A deconvolution beamforming acceleration method based on RL iterative algorithm

Abstract

The present invention discloses an acceleration method for deconvolution beamforming based on the R-L iterative algorithm, which calculates the natural directivity function PSF of a uniform linear array by using formation-related parameters. R ={PSF1, PSF2,..., PSF r}, where r represents the total number of frequency point numbers; sub-bands with unequal intervals are divided according to the bandwidth, and the natural directivity functions of the corresponding sub-band center frequencies are extracted; the conventional beamforming calculations of m beams are performed using the data of each array element of the uniform linear array to obtain the power spectra P of the m beams M ={P1, P2,..., P m}; the natural directivity function PSF after extraction of the sub-band center frequencies O ={PSF1, PSF2,..., PSF o}, and the power spectra P of M beams M ={P1, P2,..., P m} are subjected to R-L deconvolution calculation, where o represents the total number of sub-bands divided, and o < r; the results of the m beams obtained from the R-L deconvolution calculation are used as the final azimuth estimation results. The present invention overcomes the shortcomings of conventional beamforming, such as wide main lobes, high side lobes, and poor weak target detection capabilities, and at the same time solves the problem of excessive computational complexity of traditional deconvolution beamforming when processing broadband random signals.

Description

Technical Field

[0001] The present invention relates to the field of underwater acoustic array signal processing, and in particular to a deconvolution beamforming acceleration method based on RL iterative algorithm. Background Art

[0002] As a basic formation, the horizontal linear array is an important component of the sonar system. Conventional beamforming (CBF) processing is used to obtain array gain and improve target detection capability. However, the CBF main lobe beam width is wide and the side lobe level is high, which reduces the spatial resolution capability. It is easy to produce aliasing when processing multiple targets, and the weak target detection capability is weak under strong signal interference. Deconvolution beamforming is a post-processing method of CBF, which can take into account high resolution and stability. However, when the signal bandwidth is wide, the amount of calculation is large and the computational complexity is high. In addition, the deconvolution algorithm based on the Richardson-Lucy iterative method (RL) will have blurred boundaries at both ends of the beam diagram due to signal truncation, which affects practical engineering applications. Based on traditional deconvolution beamforming, the present invention uses multi-band processing and energy spectrum boundary expansion methods to obtain higher resolution while increasing the speed by 50% compared with traditional deconvolution beamforming. Summary of the invention

[0003] One of the purposes of the present invention is to provide a deconvolution beamforming acceleration method based on RL iterative algorithm to solve the shortcomings of the existing conventional beamforming in the background technology, such as wide main lobe, high side lobe, and poor weak target detection capability.

[0004] The second object of the present invention is to provide a deconvolution beamforming acceleration method based on RL iterative algorithm, while solving the problem of excessive computational complexity of traditional deconvolution beamforming when processing broadband random signals.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A deconvolution beamforming acceleration method based on RL iterative algorithm includes the following steps:

[0007] Calculate the natural directivity function (PSF) of a uniform linear array using array-related parameters R = {PSF1, PSF2, ..., PSF r}, where r represents the total number of frequency points;

[0008] Divide the bandwidth into sub-bands with unequal spacing, and extract the natural directivity function of the corresponding sub-band center frequency;

[0009] The conventional beamforming calculation of m beams is performed using the array data of the uniform linear array to obtain the power spectrum P of the m beams. M = {P1, P2, ..., P m};

[0010] The natural directivity function PSF after extracting the sub-band center frequency O = {PSF1, PSF2, ..., PSF o}、Power spectrum P of M beams M = {P1, P2, ..., P m} Perform RL deconvolution calculation, where o represents the total number of divided subbands, o<r;

[0011] The RL deconvolution calculation results are used to obtain m beam results as the final azimuth estimation results.

[0012] Preferably, the power spectrum P(cosθ) is calculated as follows:

[0013]

[0014] in,

[0015] S() is the source distribution function, B p () is the beam directivity function, K is the number of sound signals, i is the i-th sound signal, A i is the strength of the i-th sound signal, θ i is the incident azimuth of the i-th sound signal, is the differential value of the beam angle, δ() is the impulse function, N is the number of array elements, d is the array element spacing, and λ is the wavelength.

[0016] Preferably, the calculation method of the natural directivity function is:

[0017] Where T is the signal frequency, N is the number of array elements, d is the array element spacing, c is the speed of sound in water, and θ is the signal incident azimuth. is the differential value of the beam angle.

[0018] Preferably, after obtaining the natural directivity function of the corresponding sub-band center frequency point and the power spectrum P of the m beams M After that, the power spectra P of the m beams are calculated respectively. M , natural directivity function PSF after sub-band center frequency extraction O Perform period expansion to obtain the expanded power spectrum P M′ and the extended natural directivity function PSF O′ ; For the periodically extended power spectrum P M′and the natural directivity function PSF of the periodic expansion O′ Perform RL deconvolution calculation; take the middle m beam results from the RL deconvolution results as the final azimuth estimation results.

[0019] Preferably, the power spectrum P of the m beams M Doing cycle extension includes the following steps:

[0020] The power spectrum P M The left and right endpoints are taken as the starting points, and the preset number of beams l are extended outward respectively;

[0021] The expanded beam is expressed in terms of power spectrum P M The left and right endpoints are used as reference values ​​to obtain the power spectrum P of the periodic expansion. M′ = {P1, P2, ..., P m ,…,P m+2l}, P m Indicates the power value of the mth beam.

[0022] Preferably, the natural directivity function PSF after extracting the sub-band center frequency point O Doing cycle extension includes the following steps:

[0023] The natural directivity function PSF of the sub-band center frequency β ={PSF′1,PSF′2,…,PSF′ m} β The left and right endpoints of are taken as the starting point, and the preset number of beams l are extended outward respectively, β represents the number of subbands, β = {1, 2, 3, ..., o}, PSF′ m Indicates the natural directivity of the mth beam at a fixed sub-band center frequency;

[0024] The natural directivity function (PSF) of the expanded beam at the sub-band center frequency β The left and right endpoints are used as reference values ​​to obtain the extended natural directivity function PSF β′ ={PSF′1,PSF′2,...,PSF′ m , ..., PSF′ m+2l} β ;

[0025] Get the periodically extended natural directivity function PSF O = {PSF 1′ , PSF 2′ , ..., PSF β′ , ..., PSF o′}.

[0026] Preferably, the preset number l=0.25m.

[0027] Preferably, the power of the extended beam is the power spectrum P M The power value of the corresponding endpoint; the natural directivity of the extended beam is the natural directivity of the corresponding endpoint of the corresponding natural directivity function.

[0028] Preferably, the power of the extended beam is expressed in terms of power spectrum P M The reference axis of the corresponding endpoint is taken as the symmetry axis, and symmetric values ​​are taken; the natural directivity of the extended beam takes the reference axis of the corresponding endpoint of the corresponding natural directivity function as the symmetry axis, and symmetric values ​​are taken.

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] The present invention overcomes the shortcomings of conventional beamforming, such as wide main lobe, high side lobe, and poor weak target detection capability, and solves the problem of excessive computational complexity of conventional deconvolution beamforming when processing broadband random signals. The computational complexity of deconvolution is reduced through sub-band division processing; the boundary fuzzy problem caused by the use of RL iterative algorithm is solved through periodic expansion of the power spectrum, while the computational speed is further improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 This is a flowchart of the deconvolution beamforming acceleration method based on RL iterative algorithm of Example 2.

[0032] Figure 2 This is a schematic diagram of the formation layout of a horizontal uniform array.

[0033] Figure 3 It is the PSF diagram of horizontal uniform linear array at each frequency point.

[0034] Figure 4 This is the deconvolution beam output after molecular band extraction.

[0035] Figure 5 This is the deconvolution beam output image after beam spectrum boundary expansion.

[0036] Figure 6 This is a table showing the speed improvement of the deconvolution beamforming acceleration method based on RL iterative algorithm in Example 2 compared with the traditional deconvolution beamforming. DETAILED DESCRIPTION

[0037] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work belong to the protection scope of the present invention.

[0038] Embodiment 1:

[0039] Reference Figure 1 As shown in FIG. 1 , a deconvolution beamforming acceleration method based on RL iterative algorithm is mainly applied to uniform linear arrays. Taking a horizontal linear array with N array elements as an example, the horizontal uniform linear array is referred to as Figure 2 As shown, a number of array elements are arranged horizontally and spaced apart, the spacing between two adjacent array elements is d, the sound source is in the far field, the sound signal reaches each array element in the form of a plane wave, and the incident direction of the signal is θ. The deconvolution beamforming acceleration method based on the RL iterative algorithm includes the following steps.

[0040] PSF calculation steps: Use the array-related parameters to calculate the natural directivity function PSF of the uniform linear array R = {PSF1, PSF2, ..., PSF r}, where r represents the total number of frequency points.

[0041] The natural directivity function (also known as the point spread function PSF in the deconvolution algorithm) is a parameter related to the signal frequency. The specific calculation formula is: Where f is the signal frequency, N is the number of array elements, d is the array element spacing, c is the speed of sound in water, and θ is the signal incident azimuth. is the differential value of the beam angle θ.

[0042] In the present invention, frequency points are numbers given to fixed frequencies. If the frequency intervals are all 1 Hz, then 0 Hz, 1 MHz, 2 Hz, 3 Hz, 4 Hz, 5 Hz...6000 Hz are divided into 6001 wireless frequency bands according to the frequency interval of 1 Hz, and each frequency band is numbered from 1, 2, 3, 4...6001; these numbers for fixed frequencies are frequency points. The present invention calculates the natural directivity function of the uniform horizontal line array at each frequency point through the calculation formula of the natural directivity function, and finally summarizes the natural directivity functions of all frequency points to obtain the natural directivity function PSF of the uniform horizontal line array. R = {PSF1, PSF2, ..., PSF r}, PSF r The meaning is the m natural directivities corresponding to the m beams at the rth frequency point.

[0043] Molecular band extraction steps: divide the bandwidth into sub-bands with unequal spacing, and extract the natural directivity function of the corresponding sub-band center frequency.

[0044] In the present invention, bandwidth division can be performed according to actual conditions. For example, if the total analysis bandwidth is 0-6kHz, it can be divided into (0-2000).(2000-3600).(3600-4600).(4600-5200). (5200-5600).(5600-6000). The general principle of division is that the low-frequency band is narrow and the high-frequency band is wide. The width of the band is relative. For example, the low-frequency part needs to be divided into a sub-band of 2k, and the high-frequency part can be divided into a sub-band of 200.

[0045] In the present invention, bandwidth usually refers to the width of the frequency band occupied by a signal; when used to describe a channel, bandwidth refers to the maximum frequency bandwidth of a signal that can effectively pass through the channel. For analog signals, bandwidth is also called frequency bandwidth, measured in Hertz (Hz). Subband is also called sub-frequency band, which is a part of a frequency band. The unit of frequency band is Hertz (Hz), which refers to the part of the spectrum between two specific frequency limits. For a signal, the frequency band is the frequency range between the highest frequency and the lowest frequency contained in the signal.

[0046] The present invention divides the broadband signal into several narrow sub-bands of unequal width, takes a natural directivity function of a center frequency point in each sub-band, and finally obtains the natural directivity function PSF of the center frequency points of all sub-bands. O = {PSF1, PSF2, ..., PSF O}.

[0047] Beam spectrum acquisition steps: Use the array data of the uniform linear array to perform conventional beamforming calculations on m beams to obtain the power spectrum P of the m beams. M = {P1, P2, ..., P m}, P m Indicates the power value of the mth beam.

[0048] In the present invention, since there are N array elements of the uniform line array, when there are K (K ≥ 1) mutually unrelated random incident sound signals, the incident signal received by each array element can be expressed as:

[0049] Where i is the number of targets, A i is the signal strength of the i-th target, Noise n is the noise, n is the array element number;

[0050] is the phase shift vector of the i-th target in each array element, where p is defined as follows T stands for transpose, j is an imaginary number, and k is the wave number.

[0051] According to conventional beamforming theory, the steering vector S is defined as [s1, s2, ..., s N ]T , where s n =( 1 / N )exp(-jk(n-1)dcosθ i ), then the power spectrum P(cosθ) can be expressed as

[0052]

[0053] in,

[0054]

[0055] S() is the source distribution function, B p () is the beam directivity function, K is the number of sound signals, i is the i-th sound signal, A i is the strength of the i-th sound signal, θ i is the incident azimuth of the i-th sound signal, is the differential value of the beam angle θ, δ() is the impulse function, N is the number of array elements, d is the array element spacing, λ is the wavelength, and * is the convolution.

[0056] RL iterative calculation steps: natural directivity function PSF after extracting the sub-band center frequency point O = {PSF1, PSF2, ..., PSF o}、Power spectrum P of M beams M = {P1, P2, ..., P m} Perform RL deconvolution calculation, where o represents the total number of divided subbands, o<r.

[0057] In the present invention, the RL deconvolution calculation is the Richardson-Lucy algorithm, which is an iterative algorithm.

[0058] The traditional deconvolution beamforming algorithm requires the beam data of each frequency point of the CBF to be deconvolved with the PSF of the corresponding frequency point to obtain the source distribution function SS(f, cosθ) of each frequency point, and finally the source distribution function of each frequency point is added in phase to obtain From the above formula, we can see that the natural directivity function PSF has little difference in a narrow frequency band, and the higher the frequency, the wider the frequency band with approximate PSF, such as Figure 3 As shown, the present invention divides the broadband signal into several narrower sub-bands of unequal widths, and takes a natural directivity function PSF of a central frequency point in each sub-band for approximate deconvolution calculation, thereby greatly reducing the amount of calculation of the natural directivity function PSF and deconvolution.

[0059] Value selection steps: The m beam results obtained by RL deconvolution calculation are used as the final azimuth estimation results.

[0060] In the present invention, the power spectrum P of the m beams M The natural directivity function PSF after extracting the center frequency of each sub-band O = {PSF1, PSF2, ..., PSF o}Perform RL deconvolution calculation to obtain o RL deconvolution calculation results, add all the deconvolution calculation results, and the result obtained by the addition is m beam results, which is also the final output result of the m beams.

[0061] In the present invention, the power output of conventional beamforming can be regarded as the convolution of the point source scattering function of the signal and the natural directivity function of the array, that is, P M =PSF M *S M , * represents convolution calculation. When the power output and natural directivity are known, the point source scattering function corresponding to the target direction can be restored by deconvolution. M And the PSF of each frequency point M The RL iterative algorithm is used for deconvolution calculation to obtain the estimated S M value, each S M The m beam results obtained by adding the values ​​are taken as the final azimuth estimation result, and the final azimuth estimation result is also the output result of the deconvolution beamforming.

[0062] Figure 4 The beam output diagram of deconvolution beam forming after molecular band extraction is given, where the solid line is the CBF result, the dotted line is the deconvolution result of the present invention, and the dashed line is the traditional deconvolution result. It shows that the detection ability and resolution of the present method are similar to those of the traditional deconvolution method, both of which are better than CBF, and the present method has a smaller amount of calculation.

[0063] Embodiment 2:

[0064] Furthermore, the present invention discloses another deconvolution beamforming acceleration method based on RL iterative algorithm.

[0065] Step 1: PSF calculation, using the array-related parameters to calculate the natural directivity function PSF of the uniform linear array R = {PSF1, PSF2, ..., PSF r}, where r represents the total number of frequency points.

[0066] Step 2: Sub-band extraction: divide the bandwidth into sub-bands with unequal spacing and extract the natural directivity function of the corresponding sub-band center frequency.

[0067] Step 3: Beam spectrum acquisition: use the array data of the uniform linear array to perform conventional beamforming calculations on m beams and obtain the power spectrum P of the m beams. M = {P1, P2, ..., P m}, P m Indicates the power value of the mth beam.

[0068] The above steps in the second embodiment are the same as those in the first embodiment, but the second embodiment of the present invention further includes a cycle extension step, specifically:

[0069] Step 4: Calculate the power spectra P of the m beams respectively M , natural directivity function PSF after sub-band center frequency extraction O Perform period expansion to obtain the expanded power spectrum P M′ and the extended natural directivity function PSF o′ ; For the periodically extended power spectrum P M′ and the natural directivity function PSF of the periodic expansion o′ Perform RL deconvolution calculations.

[0070] In step 4, the power spectrum P of the m beams M Cycle expansion involves the following two steps:

[0071] The power spectrum P M The left and right endpoints are taken as the starting point, and 0.25m beams are extended outward respectively;

[0072] The expanded beam is expressed in terms of power spectrum P M The values ​​of the left and right endpoints are taken as the reference values ​​to obtain the power spectrum P of the periodic expansion M′ = {P1, P2, ..., P m ,…,P 1.5m}, P m Indicates the power value of the mth beam.

[0073] As one specific implementation of the power spectrum period expansion, the power spectrum P M With cosθ=±1 as the left and right endpoints, 0.25m beams are extended outward respectively. The power values ​​of the 0.25m beams extended at cosθ=-1 are all the power values ​​at cosθ=-1, and the power values ​​of the 0.25m beams extended at cosθ=1 are all the power values ​​at cosθ=1.

[0074] As another specific implementation of the power spectrum period expansion, the power spectrum P MTaking cosθ=±1 as the left and right endpoints, 0.25m beams are extended outward respectively. The curve formed by the power values ​​of the 0.25m beams extended at cosθ=-1 is symmetrical to the curve formed from cosθ=-1 to cosθ=-0.5, and the axis of symmetry is cosθ=-1. The curve formed by the power values ​​of the 0.25m beams extended at cosθ=1 is symmetrical to the curve formed from cosθ=1 to cosθ=0.5, and the axis of symmetry is cosθ=1.

[0075] When performing RL iteration on the one-dimensional CBF power spectrum, blurring will occur at the edge of the power spectrum. For a uniform linear array, this is near the angle cosθ = ±1. Traditional deconvolution beamforming uses the periodicity of the beam directivity pattern to provide a method to expand the power spectrum integration range. The specific formula is: By transforming the above formula, the boundary ambiguity problem can be moved to the vicinity of cosθ=±1.5, so that it is not affected by boundary ambiguity within the range of |cosθ|≤1 with actual physical significance. However, while this boundary expansion method solves the boundary ambiguity problem of the RL algorithm, it increases the amount of calculation for beamforming by 50%. Especially when the number of beams is large, this increase in calculation greatly affects the application of the algorithm in engineering. The present invention optimizes this expansion, and periodically expands M / 4 beams on both sides with the power spectrum value P(cosθ) near the boundary cosθ=±1, omitting the beamforming calculation process of M / 2 beams and improving the speed.

[0076] The natural directivity function PSF after extracting the sub-band center frequency point O Cycle expansion includes the following two steps:

[0077] The natural directivity function PSF of the sub-band center frequency β ={PSF′1,PSF′2,...,PSF′ m} β The left and right endpoints are taken as the starting point, and 0.25m beams are extended outward respectively. β represents the subband number, β = {1, 2, 3, ..., o}, PSF′ m Indicates the natural directivity of the mth beam at a fixed sub-band center frequency;

[0078] The natural directivity function (PSF) of the expanded beam at the sub-band center frequency β The values ​​of the left and right endpoints are taken as the reference values ​​to obtain the extended natural directivity function PSF β′ ={PSF′1,PSF′2,...,PSF′ m , ..., PSF′ 1.5m} β ;

[0079] Get the periodically extended natural directivity function PSF O′ = {PSF 1′ , PSF 2′ , ..., PSF β′ , ..., PSF o′}.

[0080] Natural directivity function PSF O It is two-dimensional, and the natural directivity function PSF of each sub-band center frequency is required β Perform period expansion. The method of period expansion is the same as that of power spectrum.

[0081] Step 5: Take the middle m beam results from the RL deconvolution results as the final azimuth estimation results.

[0082] In step 5 of the present invention, the power spectrum P after period expansion is M′ Natural directivity function PSF with each period expansion β′ Perform RL deconvolution calculation to obtain o calculation results. After adding all the calculation results, take the middle m beam results as the final azimuth estimation results.

[0083] Embodiment 2 of the present invention solves the boundary blur problem caused by the deconvolution RL algorithm without increasing the amount of beamforming calculations.

[0084] Figure 5 The deconvolution beam output diagram after the beam spectrum boundary expansion is given. The solid line is the CBF result, the dashed line is the deconvolution result of the traditional method with periodic expansion in the beam domain, and the dotted line is the deconvolution result of the CBF energy spectrum boundary expansion of the present invention. It shows that the method of the present invention is only slightly inconsistent with the traditional deconvolution beamforming at the boundary, and the detection performance and resolution are consistent with the traditional deconvolution method, which are better than CBF.

[0085] Figure 6 The comparison results of the calculation speed between the method of the present invention and the traditional deconvolution beamforming method are given, indicating that the overall calculation speed of the method of the present invention is improved by about 50% at a single frequency point, which is also the lower limit of the speed improvement of the method of the present invention. With the increase of the analysis frequency points, the speed improvement of the method of this paper also increases.

Claims

1. A deconvolution beamforming acceleration method based on RL iterative algorithm, characterized in that: The following steps are involved: Calculate the natural directivity function (PSF) of a uniform linear array using array-related parameters R = {PSF1, PSF2, ..., PSF r }, where r represents the total number of frequency points; Divide the bandwidth into sub-bands with unequal spacing, and extract the natural directivity function of the corresponding sub-band center frequency; The conventional beamforming calculation of m beams is performed using the array data of the uniform linear array to obtain the power spectrum P of the m beams. M = {P1, P2, ..., P m }; The power spectra P of m beams are separately M =(P1, P2,..., P m}, and the natural directivity function PSF after sub-band center frequency point extraction O ={PSF1, PSF2,..., PSF o} are subjected to periodic extension to obtain the extended power spectrum P M′ and the extended natural directivity function PSF O′ ; perform R-L deconvolution calculation on the periodically extended power spectrum P M′ and the periodically extended natural directivity function PSF O′ , where o represents the total number of sub-bands divided, o < r; take the middle m beam results in the R-L deconvolution result as the final azimuth estimation result; The power spectrum P of the m beams is M Doing cycle extension includes the following steps: The power spectrum P M The left and right endpoints are taken as the starting points, and the preset number of beams l are extended outward respectively; The expanded beam is expressed in terms of power spectrum P M The values ​​of the left and right endpoints are taken as the reference values ​​to obtain the power spectrum P of the periodic expansion M′ =(P1, P2, ..., P m , ..., P m+2l }, P m Indicates the power value of the mth beam.

2. The deconvolution beamforming acceleration method based on RL iterative algorithm as claimed in claim 1, characterized in that: The power spectrum P(cosθ) is calculated as follows: Among them, S() is the source distribution function, B p () is the beam directivity function, K is the number of sound signals, i is the i-th sound signal, A i is the strength of the i-th sound signal, θ i is the incident azimuth of the i-th sound signal, is the differential value of the beam angle, δ() is the impulse function, N is the number of array elements, d is the array element spacing, and λ is the wavelength.

3. The deconvolution beamforming acceleration method based on RL iterative algorithm as claimed in claim 1, characterized in that: The natural directivity function is calculated as Where f is the signal frequency, N is the number of array elements, d is the array element spacing, c is the speed of sound in water, and θ is the signal incident azimuth. is the differential value of the beam angle.

4. The deconvolution beamforming acceleration method based on RL iterative algorithm as claimed in claim 1, characterized in that: The natural directivity function PSF after extracting the sub-band center frequency point O Doing cycle extension includes the following steps: The natural directivity function PSF of the sub-band center frequency β =(PSF′1,PSF′2,...,PSF′ m } β The left and right endpoints of are taken as the starting point, and the preset number of beams l are extended outward respectively, β represents the number of subbands, β = {1, 2, 3, ..., o}, PSF′ m Indicates the natural directivity of the mth beam at a fixed sub-band center frequency; The natural directivity function (PSF) of the expanded beam at the sub-band center frequency β The values ​​of the left and right endpoints are taken as the reference values ​​to obtain the extended natural directivity function PSF β′ =(PSF′1,PSF′2,...,PSF′ m , ..., PSF′ m+2l } β ; Get the periodically extended natural directivity function PSF O′ = {PSF 1′ , PSF 2′ , ..., PSF β′ , ..., PSF o′ }.

5. The deconvolution beamforming acceleration method based on RL iterative algorithm as claimed in claim 4, characterized in that: The preset number l=0.25m.

6. The deconvolution beamforming acceleration method based on RL iterative algorithm as claimed in claim 4, characterized in that: The power of the expanded beam is the power spectrum P M The power value of the corresponding endpoint; the natural directivity of the extended beam is the natural directivity of the corresponding endpoint of the corresponding natural directivity function.

7. The deconvolution beamforming acceleration method based on RL iterative algorithm as claimed in claim 4, characterized in that: The power of the expanded beam is expressed in terms of the power spectrum P M The reference axis corresponding to the endpoint is the symmetry axis, and the values ​​are taken symmetrically; The natural directivity of the extended beam is symmetrically valued with the reference axis of the corresponding endpoint of the corresponding natural directivity function as the symmetry axis.

Images