Linear Differential Directional Microphone Array

Abstract

The apparatus and method provided by the present application relate to a linear differential directional microphone array (LDDMA), which takes into account the directivity of array elements. The LDDMA can be designed by the following steps: generating a steering vector for a linear array (LA) with preselected parameters (including parameters δ, p, θ, N and M), generating a constraint matrix based on the steering vector, reconstructing the constraint matrix based on the microphone response matrix and the steering matrix, obtaining a beamformer by applying a minimum norm solution of the constraint matrix, verifying the desired characteristics of the LA by calculating the beamformer in the desired direction, and constructing the LA based on the preselected parameters and the beamformer.

Description

Background Art

[0001] Speech enhancement technology is an integral component of many far-field pickup devices in noisy environments. Both shotgun microphones (usually a supercardioid package with a long, hollow, slotted interference tube) and microphone arrays are able to attenuate ambient noise or interference due to their high directivity. Shotgun microphones are commonly used in many applications that require low noise, such as specific cameras, conference-specific, or specific interview scenarios. Although this type of shotgun microphone can pick up sound in a certain direction in a noisy environment, making the picked up sound clearer and less noisy, they have fixed beamforming characteristics and are not tunable. In addition, the cost of designing and producing this type of microphone is relatively high. In contrast, microphone arrays with appropriate signal processing algorithms can provide a more flexible solution.

[0002] Among all microphone arrays, differential microphone arrays (DMAs) have attracted much attention in recent years. As a type of DMA, linear differential microphone arrays (LDMAs) have been widely studied, however, many of the developed LDMA designs seem to adopt omnidirectional microphones. Although a robust LDMA design can improve the white noise gain (WNG) with a minimum norm solution by using more microphone elements than the order of the LDMA, the WNG may still be relatively low, especially at low frequencies, which causes the well-known white noise amplification problem in practical implementations. In addition, the directivity factor (DF) of conventional LDMAs generally decreases with increasing frequency, and the beam pattern also tends to deform at high frequencies. BRIEF DESCRIPTION OF THE DRAWINGS

[0003] The detailed description refers to the accompanying drawings. In the drawings, the left-most digit(s) of a reference number identifies the figure in which the reference number first appears. The use of the same reference numbers in different figures indicates similar or identical items or features.

[0004] Figure 1A An example diagram of a uniform linear array (ULA) with M directional microphones is shown.

[0005] Figure 1B An example diagram of a non-uniform linear array (NULA) with M directional microphones is shown.

[0006] Figure 2 An example of a linear differential directional microphone array (LDDMA) is shown.

[0007] Figure 3A and Figure 3B The beam patterns of the second order cardioid pattern and the third order pattern at 1 kHz are shown respectively.

[0008] Figure 4A and Figure 4BThe beam patterns of the second order cardioid pattern and the third order pattern at 3 kHz are shown respectively.

[0009] Figure 5A and Figure 5B The beam patterns of the second order cardioid pattern and the third order pattern at 6 kHz are shown respectively.

[0010] Fig. 6A and Figure 6B Comparison of white noise gain (WNG) and directivity index (DI) of a 2nd-order LDDMA design for different types of microphones (p=1, p=0.5, p=0) are shown, respectively.

[0011] Fig. 7A and 7B The comparison in WNG and DI of the 3rd order LDDMA design for different types of microphones (p=1, p=0.5, p=0) is shown respectively.

[0012] Figure 8 An example process for constructing LDDMA is shown. DETAILED DESCRIPTION

[0013] A design method for a linear differential directional microphone array (LDDMA) is provided that takes into account the directivity of the array elements. Certain directional microphone elements have inherent unique properties that can outperform omnidirectional microphone elements. LDDMA can be implemented as a high-performance shotgun-type pickup device.

[0014] Omnidirectional and directional microphone elements are commonly used in the industry. Omnidirectional microphones pick up sound from all directions with equal gain, while directional microphones pick up sound mainly from certain specific directions. Mathematically, the beam pattern of a directional microphone can be expressed as u(p,θ,α)=p+(1-p)cos(θ-α)

[0015] Where θ is the angle of incidence of the sound, α is the steering direction of the microphone element, and p defines the characteristics of the directional microphone, for example, when p = 0.5, it forms the well-known cardioid beam pattern, and when p = 0, it forms a dipole beam pattern. The directional microphone can be any type of directional microphone, including omnidirectional, cardioid, dipole microphones, etc.

[0016] There are two ways to implement a directional microphone, one is a dedicated directional microphone using a single microphone box with two sound inlets, and the other is a dual omni-directional element system with appropriate digital signal processing. In terms of signal-to-noise ratio (SNR), the known dedicated directional microphone approach produces a better directional microphone than the dual omni-directional element system approach. This performance advantage of the dedicated directional microphone is mainly attributed to the signal processing that produces the directivity and acoustic processing through the front and rear sound inlets. This unique characteristic of the dedicated directional microphone can be used to achieve better performance than conventional LDMA. The dedicated directional microphone can take the form of an electret condenser microphone (ECM) or a microelectromechanical system (MEMS).

[0017] Figure 1A An example of a uniform linear array (ULA) 102 having M directional microphones 104 is shown and may be indicated generally at 100 . Figure 1B An example of a non-uniform linear array (NULA) 108 having M directional microphones 110 is shown, which may be indicated generally at 106 .

[0018] for Figure 1A For the ULA 102 in FIG. 1 , the spacing between elements is denoted as δ, and all directional microphones 104, 1 to M are directed to the right, i.e., α=0 (for simplicity, it will be omitted in the following description). Figure 1B The inter-element spacing of the NULA 108 in FIG. 1 varies relative to the first directional microphone 110 and is denoted by δ1…δ M All directional microphones 110, 1 to M are also pointed to the right. If a plane wave 112 is incident on the array 102 at an angle of incidence of θ, the steering vector d is given by:

[0019] d(ω,θ)=[p+(1-p)cosθ][1e -jωδcosθ / c … e -jω(M-1)δcosθ / c ] T , (1)

[0020] The superscript T is the transpose operator. is the imaginary unit, ω = 2πf is the angular frequency, and f is the temporal frequency. For comparison, the steering vector of a conventional ULA with an omnidirectional microphone can be expressed as:

[0021] a(ω,θ)=[1e -jωδcosθ / c … e -jω(M-1)δcosθ / c ] T , (2)

[0022] By combining the equation for the beam pattern of a directional microphone with equation (2) for a conventional ULA with an omnidirectional microphone, the steering vector d(ω,θ) can be expressed as:

[0023] d(ω,θ)=u(p,θ)a(ω,θ) (3)

[0024] The beamforming problem can be interpreted as a spatial filter that estimates the signal from the desired gaze direction and suppresses the signal from the unwanted directions by applying a complex weight vector:

[0025] h(ω)=[H1(ω)H2(ω)···H M (ω)] T . (4)

[0026] Under the given signal model, the beamformer exhibits a distortion-free response at the desired viewing direction θ = 0, i.e., d H (ω,0)h(ω)=1, where the superscript H is the conjugate-transpose operator. In other directions, the beamformer shows some distortion in the response, i.e., d H (ω,θ)h(ω)<1.

[0027] The mathematical definitions of three widely used performance metrics, namely white noise gain (WNG), beam pattern and directivity factor (DF) are shown below. WNG shows the ability of a beamformer to suppress spatially uncorrelated noise and is also the most convenient way to evaluate the sensitivity of a beamformer to some of its imperfections (e.g. sensor noise, position error, etc.). WNG is defined as: W[h(ω)] = 1 / [h H (ω)h(ω)]. The beam diagram shows the directional sensitivity of the beamformer to a plane wave 108 incident on the array 102 at an incident angle θ, as shown in FIG. Figure 1A As shown, and mathematically defined as B[h(ω),θ]=d H (ω,θ)h(ω). Frequency-invariant beam patterns are usually preferred for wideband speech processing.

[0028] The directivity factor (DF) is defined as the ratio between the array output response power in the desired steering direction and the average power in all directions, that is, DF is calculated as:

[0029] Where |B(ω,φ,θ)| is the beam pattern in spherical coordinates; θ is the azimuth angle, and φ is the elevation angle. The directivity index (DI) is defined as DI[h(ω)] = 10*log10(DF[h(ω)]).

[0030] To design the Nth-order differential beamforming problem for a ULA with directional microphones, the problem can be formulated as a system of linear equations as shown below:

[0031] R(ω,θ)h(ω)=c,(5)

[0032] Where θ is the constraint matrix R(ω,θ) of size (N+1)×M, which is given by:

[0033]

[0034] Among them, d H (ω,θ n ), n = 1, 2, ..., N, is the steering vector of length M defined in equation (1), and

[0035] θ = [0 θ1 …θ N ] T , (7)

[0036] c = [1 c1 …c N ] T , (8)

[0037] is a vector of size (N+1) containing the design parameters of the beamformer. θ (in bold letters) denotes the null position constraint vector defined in equation (7), θ1…θ N Usually define the desired null direction, and θ1…θ N are the corresponding responses in those directions, i.e. null or small values ​​of 0 if some attenuation is desired.

[0038] Combining equation (3) and equation (6), we obtain:

[0039] R(ω, θ) = U(p, θ)A(ω, θ), (9)

[0040] Among them, the steering matrix A(ω,θ) is constructed according to the steering vector a(ω,θ) as shown below:

[0041]

[0042] U(p,θ) is called the microphone response matrix and is represented as a diagonal matrix:

[0043] U(p, θ) = diag(1, u(p, θ1), …, u(p, θ N )) (11)

[0044] To maximize the WNG of array 102 and solve the linear system of equations (5), a minimum norm solution can be used to obtain the LDDMA beamformer as follows:

[0045] h(ω) = R H (ω, θ) R(ω, θ) R H (ω, θ) -1 c (12)

[0046] Among them, the LDDMA beamformer with the minimum norm solution can be identified as the same form as the LDMA beamformer.

[0047] The difference is reflected in R(ω,θ), which consists of the conventional far-field steering vector of the omnidirectional microphone and the proposed directional microphone response vector, as shown in equation (9). Combining equations (9) and (12), the LDDMA beamformer can be reformulated as:

[0048] h(ω) = A H (ω, θ) U H (p, θ) [U(p, θ) A(ω, θ) A H (ω, θ) U H (p, θ)] -1 c.(13)

[0049] This equation neatly shows the relationship between the solution of conventional LDMA and the proposed LDDMA, which extends LDMA by introducing another degree of freedom U(p,θ). In other words, LDMA is a special case of LDDMA when all microphones in equation (11) are reduced to the identity matrix when p=1, i.e., LDDMA can be used as a more general framework for designing LDMA.

[0050] Figure 2 An example of LDDMA 200 is shown. In this example, LDDMA 200 is shown to have six microphones, 202, 204, 206, 208, 210 and 212 (M=6), which are arranged linearly with an inter-element spacing of 1 cm (δ=1 cm). The types of microphones that can be used are omnidirectional (p=1), cardioid (p=0.5) and dipole (p=0). The incident angle of plane wave 214 is θ.

[0051] To evaluate the impact of different types of directional microphones (i.e., p) on the performance of the LDDMA beamformer, three commonly used microphone elements, i.e., omnidirectional (p=1), cardioid (p=0.5), and dipole (p=0), were used to form a ULA with an array configuration of δ=1 cm and M=6. The beam patterns of the two designs at 1 kHz, 3 kHz, and 6 kHz are compared, i.e., and c = [1 0 0] T The second-order cardioid and and c = [1 0 0 0] T The third-order pattern.

[0052] Figure 3A and Figure 3B The beam patterns of the second order cardioid 302 and the third order pattern 304 at 1 kHz are shown respectively, and Figure 4A and Figure 4B The beam patterns of the second order cardioid 402 and the third order pattern 404 at 3 kHz are shown respectively. It can be observed that the LDDMA beamformer at low frequencies (1 kHz and 3 kHz) matches the desired beam pattern well. Figure 5A and Figure 5B , which shows the beam patterns of the second-order cardioid line 502 and the third-order pattern 504 at 6kHz, respectively, and the beam patterns deviate from the desired beam patterns at higher frequencies (i.e., 6kHz). The LDDMA with a cardioid microphone (p=0.5) obtains the maximum sidelobe attenuation in the 2nd order design, while the LDDMA with a dipole microphone (p=0) obtains the maximum sidelobe attenuation in the 3rd order design. It should be noted that the LDDMA with p=1 becomes a conventional LDMA with an omnidirectional microphone.

[0053] Fig. 6A and Figure 6B Comparisons in WNG 602 and DI 604 are shown for 2nd order LDDMA designs with different types of microphones (p=1, p=0.5, p=0), respectively. Fig. 7A and 7B Comparisons in WNG 702 and DI 704 are shown for 3rd order LDDMA designs with different types of microphones (p=1, p=0.5, p=0), respectively.

[0054] like Fig. 6A As shown, the 2nd order LDDMA with directional microphones (p=0.5 and p=0) exhibits significantly higher WNG than conventional LDMA (omnidirectional microphone, p=1) at low frequencies, about 20 dB below 400 Hz. Figure 6B In FIG. 5 , for directional microphones with p=0.5 and p=0), their 2nd-order LDDMA has the same DI, which is higher than the DI of conventional LDMA at a high frequency of about 3 kHz.

[0055] like Fig. 7A As shown in Figure 2, the third-order LDDMA design with a dipole microphone (p=0) achieves the best WNG at low frequencies, while the conventional LDMA (p=1) is the worst. Figure 7BAs shown, at high frequencies above about 5.5 kHz, LDDMA with two directional microphones (p=0.5 for cardioid and p=0 for dipole) produces better DI than LDDMA with an omnidirectional microphone (p=1), and the type of microphone does not make much difference. At low frequencies below about 5.5 kHz, the DI is less than about 0.5 dB.

[0056] Therefore, the performance of WNG and DI of the 3rd-order design is similar to that of the 2nd-order design, that is, given the same constraints, directional microphones are more suitable than omnidirectional microphones in terms of WNG and DI performance when constructing LDMA.

[0057] Figure 8 An example process 800 for constructing LDDMA is shown. LDDMA can include uniform and non-uniform LDDMA.

[0058] In box 802, a steering vector d(ω,θ) may be generated, which is used for the proposed device (LDDMA). That is, some required parameters of LDDMA, including parameters δ, p, θ, N and M, may be preselected to generate the steering vector d(ω,θ). In box 804, a constraint matrix R(ω,θ) may be generated based on the steering vector d(ω,θ). For example, as shown in equation (9), the constraint matrix R(ω,θ) may be re-formulated based on the steering matrix and the microphone response matrix (e.g., equations (10) and (1)), respectively, and is a matrix of size (N+1) x M, where N is the order of differential beamforming of the ULA and M is the number of microphones. The microphone response matrix may be derived based on the beam pattern of the directional microphone with the properties of the sound incident angle θ, the steering direction α and the directional microphone p as described above. For example, p=1 represents an omnidirectional microphone, p=0.5 represents a cardioid microphone, and p=0 represents a dipole microphone. Although omnidirectional, cardioid, and dipole microphones are described, the directional microphone may be any type of directional microphone.

[0059] Based on the minimum norm solution, such as equation (12) for maximizing white noise gain (WNG), the LDDMA beamformer, such as h(ω) of equation (13), can be obtained at block 806. It can be seen that the beamformer h(ω) is a frequency-dependent complex-valued weight.

[0060] At box 808, the LDDMA beamformer for the desired direction at the desired frequency can be calculated and stored in the memory, and the frame-by-frame time domain sensor signal through LDDMA can be obtained at box 810. At box 812, all time domain sensor signals can be transformed into frequency domain sensor values. For each frame, the real value of the signal in the time domain becomes a complex value in the frequency domain. The transformation method used can be a short-time Fourier transform (STFT), a filter bank, a wavelet transform, etc. In the frequency domain, the LDDMA beamformer complex value weights can be loaded in vector form (LDDMA beamformer vector), and the dot product of the frequency domain sensor signal complex value and the LDDMA beamformer vector can be obtained at box 814. Then, the result of the dot product is a single complex value in the frequency domain, which can be converted into a real value in the time domain signal by the corresponding inverse transform function.

[0061] As described above, the impact of different types of directional microphones forming a ULA, such as omnidirectional (p=1), cardioid (p=0.5), and dipole (p=0), on the performance of the LDDMA beamformer can be used with different array configurations having different inter-element spacings δ for different command patterns and the number of elements M at different frequencies to evaluate the beam patterns, such as Figure 3A , 3B , 4A, 4B, 5A, 5B, 6A, 6B, 7A and 7B. The actual LDMMA can then be constructed based on the beam pattern selected from the beam patterns and the associated parameters δ, p, θ, N and M.

[0062] Some or all of the operations of the above methods may be performed by executing computer-readable instructions stored on a computer-readable storage medium, as defined below. The term "computer-readable instructions" used in the specification and claims includes routines, applications, application modules, program modules, programs, components, data structures, algorithms, etc. Computer-readable instructions may be implemented on a variety of system configurations, including single-processor or multi-processor systems, minicomputers, mainframe computers, personal computers, handheld computing devices, microprocessor-based programmable consumer electronics, combinations thereof, and the like.

[0063] Computer-readable storage media may include volatile memory (such as random access memory (RAM)) and / or non-volatile memory (such as read-only memory (ROM), flash memory, etc.). Computer-readable storage media may also include additional removable storage and / or non-removable storage, including but not limited to flash memory, magnetic storage, optical storage, and / or tape storage, which may provide non-volatile storage of computer-readable instructions, data structures, program modules, etc.

[0064] Non-transitory computer-readable storage media are examples of computer-readable media. Computer-readable media include at least two types of computer-readable media, namely computer-readable storage media and communication media. Computer-readable storage media include volatile and non-volatile, removable and non-removable media implemented in any process or technology for storing information such as computer-readable instructions, data structures, program modules or other data. Computer-readable storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (cd-rom), digital versatile disk (DVD) or other optical storage, cassettes, tapes, disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information for access by a computing device. In contrast, communication media can embody computer-readable instructions, data structures, program modules or other data in modulated data signals (e.g., carrier waves) or other transmission mechanisms. As defined herein, computer-readable storage media do not include communication media.

[0065] Computer readable instructions stored on one or more non-transitory computer readable storage media, when executed by one or more processors, can perform the above Figure 8 The operations shown. In general, computer readable instructions include routines, programs, objects, components, data structures, and the like that perform specific functions or implement specific abstract data types. The order in which the operations are described is not intended to be construed as a limitation, and any number of the described operations can be combined in any order and / or in parallel to implement the flow.

[0066] Example clauses

[0067] A. A method for constructing a linear array (LA) of microphones, comprising: generating a steering vector of the LA with preselected parameters; generating a constraint matrix based on the steering vector; reconstructing the constraint matrix based on a microphone response matrix and a steering matrix; obtaining a beamformer by applying a minimum norm solution of the constraint matrix; verifying desired characteristics of the LA by calculating the beamformer in a desired direction; and constructing the LA based on the preselected parameters and the beamformer.

[0068] B. The method of paragraph A, wherein the microphones of the LA are directional microphones and the LA is a linear differential directional microphone array (LDDMA).

[0069] C. The method as described in paragraph B, wherein the LDDMA is one of uniform LDDMA or non-uniform LDDMA.

[0070] D. The method of paragraph A, wherein the constraint matrix is ​​a matrix of size (N+1) x M, where N is the differential beamforming order of the LA and M is the number of microphones.

[0071] E. A method according to paragraph A, wherein the microphone response matrix is ​​derived based on a beam pattern of a directional microphone having a sound incident angle, a steering direction, and characteristics of the directional microphone.

[0072] F. The method of paragraph E, wherein the characteristics of the directional microphone include omnidirectional, cardioid, and dipole.

[0073] G. The method of paragraph A, wherein the beamformer is obtained by applying a minimum norm solution to the constraint matrix, wherein the constraint matrix comprises maximizing white noise gain (WNG).

[0074] H. The method of paragraph A, wherein the calculating the beamformer in the desired direction comprises calculating the beamformer in the desired direction at a desired frequency.

[0075] I. A method as described in paragraph H, wherein the beamformer that calculates the desired direction is based on time-domain frame-by-frame sensor signals received through the LA.

[0076] J. The method as described in paragraph I also includes: converting all of the time-domain frame-by-frame sensor signals into frequency-domain sensor values.

[0077] K. The method of paragraph J further includes: calculating a dot product of the frequency domain sensor value and a beamformer vector, wherein the beamformer vector is associated with a complex-valued weight of the beamformer.

[0078] L. The method of paragraph K, wherein constructing the LA based on the preselected parameters and the beamformer comprises constructing the LA based on the dot product.

[0079] M.1. A linear array (LA) comprising: a required number of microphones arranged linearly and spaced at a required microphone spacing, wherein the required number of microphones and the required microphone spacing are verified by: generating a steering vector of the LA with preselected parameters; generating a constraint matrix based on the steering vector; reconstructing the constraint matrix based on a microphone response matrix and the steering matrix; obtaining a beamformer by applying a minimum norm solution of the constraint matrix; verifying the required characteristics of the LA by calculating the beamformer in a required direction; and constructing the LA based on the preselected parameters and the beamformer.

[0080] N. The LA as described in paragraph M, wherein the microphone of the LA is a directional microphone and the LA is a linear differential directional microphone array (LDDMA).

[0081] O. The LA as described in paragraph N, wherein the LDDMA is one of uniform LDDMA or non-uniform LDDMA.

[0082] P. The LA as described in paragraph M, wherein the constraint matrix is ​​a matrix of size (N+1) x M, where N is the differential beamforming order of the LA and M is the number of microphones.

[0083] Q. The LA of paragraph M, wherein the microphone response matrix is ​​derived based on a beam pattern of a directional microphone having a sound incident angle, a steering direction, and characteristics of the directional microphone.

[0084] R. The LA of paragraph Q, wherein the characteristics of the directional microphone include omnidirectional, cardioid, and dipole.

[0085] S. The LA of paragraph M, wherein the beamformer is obtained by applying a minimum norm solution to the constraint matrix, wherein the constraint matrix comprises maximizing white noise gain (WNG).

[0086] T. The LA of paragraph M, wherein the beamformer that calculates the desired direction comprises a beamformer that calculates the desired direction at a desired frequency.

[0087] U. The LA of paragraph T, wherein the beamformer that calculates the desired direction is based on time-domain frame-by-frame sensor signals received by the LA.

[0088] V. The LA as described in paragraph U also includes: converting all time domain frame-by-frame sensor signals into frequency domain sensor values.

[0089] W. The LA of paragraph V, further comprising: calculating a dot product of a frequency domain sensor value and a beamformer vector associated with a complex-valued weight of a beamformer.

[0090] X. The LA of paragraph W, wherein constructing the LA based on the preselected parameters and the beamformer comprises constructing the LA based on the dot product.

[0091] Y. A computer-readable storage medium storing computer-readable instructions executable by one or more processors, wherein when the computer-readable instructions are executed by the one or more processors, the one or more processors are caused to perform operations including: generating a steering vector of a LA having preselected parameters; generating a constraint matrix based on the steering vector; reconstructing the constraint matrix based on a microphone response matrix and a steering matrix; obtaining a beamformer by applying a minimum norm solution of the constraint matrix; verifying desired characteristics of the LA by calculating the beamformer in a desired direction; and constructing the LA based on the preselected parameters and the beamformer.

[0092] Z. The computer-readable storage medium of paragraph Y, wherein the microphones of the LA are directional microphones and the LA is a linear differential directional microphone array (LDDMA).

[0093] AA. The computer-readable storage medium of paragraph Z, wherein the LDDMA is one of uniform LDDMA or non-uniform LDDMA.

[0094] AB. The computer-readable storage medium of paragraph Y, wherein the constraint matrix is ​​a matrix of size (N+1)xM, where N is the differential beamforming order of the LA and M is the number of microphones.

[0095] AC. A computer-readable storage medium as described in paragraph Y, wherein the microphone response matrix is ​​derived based on a beam pattern of a directional microphone having a sound incident angle, a steering direction, and characteristics of the directional microphone.

[0096] AD. The computer-readable storage medium of paragraphs AC, wherein the characteristics of the directional microphone include omnidirectional, cardioid, and dipole.

[0097] AE. The computer-readable storage medium of paragraph Y, wherein the beamformer is obtained by applying a minimum norm solution to the constraint matrix, wherein the constraint matrix comprises maximizing white noise gain (WNG).

[0098] AF. The computer-readable storage medium of paragraph Y, wherein calculating the beamformer for the desired direction comprises calculating the beamformer for the desired direction at a desired frequency.

[0099] AG. The computer-readable storage medium of paragraphs AF, wherein the beamformer that calculates the desired direction is based on time-domain frame-by-frame sensor signals received via the LA.

[0100] AH. The computer-readable storage medium of paragraphs AG further comprising: converting all of the time-domain frame-by-frame sensor signals into frequency-domain sensor values.

[0101] AI. The computer-readable storage medium of paragraphs AH further comprising: calculating a dot product of a frequency domain sensor value and a beamformer vector associated with a complex-valued weight of the beamformer.

[0102] AJ. The computer-readable storage medium of paragraph AI, wherein constructing the LA based on the preselected parameters and the beamformer comprises constructing the LA based on the dot product.

Claims

1. A method for constructing a linear array (LA) of microphones, comprising: generating a steering vector of the LA with preselected parameters; generating a constraint matrix based on the steering vector; reconstructing the constraint matrix based on the microphone response matrix and the steering matrix; Obtaining a beamformer by applying a minimum norm solution of the constraint matrix; Verify the desired characteristics of the LA by calculating the beamformer in the desired direction; as well as constructing the LA based on the preselected parameters and the beamformer; The preselected parameters include complex values ​​of frequency domain sensor signals transformed from frame-by-frame time domain sensor signals of the LA, and constructing the LA based on the preselected parameters and the beamformer includes: constructing the LA based on a dot product between the complex values ​​of the frequency domain sensor signals and a vector of the beamformer.

2. The method according to claim 1, wherein: The microphones of the LA are directional microphones, and the LA is a linear differential directional microphone array (LDDMA).

3. The method according to claim 2, characterized in that The LDDMA is one of uniform LDDMA or non-uniform LDDMA.

4. The method according to claim 1, wherein: The constraint matrix is ​​a matrix of size (N+1)×M, where N is the differential beamforming order of the LA and M is the number of the microphones.

5. The method according to claim 1, wherein: The microphone response matrix is ​​derived based on a beam pattern of a directional microphone having a sound incident angle, a steering direction, and characteristics of the directional microphone.

6. The method according to claim 5, wherein: The characteristics of the directional microphones include omnidirectional, cardioid and dipole.

7. The method according to claim 1, wherein: The beamformer is obtained by applying a minimum norm solution to the constraint matrix, the constraint matrix comprising maximizing a white noise gain (WNG).

8. The method according to claim 1, wherein: The beamformer for calculating the desired direction includes a beamformer for calculating the desired direction at a desired frequency.

9. The method according to claim 8, wherein: The beamformer for the desired direction is calculated based on the time-domain frame-by-frame sensor signals received via the LA.

10. The method according to claim 9, further comprising: All of said time-domain frame-by-frame sensor signals are converted into frequency-domain sensor values.

11. The method according to claim 10, further comprising: The dot product of the frequency domain sensor value and the beamformer vector is calculated, the beamformer vector being associated with the complex-valued weights of the beamformer.

12. The method according to claim 11, wherein: The constructing the LA based on the preselected parameters and the beamformer comprises: constructing the LA based on the dot product.

13. A linear array (LA), comprising: The desired number of microphones are linearly arranged and spaced at the desired microphone spacing, the desired number of microphones and the desired microphone spacing are verified by: generating a steering vector of the LA with preselected parameters; generating a constraint matrix based on the steering vector; reconstructing the constraint matrix based on the microphone response matrix and the steering matrix; Obtaining a beamformer by applying a minimum norm solution of the constraint matrix; Verify the desired characteristics of the LA by calculating the beamformer in the desired direction; as well as constructing the LA based on the preselected parameters and the beamformer; The preselected parameters include complex values ​​of frequency domain sensor signals transformed from frame-by-frame time domain sensor signals of the LA, and constructing the LA based on the preselected parameters and the beamformer includes: constructing the LA based on a dot product between the complex values ​​of the frequency domain sensor signals and a vector of the beamformer.

14. The LA according to claim 13, wherein: The microphones of the LA are directional microphones, and the LA is a linear differential directional microphone array (LDDMA).

15. The LA according to claim 14, wherein: The LDDMA is one of uniform LDDMA or non-uniform LDDMA.

16. The LA according to claim 13, wherein: The constraint matrix is ​​a matrix of size (N+1)×M, where N is the differential beamforming order of the LA and M is the number of the microphones.

17. The LA according to claim 13, wherein: The microphone response matrix is ​​derived based on a beam pattern of a directional microphone having a sound incident angle, a steering direction, and characteristics of the directional microphone.

18. The LA according to claim 17, wherein: The characteristics of the directional microphones include omnidirectional, cardioid and dipole.

19. The LA according to claim 13, wherein: The beamformer is obtained by applying a minimum norm solution to the constraint matrix, the constraint matrix comprising maximizing a white noise gain (WNG).

20. The LA of claim 13, wherein: The beamformer for calculating the desired direction includes a beamformer for calculating the desired direction at a desired frequency.

21. The LA according to claim 20, wherein: The beamformer for the desired direction is calculated based on the time-domain frame-by-frame sensor signals received via the LA.

22. The LA of claim 21, further comprising: All of said time-domain frame-by-frame sensor signals are converted into frequency-domain sensor values.

23. The LA of claim 22, further comprising: The dot product of the frequency domain sensor value and the beamformer vector is calculated, the beamformer vector being associated with the complex-valued weights of the beamformer.

24. The LA according to claim 23, wherein: The constructing the LA based on the preselected parameters and the beamformer comprises: constructing the LA based on the dot product.

25. A computer-readable storage medium storing computer-readable instructions executable by one or more processors, the computer-readable instructions, when executed by the one or more processors, causing the one or more processors to perform operations comprising: generating a steering vector of the LA with preselected parameters; generating a constraint matrix based on the steering vector; reconstructing the constraint matrix based on the microphone response matrix and the steering matrix; Obtaining a beamformer by applying a minimum norm solution of the constraint matrix; Verify the desired characteristics of the LA by calculating the beamformer in the desired direction; as well as constructing the LA based on the preselected parameters and the beamformer; The preselected parameters include complex values ​​of frequency domain sensor signals transformed from frame-by-frame time domain sensor signals of the LA, and constructing the LA based on the preselected parameters and the beamformer includes: constructing the LA based on a dot product between the complex values ​​of the frequency domain sensor signals and a vector of the beamformer.

26. The computer-readable storage medium of claim 25, wherein: The microphones of the LA are directional microphones, and the LA is a linear differential directional microphone array (LDDMA).

27. The computer-readable storage medium of claim 26, wherein: The LDDMA is one of uniform LDDMA or non-uniform LDDMA.

28. The computer-readable storage medium of claim 25, wherein: The constraint matrix is ​​a matrix of size (N+1)×M, where N is the differential beamforming order of the LA and M is the number of the microphones.

29. The computer-readable storage medium of claim 25, wherein: The microphone response matrix is ​​derived based on a beam pattern of a directional microphone having a sound incident angle, a steering direction, and characteristics of the directional microphone.

30. The computer-readable storage medium of claim 29, wherein: The characteristics of the directional microphones include omnidirectional, cardioid and dipole.

31. The computer-readable storage medium of claim 25, wherein: The beamformer is obtained by applying a minimum norm solution to the constraint matrix, the constraint matrix comprising maximizing a white noise gain (WNG).

32. The computer-readable storage medium of claim 25, wherein: The beamformer for calculating the desired direction includes a beamformer for calculating the desired direction at a desired frequency.

33. The computer-readable storage medium of claim 32, wherein: The beamformer for the desired direction is calculated based on the time-domain frame-by-frame sensor signals received via the LA.

34. The computer-readable storage medium of claim 33, wherein: Also includes: All of said time-domain frame-by-frame sensor signals are converted into frequency-domain sensor values.

35. The computer-readable storage medium of claim 34, wherein: Also includes: The dot product of the frequency domain sensor value and the beamformer vector is calculated, the beamformer vector being associated with the complex-valued weights of the beamformer.

36. The computer-readable storage medium of claim 35, wherein: The constructing the LA based on the preselected parameters and the beamformer comprises: constructing the LA based on the dot product.

Images