Audio noise reduction method and system for Bluetooth headset

Abstract

The invention relates to the technical field of voice enhancement, in particular to an audio noise reduction method and system for a Bluetooth headset, and the method comprises the steps: analyzing the feature condition of noise influence in a mobile scene, including the feature change condition of a signal obtained by a microphone under the condition of pedestrian voice noise interference; according to the method, the spectral subtraction factor of the spectral subtraction method is adjusted in a targeted manner by integrating the characteristic difference conditions of the audios obtained by the reference microphone and the main microphone when the audios are moved to different scenes, so that the filtering error occurring during the audio noise reduction of the Bluetooth headset in the moving scene is avoided, and the audio noise reduction level in the call process of the Bluetooth headset is further improved.

Description

Technical Field

[0001] This application relates to the field of voice enhancement technology, specifically to an audio noise reduction method and system for Bluetooth headsets. Background Technology

[0002] With the rapid development of social communication technologies and the widespread adoption of modern fast-paced lifestyles, mobile communication has become an indispensable part of daily life. Especially in outdoor environments, the demand for Bluetooth headsets is gradually increasing due to privacy concerns during calls. Against the backdrop of continuous advancements in wireless audio technology, call noise reduction in Bluetooth headsets has become one of the core technologies for improving voice communication quality. The importance of this function is reflected in two aspects: First, users generally demand a clear call experience, especially in high-noise environments such as traffic, offices, and outdoors, where environmental noise can significantly interfere with voice signal transmission, making it difficult for both parties to clearly hear each other. Second, noise itself has a particularly severe impact on speech intelligibility; large and continuous noise fluctuations can completely drown out human voices. To address this challenge, modern Bluetooth headsets employ technologies such as multi-microphone arrays, CVC call noise reduction, and ENC environmental noise reduction. Through the collaborative work of hardware and software, they accurately capture human voices and effectively suppress noise, thereby ensuring a clear, natural, and stable call experience even in complex environments, meeting the high standards of call quality required by modern mobile communication.

[0003] When making mobile calls using Bluetooth headsets, changes in ambient noise, such as background noise like pedestrian conversations, can affect the audio quality of the call. Traditional Bluetooth headsets often do not adequately consider noise differences in complex environments when performing audio signal noise reduction, especially the changes in noise interference intensity when the scene changes. This lack of consideration for differences in noise interference intensity prevents Bluetooth headsets from achieving ideal noise reduction effects during audio processing. Summary of the Invention

[0004] In view of the above, it is necessary to provide an audio noise reduction method and system for Bluetooth headsets to solve the above problems.

[0005] The first aspect of this application provides an audio noise reduction method for Bluetooth headphones, the method comprising: The Bluetooth headset reference microphone is used to collect external noise signals, while the main microphone collects mixed audio signals. The signal within a preset window is divided into a preset number of Mel sub-bands. Based on the peak fluctuation characteristics of each sub-band in each window and the difference change characteristics with adjacent valley points, the characteristic fluctuation index of each sub-band in each window is determined. The distribution density of peak points and the change characteristics of the amplitude corresponding to the peak points in each sub-band in each window are analyzed. Combined with the characteristic fluctuation index, the noise impact index of each sub-band in each window is determined. The noise impact index of the external noise signal and the mixed audio signal in each window and each sub-band is compared to obtain the dynamic adjustment index of the external noise signal in each window and each sub-band. Based on the dynamic adjustment index, the optimal over-subtraction factor of the external noise signal of the reference microphone in each window and each sub-band is obtained. The mixed audio signal is filtered and synthesized by spectral subtraction to obtain the noise-reduced audio.

[0006] Specifically, the process of determining the characteristic fluctuation index of each sub-band in each window is as follows: For each window and each sub-band, obtain the first-order forward difference value and the first-order backward difference value for each peak point; The disturbance degree of the first-order forward difference value corresponding to all peak points of each sub-band in each window is positively fused with the disturbance degree of the first-order backward difference value to determine the fluctuation coefficient of each sub-band in each window; The degree of disorder in the distribution of the difference between each peak and the valley values ​​to the left and right is analyzed to obtain the peak-valley characteristic coefficients of each sub-band in each window; The product of the fluctuation coefficient of each sub-band in each window and the peak-valley characteristic coefficient is used as the characteristic fluctuation index of each sub-band in each window.

[0007] The specific process for determining the fluctuation coefficient of each sub-band in each window is as follows: For each window and each sub-band, calculate the permutation entropy of the first-order forward difference value and the permutation entropy of the first-order backward difference value of all peak points, and use the product of the two permutation entropies as the fluctuation coefficient of each window and each sub-band.

[0008] Specifically, the step of obtaining the peak-valley characteristic coefficients of each sub-band of each window is as follows: Calculate the difference between each peak and its adjacent valley values, and calculate the mean of all the differences for each peak to obtain the mean peak-valley difference for each peak. Use the permutation entropy of the mean peak-valley difference of all peaks in each sub-band as the peak-valley characteristic coefficient for each sub-band of each window.

[0009] Specifically, determining the noise impact index for each sub-band of each window involves: Calculate the ratio between the number of amplitude data points corresponding to the peak points in each sub-band and the total number of amplitude data points in that sub-band; The amplitude of each peak point in each sub-band of each window is compared with the average peak-to-valley difference. All comparison results obtained in each sub-band of each window are averaged and positively integrated with the ratio and the characteristic fluctuation index to obtain the noise impact index of each sub-band of each window.

[0010] The specific formula for the noise impact index of each sub-band in each window is as follows: In the formula, This represents the noise impact index of subband i in window a. This represents the total number of peak points in subband i within window a. This indicates the i-th subband in window a. The amplitude at each peak point This indicates the i-th subband in window a. The average peak-to-valley difference of each peak point; This represents the ratio between the number of amplitude data points corresponding to the peak point in sub-band i and the total number of amplitude data points in that sub-band. This represents the characteristic fluctuation index of subband i in window a.

[0011] Specifically, the dynamic adjustment index for each sub-band of each window of the external noise signal is obtained as follows: The difference between the noise impact index of the external noise signal and the mixed audio signal in each window and each sub-band is calculated, and then multiplied with the noise impact index of the mixed noise signal to obtain the dynamic adjustment index of the external noise signal in each window and each sub-band.

[0012] Specifically, obtaining the optimal over-attenuation factor for each sub-band of the external noise signal from the reference microphone for each window involves: Calculate the sum of the natural number 1 and the normalized dynamic adjustment index, and multiply it by the preset initial over-reduction factor to obtain the optimal over-reduction factor.

[0013] The process of filtering and synthesizing the mixed audio signal using spectral subtraction to obtain the denoised audio is as follows: the mixed audio signal is filtered and denoised using spectral subtraction, and the remaining key human voice sub-bands after denoising and enhancement are output. All the remaining key human voice sub-bands are then fused to obtain the filtered key human voice audio.

[0014] Secondly, embodiments of this application also provide an audio noise reduction system for Bluetooth headsets, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of any of the methods described above.

[0015] This application has at least the following beneficial effects: This application can accurately distinguish between noise sources and target audio by collecting external noise and mixed audio signals separately; the external noise signal acquired by the reference microphone serves as the basis for the noise model, aiding subsequent noise reduction processing; dividing the signal within a preset window into a preset number of Mel sub-bands helps to analyze the frequency characteristics of the audio signal more precisely; the use of the Mel scale can simulate the perceptual characteristics of the human ear, allowing for different noise reduction processing for different frequency bands, thus improving the targeting and efficiency of noise reduction; based on the peak fluctuation characteristics of each sub-band in each window and the difference in variation characteristics with adjacent valley points, characteristic fluctuation indices are determined; by analyzing the difference between the peak and valley values ​​of each sub-band, the signal variation can be judged. This effectively identifies frequency bands heavily affected by noise, allowing for focused suppression of these noise components in subsequent processing, reducing misidentification and noise residue. Analyzing the peak distribution density and corresponding amplitude variations in each sub-band of each window, combined with characteristic fluctuation indices, determines noise impact indicators. By analyzing the density and variations of peak distribution, noise impact indicators can be further refined, assessing the interference intensity of noise signals in each sub-band. This helps improve the accuracy of noise suppression, especially in dynamic noise environments such as pedestrian or traffic noise. Comparing the noise impact indicators of external and mixed signals allows for dynamic adjustment of noise suppression intensity. This ensures that noise reduction processing can flexibly adapt to different noise environments, avoiding over- or under-noise reduction. By dynamically adjusting the over-subtraction factor, spectral subtraction effectively reduces noise without compromising the target audio quality. Spectral subtraction suppresses noise through estimation, preserving the main characteristics of the original audio signal while removing external noise, avoiding filtering errors that occur when performing Bluetooth headset audio noise reduction in mobile scenarios, further improving the audio noise reduction level during Bluetooth headset calls. Attached Figure Description

[0016] Figure 1 A block diagram illustrating an audio noise reduction method for Bluetooth headsets, provided as an embodiment of this application; Figure 2 A flowchart illustrating the acquisition of characteristic fluctuation indicators provided in one embodiment of this application. Detailed Implementation

[0017] In the description of the embodiments in this application, the words "exemplary," "or," and "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design scheme described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of the words "exemplary," "or," and "for example" is intended to present the relevant concepts in a specific manner.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in this application's specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the application.

[0019] It should also be noted that the terms "first" and "second" in this application and its accompanying drawings are used to distinguish similar objects, rather than to describe a specific order or sequence. The methods disclosed in the embodiments of this application or the methods shown in the flowcharts include one or more steps for implementing the method. Without departing from the scope of protection of this application, the execution order of multiple steps can be interchanged, and some steps can also be deleted.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0021] The following description, in conjunction with the accompanying drawings, details a specific solution for an audio noise reduction method and system for Bluetooth headsets provided in this application.

[0022] Please see Figure 1 The diagram illustrates a flowchart of an audio noise reduction method for Bluetooth headsets according to an embodiment of this application, the method comprising: Step 1: Use the Bluetooth headset's reference microphone to collect external noise signals, and the main microphone to collect mixed audio signals.

[0023] In the audio noise reduction operation during calls using a Bluetooth headset, the audio signal is first acquired in real time through the Bluetooth headset: using the headset's dual-microphone system, the external reference microphone acquires the external ambient noise signal in real time, while the main microphone acquires the mixed audio signal of speech and noise. In this embodiment, the sampling rate for signal acquisition is set to 16kHz, which can be adjusted by the implementer according to the actual situation.

[0024] Step 2: Divide the signal within the preset window into a preset number of Mel sub-bands. Based on the peak fluctuation characteristics of each sub-band in each window and the difference in variation characteristics with adjacent valley points, determine the characteristic fluctuation index of each sub-band in each window. Analyze the distribution density of peak points and the variation characteristics of the amplitude corresponding to the peak points in each sub-band in each window. Combined with the characteristic fluctuation index, determine the noise impact index of each sub-band in each window.

[0025] The real-time acquired signal is filtered. Considering that when making calls using Bluetooth headsets, conventional spectral subtraction is effective for background noise with relatively stable frequency and intensity (steady-state noise), such as air conditioner noise, fan noise, and engine hum, it is often used as a filtering and noise reduction algorithm for Bluetooth headset calls.

[0026] Furthermore, when using Bluetooth headsets for calls in mobile environments, the intensity of ambient noise fluctuates randomly due to changes in the scene as the person walks. To avoid noise residue or the introduction of musical noise during the filtering and denoising process, it is necessary to further optimize and adjust the over-subtraction factor in the spectral subtraction by combining the signal characteristics acquired by the dual-microphone system, thereby improving the denoising effect and ensuring voice call quality.

[0027] In this application, when using spectral subtraction, a 25ms window is used, and the overlap rate of adjacent windows is set to 50% to ensure the continuity of subsequent signal synthesis. Next, the signal within each window is processed using a Fast Fourier Transform to obtain the corresponding spectrum. Using the spectrum of each window as input, a Mel filter bank is used to divide the signal into 30 Mel subbands. The Mel filter is a well-known technique, and its specific operation steps will not be elaborated further. Since the signal sampling rate is 16kHz, based on the Nyquist sampling theorem, the effective frequency range of the signal is [0, 8000]Hz.

[0028] Furthermore, considering that the main microphone on the headphones is closer to the population and usually captures strong key human voices and weak ambient noise, while the reference microphone is usually farther away from the population and captures audio mainly consisting of weak key human voices and strong ambient noise, when using spectral subtraction for noise reduction, the over-subtraction factor is mainly used to adjust the noise spectrum amplitude estimated from the reference microphone. Therefore, it is crucial to analyze the noise characteristics in the acquired signal during the over-subtraction factor adjustment process.

[0029] Based on the above analysis, the AMPD (Automatic Multiscale-based Peak Detection) peak-finding algorithm is used to obtain the peak points corresponding to the spectral peaks in each sub-band of each window. Simultaneously, the signal is inverted, and the AMPD peak-finding algorithm is used to obtain all the valley points corresponding to each sub-band of each window. Further, the first-order backward difference and first-order forward difference values ​​corresponding to each peak point in the sub-band are obtained. The first-order backward and forward differences are existing technologies, and their specific calculation processes will not be elaborated here. Due to the influence of noise, the spectral peaks exhibit chaotic fluctuations, and the steepness on both sides of each peak varies randomly, resulting in large fluctuations in the first-order backward and forward difference values ​​corresponding to the peaks. Therefore, the first-order forward and first-order backward difference values ​​corresponding to each peak point in each sub-band of the window are calculated, and the permutation entropy of the first-order forward difference value and the permutation entropy of the first-order backward difference value for all peak points are calculated separately. The embedding dimension is set to 3, the time delay is set to 1, and the product of the two permutation entropies is... , as the fluctuation coefficient for each sub-band of each window, is used to represent the variation in the steepness of the peak.

[0030] Under the influence of noise, the valley values ​​in the spectrum are randomly increased, while the peak values ​​fluctuate relatively little, resulting in a relatively large overall fluctuation in the peak-valley difference. Therefore, the difference between each peak value and the valley values ​​to its left and right are calculated, and the mean of all the differences for each peak value is calculated to obtain the mean peak-valley difference for each peak point. The permutation entropy of the mean peak-valley difference of all peak values ​​in each sub-band is obtained and denoted as the peak-valley feature coefficient of each sub-band of each window; where the embedding dimension is set to 3 and the time delay is set to 1.

[0031] Based on the above analysis, we analyze sub-band i in window a: the fluctuation coefficient obtained from the sub-band is denoted as... The peak-valley characteristic coefficient is denoted as Constructing a characteristic fluctuation index This is used to characterize the internal peak steepness and peak-valley difference fluctuation degree caused by noise interference in the sub-band corresponding to each window. The flowchart for obtaining the characteristic fluctuation index is as follows: Figure 2 As shown.

[0032] Furthermore, the ratio of the number of amplitude data points corresponding to the peak points in each sub-band to the total number of amplitude data points in the sub-band is calculated. This is used to represent the peak density caused by noise. Then, a noise impact index is constructed to characterize the changes in internal features of the target window's corresponding sub-band after being affected by noise. The specific formula is as follows: In the formula, This represents the noise impact index of subband i in window a. This represents the total number of peak points in subband i within window a. This indicates the i-th subband in window a. The amplitude at each peak point This indicates the i-th subband in window a. The average peak-to-valley difference of each peak point; among which... This represents the comparison result between the amplitude of each peak point in each sub-band of each window and the average peak-to-valley difference; This represents the result of averaging all comparison results obtained for each sub-band of each window and positively fusing them with the ratio and the characteristic fluctuation index.

[0033] Formula Principle: Existing technologies typically use direct calculation of peak-to-valley differences under noise influence to represent noise interference. However, this doesn't account for variations in the steepness of spectral peaks in sub-bands under different scenarios, particularly when pedestrians or human voices are present. Furthermore, the frequency of peak occurrence may increase. This variation can lead to errors in analyzing noise-affected sub-bands, thus impacting the filtering effect during Bluetooth headset call audio noise reduction. Therefore, this application analyzes the peak-to-valley differences, peak frequency, and peak steepness fluctuations within sub-bands under noise influence to reduce subsequent filtering errors.

[0034] When noise interference is severe, the sub-band corresponding to the window will be affected by pedestrian or human voice noise, causing irregular fluctuations in the steepness of the spectral peaks. At the same time, the peak-to-valley difference will also fluctuate randomly due to noise interference. In this case, the obtained noise impact index may be large. Conversely, when the noise impact is smaller, the signal exhibits regular fluctuations, with a larger peak-to-valley difference and a more regular overall pattern, resulting in a smaller obtained noise impact index.

[0035] Step 3: Compare the noise impact index of the external noise signal and the mixed audio signal in each window and each sub-band to obtain the dynamic adjustment index of the external noise signal in each window and each sub-band; based on the dynamic adjustment index, obtain the optimal over-subtraction factor of the external noise signal of the reference microphone in each window and each sub-band, and use spectral subtraction to filter and synthesize the mixed audio signal to obtain the noise-reduced audio.

[0036] In mobile scenarios, during Bluetooth headset calls, the main microphone primarily captures the audio signal combining key human voices and ambient noise, while the reference microphone mainly captures ambient noise. In quiet environments with minimal noise interference, the main microphone's signal primarily consists of the key human voice during the call. Simultaneously, because sound travels through the air, the reference microphone also captures the key human voice, leading to similar audio characteristics. In such cases, a smaller over-attenuation factor should be used to prevent excessive filtering and resulting music noise. However, in situations with significant noise interference, the reference microphone collects external noise entering the ear, while the main microphone, being closer to the ear, captures the key human voice along with ambient noise. This results in different audio characteristics between the two microphones, requiring a larger over-attenuation factor for effective noise filtering.

[0037] Therefore, taking sub-band i in window a as an example from the audio acquired by the main microphone and the reference microphone at the same time, a dynamic adjustment index is constructed by combining the feature difference values ​​of the sub-bands in the corresponding windows of the audio acquired by the reference microphone and the main microphone. ,in, This represents the noise impact index of subband i in window a during audio acquisition using the reference microphone. This represents the noise impact index of subband i in window a within the audio acquired by the main microphone. Under relatively quiet environmental conditions, the acquired... and All are relatively small, and the overall acquisition The smaller the value, the better, as the environmental noise becomes more severe. and All are relatively large, and the overall acquisition The value is relatively large.

[0038] After analyzing and acquiring the signal characteristics, considering that the noise level may change with the scene when making a call using a Bluetooth headset in a mobile scenario, when using spectral subtraction for call audio noise reduction, in a relatively quiet environment, using a large over-subtraction factor may cause the effective signal to be misjudged as noise and filtered out, resulting in distorted call voice and hollow and harsh sound; using a small over-subtraction factor may result in insufficient noise reduction when the noise is severe.

[0039] Based on the above analysis, it is necessary to obtain the corresponding optimal over-attenuation factor according to the noise situation during the call, adaptively adjust the over-attenuation factor starting from the minimum value, normalize the obtained dynamic adjustment index, and obtain the optimal over-attenuation factor for sub-band i in window a of the audio obtained from the reference microphone in real time: In the formula, This represents the optimal over-subtraction factor for subband i in the target signal a. This represents the dynamically adjusted indicator after normalization. This represents the preset initial over-reduction factor, which is set to 3 in this embodiment.

[0040] Formula principle: In existing technologies, the over-subtraction factor of spectral subtraction is usually set based on expert experience, but it does not take into account the different degrees of fluctuation in noise intensity in mobile scenarios during actual processing. This leads to a large error in the noise reduction process of Bluetooth headset call audio. Based on this, this application fully combines the characteristic fluctuation changes that occur inside the audio signal in the presence of background noise and the characteristic differences between the two microphones to obtain the optimal over-subtraction factor for the subband under the audio target window, thereby reducing the noise reduction error of Bluetooth headset call audio.

[0041] In the signal enhancement module of the Bluetooth headset, the mixed audio acquired in real time by the reference microphone and the mixed audio acquired in real time by the main microphone are segmented and differentiated into sub-bands through overlapping windows. These sub-bands are then used as inputs for spectral subtraction. Using the best over-subtraction factor acquired in real time, and setting the spectral base parameter (lower spectral limit) to 0.005, spectral subtraction is used to filter and denoise the signal. The output is the main microphone sub-band of the remaining key human voice after filtering and denoising enhancement. Spectral subtraction is a well-known technology in the field, and the detailed operation steps will not be elaborated further. All remaining key human voice main microphone sub-bands are fused, and all overlapping parts of the windows are fused into the filtered key human voice audio by averaging. The call audio enhanced in real time by the signal enhancement module of the Bluetooth headset is then transmitted uplink to a communication device (such as a smartphone or tablet). The communication device then transmits the filtered and enhanced audio to the receiving device via 5G communication, thus completing the audio noise reduction method for the Bluetooth headset during a call.

[0042] Based on the same inventive concept as the above methods, this application also provides an audio noise reduction system for Bluetooth headsets, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of any of the above methods.

[0043] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than that shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. In the descriptions corresponding to the flowcharts and block diagrams in the accompanying drawings, the operations or steps corresponding to different blocks may also occur in a different order than disclosed in the description; sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. Each block in a block diagram and / or flowchart, and combinations of blocks in a block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0044] It will be apparent to those skilled in the art that this application is not limited to the details of the exemplary embodiments described above, and that this application can be implemented in other specific forms without departing from its essential characteristics. Therefore, the embodiments described above should be considered exemplary and non-limiting in all respects; modifications to the technical solutions described in the foregoing embodiments, or equivalent substitutions of some technical features, without causing the essence of the corresponding technical solutions to deviate from the scope of the technical solutions in the embodiments of this application, should all be included within the protection scope of this application.

Claims

1. An audio noise reduction method for Bluetooth headphones, characterized in that, The method includes: The Bluetooth headset reference microphone is used to collect external noise signals, while the main microphone collects mixed audio signals. The signal within a preset window is divided into a preset number of Mel sub-bands. Based on the peak fluctuation characteristics of each sub-band in each window and the difference change characteristics with adjacent valley points, the characteristic fluctuation index of each sub-band in each window is determined. The distribution density of peak points and the change characteristics of the amplitude corresponding to the peak points in each sub-band in each window are analyzed. Combined with the characteristic fluctuation index, the noise impact index of each sub-band in each window is determined. The noise impact index of the external noise signal and the mixed audio signal in each window and each sub-band is compared to obtain the dynamic adjustment index of the external noise signal in each window and each sub-band. Based on the dynamic adjustment index, the optimal over-subtraction factor of the external noise signal of the reference microphone in each window and each sub-band is obtained. The mixed audio signal is filtered and synthesized by spectral subtraction to obtain the noise-reduced audio.

2. The audio noise reduction method for Bluetooth headsets as described in claim 1, characterized in that, The process of determining the characteristic fluctuation index of each sub-band in each window is as follows: For each window and each sub-band, obtain the first-order forward difference value and the first-order backward difference value for each peak point; The disturbance degree of the first-order forward difference value corresponding to all peak points of each sub-band in each window is positively fused with the disturbance degree of the first-order backward difference value to determine the fluctuation coefficient of each sub-band in each window; The degree of disorder in the distribution of the difference between each peak and the valley values ​​to the left and right is analyzed to obtain the peak-valley characteristic coefficients of each sub-band in each window; The product of the fluctuation coefficient of each sub-band in each window and the peak-valley characteristic coefficient is used as the characteristic fluctuation index of each sub-band in each window.

3. The audio noise reduction method for Bluetooth headsets as described in claim 2, characterized in that, The specific process for determining the fluctuation coefficient of each sub-band in each window is as follows: For each window and each sub-band, calculate the permutation entropy of the first-order forward difference value and the permutation entropy of the first-order backward difference value of all peak points, and use the product of the two permutation entropies as the fluctuation coefficient of each window and each sub-band.

4. The audio noise reduction method for Bluetooth headsets as described in claim 2, characterized in that, The specific steps for obtaining the peak-valley characteristic coefficients of each sub-band in each window are as follows: Calculate the difference between each peak and its adjacent valley values, and calculate the mean of all the differences for each peak to obtain the mean peak-valley difference for each peak. Use the permutation entropy of the mean peak-valley difference of all peaks in each sub-band as the peak-valley characteristic coefficient for each sub-band of each window.

5. The audio noise reduction method for Bluetooth headsets as described in claim 4, characterized in that, The determination of the noise impact index for each sub-band of each window is specifically as follows: Calculate the ratio between the number of amplitude data points corresponding to the peak points in each sub-band and the total number of amplitude data points in that sub-band; The amplitude of each peak point in each sub-band of each window is compared with the average peak-to-valley difference. All comparison results obtained in each sub-band of each window are averaged and positively integrated with the ratio and the characteristic fluctuation index to obtain the noise impact index of each sub-band of each window.

6. The audio noise reduction method for Bluetooth headsets as described in claim 5, characterized in that, The specific formula for the noise impact index of each sub-band in each window is as follows: In the formula, This represents the noise impact index of subband i in window a. This represents the total number of peak points in subband i within window a. This indicates the i-th subband in window a. The amplitude at each peak point This indicates the i-th subband in window a. The average peak-to-valley difference of each peak point; This represents the ratio between the number of amplitude data points corresponding to the peak point in sub-band i and the total number of amplitude data points in that sub-band. This represents the characteristic fluctuation index of subband i in window a.

7. The audio noise reduction method for Bluetooth headsets as described in claim 1, characterized in that, The dynamic adjustment index for each sub-band of each window of the external noise signal is obtained as follows: The difference between the noise impact index of the external noise signal and the mixed audio signal in each window and each sub-band is calculated, and then multiplied with the noise impact index of the mixed noise signal to obtain the dynamic adjustment index of the external noise signal in each window and each sub-band.

8. The audio noise reduction method for Bluetooth headsets as described in claim 1, characterized in that, The acquisition of the optimal over-attenuation factor for each sub-band of the external noise signal of the reference microphone for each window specifically involves: Calculate the sum of the natural number 1 and the normalized dynamic adjustment index, and multiply it by the preset initial over-reduction factor to obtain the optimal over-reduction factor.

9. The audio noise reduction method for Bluetooth headsets as described in claim 1, characterized in that, The process of filtering and synthesizing the mixed audio signal using spectral subtraction to obtain the denoised audio is as follows: the mixed audio signal is filtered and denoised using spectral subtraction, and the remaining key vocal sub-bands after denoising and enhancement are output. All the remaining key vocal sub-bands are then merged to obtain the filtered key vocal audio.

10. An audio noise reduction system for Bluetooth headsets, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1-9.

Images