AI noise reduction device based on directional pickup microphone
By combining a directional pickup microphone and an embedded AI processing module, the problems of poor noise reduction and inaccurate voice recognition in complex noise scenes in existing technologies are solved, achieving efficient and accurate noise reduction effects while reducing costs and delays.
Patent Information
- Application Number
- CN202510760225.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-09
- Publication Date
- 2025-09-05
AI Technical Summary
Existing technologies have difficulty achieving accurate and effective noise reduction when faced with sudden and non-steady-state noise in complex and changeable daily scenes, and are also lacking in the accuracy of human voice recognition, resulting in low output sound purity and poor audibility.
It adopts a combination of directional pickup microphone and embedded AI processing module, collects sound signals in a specific direction through a dipole microphone, and uses AI noise reduction algorithm to perform adaptive noise reduction, accurately identify human voices, and combines physical noise reduction and AI algorithm to form a dual noise reduction mechanism to achieve efficient noise reduction.
It significantly improves the noise reduction effect, can identify and process various noises in real time, provide more accurate noise reduction effects, reduce costs and data delays, and improve processing efficiency.
Smart Images
Figure CN120600037A_ABST
Abstract
Description
Technical Field
[0001] The embodiments of the present invention relate to the technical field of sound noise reduction, and in particular to an AI noise reduction device based on a directional pickup microphone. Background Art
[0002] Currently, there are many noise reduction methods in the field of sound noise reduction, such as traditional microphone array noise reduction technology, which relies on multiple microphones to form an array and uses principles such as beamforming to process sound signals. By adjusting the phase, amplitude and other parameters of the sound signals collected by each microphone, it suppresses noise and enhances the target sound (such as the human voice). However, this technology is not very effective when dealing with complex, sudden and non-steady-state noises in daily scenes. It is difficult to quickly and accurately reduce short-term high-decibel noises such as sudden vehicle horns and the sound of falling objects, resulting in the final output sound still being mixed with obvious noise interference.
[0003] Digital signal processing noise reduction technology is also a commonly used method. It processes sound signals based on fixed filtering algorithms, such as adaptive filtering, and removes noise frequency band signals according to preset filtering parameters. However, due to its lack of deep recognition of the essential characteristics of sound content (especially human voice characteristics), it is difficult to accurately distinguish between human voices and noise in complex and changing daily environmental noise environments. Misjudgment is prone to occur, either filtering out part of the human voice as noise or failing to effectively remove noise, resulting in unsatisfactory sound quality after noise reduction.
[0004] To sum up, the existing technology as a whole has obvious shortcomings. On the one hand, the noise reduction effect is limited. Faced with various complex and changeable sudden and non-steady-state noises in daily scenes, it is difficult to achieve accurate and effective noise reduction, resulting in low purity of the output sound, affecting subsequent applications; on the other hand, there are deficiencies in the accuracy of human voice recognition. It cannot accurately recognize human voices well and is prone to misjudgment, which in turn affects the audibility of the sound and the accuracy of related voice applications. Summary of the Invention
[0005] An embodiment of the present invention provides an AI noise reduction device based on a directional pickup microphone to solve the above technical problems.
[0006] An AI noise reduction device based on a directional pickup microphone, comprising:
[0007] Directional pickup microphone, using a dipole microphone, is used to collect sound analog signals in a specific direction, including human voice and environmental noise;
[0008] AD conversion module, used to convert the collected sound analog signal into a sound digital signal;
[0009] Embedded AI processing module, which uses built-in AI noise reduction algorithms to extract features from digital sound signals and processes the extracted features based on noise models to achieve adaptive noise reduction and accurate voice recognition.
[0010] The speaker is used to output the recognized human voice signal.
[0011] Optionally, the embedded AI processing module can reduce noise caused by wind blowing onto the microphone diaphragm.
[0012] Optionally, there is a sound inlet on each of the two opposite sides of the dipole microphone diaphragm, for respectively suppressing the sound signals on the two sides.
[0013] Optionally, the dipole microphone uses a dipole MEMS capacitive microphone to enhance signal sensitivity in the 0° and 180° directions and suppress sound signals in the 90° and 270° side directions.
[0014] Optionally, the embedded AI processing module includes a DSP, an MCU, and a storage unit;
[0015] Among them, DSP is used to run AI noise reduction algorithm;
[0016] The MCU has built-in noise models and parameters for various everyday scenarios. Each model is pre-trained for deep learning using a large number of noise samples from everyday scenarios.
[0017] The storage unit is used to store updated parameters.
[0018] Optionally, when the embedded AI processing module is working, the MCU reads the parameters in the storage unit, the DSP calls and runs the AI noise reduction algorithm saved in the MCU, and the DSP is connected to the AD conversion module to process the received sound signal through the internal AI noise reduction algorithm.
[0019] Optionally, when the embedded AI processing module is working, the DSP calls the noise model and corresponding parameters that have been trained through deep learning and stored in the MCU, extracts features of the sound digital signal, including the spectral features and time domain features of the sound, and analyzes the extracted features through a neural network to accurately distinguish between noise and human voices; for the identified noise, noise reduction processing is performed based on the noise model and parameters, including sudden non-steady-state noise in the conference room scenario, and wind noise in the wind noise scenario and noise caused by wind blowing on the microphone diaphragm.
[0020] Optionally, when the embedded AI processing module is working, it uses a corresponding noise reduction algorithm to perform noise reduction according to the identified noise type and intensity to obtain a pure human voice signal.
[0021] Optionally, the speaker is a high-fidelity speaker, and the corresponding frequency range can restore the sound quality of the human voice signal.
[0022] In summary, the embodiment of the present invention provides an AI noise reduction device based on a directional pickup microphone. Compared with traditional noise reduction technology, the device of this embodiment has significant advantages. Traditional noise reduction technology usually uses fixed filters or simple signal processing algorithms, which are difficult to adapt to complex noise scenarios. For example, when dealing with sudden noise, traditional technologies often make misjudgments or have poor noise reduction effects. However, this device synergizes physical noise reduction with AI algorithm noise reduction, and forms a dual noise reduction mechanism through a microphone that can collect noise in a specific direction and AI algorithm noise reduction, which greatly improves the noise reduction effect. It can also identify and process various noises in real time, providing more accurate noise reduction effects. At the same time, the device adopts a low-latency and low-cost architecture based on hardware collaborative optimization, coupling the microphone with the embedded AI processing module. The AI noise reduction algorithm runs directly on the hardware, avoiding running the algorithm in the cloud, reducing costs and data delays, and improving processing efficiency. BRIEF DESCRIPTION OF THE DRAWINGS
[0023] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0024] Figure 1 1 is a schematic structural diagram of an AI noise reduction device based on a directional pickup microphone provided by an embodiment of the present invention;
[0025] Figure 2 1 is a schematic diagram of the appearance of an AI noise reduction device based on a directional pickup microphone provided by an embodiment of the present invention;
[0026] Figure 3 This is a schematic diagram of the structure of the directional pickup microphone provided by the embodiment of the present invention. Figure 1 ;
[0027] Figure 4 This is a schematic diagram of the structure of the directional pickup microphone provided by the embodiment of the present invention. Figure 2 ;
[0028] Figure 5 This is a test chart of microphone recognition accuracy under different environmental noises provided by an embodiment of the present invention;
[0029] Figure 6 Schematic diagram of the structure of the embedded AI processing module provided by an embodiment of the present invention;
[0030] Among them, the accompanying drawings are marked as follows:
[0031] 1-diaphragm, 2-first pickup hole, 3-second pickup hole. DETAILED DESCRIPTION
[0032] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention are described clearly and completely below. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are also within the scope of protection of the present invention.
[0033] In the description of the present invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings and are intended solely to facilitate and simplify the description of the present invention. They are not intended to indicate or imply that the devices or components referred to must have, be constructed, or operate in a specific orientation, and therefore should not be construed as limitations on the present invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.
[0034] In the description of the present invention, it should also be noted that, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood broadly. For example, they may refer to fixed, detachable, or integral connections; mechanical or electrical connections; direct or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on the specific circumstances.
[0035] Figure 1 is a structural diagram of an AI noise reduction device based on a directional pickup microphone provided by an embodiment of the present invention. Figure 2 The figure is a schematic diagram of the appearance of this device. Figure 1 As shown in the figure, the device includes a directional pickup microphone, an AD conversion module, an embedded AI processing module, and a speaker. These components work together to achieve efficient noise reduction of sound signals.
[0036] Among them, the directional pickup microphone uses a single dipole microphone to collect sound analog signals in a specific direction, including human voices and environmental noise.
[0037] Preferably, the microphone adopts a dipole microphone, and there is a sound inlet on each side of the diaphragm, which can suppress the sound signals on both sides. Exemplarily, a dipole MEMS capacitive microphone can be used, that is, it points at angles of 0° and 180°. Since there is a sound inlet on each side of the diaphragm, the sensitivity is highest in the 0° and 180° directions, and it can significantly suppress the sound from the 90° and 270° side directions. This type of microphone has a high signal-to-noise ratio and can collect sound from a specific direction, which can suppress environmental noise and provide high-quality training data for subsequent AI noise reduction algorithms.
[0038] For more details, see Figure 3 and Figure 4 The dipole MEMS condenser microphone is provided with a diaphragm 1, and a first sound pickup hole 2 and a second sound pickup hole 3 are provided at the top and bottom of the pickup microphone. The center line of the diaphragm 1 coincides with the center lines of the first sound pickup hole 2 and the second sound pickup hole 3. The first sound pickup hole 2 and the second sound pickup hole 3 are respectively located on both sides of the diaphragm 1, and the first sound pickup hole 2 and the second sound pickup hole 3 are at equal distances from the diaphragm 1. The radius of the first sound pickup hole 2 is equal to the radius of the second sound pickup hole 3. Through the above structure, see Figure 3 , when the sound signal comes from the upper or lower side, that is, 0° or 180°, it is collected normally; see Figure 4 When a sound signal arrives from the left or right, that is, at 90° or 270°, the sound is split into two paths, reaching diaphragm 1 simultaneously through the top and bottom holes. At this point, the two identical sound signals are superimposed and at least partially canceled out. This microphone is more sensitive to sound signals at 0° and 180°, achieving roughly directional sound acquisition.
[0039] Figure 5 This is a test chart of microphone recognition accuracy under different environmental noise conditions. Figure 5 As shown in the figure, the microphone has a certain suppressive effect on environmental noise. A sound source with an amplitude of about 70dB is fixed 0.5 meters away from the microphone. Multiple sound sources are used to simulate environmental noise. The multiple sound sources are 0.1 meters, 0.5 meters, and 1 meter away from the microphone respectively. It can be seen from the figure that:
[0040] The ambient noise level was 60dB, which is lower than the speech amplitude, and the accuracy of the three distances remained at 100%;
[0041] The ambient noise level is 70dB, the speech amplitude is comparable, the accuracy is 87% at 1 meter, and 100% at other distances;
[0042] The ambient noise level is 85dB, which is greater than the speech amplitude. The accuracy is 100% at 0.1 meters and less than 85% at other distances.
[0043] Corresponding to a typical distance of 0.5 meters and an ambient noise level of no more than 70dB, it can maintain 100% accuracy.
[0044] The AD conversion module is used to convert the collected sound analog signals into sound digital signals. Specifically, the analog sound signals collected by the microphone are converted into digital signals through the AD conversion module for subsequent processing by the embedded AI processing module.
[0045] The AD conversion module preferably uses a high-precision analog-to-digital conversion chip, and its sampling rate can be adjusted according to actual needs. For example, for voice communication scenarios, the sampling rate can be set to 16kHz to meet the processing requirements of voice signals. This module can accurately quantize analog signals into digital signals, ensuring that the signals are not distorted during the conversion process.
[0046] The embedded AI processing module is used to use the built-in AI noise reduction algorithm to extract features from the sound digital signal, and process the extracted features based on the noise model to achieve adaptive noise reduction and accurately identify human voices.
[0047] Preferably, Figure 6 As shown, the embedded AI processing module includes:
[0048] DSP (high-performance digital signal processor): runs AI noise reduction algorithm;
[0049] MCU (control unit): Built-in noise models and parameters for hundreds of thousands of everyday scenarios. The models are pre-trained through deep learning using a large number of noise samples from everyday scenarios.
[0050] Storage unit: used to store updated parameters.
[0051] During operation, the MCU reads the parameters in the storage unit, the DSP calls and runs the AI noise reduction algorithm stored in the MCU, and the DSP is connected to the AD conversion module to process the received sound signal through the internal AI noise reduction algorithm.
[0052] In a specific embodiment, after the digital sound signal enters the embedded AI processing module, within the module, the DSP calls the noise model and corresponding parameters that have been stored in the MCU and have been trained through deep learning, and begins to process the sound signal. Feature extraction is first performed, such as extracting the spectral features and time domain features of the sound, and analyzing these features through a neural network to accurately distinguish between noise and human voices. For the identified noise, noise reduction processing is performed based on the noise model and parameters. For example, for sudden non-steady-state noise such as someone coughing in the conference room, it can be quickly and effectively suppressed. In addition, for application scenarios with wind noise, in addition to suppressing wind noise, this module can also eliminate noise caused by wind blowing on the microphone diaphragm.
[0053] Furthermore, neural network-based noise and voice recognition is a key component of the embedded AI processing module. It can effectively handle a variety of complex noise scenarios, including steady-state noise (such as air conditioning), sudden noise (such as breaking glass), and non-steady-state noise (such as crowd noise). During the noise reduction process, the embedded AI processing module applies the corresponding noise reduction algorithm based on the identified noise type and intensity. After noise reduction, a pure human voice signal is obtained.
[0054] To improve the performance of the embedded AI processing module, a high-performance digital signal processor (DSP) is used in this embodiment. The DSP has strong computing power and low power consumption, and can process large amounts of audio signal data in real time.
[0055] The speaker is used to output the recognized human voice signal. The noise-reduced human voice signal output by the embedded AI processing module is played through the speaker. In this embodiment, the speaker is a high-fidelity speaker with a wide frequency response range, capable of accurately restoring the sound quality of the human voice signal.
[0056] In summary, the embodiment of the present invention provides an AI noise reduction device based on a directional pickup microphone. Compared with traditional noise reduction technology, the device of this embodiment has significant advantages. Traditional noise reduction technology usually uses fixed filters or simple signal processing algorithms, which are difficult to adapt to complex noise scenarios. For example, when dealing with sudden noise, traditional technologies often make misjudgments or have poor noise reduction effects. However, this device synergizes physical noise reduction with AI algorithm noise reduction, and forms a dual noise reduction mechanism through a microphone that can collect noise in a specific direction and AI algorithm noise reduction, which greatly improves the noise reduction effect. It can also identify and process various noises in real time, providing more accurate noise reduction effects. At the same time, the device adopts a low-latency and low-cost architecture based on hardware collaborative optimization, coupling the microphone with the embedded AI processing module. The AI noise reduction algorithm runs directly on the hardware, avoiding running the algorithm in the cloud, reducing costs and data delays, and improving processing efficiency.
[0057] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the technical solutions of the embodiments of the present invention.
Claims
1. An AI noise reduction device based on a directional pickup microphone, characterized in that: include: Directional pickup microphone, using a dipole microphone, is used to collect sound analog signals in a specific direction, including human voice and environmental noise; AD conversion module, used to convert the collected sound analog signal into a sound digital signal; Embedded AI processing module, which uses built-in AI noise reduction algorithms to extract features from digital sound signals and processes the extracted features based on noise models to achieve adaptive noise reduction and accurate voice recognition. The speaker is used to output the recognized human voice signal.
2. The AI noise reduction device based on a directional pickup microphone according to claim 1, characterized in that: The embedded AI processing module is capable of reducing noise caused by wind blowing onto the microphone diaphragm.
3. The AI noise reduction device based on a directional pickup microphone according to claim 1, characterized in that: There is a sound inlet on each of the two opposite sides of the dipole microphone diaphragm, which is used to suppress the sound signals on the two sides respectively.
4. The AI noise reduction device based on a directional pickup microphone according to claim 3, characterized in that: The dipole microphone adopts a dipole MEMS capacitor microphone, which enhances the signal sensitivity in the directions of 0° and 180° and suppresses the sound signals in the 90° and 270° side directions.
5. The AI noise reduction device based on a directional pickup microphone according to claim 1, characterized in that: The embedded AI processing module includes a DSP, an MCU, and a storage unit; Among them, DSP is used to run AI noise reduction algorithm; The MCU has built-in noise models and parameters for various everyday scenarios. Each model is pre-trained for deep learning using a large number of noise samples from everyday scenarios. The storage unit is used to store updated parameters.
6. The AI noise reduction device based on a directional pickup microphone according to claim 5, characterized in that: When the embedded AI processing module is working, the MCU reads the parameters in the storage unit, the DSP calls and runs the AI noise reduction algorithm saved in the MCU, and the DSP is connected to the AD conversion module to process the received sound signal through the internal AI noise reduction algorithm.
7. The AI noise reduction device based on a directional pickup microphone according to claim 5, characterized in that: When the embedded AI processing module is working, the DSP calls the noise model and corresponding parameters stored in the MCU that have been trained through deep learning, extracts features of the sound digital signal, including the spectral features and time domain features of the sound, and analyzes the extracted features through a neural network to accurately distinguish between noise and human voices; for the identified noise, noise reduction processing is performed based on the noise model and parameters, including sudden non-steady-state noise in the conference room scenario, wind noise in the wind noise scenario, and noise caused by wind blowing on the microphone diaphragm.
8. The AI noise reduction device based on a directional pickup microphone according to claim 1, characterized in that: When the embedded AI processing module is working, it uses the corresponding noise reduction algorithm to reduce noise according to the identified noise type and intensity to obtain a pure human voice signal.
9. The AI noise reduction device based on a directional pickup microphone according to claim 1, characterized in that: The loudspeaker is a high-fidelity loudspeaker, and the frequency response range can restore the sound quality of the human voice signal.