimation. The minimum element spacing of the two-dimensional planar nested array is half a wavelength. The Archimedean spiral array has a spiral spacing parameter of 2.5, a spiral rotation range of (0, 8π), and a step size of 0.1π. The involute spiral linear array has a maximum aperture of 100, a base circle radius of 10, 9 microphone rings, and 9 spiral arms. The annural array has a base circle radius of 10, 9 microphone rings, and 10 spiral arms. The array element positions of the Archimedean spiral array, annural array, involute spiral linear array, and nested array are shown below. Figure 4 As shown in Figures (a)-(e).

[0104] 1.1 Comparison of Single Sound Source Localization Results

[0105] This experiment compares the scenario of single sound source localization. The sound source is positioned in the (0°, 45°) direction, with 1000 snapshots and a signal-to-noise ratio of 10dB. The spatial power spectrum estimation results for the Archimedean spiral array, annular array, involute spiral array, and two-dimensional nested arrays with 81 and 36 elements are as follows: Figure 5 As shown in Figures (a)-(e). First, from... Figure 5 As can be seen from Figures (a)-(d), with the same number of array elements, the main lobe of the spatial spectrum of the differential array localization based on nested arrays is smaller than that of the spatial spectrum of other array localization arrays; secondly, from Figure 5 Figure (e) shows that when the number of elements in the nested array is much smaller than that of other physical arrays, the main lobe of the spatial spectrum of the differential array positioning is still smaller than that of the spatial spectrum of other physical array positioning. Finally, the positioning based on the differential array of the nested array has almost no side lobe interference in the spatial spectrum, while the spatial spectrum of the positioning based on other physical arrays has a lot of side lobes.

[0106] 1.2 Comparison of Multi-Source Localization Results

[0107] In practical applications, the simultaneous occurrence of multiple sound sources is very common. Especially when detecting multiple sound sources that are close together, the resolution of acoustic imaging localization is crucial. The second experiment in this application sets up a localization scenario for two close sound sources, with the source signal positions set at (0°, 42°) and (5°, 48°) directions, a sampling snapshot count of 1000, and a signal-to-noise ratio of 10dB. The spatial power spectrum estimation and acoustic imaging simulation results for Archimedean spiral arrays, annular arrays, involute spiral arrays, and two-dimensional nested arrays with 81 and 36 elements are shown below. Figure 6 (a)-(e) and Figure 7 As shown in (a)-(e).

[0108] In terms of spatial spectrum, firstly from Figure 6 The spatial spectrum simulation results of figures (a) to (d) show that, under the same number of array elements, the spatial spectrum based on the physical array has two large main lobes and many side lobe interferences in the power spectrum. In contrast, the power spectrum of the differential array acoustic imaging based on nested arrays has a smaller main lobe and almost no side lobe interference. Secondly, from... Figure 6 The simulation results in Figure (e) show that even when the number of elements in the nested array is much smaller than that of other physical arrays, the two main lobes of the spatial spectrum image based on the nested array difference array are still smaller than those of other physical arrays; finally, from Figure 6 Figures (d) and (e) show that the 36-element nested array has fewer sidelobe interferences near the main peak, while the 81-element nested array has almost no sidelobe interferences near the main peak. As the number of elements in the nested array increases, its main lobe width is smaller and its positioning resolution is higher.

[0109] In terms of acoustic imaging, firstly from Figure 7 The simulation results in figures (a) to (d) show that, with the same number of array elements, the acoustic imaging based on the physical array exhibits two indistinguishable sound sources, resulting in interference around the perimeter of the image. In contrast, the acoustic imaging based on the nested differential array exhibits two independent sound sources. Secondly, as shown in figure (e) of (7), even when the number of array elements in the nested array is much smaller than that of other physical arrays, the acoustic imaging results based on the nested differential array are still better. Finally, from... Figure 7 As can be seen from (d) and (e), there is a low valley connecting the two main peaks of the 36-element nested array, while the acoustic imaging results of the 81-element nested array are two completely independent sound sources with almost no sidelobe interference. Moreover, as the number of array elements increases and the array aperture increases, the acoustic imaging resolution of the differential array based on the nested array is higher, and it can better locate the sound source.

[0110] The above simulation results show that, compared with existing acoustic imaging based on physical array elements, the virtual array beamforming algorithm based on the two-dimensional planar nested array in the differential domain has good localization effect on both single and multiple noise signals. Compared with acoustic imaging based on the same physical array elements, the method proposed in this application has a narrower main lobe, higher resolution, and can distinguish two close sound sources. Even when using a nested array with far fewer array elements than other physical arrays, the method proposed in this application still has a lower main lobe width and can still locate nearby sound sources, with strong noise resistance and high spatial resolution.

Claims

1. A method of high resolution imaging of a planar microphone array in differential domain, characterized in that, The method comprises: Step 1, using a plurality of microphones to form a typical planar sparse array to collect sound signals to obtain array receiving signals, and calculating a covariance matrix of the array receiving signals; Step 2, vectorizing the covariance matrix to obtain a single-snapshot receiving signal vector of a differential common array, the differential common array containing partially overlapping array element positions; specifically comprising: Vectorizing the covariance matrix of the receiving signals to obtain a single-snapshot observation model of a new virtual array: （2） where , , is a column vector with 1 at the mth position and 0 at other positions, and denote the KR product and the Kronecker product, respectively, is the total number of sources; , , , denote the equivalent received signal, array manifold, signal source and noise signal of the virtual array, respectively; the virtual array steering vector obtained according to the virtual array observation model is , and the virtual array element position is given by , wherein are the position vectors of the mth and nth elements, respectively, and M is the number of elements of the original array; the obtained set is the differential common array of the original array; the observation vector of the differential common array corresponding to formula (2) contains elements with repeated element positions.​ Step 3, performing mean value processing on the receiving signals corresponding to the array elements with the same position in the differential common array, and then rearranging the single-snapshot receiving signal vector of the differential common array in the order of increasing coordinate positions to equivalently obtain a single-snapshot receiving signal vector of a differential common array without repeated array elements; Step 4, taking the single-snapshot receiving signal of the differential common array without repeated array elements as an input signal of beam forming, using a beam forming algorithm to complete power scanning in all directions, performing acoustic imaging positioning on a target sound source, and obtaining an acoustic imaging map.

2. The method of high-resolution imaging of a planar microphone array of differential domains according to claim 1, characterized in that, The step 1 of calculating the covariance matrix of the array receiving signals specifically comprises: Assume that the original array received signal vector is According to the assumption that the signal and the noise are not correlated with each other, the covariance matrix of the array received signal can be expressed as （1） wherein is the covariance matrix of the signals, is the power of the th signal, is the power of the noise, is the array flow pattern matrix, is the array steering vector.

3. The method of high-resolution imaging of a planar microphone array of differential domains according to claim 1, characterized in that, The step 3 specifically comprises: The received signals corresponding to the same array element positions in formula (2) are processed by averaging, and are arranged in ascending order of differential array element positions, and the equivalent received signals are removed To , wherein , is a vector with 1 in the middle element and 0 elsewhere. The differential common array steering vector after the differential array element position rearrangement is denoted as , wherein is the number of differential co-array elements after removing redundancies, , represents the position of the th differential co-array element.

4. The method of claim 1, wherein, The step 4 specifically comprises: Record is the weight vector, where is the weight coefficient corresponding to the th difference co-array element pair, The beamforming output expression based on the difference co-array is （3） wherein is the equivalent received signal of the th difference coarray element; the expression of the weight coefficient is , the time delay , is the scanning azimuth and elevation angle, whose ranges are , are the index values of the grid points of the scanning azimuth and elevation angle, respectively; The output power of the differential common array beam forming can be obtained as （4） wherein is the covariance matrix of the difference coarray output signals; The output power of the virtual array beam forming in each direction can be obtained as the acoustic imaging map of the sound source.

5. A high-resolution imaging device of a planar microphone array in differential domain, characterized in that, The computer readable storage medium stores computer executable instructions for making the computer execute the steps of the high-resolution imaging method of the differential domain planar microphone array according to any one of claims 1-4. The computer readable storage medium stores computer executable instructions for making the computer execute the steps of the high-resolution imaging method of the differential domain planar microphone array according to any one of claims 1-4. The covariance matrix calculation unit is configured to use a plurality of microphones to form a typical planar sparse array to collect sound signals to obtain array receiving signals, and calculate a covariance matrix of the array receiving signals; The vectorization unit is configured to vectorize the covariance matrix to obtain a single-snapshot receiving signal vector of a differential common array, the differential common array containing partially overlapping array element positions; The mean value processing and vector rearrangement unit is configured to perform mean value processing on the receiving signals corresponding to the array elements with the same position in the differential common array, and then rearrange the single-snapshot receiving signal vector of the differential common array in the order of increasing coordinate positions to equivalently obtain a single-snapshot receiving signal vector of a differential common array without repeated array elements; 6. A computer-readable storage medium, characterized in that, The acoustic imaging map acquisition unit is configured to take the single-snapshot receiving signal of the differential common array without repeated array elements as an input signal of beam forming, use a beam forming algorithm to complete power scanning in all directions, perform acoustic imaging positioning on a target sound source, and obtain an acoustic imaging map. The computer readable storage medium stores computer executable instructions for making the computer execute the steps of the high-resolution imaging method of the differential domain planar microphone array according to any one of claims 1-4.

Images