Multi-level microphone array stacking device and method

Abstract

The invention relates to the technical field of modular connection of electronic equipment, and discloses a multi-level microphone array stacking device and method.The multi-level microphone array stacking device comprises microphone bodies arranged in a multi-layer stacking mode, every two adjacent microphone bodies are rotationally connected, and the side edges of the microphone bodies are provided with a plurality of identifiers used for identifying the stacking state of the equipment. According to the invention, the microphone bodies are stacked in a modularized manner, 2-4 layers of equipment are supported to be stacked, and each layer of equipment comprises eight MEMS microphone array units, so that the requirements of more installation environments can be met while the appearance effect and the installation convenience are ensured; and during superposition, a sound field model is reconstructed through an asymmetric spherical harmonic function correction algorithm, so that more excellent acoustic characteristics are obtained. Through a core scheme of dynamic geometric modeling and deep learning sound field reconstruction, the problem of geometric suitability of a rotary stacking array is solved, and high-precision and extensible three-dimensional space audio capture is realized in combination with modular hardware design and adaptive signal processing.

Description

Technical Field

[0001] The present invention relates to the technical field of modular connection of electronic equipment, and in particular to a multi-level microphone array superposition device and method. Background Art

[0002] Modular connection technology of electronic devices enables flexible combination and expansion of hardware components through standardized interfaces and protocols, and is a key technology for building reconfigurable systems.

[0003] In existing technologies, spatial audio capture solutions using traditional microphone arrays are primarily designed around uniform geometric structures (e.g., spherical, linear, or circular arrays) and rely on least squares-based filter matrix optimization. Their technical solutions and limitations are as follows:

[0004] Ambisonics coding scheme under geometric constraints; the current mainstream Ambisonics (high-fidelity stereo sound replication) technology achieves sound field reconstruction through spherical harmonics (SHs) matching, but it is only applicable to uniformly distributed spherical / circular arrays.

[0005] Adaptive spatial filtering and noise suppression: Beamforming (such as the MVDR algorithm (minimum variance distortionless response)) and adaptive filters are used to achieve noise suppression, but they face challenges in dynamic sound source tracking and complex sound fields.

[0006] Limitations of hardware architecture: Existing microphone arrays mostly adopt a fixed layout (such as a 4-8 channel planar or spherical array) and lack a modular stacking design.

[0007] Insufficient coordination between algorithms and hardware: Signal processing (such as DNN (deep neural network) encoding) and hardware design are often separated.

[0008] The core issues that the current solution needs to address can be summarized into the following four categories:

[0009] 1. Limited sound field reconstruction accuracy due to insufficient geometric adaptability

[0010] Traditional microphone arrays (such as uniform linear arrays and spherical arrays) rely on fixed geometric structures and are not suitable for stacking or compact stereoscopic layouts. Stackable designs can lead to non-uniform spatial distribution, disrupting the matching conditions of spherical harmonics, resulting in increased sound source localization errors and spatial aliasing (distortion of high-frequency signals).

[0011] 2. Robustness Defects in Dynamic Sound Field Environments

[0012] Existing beamforming (such as MVDR) and TDOA positioning technologies face bottlenecks in tracking moving sound sources, separating multiple sound sources, and suppressing non-stationary noise. For example, adaptive filtering algorithms rely on preset sound source parameters and struggle to adapt in real time to delay variations caused by geometric changes in stacked arrays. Traditional reverberation suppression algorithms (such as GCC-PHAT) also experience performance degradation in complex reflective environments.

[0013] 3. Insufficient compatibility between hardware scalability and modular design

[0014] Existing arrays are mostly fixed layouts (e.g., 4-8 channel planar arrays), making it difficult to expand the channel count through stacking to support high-order ambisonics (HOA). For example, stacked microphone structures in mobile terminals suffer from poor sealing, circuit interference, and low resonant frequency, leading to crosstalk between channels and reduced signal-to-noise ratio. Furthermore, stacking multiple modules presents issues with synchronization errors (clock drift) and power optimization (multi-channel real-time processing).

[0015] 4. Lack of collaborative optimization of algorithms and hardware

[0016] Traditional signal processing (such as beamforming and spatial spectrum estimation) is not optimized for the non-uniform characteristics of stacked arrays. For example, TDOA positioning relies on precise geometric parameters, but the dynamic combination of stacked modules can cause the positioning model to fail. Existing deep learning models (such as U-Net) do not incorporate the multi-layer sound field characteristics of stacked arrays, resulting in insufficient 3D audio reconstruction accuracy.

[0017] In summary, the core problems of existing technical defects are: poor geometric adaptability (dependence on regular arrays), insufficient robustness in dynamic environments (difficulty in handling moving sound sources / noise), and limited hardware scalability (lack of modular stacking design).

[0018] Therefore, we need to propose a multi-level microphone array superposition device and method to break through the geometric constraints and dynamic environment limitations of traditional microphone arrays and provide a high-precision and scalable spatial audio acquisition solution for professional fields. Summary of the Invention

[0019] The purpose of the present invention is to provide a multi-level microphone array superposition device and method, which stacks microphone bodies in a modular manner and supports 2-4 layers of device stacking. Each layer of equipment contains 8 MEMS microphone array units, which can meet the requirements of more installation environments while ensuring the appearance effect and installation convenience; during superposition, the asymmetric spherical harmonic function correction algorithm is used to reconstruct the sound field model to obtain more outstanding acoustic characteristics, so as to solve the problems raised in the above background technology.

[0020] To achieve the above-mentioned object, the present invention provides the following technical solution: a multi-level microphone array stacking device, comprising a plurality of stacked microphone bodies, wherein two adjacent microphone bodies are rotatably connected to each other, a fixed end plate is provided on the top surface of the microphone body, and a rotating plate is provided on the bottom of the microphone body, and the two microphone bodies stacked up and down are connected by the fixed end plate and the rotating plate;

[0021] The interior of the microphone body is provided with a plurality of microphones distributed in an annular manner with equal distances, and the side of the microphone body is provided with a plurality of identifications for identifying the superposition status of the device, and different numbers of channel recording modes are enabled according to the different superposition states of the identified devices.

[0022] Preferably, the fixed end disk is provided with a first magnet, a first anti-mistake boss and a female end connector, and the rotating disk is provided with a second magnet, a second anti-mistake boss and a male end connector.

[0023] Preferably, the first magnet is magnetically connected to the second magnet, the first anti-mistake boss is aligned with the second anti-mistake boss, and the fixed end disk and the rotating disk are tightly fitted together by magnetic attraction.

[0024] Preferably, eight microphones are provided, and the microphones are configured as omnidirectional MEMS microphones, and the microphones are fixed inside the microphone body via a magnetic interface.

[0025] Preferably, each microphone is provided with an independent cavity isolated from each other, the sound hole depth of the independent cavity is less than 1.5 mm and the diameter is greater than 2 mm, and the signal-to-noise ratio of the microphone is greater than 74 dB.

[0026] Preferably, each of the microphone bodies is provided with four marks, and the four marks include mark 2, mark 3, and mark 4 located at the upper end of the side of the microphone body, and mark 1 located at the lower end of the side of the microphone body.

[0027] Preferably, the mark 1 on the upper microphone body is aligned with any one of the marks 2, 3 or 4 on the adjacent microphone body located below.

[0028] Based on the multi-level microphone array superposition device described above, the present invention also provides a multi-level microphone array superposition method, including the following process:

[0029] A1. Stack multiple microphone bodies together, and automatically attach the two adjacent microphone bodies above and below to each other through magnetic attraction.

[0030] A2. Rotate the upper microphone body and align Mark 1 on the upper microphone body with Mark 2, Mark 3, or Mark 4 on the lower microphone body.

[0031] A3. The microphone body automatically recognizes the superposition state and enables multi-channel recording mode.

[0032] Preferably, this method is applied to capture spatial audio, and the process of spatial audio capture is as follows:

[0033] S1. Initialization calibration:

[0034] Acoustic pulse calibration: emit pink noise, collect the impulse response of each microphone, and establish the mapping relationship between stacking geometry parameters and sound field;

[0035] Deep learning model loading: pre-trained U-Net+GRU model adapted to the current number of stacked layers;

[0036] S2. Real-time processing:

[0037] Sensor data fusion: stacking coordinates and dynamically updating the spherical harmonics weight matrix;

[0038] Frequency band processing: low-frequency signals are directly synthesized, and high-frequency signals are optimized for beam directivity through the RLS algorithm;

[0039] S3, Adaptive calibration:

[0040] Periodic cross-correlation analysis: performs inter-channel delay compensation every 10 seconds to correct stacking mechanical errors;

[0041] Abnormal detection: When an angle deviation >1° or a signal-to-noise ratio drop >3dB is detected, the recalibration process is triggered;

[0042] S4. Performance verification and optimization: including

[0043] Tests in a static environment: verify the impact of the number of stacking layers on sound source localization error;

[0044] Testing in a dynamic environment: Simulating a moving sound source (speed 0.5m / s) and background noise (SNR = 10dB) to test tracking accuracy;

[0045] Positioning error optimization;

[0046] and high frequency aliasing suppression.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] 1. The present invention adopts modular stacking of microphone bodies and supports 2-4 layers of device stacking. Each layer of equipment contains 8 MEMS microphone array units, which can meet the requirements of more installation environments while ensuring the appearance effect and installation convenience. During stacking, the asymmetric spherical harmonic function correction algorithm is used to reconstruct the sound field model to obtain more outstanding acoustic characteristics.

[0049] 2. This invention solves the geometric adaptability problem of rotating stacked arrays through the core solution of dynamic geometric modeling + deep learning sound field reconstruction. Combining modular hardware design with adaptive signal processing, it realizes high-precision and scalable three-dimensional spatial audio capture.

[0050] 3. The present invention supports stacking of any number of layers and fixed-angle rotation, reducing the sound field reconstruction error by >40%; combined with frequency-band hybrid beamforming + time synchronization algorithm, the signal-to-noise ratio is improved by >15dB; the modular design is compatible with high-order ambisonics (HOA), the number of channels can be expanded to 32, and the hardware compatibility is strong. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a schematic structural diagram of the top surface of the microphone body of the present invention;

[0052] Figure 2 Schematic diagram of the distribution of microphones of the present invention;

[0053] Figure 3 This is a schematic structural diagram of the bottom surface of the microphone body of the present invention;

[0054] Figure 4 is a cross-sectional view of the microphone body of the present invention;

[0055] Figure 5 This is a schematic diagram of the structure of two layers of microphone bodies stacked together in the present invention;

[0056] Figure 6 This is a schematic diagram of the structure of three stacked microphone bodies in Example 2 of the present invention;

[0057] Figure 7 This is a schematic diagram of the structure of four stacked microphone bodies in Example 3 of the present invention.

[0058] In the figure: 1. Microphone body; 2. Fixed end plate; 21. Magnet 1; 22. Anti-fouling boss 1; 23. Female connector; 3. Label; 4. Microphone; 5. Magnet 2; 6. Anti-fouling boss 2; 7. Male connector; 8. Rotating plate. DETAILED DESCRIPTION

[0059] 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 making creative efforts are within the scope of protection of the present invention.

[0060] See also Figure 1-5 The present invention provides a technical solution: a multi-level microphone array superposition device, comprising microphone bodies 1 stacked in multiple layers, two adjacent microphone bodies 1 being rotatably connected, the inter-layer rotation angle being set by a limiting structure during stacking, and the preset minimum step angle during rotation being 7.5°, the top surface of the microphone body 1 being provided with a fixed end plate 2, the bottom of the microphone body 1 being provided with a rotating plate 8, and the two microphone bodies 1 stacked up and down being connected via the fixed end plate 2 and the rotating plate 8;

[0061] Optionally, the limiting structure includes a limiting ring arranged on the top surface of the microphone body 1 and a protrusion fixedly arranged on the bottom surface of the microphone body 1. The upper surface of the limiting ring is set as a continuous wavy curved surface, and the limiting ring is movably arranged. When the upper microphone body 1 is rotated, the protrusion gradually moves from the lowest point of the curved surface to the highest point of the curved surface on the limiting ring, and then moves from the highest point to the lowest point, thereby realizing a minimum step angle rotation of the microphone body 1.

[0062] A clock unit and a main control unit are also provided inside the microphone body 1. The main control unit coordinates and synchronizes the clock units in each microphone body 1 through the PTP protocol, so that the time jitter is less than 1μs. The outer shell of the microphone body 1 adopts an aluminum shielding cover to reduce circuit crosstalk (signal-to-noise ratio improvement >10dB).

[0063] The interior of the microphone body 1 is provided with a plurality of microphones 4 distributed in an annular manner at equal intervals. The side of the microphone body 1 is provided with a plurality of identification marks 3 for identifying the device superposition state. Different numbers of channel recording modes are enabled by different identified device superposition states.

[0064] The fixed end disk 2 is provided with a magnet 21, an anti-mistake boss 22 and a female end connector 23, and the rotating disk 8 is provided with a magnet 2 5, an anti-mistake boss 2 6 and a male end connector 7.

[0065] The foolproof boss is used to constrain the angular deviation of the microphone body 1 during rotation.

[0066] Magnet 1 21 is magnetically connected to magnet 2 5 , anti-mistake boss 1 22 is aligned with anti-mistake boss 2 6 , and the fixed end disk 2 and the rotating disk 8 are tightly fitted together by magnetic attraction.

[0067] There are eight microphones 4 , each of which is an omnidirectional MEMS microphone. The microphone 4 is fixed inside the microphone body 1 through a magnetic interface.

[0068] Each microphone 4 is provided with an independent cavity isolated from each other. The sound hole of the independent cavity has a depth of less than 1.5 mm and a diameter greater than 2 mm. The signal-to-noise ratio of the microphone 4 is greater than 74 dB.

[0069] Each microphone body 1 is provided with four marks 3 , including mark 2, mark 3, and mark 4 located at the upper end of the side of the microphone body 1 , and mark 1 located at the lower end of the side of the microphone body 1 .

[0070] The mark 1 on the upper microphone body 1 is aligned with any one of the marks 2, 3 or 4 on the adjacent microphone body 1 below.

[0071] Based on the multi-level microphone array superposition device described above, the present invention also provides a multi-level microphone array superposition method, including the following process:

[0072] A1. Stack multiple microphone bodies 1 together, and automatically attach two adjacent microphone bodies 1 to each other through magnetic attraction.

[0073] A2. Rotate the upper microphone body 1 to align the mark 1 on the upper microphone body 1 with the mark 2, mark 3, or mark 4 on the lower microphone body 1;

[0074] A3. Microphone body 1 automatically recognizes the superposition state and enables multi-channel recording mode.

[0075] Take the stacking of two microphone bodies 1 as an example: the microphone bodies 1 from bottom to top are respectively recorded as device one and device two, and device two is placed on device one, automatically adsorbed by magnetic attraction, and the male connector 7 on device two is tightly matched with the female connector 23 on device one, and device two is rotated to align the mark one of device two with the mark two on device one, and then the superposition state is automatically identified, and the 16-channel recording mode is enabled.

[0076] This method is applied to the capture of spatial audio. The process of spatial audio capture is as follows:

[0077] S1. Initialization calibration:

[0078] Acoustic pulse calibration: emit pink noise, collect the impulse response of each microphone, and establish the mapping relationship between stacking geometry parameters and sound field;

[0079] Deep learning model loading: pre-trained U-Net+GRU model adapted to the current number of stacked layers;

[0080] S2. Real-time processing:

[0081] Sensor data fusion: stacking coordinates and dynamically updating the spherical harmonics weight matrix;

[0082] Frequency band processing: low-frequency signals are directly synthesized, and high-frequency signals are optimized for beam directivity through the RLS algorithm;

[0083] S3, Adaptive calibration:

[0084] Periodic cross-correlation analysis: performs inter-channel delay compensation every 10 seconds to correct stacking mechanical errors;

[0085] Abnormal detection: When an angle deviation >1° or a signal-to-noise ratio drop >3dB is detected, the recalibration process is triggered;

[0086] S4. Performance verification and optimization: including

[0087] Tests in a static environment: verify the impact of the number of stacking layers on sound source localization error;

[0088] Testing in a dynamic environment: Simulating a moving sound source (speed 0.5m / s) and background noise (SNR = 10dB) to test tracking accuracy;

[0089] Positioning error optimization;

[0090] and high frequency aliasing suppression.

[0091] This invention solves the geometric adaptability problem of rotating stacked arrays through the core solution of dynamic geometric modeling and deep learning sound field reconstruction. Combining modular hardware design with adaptive signal processing, it achieves high-precision and scalable three-dimensional spatial audio capture. Its technical advantages include:

[0092] 1. Geometric robustness: supports arbitrary stacking of layers and fixed-angle rotation, reducing sound field reconstruction error by >40%;

[0093] 2. Dynamic anti-interference: frequency-band hybrid beamforming + time synchronization algorithm, signal-to-noise ratio improved by >15dB;

[0094] 3. Hardware compatibility: The modular design is compatible with High-Order Ambisonics (HOA), and the number of channels can be expanded to 32.

[0095] The present invention stacks the microphone body 1 in a modular manner and supports 2-4 layers of device stacking. Each layer of device contains 8 MEMS microphone array units, which can meet the requirements of more installation environments while ensuring the appearance effect and installation convenience. During stacking, the asymmetric spherical harmonic function correction algorithm is used to reconstruct the sound field model to obtain more outstanding acoustic characteristics.

[0096] The microphone array superposition of the present invention is achieved through the following technologies:

[0097] 1. Dynamic geometry modeling and spherical harmonic function correction

[0098] Real-time geometric parameter perception: A magnetic induction device is integrated in each stacked unit to sense the rotation angle between stacked layers (such as the fixed rotation angle of each layer) in real time, thereby determining the microphone position of each unit in three-dimensional space.

[0099] Spherical harmonic function correction: Based on the three-dimensional uniform geometric distribution after stacking, the cylindrical body is closed, and the expansion order of the spherical harmonic function is adjusted by dynamic fitting (such as expanding from first-order to second-order Ambisonics). A frequency-band phase compensation factor is introduced to suppress spatial aliasing caused by rotational stacking.

[0100] This technology combines fixed angle parameters of rotation stacking, presets a spherical harmonic weight matrix library with different numbers of layers, and dynamically matches the optimal model through sensor data during runtime.

[0101] 2. Adaptive Beamforming and Phase Synchronization

[0102] Frequency-band hybrid beamforming: The delay-and-add method is used in the low-frequency band (<2kHz), leveraging the three-dimensional symmetry formed by stacked rotation to enhance directivity. The RLS adaptive beamforming algorithm is used in the high-frequency band (>2kHz), optimizing weights in real time to suppress phase ambiguity caused by rotational stacking.

[0103] Multi-level phase synchronization: PTP (Precise Time Protocol) and fractional delay filters are used to compensate for delay errors caused by rotation angle differences between stacked layers.

[0104] This technology pre-calculates the delay compensation parameters between units in different layers based on the fixed angle rule of rotational stacking, reducing the real-time computing load.

[0105] 3. Deep Learning-Driven Sound Field Reconstruction

[0106] Multi-scale sound field feature fusion:

[0107] Input layer: stacked multi-channel signals (8 channels / layer) and geometric parameters (rotation angle, coordinates);

[0108] Network architecture: A U-Net+GRU hybrid model (convolutional neural network (CNN) and gated recurrent unit (GRU)) is used. The CNN branch extracts spatial-frequency features, and the GRU branch models time series correlation.

[0109] End-to-end sound field mapping: Output is a modified B-Format signal (B-Format signal is a multi-channel audio format used in Ambisonics technology to store and process three-dimensional sound field information). It supports dynamic adjustment of the number of stacking layers (for example, a three-layer stereo array corresponds to 24 channels).

[0110] This technology uses training data to simulate different rotational stacking forms, enhancing the model's generalization ability to geometric changes.

[0111] 4. Modular Hardware Architecture and Anti-interference Design

[0112] Rotatable stackable module design: Each layer of modules uses a magnetic interface + pogoPIn, supports 360° free rotation and locks the preset angle (such as 3.75° and 7.5° step); the microphone layout refers to the six-microphone ring array configuration (radius 35mm, horizontal plane angle <10°) to optimize near-field pickup performance.

[0113] Anti-crosstalk and sealed design: Each microphone is isolated in an independent cavity, with a sound hole depth of less than 1.5mm and a diameter of more than 2mm to avoid cavity resonance; the silicone sleeve provides shock absorption and sealing, reducing the impact of stacked mechanical vibration on the signal.

[0114] This technology transmits data between modules via a high-speed I2S bus, and the main control unit integrates the signals in real time and processes them synchronously.

[0115] Example 2

[0116] See also Figure 6 The same as Example 1 will not be repeated here, except that

[0117] Take three microphone bodies 1 stacked as an example: the microphone bodies 1 from bottom to top are respectively denoted as device 1, device 2, and device 3;

[0118] Place device 2 on device 1, and automatically attach them by magnetic attraction. The male connector 7 on device 2 fits tightly with the female connector 23 on device 1. Rotate device 2 to align the mark 1 on device 2 with the mark 3 on device 1.

[0119] Place device 3 on device 2, and automatically attach them by magnetic attraction. The male connector 7 on device 3 fits tightly with the female connector 23 on device 2. Rotate device 3 to align the mark 1 on device 3 with the mark 3 on device 2.

[0120] Then it automatically recognizes the superposition status and enables 24-channel recording mode.

[0121] Example 3

[0122] See also Figure 7The same points as in Example 1 and Example 2 are not described in detail.

[0123] Take four microphone bodies 1 stacked as an example: the microphone bodies 1 from bottom to top are respectively denoted as device 1, device 2, device 3, and device 4;

[0124] Place device 2 on device 1, and automatically attach them by magnetic attraction. The male connector 7 on device 2 fits tightly with the female connector 23 on device 1. Rotate device 2 to align the marker 1 on device 2 with the marker 4 on device 1.

[0125] Place device 3 on device 2, and automatically attach them by magnetic attraction. The male connector 7 on device 3 fits tightly with the female connector 23 on device 2. Rotate device 3 to align the mark 1 on device 3 with the mark 4 on device 2.

[0126] Place device 4 on device 3, and automatically attach them by magnetic attraction. The male connector 7 on device 4 and the female connector 23 on device 3 fit tightly together. Rotate device 4 to align the marker 1 on device 4 with the marker 4 on device 2.

[0127] Then it automatically recognizes the superposition status and enables 32-channel recording mode.

[0128] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A multi-level microphone array superposition device, characterized in that: The invention comprises microphone bodies (1) stacked in multiple layers, wherein two adjacent microphone bodies (1) are rotatably connected to each other, a fixed end plate (2) is provided on the top surface of the microphone body (1), a rotating plate (8) is provided on the bottom of the microphone body (1), and the two microphone bodies (1) stacked up and down are connected via the fixed end plate (2) and the rotating plate (8); The microphone body (1) is provided with a plurality of microphones (4) distributed in an annular manner with equal spacing therebetween, and the side of the microphone body (1) is provided with a plurality of identification marks (3) for identifying the superposition state of the device, and different numbers of channel recording modes are enabled according to the different superposition states of the identified devices.

2. The multi-level microphone array superposition device according to claim 1, characterized in that: The fixed end disk (2) is provided with a first magnet (21), a first anti-misalignment boss (22) and a female end connector (23), and the rotating disk (8) is provided with a second magnet (5), a second anti-misalignment boss (6) and a male end connector (7).

3. The multi-level microphone array superposition device according to claim 2, characterized in that: The magnet one (21) is magnetically connected to the magnet two (5), the anti-mistake boss one (22) is aligned with the anti-mistake boss two (6), and the fixed end disk (2) and the rotating disk (8) are tightly fitted by magnetic attraction.

4. The multi-level microphone array superposition device according to claim 1, characterized in that: Eight microphones (4) are provided, and the microphones (4) are configured as omnidirectional MEMS microphones. The microphones (4) are fixed inside the microphone body (1) via a magnetic interface.

5. The multi-level microphone array superposition device according to claim 1, characterized in that: Each microphone (4) is provided with an independent cavity isolated from each other, the sound hole depth of the independent cavity is less than 1.5 mm and the diameter is greater than 2 mm, and the signal-to-noise ratio of the microphone (4) is greater than 74 dB.

6. The multi-level microphone array superposition device according to claim 1, characterized in that: Each microphone body (1) is provided with four marks (3), and the four marks (3) include mark 2, mark 3, and mark 4 located at the upper end of the side of the microphone body (1), and mark 1 located at the lower end of the side of the microphone body (1).

7. The multi-level microphone array superposition device according to claim 6, characterized in that: The mark 1 on the upper microphone body (1) is aligned with any one of the mark 2, mark 3 or mark 4 on the adjacent microphone body (1) located below.

8. A multi-level microphone array superposition method, based on a multi-level microphone array superposition device according to any one of claims 1 to 7, characterized in that: The process includes the following: A1. Stacking a plurality of microphone bodies (1) together, and automatically adsorbing two adjacent microphone bodies (1) above and below each other by magnetic attraction; A2. Rotate the upper microphone body (1) to align the mark 1 on the upper microphone body (1) with the mark 2, mark 3, or mark 4 on the lower microphone body (1); A3. The microphone body (1) automatically recognizes the superposition state and enables the multi-channel recording mode.

9. The multi-level microphone array superposition method according to claim 8, applied to capturing spatial audio, is characterized in that: The process of spatial audio capture is as follows: S1. Initialization calibration: Acoustic pulse calibration: emit pink noise, collect the impulse responses of each microphone, and establish the mapping relationship between stacking geometry parameters and sound field; Deep learning model loading: pre-trained U-Net+GRU model adapted to the current number of stacked layers; S2, real-time processing: Sensor data fusion: stacking coordinates and dynamically updating the spherical harmonics weight matrix; Frequency band processing: low-frequency signals are directly synthesized, and high-frequency signals are optimized for beam directivity through the RLS algorithm; S3, Adaptive calibration: Periodic cross-correlation analysis: performs inter-channel delay compensation every 10 seconds to correct stacking mechanical errors; abnormal Detection: When an angle deviation >1° or a signal-to-noise ratio drop >3dB is detected, the recalibration process is triggered; S4. Performance verification and optimization: including Tests in a static environment: verify the impact of the number of stacking layers on sound source localization error; Testing in a dynamic environment: Simulating a moving sound source (speed 0.5m / s) and background noise (SNR = 10dB) to test tracking accuracy; Positioning error optimization; and high frequency aliasing suppression.

Images