Intelligent pickup method and system of microphone

Abstract

The invention relates to the technical field of microphone pickup, and discloses an intelligent pickup method and system of a microphone, and the method comprises the steps: respectively picking up an audio signal through a plurality of microphone units in a microphone array, carrying out the analysis of the audio signal to obtain a human voice signal and a background signal, and carrying out the collection of a voice signal based on the human voice signal and the background signal picked by each microphone unit. According to the technical scheme, the orientation angle of each microphone unit is adjusted, so that the microphone array is adjusted to the optimal pickup state, key frequency band recognition is performed on the human voice signal collected in the optimal pickup state, and voice enhancement processing is performed on the human voice signal according to the recognition result, so that the problem of poor pickup tracking capability of a microphone in the prior art is solved.

Description

Technical Field

[0001] This invention relates to the technical field of microphone pickup, and more particularly to an intelligent microphone pickup method and system. Background Technology

[0002] Microphone arrays consist of multiple microphone units. Through proper layout and signal processing, they can achieve a wider pickup range and more accurate sound source localization. Fixed-direction microphones can only pick up sound from a specific direction well. When the target sound source changes position, their pickup effect will be greatly reduced. If the target sound source is not directly in front of the microphone, the sound signal will be attenuated and distorted, affecting the final pickup quality. In some scenarios where it is necessary to track moving sound sources, such as stage performances and live sports events, the limitations of fixed-direction microphones are even more obvious. Summary of the Invention

[0003] The purpose of this invention is to provide an intelligent microphone pickup method and system, which aims to solve the problem of poor microphone tracking pickup capability in the prior art.

[0004] The present invention is implemented as follows: In a first aspect, the present invention provides an intelligent sound pickup method for a microphone, comprising: Audio signals are picked up by multiple microphone units in the microphone array, and the human voice signal and background signal are obtained by analysis; Based on the human voice signal and background signal picked up by each microphone unit, the orientation angle of each microphone unit is adjusted to adjust the microphone array to the optimal sound pickup state; The key frequency bands of the human voice signal acquired under optimal sound pickup conditions are identified, and the human voice signal is subjected to speech enhancement processing based on the identification results.

[0005] In a second aspect, the present invention provides an intelligent microphone pickup system for implementing the intelligent microphone pickup method described in any one of the first aspects, comprising: The signal acquisition module is used to pick up audio signals through multiple microphone units in the microphone array and analyze them to obtain human voice signals and background signals. An angle adjustment module is used to adjust the orientation angle of each microphone unit based on the human voice signal and background signal picked up by each microphone unit, so as to adjust the microphone array to the optimal sound pickup state; The voice enhancement module is used to identify key frequency bands of the human voice signal collected under optimal sound pickup conditions, and to perform voice enhancement processing on the human voice signal based on the recognition results.

[0006] This invention provides an intelligent microphone pickup method, which has the following beneficial effects: This invention can accurately separate human voice from background signals, has a good noise reduction effect, can adjust the microphone orientation to achieve the best sound pickup state, improve sound pickup accuracy, and can also identify key frequency bands to enhance speech, making the sound pickup clearer and more natural, and improving the overall audio quality. Attached Figure Description

[0007] Figure 1 This is a schematic diagram illustrating the steps of an intelligent microphone pickup method provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of an intelligent microphone pickup system provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the operation of an intelligent microphone pickup system provided in an embodiment of the present invention. Detailed Implementation

[0008] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0009] The implementation of the present invention will be described in detail below with reference to specific embodiments.

[0010] Reference Figure 1 , Figure 2 , Figure 3 The diagram shows a preferred embodiment of the present invention.

[0011] In a first aspect, the present invention provides an intelligent sound pickup method for a microphone, comprising: S1: Audio signals are picked up by multiple microphone units in the microphone array, and the human voice signal and background signal are obtained by parsing. S2: Based on the human voice signal and background signal picked up by each microphone unit, adjust the orientation angle of each microphone unit to adjust the microphone array to the optimal sound pickup state; S3: Identify key frequency bands in the human voice signal acquired under optimal sound pickup conditions, and perform speech enhancement processing on the human voice signal based on the identification results.

[0012] Specifically, in step S1 of the embodiment provided by the present invention, audio signals are picked up by multiple microphone units in a microphone array to obtain unit audio signals for each microphone unit. Each microphone unit collects surrounding sounds at its location, converting sound energy into electrical energy to obtain its own unit audio signal. These signals contain all sound information in the environment, including human voices and background noise. The pickup range and effect of a single microphone are limited, while a microphone array, composed of multiple microphone units, can simultaneously pick up audio signals from different positions and angles, increasing the coverage and information content of the pickup, which helps to capture the sound in the environment more comprehensively and improves the accuracy of subsequent analysis and processing.

[0013] More specifically, based on the relative position and orientation angle of each microphone unit in the microphone array, the cross-correlation characteristics of the target sound source picked up by each microphone unit are analyzed. Since the positions and orientations of different microphone units are different, the target sound source signals they receive will have time differences and phase differences. By analyzing these relative positions and orientation angles, the cross-correlation characteristics of the target sound source picked up by each microphone unit can be calculated. The cross-correlation characteristics reflect the similarity and correlation between the signals received by different microphone units, which can help determine which signals come from the same target sound source (such as human voice). The sound emitted by the target sound source (such as human voice) will have certain time and phase differences when it propagates to different microphone units. By analyzing the cross-correlation characteristics, these differences can be used to determine the source and propagation path of the sound, thereby better distinguishing human voice from background signals. The background noise received by microphones in different positions may vary greatly, but human voice signals from the same target sound source have a certain correlation. The cross-correlation characteristics can highlight this correlation and help identify human voice signals.

[0014] More specifically, the step of identifying human voice signals in each unit audio signal based on the cross-correlation characteristics to obtain the human voice signal and background signal of each microphone unit can be further subdivided into: performing correlation analysis on the signal content of each unit audio signal to parse each unit audio signal into several segments of audio data and corresponding correlation information; evaluating the correlation information of each segment of audio data based on the cross-correlation characteristics; and dividing each audio data into human voice signal and background signal based on the evaluation results.

[0015] More specifically, each unit audio signal is divided into several audio data segments according to certain rules (such as time windows), and the correlation between each segment is analyzed to obtain the corresponding correlation information. This is done to analyze the characteristics of the audio signal in more detail, making it easier to accurately distinguish between human voice and background signals later. Using the previously obtained cross-correlation features, the correlation information of each audio data segment is evaluated. If a segment of audio data has a high correlation among multiple microphone units and matches the characteristics of the target sound source (human voice), it is classified as a human voice signal; otherwise, it is classified as a background signal. By segmenting and analyzing the correlation of the unit audio signals, the complex audio signal can be decomposed into smaller segments of data that are easier to process. Then, these smaller segments of data are evaluated based on the cross-correlation features, which can more accurately determine whether each segment of data belongs to human voice or background signal, achieving effective separation of the audio signal and providing a clean human voice signal for subsequent sound pickup optimization and speech processing.

[0016] Specifically, in step S2 of the embodiment provided by the present invention, the voice signal is traced and located based on the relative position and orientation angle of each microphone unit in the microphone array to obtain fuzzy location information of the target sound source emitting the voice signal. Using the time difference, phase difference, and other information of the voice signal received by each microphone unit, combined with their relative position and orientation angle in the array, the source of the voice signal is initially located using relevant algorithms (such as time difference of arrival localization algorithm). Since this localization may have some error, the obtained information is fuzzy location information of the target sound source. Initially, it is difficult to directly and accurately determine the location of the target sound source by only using the voice signal picked up by each microphone unit. Using the relative position and orientation angle of the microphone units to trace and locate the voice signal can initially narrow down the possible range of the target sound source, providing a basis for subsequent fine-tuning. The fuzzy location information can provide a general direction for subsequent specific localization prediction, reduce unnecessary search range, and improve localization efficiency.

[0017] More specifically, the orientation angles of each microphone unit are adjusted multiple times based on fuzzy positioning information. Corresponding voice and background signals are collected based on each adjusted orientation angle. Based on the fuzzy positioning information, the more specific location of the target sound source is inferred, generating multiple predicted positioning information sets to expand the search range and improve positioning accuracy. For each predicted positioning information set, combined with the relative position of the microphone unit, the theoretical orientation angle and state of each microphone unit at that predicted position are calculated to achieve optimal sound pickup. According to the theoretical optimal sound pickup state, the orientation angles of each microphone unit are adjusted sequentially, and corresponding voice and background signals are collected after each adjustment for subsequent analysis and comparison. Since fuzzy positioning information is not precise enough, generating multiple predicted positioning information sets can cover more possible locations of the target sound source, increasing the possibility of finding the accurate location. Different predicted positioning information sets correspond to different optimal sound pickup states. By analyzing the theoretical optimal sound pickup state, a basis can be provided for adjusting the orientation angles of the microphone units, ensuring that each adjustment is made in a direction more conducive to sound pickup. By adjusting the orientation angles multiple times and collecting corresponding signals, more data can be collected, providing richer information for subsequent precise positioning analysis and improving positioning accuracy.

[0018] More specifically, based on the adjusted orientation angles and the relative positions of each microphone unit within the microphone array, the optimal pickup angle for each microphone unit in the array is analyzed. Then, using the human voice signals received by each microphone unit, combined with the adjusted orientation angles and relative positions, the target sound source is located and predicted, yielding location prediction information after each adjustment. By integrating multiple location predictions with the initial fuzzy location information, data fusion and optimization algorithms are used to perform a more precise location analysis of the target sound source, obtaining its accurate location information. Based on the precise location information of the target sound source, combined with the relative positions of the microphone units, the optimal pickup angle for each microphone unit is calculated. The microphone unit is positioned at the optimal angle to maximize the reception of human voice signals and minimize background interference. Each adjustment of the orientation angle is followed by source localization to obtain localization prediction information under different conditions. This information reflects the characteristics of the target sound source at different angles, which helps to understand the location of the target sound source more comprehensively. By combining multiple localization prediction information and fuzzy localization information, the error of a single localization can be eliminated, the accuracy of localization can be improved, and the precise location information of the target sound source can be obtained. Based on the precise localization information, the optimal pickup angle is determined, which enables the microphone unit to be aligned with the target sound source to the greatest extent, reducing background noise interference and improving the quality of human voice signal acquisition.

[0019] More specifically, see reference Figure 3The orientation angle of each microphone unit is adjusted to the optimal pickup angle, so that the microphone array is in the best pickup state. According to the calculated optimal pickup angle, the actual orientation angle of each microphone unit is adjusted so that the entire microphone array is in the best pickup state. At this time, the target sound source emitting the human voice signal is in the optimal pickup area of ​​each microphone unit, and therefore in the optimal pickup area of ​​the microphone array. This can improve the quality and effect of the pickup. Adjusting the orientation angle of the microphone unit to the optimal pickup angle can ensure that the entire microphone array is in the best pickup state, making the collected human voice signal clearer and more accurate, and improving the effect of subsequent audio processing.

[0020] Specifically, in step S3 of the embodiment provided by the present invention, the voice signals collected at continuous moments are arranged into an audio data stream, and the voice frequency characteristics of the audio data at each moment are analyzed. Based on the analysis results, weights are assigned to each frequency band range of the voice signals, generating weight parameters for the voice signals in each frequency band range. The discrete voice signals collected at continuous moments are arranged sequentially to form a continuous audio data stream. This integrates the discrete audio data, facilitating subsequent overall frequency characteristic analysis. Frequency analysis is performed on the audio data at each moment in the audio data stream to understand the distribution and characteristics of the voice at different frequencies. For example, the fundamental frequency of the voice is usually within a certain range, and different speech content and pronunciation methods will cause changes in frequency components. Based on the analyzed voice frequency characteristics, corresponding weights are assigned to different frequency band ranges, and the frequency components in... In human speech, the more critical frequency bands that best reflect the characteristics of the voice are assigned higher weights, while relatively less important frequency bands are assigned lower weights. This process generates weight parameters for each frequency band range. By arranging the voice signals from consecutive moments into an audio data stream, it is possible to integrate scattered audio information and perform frequency characteristic analysis from a holistic perspective. This allows for a more comprehensive understanding of the temporal and frequency variations of the voice. Analyzing the frequency characteristics of the voice helps us grasp the energy distribution of the voice at different frequencies. Different speech content and pronunciation methods lead to differences in frequency components; analyzing frequency characteristics can capture these differences, providing a basis for subsequent frequency band weight allocation. Assigning weights to different frequency bands can highlight those that are more important to the expression of the voice. In subsequent processing, key frequency bands can be enhanced in a targeted manner based on the weight parameters, improving the quality and clarity of the speech.

[0021] More specifically, the key frequency band features of the human voice signal are obtained by analyzing the weight parameters of the human voice signal in each frequency band range. Based on these key frequency band features, the parameters of a preset boosting strategy are adjusted to perform gain processing on a specified amplitude range of the human voice signal. By analyzing the weight parameters of each frequency band, it is determined which frequency bands are most important for the expression of the human voice; these are the key frequency bands. Key frequency bands typically contain important information such as the fundamental frequency and formants of the human voice, and have a significant impact on the clarity and intelligibility of speech. Based on the key frequency band features, the parameters of a preset speech enhancement boosting strategy are adjusted. The boosting strategy typically specifies… How to process signals in different frequency bands, such as through gain and filtering, and adjust parameters to make the enhancement strategy more suitable for the characteristics of the currently acquired human voice signal? According to the adjusted enhancement strategy, a specified gain is applied to a specified amplitude range of the human voice signal. Within key frequency bands, the signal amplitude is appropriately increased to highlight the characteristics of the human voice; while for non-key frequency bands or frequency bands with more noise, a smaller gain or attenuation can be applied to enhance the speech and suppress noise. Analyzing the characteristics of key frequency bands can identify which frequency bands contain key information about the human voice, which is crucial for speech recognition and understanding. By focusing on key frequency bands, speech enhancement processing can be performed more effectively, avoiding indiscriminate processing of the entire audio signal and improving processing efficiency. Adjusting the preset enhancement strategy parameters according to the characteristics of key frequency bands makes speech enhancement processing more personalized and precise. Different human voice signals have different frequency characteristics. By adjusting the parameters, the enhancement strategy can better adapt to the current audio signal and achieve the best enhancement effect. Gain processing on a specified signal amplitude range can highlight the signal in the key frequency band and enhance the clarity and intelligibility of the speech. At the same time, appropriate processing of non-key frequency bands can suppress background noise and unnecessary frequency components, further improving the quality of the speech.

[0022] This invention provides an intelligent microphone pickup method, which has the following beneficial effects: This invention can accurately separate human voice from background signals, has a good noise reduction effect, can adjust the microphone orientation to achieve the best sound pickup state, improve sound pickup accuracy, and can also identify key frequency bands to enhance speech, making the sound pickup clearer and more natural, and improving the overall audio quality.

[0023] Preferably, the step of picking up audio signals through multiple microphone units in a microphone array and analyzing them to obtain human voice signals and background signals includes: S11: Audio signals are picked up by multiple microphone units in the microphone array to obtain the unit audio signal of each microphone unit; S12: Based on the relative position and orientation angle of each microphone unit in the microphone array, analyze the cross-correlation characteristics of each microphone unit in picking up the target sound source; S13: Based on the cross-correlation characteristics, identify the human voice signal of each unit audio signal to obtain the human voice signal and background signal of each microphone unit.

[0024] Specifically, audio signals are picked up by multiple microphone units in the microphone array to obtain unit audio signals for each microphone unit. Each microphone unit independently collects sound from the surrounding environment. The transducer inside the microphone (such as electret capacitor) converts the change in sound pressure into a change in electrical signal. Since each microphone unit is located in a different position in space, the intensity and phase of the sound signals they receive may differ. Finally, each microphone unit outputs a unit audio signal that contains all sound information in the environment (including human voices and background signals).

[0025] More specifically, the sound emitted from the target sound source travels to different microphone units at different times and along different paths. By analyzing the cross-correlation characteristics, these time differences and phase differences can be determined, thereby locating the target sound source. For example, based on the peak position of the cross-correlation function, the time delay of the sound reaching different microphone units can be calculated, and the approximate direction of the target sound source can be inferred. The cross-correlation characteristics can reflect whether the signals received by different microphone units come from the same target sound source. If the signals received by two microphone units are highly correlated, it means that they are likely receiving signals from the same target sound source; otherwise, they may be signals from different sources.

[0026] More specifically, based on the relative position and orientation angle of each microphone unit in the microphone array, the cross-correlation characteristics of the target sound source picked up by each microphone unit are analyzed. First, the specific coordinate position of each microphone unit in the array and their orientation angle are determined. This information can be predetermined and recorded during the design and installation of the microphone array. For the unit audio signals of any two microphone units, the cross-correlation function is used to measure the correlation between them. The cross-correlation function is a function of time delay, which describes the similarity of two signals under different time delays. By performing cross-correlation calculations on the signals of different microphone units, a series of cross-correlation values ​​can be obtained.

[0027] More specifically, features are extracted from the calculated cross-correlation function, such as the peak position and peak size of the cross-correlation function. These features reflect the time difference and similarity between the target sound source signals received by different microphone units. The sound emitted by the target sound source travels to different microphone units at different times and along different paths. By analyzing the cross-correlation features, these time differences and phase differences can be determined, thereby locating the target sound source. For example, based on the peak position of the cross-correlation function, the time delay of the sound reaching different microphone units can be calculated, and the approximate direction of the target sound source can be inferred. The cross-correlation features can reflect whether the signals received by different microphone units come from the same target sound source.

[0028] More specifically, based on the cross-correlation characteristics, the human voice signal of each unit audio signal is identified to obtain the human voice signal and background signal of each microphone unit. Correlation analysis of the signal content of each unit audio signal is performed to parse each unit audio signal into several audio data segments and corresponding correlation information. Each unit audio signal is segmented according to a certain time window (e.g., tens of milliseconds) to obtain several audio data segments. This is done to facilitate local analysis of the signal. For each audio data segment, its correlation with the audio data of other microphone units in the corresponding time period is calculated. The strength of the correlation can be measured by indicators such as the correlation coefficient.

[0029] More specifically, the correlation information of each audio data segment is evaluated based on the cross-correlation characteristics. Based on the evaluation results, each audio data segment is divided into human voice signals and background signals. Based on the previously obtained cross-correlation characteristics, an evaluation criterion is set to determine whether a certain audio data segment belongs to human voice signals or background signals. For example, if a certain audio data segment has a high correlation among multiple microphone units and the peak value of the cross-correlation function matches the characteristics of the target sound source, then the audio data segment is considered to contain human voice signals; otherwise, it is considered to be a background signal. According to the evaluation criterion, each audio data segment is classified into human voice signals or background signals. Finally, the unit audio signal of each microphone unit is analyzed into two parts: human voice signals and background signals.

[0030] More specifically, in practical applications, there are often various background noises in the environment. Separating the human voice signal from the background signal can effectively improve the clarity and intelligibility of speech, providing a cleaner speech signal for subsequent speech processing (such as speech recognition, speech enhancement, etc.). Since the sound emitted by the target sound source has a strong correlation between different microphone units, while background noise is usually random and uncorrelated, the human voice signal and background signal can be distinguished more accurately by evaluating the correlation of audio data.

[0031] Preferably, the step of identifying human voice signals in the audio signals of each unit based on the cross-correlation characteristics to obtain the human voice signal and background signal of each microphone unit includes: S131: Perform correlation analysis on the signal content of each unit audio signal to parse each unit audio signal into several segments of audio data and corresponding correlation information; S132: Evaluate the correlation information of each segment of audio data according to the cross-correlation characteristics, so as to divide each segment of audio data into human voice signal and background signal according to the evaluation results.

[0032] Specifically, a suitable time window length, such as 20-50 milliseconds, is selected to segment the unit audio signal acquired by each microphone unit. A sliding window approach can be used, sliding across the audio signal in steps of a certain size (e.g., 10 milliseconds) to cut the signal into a series of continuous and partially overlapping audio segments. For each audio segment, its characteristic parameters are extracted, such as time-domain features (e.g., energy, zero-crossing rate) and frequency-domain features (e.g., spectrum, cepstral coefficients). These features reflect different characteristics of the audio signal and are helpful for subsequent correlation analysis.

[0033] More specifically, the correlation between audio data from different microphone units within the same time segment is calculated. Commonly used correlation calculation methods include cross-correlation function and Pearson correlation coefficient. The calculated correlation information is then associated with the corresponding audio segments to form the correlation information for each audio data segment. For example, information such as the peak value and peak position of the cross-correlation function is recorded. This information will be used for subsequent evaluation.

[0034] More specifically, audio signals are continuously changing over time. Segmenting them allows us to break down complex overall signals into relatively simple local signals, facilitating detailed feature analysis and correlation calculations for each local segment. Different audio segments may have different characteristics, such as different syllables in speech or different states of background noise. Segmentation can better capture these changes. Extracting features from audio data can transform the original representation of the audio signal in the time or frequency domain into a more representative feature vector, which facilitates correlation calculations. Correlation calculations can measure the similarity between audio signals received by different microphone units. Since the sound emitted by the target sound source has a certain correlation between different microphone units, while background noise is usually random and uncorrelated, correlation information can serve as an important basis for distinguishing human voice signals from background signals. Associating correlation information with corresponding audio segments allows for convenient retrieval and analysis of this information in subsequent evaluation processes, ensuring the accuracy and effectiveness of the evaluation.

[0035] More specifically, the correlation information of each audio data segment is evaluated based on the cross-correlation characteristics. Based on the evaluation results, each audio data segment is divided into human voice signals and background signals. Based on the cross-correlation characteristics obtained from the previous analysis, combined with actual application scenarios and experience, a standard for evaluating whether audio data is a human voice signal is set. For example, a threshold for the peak value of the cross-correlation function and a range of peak positions can be set. Generally speaking, if the peak value of the cross-correlation function between audio segments of different microphone units is high and the peak position matches the propagation characteristics of the target sound source, then the audio segment is considered to be more likely to contain a human voice signal.

[0036] More specifically, for each audio data segment and its correlation information, a judgment is made according to the set evaluation criteria. If the correlation information of a certain audio data segment meets the evaluation criteria for human voice signals, it is marked as a human voice signal; otherwise, it is marked as a background signal. Based on the marking results, all audio segments of each microphone unit are integrated into human voice signals and background signals respectively. That is, the audio segments marked as human voice signals are spliced ​​together to form the human voice signal of the microphone unit; the audio segments marked as background signals are spliced ​​together to form the background signal.

[0037] More specifically, setting evaluation criteria provides clear rules for determining whether audio data constitutes a human voice signal, making the classification process objective and repeatable. Through analysis of cross-correlation characteristics and empirical summarization, reasonable evaluation criteria can be determined, improving classification accuracy. Based on the evaluation results, audio data can be divided into human voice signals and background signals, effectively extracting the desired human voice signal from complex environmental sounds and removing background noise interference. This is of great significance for improving the performance of speech recognition, voice communication, and other systems; a pure human voice signal can improve the system's recognition accuracy and communication quality.

[0038] Preferably, the step of adjusting the orientation angle of each microphone unit based on the human voice signal and background signal picked up by each microphone unit to adjust the microphone array to the optimal sound pickup state includes: S21: Based on the relative position and orientation angle of each microphone unit in the microphone array, perform source localization processing on the human voice signal to obtain fuzzy localization information of the target sound source emitting the human voice signal; S22: Based on the fuzzy positioning information, the orientation angle of each microphone unit is adjusted multiple times, and the corresponding human voice signal and background signal are collected based on the orientation angle of each adjustment. S23: Based on the orientation angle of each adjustment and the relative position of each microphone unit in the microphone array, the optimal pickup angle of each microphone unit in the microphone array is obtained by analysis; S24: Adjust the orientation angle of each microphone unit to the optimal pickup angle so that the microphone array is adjusted to the optimal pickup state.

[0039] Specifically, the three-dimensional coordinate position and orientation angle of each microphone unit in the microphone array are determined. This information is usually determined and recorded during the design and installation of the microphone array. Features such as arrival time, phase, and intensity are extracted from the human voice signals picked up by each microphone unit. The human voice signals received by microphones at different positions and orientations will have differences in these features. By comparing the time difference of the same human voice signal received by different microphone units and combining it with the speed of sound in the air, the distance relationship between the target sound source and each microphone is initially estimated. Combining the above information, geometric relationships and algorithms (such as positioning algorithms based on time difference of arrival) are used to obtain the approximate location range of the target sound source, i.e., fuzzy positioning information. This range is an area, not a precise point.

[0040] More specifically, initially, it is difficult to directly determine the precise location of the target sound source by relying solely on the human voice signal received from the microphone. By using the relative position and orientation angle of the microphone unit and the characteristics of the human voice signal for source tracing and localization, the possible range of the target sound source can be initially narrowed down, providing a basis for subsequent fine-tuning. The fuzzy localization information can provide a general direction for subsequent specific localization prediction, reducing unnecessary search range and improving localization efficiency.

[0041] More specifically, based on fuzzy positioning information, the more specific location of the target sound source is inferred, generating several predicted positioning information. For example, multiple predicted points can be generated at certain angles and distance intervals, centered on the fuzzy positioning area. For each predicted positioning information, combined with the relative position of the microphone unit, the theoretical orientation angle and state of each microphone unit at the predicted position are analyzed using acoustic models and algorithms to achieve the best sound pickup effect. The best sound pickup effect usually means that the human voice signal can be received to the maximum extent while reducing the interference of background noise. According to the theoretical best sound pickup state of each predicted positioning information, the orientation angle of each microphone unit is adjusted sequentially. After each adjustment, the microphone array is allowed to collect audio signals for a period of time, and the human voice signal and background signal of each microphone unit after the adjustment are obtained respectively.

[0042] More specifically, fuzzy positioning information is not precise enough. By generating multiple predicted positioning information, more possible locations of the target sound source can be covered, increasing the likelihood of finding the accurate location. Different predicted positioning information corresponds to different optimal sound pickup states. By analyzing the theoretical optimal sound pickup state and adjusting the orientation angle of the microphone units multiple times, the microphone array can gradually approach the optimal sound pickup state, improving the sound pickup quality. Multiple adjustments and signal acquisitions can collect more data, providing richer information for subsequent precise positioning analysis, which helps to more accurately determine the location of the target sound source and the optimal sound pickup angle of each microphone unit.

[0043] More specifically, based on the adjusted orientation angle and the relative position of the microphone units, the acquired voice signals are re-processed for source localization to obtain prediction information for the target sound source. This step is similar to the source localization in the first step, but this time the adjusted orientation angle and the newly acquired signal are used, resulting in a more accurate localization result. Combining the prediction information and the fuzzy localization information obtained in the first step, data fusion and optimization algorithms (such as Kalman filtering and particle filtering) are used to perform precise localization analysis on the target sound source, obtaining its precise localization information, i.e., its accurate three-dimensional coordinate position. Based on the precise localization information of the target sound source, combined with the relative position of the microphone units, the optimal pickup angle of each microphone unit is analyzed through geometric calculations and acoustic principles. The optimal pickup angle should make the main axis of the microphone as aligned as possible with the target sound source to maximize the reception of the voice signal.

[0044] More specifically, by repeatedly adjusting and acquiring signals, and tracing the source of each acquired signal, location prediction information under different conditions can be obtained. By combining this information and performing precise location analysis, the error of a single location can be eliminated, and the accuracy of target sound source location can be improved. The accurate target sound source position is the basis for determining the optimal pickup angle of each microphone unit. The optimal pickup angle calculated based on the precise location information can ensure that the microphone unit is aligned with the target sound source to the greatest extent, reduce background noise interference, and improve the acquisition quality of human voice signals.

[0045] More specifically, based on the calculated optimal pickup angle of each microphone unit, the actual orientation angle of each microphone unit is adjusted through mechanical devices or electronic control to achieve the optimal pickup angle. After adjustment, the audio signal is collected again, and the quality of the collected human voice signal is evaluated. If the pickup effect is still not ideal, the optimal pickup angle can be fine-tuned according to the evaluation results until a satisfactory pickup effect is achieved.

[0046] More specifically, adjusting the orientation angle of the microphone unit to the optimal pickup angle ensures that the entire microphone array is in the best pickup state, making the collected human voice signal clearer and more accurate, effectively suppressing background noise, and improving the effect of subsequent audio processing. By collecting the adjusted audio signal and performing quality evaluation, it can be verified whether the adjustment has achieved the expected effect. If there are any deficiencies, fine-tuning can be made in time to further optimize the pickup effect and ensure that the microphone array is always in the best working state.

[0047] Preferably, the step of adjusting the orientation angle of each microphone unit multiple times based on the fuzzy positioning information includes: S221: Based on the fuzzy positioning information, perform specific positioning prediction of the target sound source to generate several predicted positioning information; S222: Based on the relative positions of each microphone unit in the microphone array, the theoretical optimal sound pickup state corresponding to each of the predicted positioning information is obtained through analysis; S223: Based on the theoretical optimal sound pickup state corresponding to each of the predicted positioning information, the orientation angle of each microphone unit is adjusted multiple times in sequence.

[0048] Specifically, based on the approximate area determined by the fuzzy positioning information, a prediction range containing the possible location of the target sound source is delineated according to the actual scene and accuracy requirements. For example, if the fuzzy positioning information shows that the target sound source is within a circular area, the radius of this circular area can be appropriately expanded to ensure that the target sound source is within the prediction range. Prediction points are generated within the prediction range according to certain rules. These prediction points are the prediction positioning information. Common rules include equal angular intervals and equal distance intervals. For example, prediction points are generated at 10° angular intervals on the circumference of a circle with the center of the fuzzy positioning area as the center; or prediction points are generated according to a certain grid spacing within a rectangular prediction area. According to the set prediction rules, the three-dimensional coordinate position of each prediction point is calculated to form several prediction positioning information. These prediction positioning information represent the specific location where the target sound source may exist.

[0049] More specifically, fuzzy location information is only a general area, and the exact location of the target sound source is still unclear. By generating predictive location information, the search range can be further refined, and the possible location of the target sound source can be specified to a series of predictive points, thereby improving the accuracy of subsequent location. Generating multiple predictive location information according to certain rules can ensure that the possible location of the target sound source is covered as comprehensively as possible within the prediction range. This can avoid missing the actual location of the target sound source and increase the possibility of finding the accurate location.

[0050] More specifically, based on the relative positions of each microphone unit in the microphone array, the theoretical optimal pickup state for each predicted location information is analyzed. Based on acoustic principles, an acoustic model describing sound propagation in space and microphone pickup characteristics is established. This model considers factors such as sound propagation attenuation, reflection, and diffraction, as well as the microphone's directivity and sensitivity. For each predicted location information, it is taken as the assumed location of the target sound source. Combining the relative positions of the microphone units and the acoustic model, the theoretical pickup parameters of each microphone unit at that location are calculated, such as the received sound intensity, phase, and frequency response. Based on the calculated theoretical pickup parameters, with the goal of maximizing the reception of the target sound source signal and minimizing background noise interference, the theoretical optimal pickup state of each microphone unit under the predicted location information is determined, including the optimal orientation angle and gain settings.

[0051] More specifically, different target sound source locations correspond to different optimal sound pickup states. By establishing acoustic models and calculating theoretical sound pickup parameters, the optimal sound pickup method for each microphone unit under each predicted positioning information can be determined. This allows for maximizing the reception of target sound source signals and improving sound pickup quality during subsequent adjustments. The theoretical optimal sound pickup state provides a clear basis for adjusting the orientation angle of the microphone unit. Adjusting according to these theoretical states can make the adjustment process more scientific and reasonable, avoiding inefficiency and poor results caused by blind adjustments.

[0052] More specifically, the generated predicted positioning information can be sorted in a certain logical order, such as starting from the predicted point closest to the fuzzy positioning center, and then processing the theoretical optimal sound pickup state corresponding to each predicted positioning information in turn. Based on the theoretical optimal sound pickup state corresponding to each predicted positioning information, the orientation angle of each microphone unit is adjusted by mechanical devices or electronic control methods to achieve the orientation angle under the theoretical optimal sound pickup state. After each adjustment, the actual adjustment angle, adjustment time, and corresponding predicted positioning information of each microphone unit are recorded for subsequent analysis and comparison.

[0053] More specifically, by sequentially adjusting the theoretically optimal sound pickup state corresponding to each predicted positioning information and collecting the corresponding audio signals, the sound pickup effect at different locations can be compared and evaluated. Ultimately, the actual location of the target sound source and the corresponding optimal sound pickup angle can be found, enabling the microphone array to reach its optimal sound pickup state. The information recorded after each adjustment can serve as important data for subsequent analysis. By analyzing this data, the relationship between different predicted positioning information and actual sound pickup effect can be summarized, further optimizing the positioning and adjustment strategy and improving the performance of the entire system.

[0054] Preferably, the step of analyzing and obtaining the optimal pickup angle of each microphone unit in the microphone array based on the orientation angle of each adjustment and the relative position of each microphone unit in the microphone array includes: S231: Based on the orientation angle of each adjustment and the relative position of each microphone unit in the microphone array, the source localization processing of each collected human voice signal is performed to obtain the localization prediction information of the target sound source for each time. S232: Combining the positioning prediction information and the fuzzy positioning information, perform precise positioning analysis on the target sound source to obtain precise positioning information of the target sound source; S233: Based on the precise positioning information, the optimal pickup angle of each microphone unit is analyzed to obtain the optimal pickup angle of each microphone unit.

[0055] Specifically, features are extracted from the human voice signals collected by each microphone unit after each adjustment of the orientation angle. Common features include the arrival time, phase, and amplitude of the signal. These features can reflect the process information of sound propagation from the target sound source to each microphone unit. Using the time difference of the same human voice signal received by different microphone units, combined with the speed of sound in the air, the propagation time difference of sound from the target sound source to each microphone unit is calculated. This can be achieved by comparing the arrival time of the signals from each microphone unit. Based on the relative position of each microphone unit, the adjusted orientation angle, and the calculated propagation time difference, appropriate positioning algorithms (such as the time difference of arrival (TDOA) positioning algorithm, the phase difference-based positioning algorithm, etc.) are used to locate the target sound source. Through these algorithms, the positioning prediction information of the target sound source after each adjustment can be obtained. This information is usually represented as the approximate position coordinates of the target sound source in space.

[0056] More specifically, the acquired voice signal changes each time the orientation angle of the microphone unit is adjusted. By performing source localization processing on each acquired signal, the localization prediction information of the target sound source under different adjustment states can be obtained. This information can gradually refine the possible location range of the target sound source, providing more data support for subsequent accurate localization. Different orientation angles will affect the characteristics of the sound signal received by the microphone unit. Performing localization processing on the signals after each adjustment can fully take into account the impact of this angle change on the localization result, making the localization more accurately reflect the location information of the target sound source under different conditions.

[0057] More specifically, the positioning prediction information obtained from each adjustment and the initial fuzzy positioning information are fused. Data fusion methods such as weighted averaging, Kalman filtering, and particle filtering can be used. Appropriate weights are assigned according to the reliability and accuracy of different information. These information are then combined, and the fused data is subjected to error correction and optimization. Since various errors may exist in the positioning process (such as measurement errors, environmental interference, etc.), some optimization algorithms are needed to reduce the impact of these errors. For example, the least squares method can be used to fit and optimize the positioning results to make the positioning results more accurate. After data fusion and error correction optimization, the precise positioning information of the target sound source is obtained, that is, the precise position coordinates of the target sound source in space.

[0058] More specifically, single location prediction information may contain certain errors and uncertainties. By fusing the location prediction information from each iteration with the initial fuzzy location information, we can comprehensively utilize the advantages of multiple types of information, complement and verify each other, thereby improving the accuracy and reliability of target sound source localization. The data fusion and error correction optimization process can effectively reduce the impact of various errors on the localization results. By weighting and optimizing different information, the localization results can be made closer to the true location of the target sound source.

[0059] More specifically, based on the precise location information of the target sound source and the relative positions of each microphone unit in the microphone array, the geometric relationship between the target sound source and each microphone unit is determined, including distance and angle. Considering the directional characteristics of each microphone unit, different types of microphones have different directional patterns. Based on the directional characteristics of the microphones and the position of the target sound source, the optimal orientation angle of each microphone unit to receive the target sound source signal to the greatest extent is determined. Through geometric calculations and acoustic principles, combined with the geometric relationship and microphone directivity determined above, the optimal pickup angle of each microphone unit is calculated. This angle enables the main axis direction of the microphone unit to be aligned with the target sound source as much as possible, thereby maximizing the reception of human voice signals while reducing background noise interference.

[0060] More specifically, accurate target sound source location is the basis for determining the optimal pickup angle. Based on precise positioning information and the directional characteristics of the microphone, the optimal orientation angle of each microphone unit when receiving the target sound source signal to the maximum extent can be calculated, thereby optimizing the pickup effect, improving the quality of the human voice signal, and aligning the main axis of the microphone unit with the target sound source to receive the human voice signal to the maximum extent, while reducing background noise interference from other directions. This can improve the clarity and intelligibility of speech, providing a better foundation for subsequent speech processing and applications.

[0061] Preferably, the analysis of the optimal pickup angle further includes: S201: Perform quality assessment on the human voice signals collected in each instance, and based on the results of the quality assessment and the orientation angle when collecting human voice signals in each instance, predict the sound pickup quality of the optimal sound pickup angle obtained from the analysis, and generate the expected sound pickup quality of the optimal sound pickup angle. S202: After the microphone array is adjusted to the optimal sound pickup state, the sound pickup quality of the acquired human voice signal is evaluated according to the expected sound pickup quality to determine whether the optimal sound pickup angle needs to be adjusted.

[0062] Specifically, the metrics used to evaluate the quality of human voice signals are determined. Common metrics include signal-to-noise ratio (SNR), intelligibility index (AI), and speech intelligibility index (STI). SNR reflects the ratio of useful signal to noise in a human voice signal, while AI and STI focus on evaluating the clarity and intelligibility of speech. For each acquired human voice signal, the selected quality evaluation metrics are calculated. For example, when calculating SNR, the useful signal and noise components in the human voice signal need to be separated, and then their power ratio is calculated. Based on the orientation angle and the corresponding quality evaluation metric values ​​for each acquired human voice signal, a predictive model is established. Machine learning algorithms (such as linear regression, decision trees, neural networks, etc.) can be used to construct this model, which describes the relationship between orientation angle and sound pickup quality. The optimal sound pickup angle obtained from the analysis is input into the established predictive model, and the model is used to predict the sound pickup quality at the optimal sound pickup angle, generating the expected sound pickup quality for the optimal sound pickup angle, which is presented in the form of quality evaluation metric values.

[0063] More specifically, before adjusting the microphone array to the optimal pickup angle, by evaluating the quality of each collected voice signal and establishing a predictive model, the pickup quality at the optimal pickup angle can be predicted in advance. This helps to understand the possible effect before actual adjustment, providing a basis for subsequent judgment and decision-making. The predicted expected pickup quality can help optimize the adjustment strategy of the optimal pickup angle. If the prediction results show that the expected pickup quality of a certain optimal pickup angle is not ideal, the analysis process of the optimal pickup angle can be reflected on and adjusted in time, avoiding the discovery of poor results only after actual adjustment.

[0064] More specifically, the orientation angle of each microphone unit is adjusted to the optimal pickup angle, so that the microphone array is in the best pickup state. Then, human voice signals are collected for a period of time. According to the previously selected quality evaluation index, the quality of the collected human voice signals in the optimal pickup state is evaluated, and the actual pickup quality index value is calculated. The actual pickup quality index value is compared with the previously predicted expected pickup quality index value. An allowable error range can be set. If the deviation between the actual value and the expected value is within the error range, the current optimal pickup angle is considered appropriate. If the deviation exceeds the error range, the optimal pickup angle needs to be adjusted. Based on the comparison results, a decision is made on whether to adjust the optimal pickup angle. If adjustment is needed, the optimal pickup angle can be fine-tuned according to the actual situation. Then, the above steps are repeated until the actual pickup quality achieves a satisfactory effect.

[0065] More specifically, various factors in the actual environment (such as changes in ambient noise, slight movements of the target sound source, etc.) may affect the sound pickup effect. By comparing the actual sound pickup quality with the expected sound pickup quality, it is possible to promptly determine whether the actual sound pickup effect meets expectations, ensuring that the final acquired human voice signal has high quality. Based on the comparison results, it is possible to decide whether to adjust the optimal sound pickup angle so that the microphone array can adapt to changes in the actual environment. If the actual sound pickup quality is not ideal, timely adjustments can keep the microphone array in optimal working condition, improving the stability and reliability of sound pickup.

[0066] Preferably, the steps of identifying key frequency bands in the human voice signal acquired under optimal sound pickup conditions and performing speech enhancement processing on the human voice signal based on the identification results include: S31: Arrange the human voice signals collected at continuous moments into an audio data stream, and analyze the human voice frequency characteristics of the audio data at each moment in the audio data stream. Based on the analysis results, assign weights to the human voice signals in each frequency band range to generate weight parameters of the human voice signals in each frequency band range. S32: Based on the weight parameters of the human voice signal in each frequency band range, the key frequency band characteristics of the human voice signal are obtained. Based on the key frequency band characteristics, the parameters of the preset boosting strategy are adjusted to perform a specified gain processing on the specified signal amplitude range of the human voice signal.

[0067] Specifically, discrete human voice signals collected at continuous moments are arranged sequentially in time to form a continuous audio data stream. This process is similar to connecting a series of discrete points into a continuous curve so that the entire human voice signal can be uniformly analyzed and processed later. The audio data stream is divided into several shorter frames, each of which usually contains tens to hundreds of sampling points. The purpose of framing is to perform stationarity analysis on the audio signal within a shorter time period, because the frequency characteristics of the audio signal are relatively stable within a shorter time period.

[0068] More specifically, a window function (such as a Hamming window or Heining window) is applied to each frame of audio data to reduce spectral leakage. The window function gradually attenuates the signal at both ends of the frame, making the signal more concentrated in the frequency domain and facilitating accurate analysis of its frequency characteristics. Spectral analysis is then performed on each frame of windowed audio data. A common method is to use a Fast Fourier Transform (FFT) to convert the time-domain signal into a frequency-domain signal. Through spectral analysis, the energy distribution of each frame of audio data at different frequencies can be obtained, i.e., the frequency characteristics of the human voice. The entire audio frequency band can be divided into several frequency ranges; for example, it can be divided into multiple sub-bands according to a logarithmic scale. This division method allows for a more detailed analysis of the human voice characteristics in different frequency regions.

[0069] More specifically, based on the importance of each frequency band in human voice expression and combined with the frequency characteristic analysis results, corresponding weights are assigned to each frequency band range. For example, frequency bands containing the fundamental frequency and formants of the human voice usually have a greater impact on the clarity and intelligibility of speech and are assigned higher weights; while some frequency bands with more noise or less contribution to speech expression are assigned lower weights. This ultimately generates weight parameters for the human voice signal in each frequency band range.

[0070] More specifically, arranging continuous human voice signals into an audio data stream allows for the integration of scattered audio information, enabling frequency characteristic analysis from a holistic perspective. This provides a more comprehensive understanding of the temporal and frequency variations of human voice. While audio signals are non-stationary in the time domain, they can be approximated as stationary signals over shorter time periods. Framing and windowing can divide the audio signal into relatively stable segments, facilitating accurate analysis of its frequency characteristics and avoiding analysis errors caused by signal non-stationarity. Through frequency band division and weight allocation, frequency bands that are more important to human voice expression can be highlighted. In subsequent processing, key frequency bands can be targeted for enhancement based on weight parameters, improving speech quality and clarity.

[0071] More specifically, the weighting parameters of each frequency band are analyzed to identify the frequency bands with higher weights as key frequency bands. Key frequency bands typically contain important characteristic information of human voice, such as fundamental frequency and formants, and play a crucial role in speech recognition and understanding. Based on the identified key frequency band characteristics, the parameters of the pre-set speech enhancement strategy are adjusted. The enhancement strategy usually specifies how to process signals from different frequency bands, such as gain and filtering. For example, if the key frequency bands are concentrated in a specific frequency range, the gain parameters within that frequency range can be increased to highlight the signals of these frequency bands. According to the adjusted enhancement strategy, the specified amplitude range of the human voice signal is subjected to gain processing of a specified amplitude. Frequency domain filtering can be used to enhance the signal of the key frequency band in the frequency domain. Then, the processed frequency domain signal is converted back to the time domain signal through inverse fast Fourier transform (IFFT). This can enhance the signal amplitude of the key frequency band without changing the signal phase, while suppressing noise in non-key frequency bands, thereby achieving the effect of enhancing speech.

[0072] More specifically, identifying key frequency bands clarifies which bands contain crucial information about human voice, essential for speech recognition and comprehension. By focusing on key frequency bands, speech enhancement processing can be performed more effectively, avoiding indiscriminate processing of the entire audio signal and improving processing efficiency. Adjusting preset enhancement strategy parameters based on the characteristics of key frequency bands makes speech enhancement processing more personalized and precise. Different human voice signals have different frequency characteristics; adjusting parameters allows the enhancement strategy to better adapt to the characteristics of the currently acquired human voice signal, achieving the best enhancement effect. Applying gain processing to a specified signal amplitude range can highlight the signal in key frequency bands and enhance speech features. At the same time, applying small-amplitude gain or attenuation processing to non-key frequency bands or frequency bands with more noise can suppress background noise, further improving speech quality and intelligibility.

[0073] Reference Figure 2 As shown, in a second aspect, the present invention provides an intelligent microphone pickup system for implementing the intelligent microphone pickup method described in any one of the first aspects, comprising: The signal acquisition module is used to pick up audio signals through multiple microphone units in the microphone array and analyze them to obtain human voice signals and background signals. An angle adjustment module is used to adjust the orientation angle of each microphone unit based on the human voice signal and background signal picked up by each microphone unit, so as to adjust the microphone array to the optimal sound pickup state; The voice enhancement module is used to identify key frequency bands of the human voice signal collected under optimal sound pickup conditions, and to perform voice enhancement processing on the human voice signal based on the recognition results.

[0074] In this embodiment, the specific implementation of each module in the above system embodiment is described in the above method embodiment, and will not be repeated here.

[0075] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for intelligent sound pickup using a microphone, characterized in that, include: Audio signals are picked up by multiple microphone units in the microphone array, and the human voice signal and background signal are obtained by analysis; Based on the human voice signal and background signal picked up by each microphone unit, the orientation angle of each microphone unit is adjusted to adjust the microphone array to the optimal sound pickup state; The key frequency bands of the human voice signal acquired under optimal sound pickup conditions are identified, and the human voice signal is subjected to speech enhancement processing based on the identification results.

2. The intelligent sound pickup method for a microphone as described in claim 1, characterized in that, The steps of picking up audio signals from multiple microphone units in a microphone array and analyzing them to obtain the human voice signal and background signal include: Audio signals are picked up by multiple microphone units in the microphone array to obtain the unit audio signal of each microphone unit; Based on the relative position and orientation angle of each microphone unit in the microphone array, the cross-correlation characteristics of each microphone unit in picking up the target sound source are analyzed. Based on the cross-correlation characteristics, the human voice signal of each unit audio signal is identified to obtain the human voice signal and background signal of each microphone unit.

3. The intelligent microphone pickup method as described in claim 2, characterized in that, The step of identifying human voice signals in each of the unit audio signals based on the cross-correlation characteristics to obtain the human voice signal and background signal of each microphone unit includes: Correlation analysis of the signal content of each unit audio signal is performed to parse each unit audio signal into several segments of audio data and corresponding correlation information; The correlation information of each audio data segment is evaluated based on the cross-correlation characteristics, so as to divide each audio data segment into human voice signal and background signal according to the evaluation results.

4. The intelligent sound pickup method for a microphone as described in claim 1, characterized in that, The step of adjusting the orientation angle of each microphone unit to optimize the microphone array for optimal sound pickup, based on the human voice signal and background signal picked up by each microphone unit, includes: Based on the relative position and orientation angle of each microphone unit in the microphone array, the human voice signal is subjected to source localization processing to obtain fuzzy localization information of the target sound source emitting the human voice signal. Based on the fuzzy positioning information, the orientation angle of each microphone unit is adjusted multiple times, and the corresponding human voice signal and background signal are collected based on the orientation angle of each adjustment. Based on the orientation angle of each adjustment and the relative position of each microphone unit in the microphone array, the optimal pickup angle of each microphone unit in the microphone array is analyzed and obtained. The orientation angle of each microphone unit is adjusted to the optimal pickup angle so that the microphone array is adjusted to the optimal pickup state.

5. The intelligent microphone pickup method as described in claim 4, characterized in that, The step of adjusting the orientation angle of each microphone unit multiple times based on the fuzzy positioning information includes: Based on the fuzzy positioning information, the target sound source is specifically located and predicted to generate several predicted positioning information. Based on the relative positions of each microphone unit in the microphone array, the theoretical optimal sound pickup state corresponding to each of the predicted positioning information is analyzed and obtained. Based on the theoretically optimal sound pickup state corresponding to each of the predicted positioning information, the orientation angle of each microphone unit is adjusted multiple times in sequence.

6. The intelligent sound pickup method for a microphone as described in claim 4, characterized in that, The steps for analyzing and determining the optimal pickup angle for each microphone unit in the microphone array based on the adjusted orientation angle and the relative position of each microphone unit in the microphone array include: Based on the orientation angle of each adjustment and the relative position of each microphone unit in the microphone array, the human voice signals collected each time are subjected to source localization processing to obtain the localization prediction information of the target sound source for each time. By combining the positioning prediction information and the fuzzy positioning information, the target sound source is precisely located and analyzed to obtain the precise positioning information of the target sound source. Based on the precise positioning information, the optimal pickup angle of each microphone unit is analyzed to obtain the optimal pickup angle of each microphone unit.

7. The intelligent sound pickup method for a microphone as described in claim 6, characterized in that, The analysis of the optimal pickup angle also includes: The quality of each collected human voice signal is evaluated, and based on the results of the quality evaluation and the orientation angle when the human voice signal is collected, the sound pickup quality of the optimal sound pickup angle is predicted, and the expected sound pickup quality of the optimal sound pickup angle is generated. After the microphone array is adjusted to the optimal sound pickup state, the sound pickup quality of the acquired human voice signal is evaluated according to the expected sound pickup quality to determine whether the optimal sound pickup angle needs to be adjusted.

8. The intelligent sound pickup method for a microphone as described in claim 1, characterized in that, The steps of identifying key frequency bands in the human voice signal acquired under optimal sound pickup conditions, and performing speech enhancement processing on the human voice signal based on the identification results, include: The human voice signals collected at continuous moments are arranged into an audio data stream, and the human voice frequency characteristics of the audio data at each moment are analyzed. Based on the analysis results, the weights of the human voice signals in each frequency band range are assigned, and the weight parameters of the human voice signals in each frequency band range are generated. Based on the weight parameters of the human voice signal in each frequency band range, the key frequency band characteristics of the human voice signal are obtained. Based on the key frequency band characteristics, the parameters of the preset boosting strategy are adjusted to perform a specified gain processing on the specified signal amplitude range of the human voice signal.

9. An intelligent microphone pickup system, characterized in that, A method for implementing the intelligent sound pickup method for a microphone according to any one of claims 1-8 includes: The signal acquisition module is used to pick up audio signals through multiple microphone units in the microphone array and analyze them to obtain human voice signals and background signals. An angle adjustment module is used to adjust the orientation angle of each microphone unit based on the human voice signal and background signal picked up by each microphone unit, so as to adjust the microphone array to the optimal sound pickup state; The voice enhancement module is used to identify key frequency bands of the human voice signal collected under optimal sound pickup conditions, and to perform voice enhancement processing on the human voice signal based on the recognition results.

Images