A talk active noise reduction method and device for a Bluetooth earphone

By acquiring the distribution of environmental noise weights and signal processing efficiency, and combining short-time Fourier transform to separate user voice from background voice, the noise reduction depth and resource allocation are dynamically adjusted, and the clock synchronization of the noise reduction chip is optimized. This solves the problem of unstable call noise reduction in Bluetooth headsets in complex environments, achieving stable call quality and safe device operation.

CN123123777APending Publication Date: 2026-11-17SHENZHEN SHIKE JINHUA ELECTRONICS
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611481056.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-22
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing Bluetooth headsets have unstable noise reduction performance in complex dynamic environments, making it difficult to accurately distinguish between user voice and background noise in real time, resulting in poor noise reduction performance and affecting call clarity and continuity.

Method used

By obtaining the preliminary distribution of environmental noise weights and signal processing efficiency, and combining short-time Fourier transform to separate user speech from background voice, the noise reduction depth and resource allocation are dynamically adjusted, the clock synchronization of the noise reduction chip is optimized, and the device status is monitored in real time to achieve accurate call noise reduction.

Benefits of technology

It improves the stability of call noise reduction in Bluetooth headsets under complex dynamic environments, reduces the occurrence of accidental cancellation of user voice messages, enhances call clarity and continuity, and rationally allocates device resources to extend battery life and ensure operational safety.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123123777A_ABST
    Figure CN123123777A_ABST
Patent Text Reader

Abstract

The application relates to the fields of audio processing and wireless communication technology, and discloses a call active noise reduction method and device of a Bluetooth earphone. The method comprises the following steps: acquiring a preliminary distribution of an environmental noise weight and a signal processing efficiency; separating user voice and background human voice according to the preliminary distribution to obtain a user voice signal; extracting a call state feature and analyzing the call state feature to obtain a preliminary parameter of noise reduction depth and a real-time load result; adjusting resource allocation and a switching frequency, determining priority and a delay parameter; optimizing a noise reduction chip clock synchronization to obtain a frequency adjustment amplitude and a power consumption rule; obtaining an electric quantity curve and state feedback through current, temperature monitoring and noise analysis; and adjusting a protection response and the noise reduction depth to obtain a noise reduction output if overload and temperature rise exceed a standard. The method can realize stable and accurate noise reduction in a complex environment, and takes into account power consumption and safety.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the fields of audio processing and wireless communication technology, and in particular to a method and apparatus for active noise cancellation during Bluetooth headset calls. Background Technology

[0002] In the field of modern audio equipment, with the rapid development of wireless communication technology and the widespread adoption of smart terminals, Bluetooth headsets, with their wireless and portable characteristics, have become an important device for people's daily calls and audio listening. Call noise reduction technology, as a core indicator of Bluetooth headset performance, effectively suppresses background noise in complex environments, ensuring clear transmission of the user's voice and meeting the user's need for high-quality calls in noisy places such as public areas and commuting scenarios.

[0003] Currently, existing Bluetooth headset noise reduction methods have significant limitations in practical applications. A common approach is to collect audio signals through a microphone and use fixed filtering algorithms to separate noise from the user's speech, thereby achieving noise reduction. For example, some solutions use a preset noise model to process the collected audio, attenuating the noise signal through filters with fixed parameters. However, this method performs poorly in dynamically changing noise environments. Due to the diversity and variability of environmental noise, such as the intermingling or rapid switching of crowd conversations, vehicle noise, and wind noise, existing technologies struggle to accurately distinguish user speech from background noise in real time. This results in unstable noise reduction effects, often leading to the accidental cancellation of user speech, affecting call continuity and clarity.

[0004] In summary, existing technologies suffer from unstable noise reduction performance in Bluetooth headsets during calls under complex and dynamic conditions. Summary of the Invention

[0005] This invention provides a method and apparatus for active noise cancellation during Bluetooth headset calls, which solves the problem of unstable call noise cancellation effect of existing Bluetooth headsets in complex dynamic environments, and achieves stable and accurate call noise cancellation in complex dynamic noise environments, thereby improving the user's call experience.

[0006] In a first aspect, to solve the above-mentioned technical problems, the present invention provides an active noise cancellation method for Bluetooth headset calls, comprising: Obtain the preliminary distribution of environmental noise weights and signal processing efficiency; Based on the preliminary distribution of environmental noise weights and signal processing efficiency, user speech and background voice are separated to obtain the user speech signal; Based on the user's voice signal, extract the user's call status features, perform status judgment and signal analysis, and obtain preliminary parameters for noise reduction depth adjustment and real-time load detection results. Based on the preliminary parameters of the noise reduction depth adjustment and the real-time load detection results, adjust the resource allocation order and switching frequency, and determine the resource allocation priority and frequency switching delay parameters. Based on the resource allocation priority and the frequency switching delay parameter, combined with the user call status characteristics, the clock synchronization of the noise reduction chip is adjusted and optimized to obtain the adjusted chip clock frequency and power consumption allocation rules. Based on the adjusted chip clock frequency and the power consumption allocation rules, the power consumption curve and real-time status feedback data are obtained through current monitoring, temperature threshold control and noise analysis. Based on the power consumption curve and real-time status feedback, status analysis is performed. If an overload interruption is detected and the temperature rise rate exceeds the preset temperature change threshold, the protection response speed and noise reduction depth are adjusted to obtain the noise reduction output of the Bluetooth headset.

[0007] In one optional implementation, obtaining the preliminary distribution of environmental noise weights and signal processing efficiency includes: Ambient audio data is collected through the microphone array of a Bluetooth headset, and the ambient audio data is digitally processed using a preset sampling rate to obtain the original audio data. The original audio data is decomposed into time and frequency using short-time Fourier transform to generate the spectral distribution of the audio signal; If there are frequency components in the spectrum distribution that are higher than the preset spectrum component threshold, the environmental noise weight is obtained by weighted calculation. By jointly analyzing the environmental noise weights and the spectral distribution, and adjusting the signal processing parameters based on the original audio time-frequency characteristics, a preliminary distribution of signal processing efficiency is obtained.

[0008] In one optional implementation, the step of separating user speech from background voice based on the preliminary distribution of environmental noise weights and signal processing efficiency to obtain the user speech signal includes: The input audio is acquired and decomposed into a first spectral signal through Fourier transform. Based on the preliminary distribution of environmental noise weights and signal processing efficiency, the target frequency band for noise feature extraction is determined. The noise features are extracted from the target frequency band of the first spectral signal using short-time Fourier transform to obtain the first noise weight; Based on the first noise weight and referring to the preliminary distribution of the signal processing efficiency, the filtering parameters are adjusted to filter the first spectral signal to obtain the first separated signal; Extract user speech features from the first separated signal, separate background human voice, and obtain a second separated signal containing signal amplitude compensation. If the speech clarity of the second separated signal is lower than a preset speech clarity threshold, the signal amplitude compensation amount is adjusted by dynamic range compression to obtain the user's speech signal.

[0009] In one optional implementation, the step of extracting user call state features based on the user's voice signal, performing state judgment and signal analysis, and obtaining preliminary parameters for noise reduction depth adjustment and real-time load detection results includes: Real-time load detection data is obtained through the built-in sensors in the headphones; Spectral data is obtained from the user's voice signal, and the spectral data is decomposed by short-time Fourier transform to extract the main frequency band energy distribution and obtain the first energy distribution feature. Based on the first energy distribution characteristics, the degree of overlap between the main frequency band energy and the background human voice frequency is calculated to obtain the first overlap characteristic; If the first overlap feature is greater than the preset call state classification threshold, then the first energy distribution feature and the first overlap feature are classified to obtain the first call state classification result. Based on the first call status classification result and combined with the real-time load detection data, the classification threshold is dynamically optimized, and the call status is reclassified to obtain the current call status. Based on the current call status and the first overlap feature, the dynamic gain requirements of the main frequency band energy concentration and noise reduction are analyzed to obtain preliminary parameters for noise reduction depth adjustment, and the real-time load detection results are recorded simultaneously.

[0010] In one optional implementation, the step of adjusting the resource allocation order and switching frequency based on the preliminary parameters adjusted according to the noise reduction depth and the real-time load detection results, and determining the resource allocation priority and frequency switching delay parameters, includes: Obtain user call duration data; Based on the call duration data, combined with the preliminary parameters of the noise reduction depth adjustment and the real-time load detection results, power consumption requirements are assessed to obtain power consumption constraints. If the power consumption constraint exceeds the preset power consumption constraint threshold, the resource allocation is prioritized to obtain the resource allocation priority. Based on the resource allocation priority, the bandwidth utilization rate is analyzed to obtain bandwidth utilization data, and the link switching frequency is dynamically adjusted to obtain the frequency switching delay parameter.

[0011] In one optional implementation, the step of adjusting and optimizing the clock synchronization of the noise reduction chip based on the resource allocation priority and the frequency switching delay parameter, combined with the user call state characteristics, to obtain the adjusted chip clock frequency and power consumption allocation rules, includes: Based on the resource allocation priority and combined with the main frequency band energy distribution in the user call status characteristics, calculate the energy distribution data and perform dynamic frequency adjustment to generate frequency adjustment parameters and determine the chip clock synchronization scheme. Based on the chip clock synchronization scheme and the frequency adjustment parameters, the frequency switching delay is optimized to obtain the adjusted chip clock frequency. Based on the adjusted chip clock frequency, power allocation data is calculated, and signal quality is adjusted in conjunction with the noise reduction chip hardware operating parameters and audio signal processing requirements to determine power allocation rules.

[0012] In one optional implementation, the step of obtaining the power consumption curve and real-time status feedback data based on the adjusted chip clock frequency and the power allocation rule, through current carrying capacity monitoring, temperature threshold control, and noise analysis, includes: Acquire chip temperature sensor data and overload current monitoring data; Based on the adjusted chip clock frequency and the power consumption allocation rule, the overlap between the frequency of the chip temperature sensor data and the background human voice frequency is analyzed to generate the initial frequency adjustment amplitude and obtain the chip temperature sensor data. The chip temperature sensor data is compared with a preset chip safety temperature threshold. If the temperature exceeds the safety temperature threshold, the initial frequency adjustment amplitude is reduced to obtain the temperature-optimized frequency adjustment amplitude. Based on the frequency adjustment amplitude optimized by the temperature, noise suppression processing is performed on the overlap of the background human voice frequencies to optimize the noise reduction depth and obtain real-time status feedback data. The overload current monitoring data is compared with the preset chip safe current threshold. If the monitored current exceeds the chip safe current threshold, the frequency adjustment range after temperature optimization is further reduced to obtain the final frequency adjustment range. Based on the final frequency adjustment amplitude, the power consumption per unit time is calculated in conjunction with the power consumption allocation rules. Power consumption data is recorded in time series to generate a power consumption curve. The power consumption curve and the real-time status feedback data are then integrated as the result.

[0013] In one optional implementation, the step of performing state analysis based on the power consumption curve and real-time status feedback data, and adjusting the protection response speed and noise reduction depth to obtain the noise reduction output of the Bluetooth headset if an overload interruption is detected and the temperature rise rate exceeds a preset temperature change threshold, includes: Based on the power consumption curve and the real-time status feedback data, collect device status parameters and obtain overload interruption trigger signal and temperature rise rate data. If an overload interruption is detected and the rate of temperature rise exceeds the preset temperature change threshold, the protection response speed is adjusted by using a dynamic threshold to obtain the adjusted response speed parameters. Based on the adjusted response speed parameters, the noise reduction depth is adjusted to obtain the optimized noise reduction depth parameters. By optimizing the noise reduction depth parameters and data stream stability processing, the signal processing efficiency is improved, resulting in the final noise reduction output of the Bluetooth headset.

[0014] In a second aspect, the present invention provides an active noise cancellation device for Bluetooth headset calls, comprising: The environmental feature acquisition module is used to obtain the preliminary distribution of environmental noise weights and signal processing efficiency; The speech separation processing module is used to separate user speech from background human voice based on the preliminary distribution of environmental noise weights and signal processing efficiency to obtain the user speech signal. The call status analysis module is used to extract user call status features based on the user's voice signal, perform status judgment and signal analysis, and obtain preliminary parameters for noise reduction depth adjustment and real-time load detection results. The resource scheduling module is used to adjust the resource allocation order and switching frequency based on the preliminary parameters of the noise reduction depth adjustment and the real-time load detection results, and to determine the resource allocation priority and frequency switching delay parameters. The chip optimization module is used to adjust and optimize the clock synchronization of the noise reduction chip based on the resource allocation priority and the frequency switching delay parameter, combined with the user call status characteristics, to obtain the adjusted chip clock frequency and power consumption allocation rules. The status feedback module is used to obtain the power consumption curve and real-time status feedback data based on the adjusted chip clock frequency and the power consumption allocation rules, through current monitoring, chip temperature threshold control and noise analysis. The noise reduction output module is used to perform state analysis based on the power consumption curve and real-time status feedback data. If an overload interruption is detected and the temperature rise rate exceeds the preset temperature change threshold, the protection response speed for overload interruption and abnormal temperature rise, as well as the noise reduction depth, are adjusted to obtain the noise reduction output of the Bluetooth headset.

[0015] Thirdly, the present invention also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the active noise cancellation method for Bluetooth headsets described in any of the above.

[0016] Compared with the prior art, the present invention has the following beneficial effects: (1) By obtaining the environmental noise weight and the preliminary distribution of signal processing efficiency, and combining the short-time Fourier transform and other technologies, the present invention can accurately identify the dynamically changing noise environment, avoid the problem that the fixed filtering algorithm in the prior art is difficult to adapt to complex noise scenarios, reduce the situation of false cancellation of user voice, improve the clarity and continuity of the call, and improve the stability of the call noise reduction effect of Bluetooth headset in complex dynamic environments. (2) The present invention dynamically adjusts the noise reduction depth parameters based on the user's call status characteristics, and optimizes resource allocation and chip clock synchronization by combining real-time load detection results. While ensuring the noise reduction effect, it reasonably allocates equipment resources and controls power consumption, which solves the problem of difficulty in balancing power consumption and noise reduction effect in the existing noise reduction scheme and extends the battery life of Bluetooth headsets. (3) This invention monitors the operating status of the device in real time by monitoring overload current, controlling temperature threshold and adjusting dynamic protection response. When an overload interruption or temperature abnormality is detected, the protection strategy is adjusted in time to avoid damage to the device due to overload or high temperature, thereby improving the safety and stability of Bluetooth headset operation and further protecting the user's call experience. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the active noise cancellation method for Bluetooth headsets provided in the first embodiment of the present invention. Figure 2 This is a schematic diagram of the active noise cancellation device for Bluetooth headsets provided in the second embodiment of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Reference Figure 1 The first embodiment of the present invention provides a method for active noise cancellation during Bluetooth headset calls, comprising the following steps: S101, Obtain the preliminary distribution of environmental noise weights and signal processing efficiency; S102, based on the preliminary distribution of the environmental noise weight and signal processing efficiency, separate the user's voice from the background voice to obtain the user's voice signal; S103, Based on the user's voice signal, extract the user's call state features, perform state judgment and signal analysis, and obtain preliminary parameters for noise reduction depth adjustment and real-time load detection results; S104, Based on the preliminary parameters of the noise reduction depth adjustment and the real-time load detection results, adjust the resource allocation order and switching frequency, and determine the resource allocation priority and frequency switching delay parameters; S105, based on the resource allocation priority and the frequency switching delay parameter, and combined with the user call status characteristics, adjust and optimize the clock synchronization of the noise reduction chip to obtain the adjusted chip clock frequency and power consumption allocation rules. S106, based on the adjusted chip clock frequency and the power consumption allocation rule, the power consumption curve and real-time status feedback data are obtained through current monitoring, chip temperature threshold control and noise analysis. S107. Based on the power consumption curve and real-time status feedback data, perform status analysis. If an overload interruption is detected and the temperature rise rate exceeds the preset temperature change threshold, adjust the overload interruption and abnormal temperature rise protection response speed and noise reduction depth to obtain the noise reduction output of the Bluetooth headset.

[0020] In step S101, the preliminary distribution of environmental noise weights and signal processing efficiency is obtained.

[0021] In one optional implementation, obtaining the preliminary distribution of environmental noise weights and signal processing efficiency includes: Ambient audio data is collected through the microphone array of a Bluetooth headset, and the ambient audio data is digitally processed using a preset sampling rate to obtain the original audio data. The original audio data is decomposed into time and frequency using short-time Fourier transform to generate the spectral distribution of the audio signal; If there are frequency components in the spectrum distribution that are higher than the preset spectrum component threshold, the environmental noise weight is obtained by weighted calculation. By jointly analyzing the environmental noise weights and the spectral distribution, and adjusting the signal processing parameters based on the original audio time-frequency characteristics, a preliminary distribution of signal processing efficiency is obtained.

[0022] It's worth noting that when collecting ambient audio data via the Bluetooth headset's microphone array, a spatial three-dimensional acquisition layout is formed by 2-4 microphones in the left and right earbud units and the charging case, ensuring comprehensive capture of ambient noise. The preset sampling rate is dynamically selected based on scenario requirements; 8kHz is used for everyday scenarios to balance performance and power consumption, while 16kHz is increased for high-fidelity scenarios to retain more high-frequency noise details. The acquired analog audio signal undergoes 16-bit quantization processing, with each audio frame length set to 20ms and a frame shift of 10ms, ensuring signal continuity while avoiding data redundancy. For example, in a noisy street environment, the 4-microphone array can simultaneously capture multi-directional noise such as vehicle engine sounds and pedestrian conversations, while the 16kHz sampling rate accurately records the high-frequency details of braking sounds.

[0023] When using short-time Fourier transform for time-frequency decomposition, a Hanning window function is selected, with a window length set to 20ms to match the length of each audio frame, avoiding signal truncation distortion within the frame. The original audio data is divided into continuous audio frames according to a rule of 20ms / frame and a 10ms frame shift. Hanning window weighting is applied to each frame of audio data—the Hanning window weights the amplitude of each sampling point within the frame, reducing the abrupt changes in signal at frame edges. After weighting, a Fourier transform is performed on each frame of data, converting the time-domain signal into a two-dimensional "time-frequency" spectrum matrix. Rows represent time frames (50 frames per second), and columns represent frequency points (2048 frequency points per frame at a 48kHz sampling rate, covering the 0-24kHz frequency band). For example, in a coffee shop scenario, the spectrum matrix can clearly show the difference in the distribution of the low-frequency rumble (100-300Hz) of the coffee machine and the mid-frequency signal (500-2kHz) of conversation at the next table on the time axis.

[0024] The preset spectral component thresholds in the preset spectral distribution were calibrated through extensive experiments, set to -35dB for mid-to-high frequency scenarios and -40dB for low-frequency noise scenarios. For frequency components exceeding the preset spectral component thresholds, the environmental noise weighting method was used to calculate the environmental noise weight, as shown in the following formula: in, This represents the environmental noise weight of the i-th frequency band, used to characterize the contribution of noise in that frequency band to overall call interference. This represents the noise energy in the i-th frequency band. This represents the total noise energy across the entire frequency band, calculated as the sum of the energy across all frequency bands. For example, if low-frequency noise (energy above the 65dB low-frequency threshold) is detected from an office air conditioner at 400Hz, its corresponding frequency band energy... Accounting for the total energy of noise across the entire frequency band Substituting 25% into the formula, we can calculate the noise weight for that frequency band. =0.25, and finally integrate the calculation results of all effective frequency bands to form a weighted distribution table of noise contribution of each frequency band.

[0025] When jointly analyzing noise weight distribution and spectral distribution, differentiated filtering strategies are adopted based on the differences in noise weight across different frequency bands: for frequency bands with high noise weights, such as 50-200Hz in subway scenarios, a 128th-order filter is used to enhance noise suppression; while for frequency bands with low noise weights, such as 3-8kHz, a 64th-order filter is used to reduce chip computation while ensuring basic noise reduction. Here, signal processing efficiency is used to measure the noise suppression effect that computational resources can bring to different frequency bands. Its purpose is to prioritize allocating computational resources to the frequency band with the highest 'cost-effectiveness' in subsequent processing. Signal processing efficiency is calculated by dividing the historical average noise reduction performance of the frequency band by its historical average computational load. Specifically, this is achieved by running the device in a typical noise scenario in a laboratory, statistically analyzing the average data afterward, and pre-setting this average data in the device's firmware as a lookup table. During real-time operation, this ratio does not need to be calculated in real time; instead, this pre-set "efficiency" lookup table is directly invoked to guide resource allocation. Taking a shopping mall environment as an example, for crowd noise in the 300-800Hz range (with a noise weight of 0.4), a table was consulted and it was found that the signal processing efficiency was 0.9dB / MIPS, providing reference data for subsequent resource allocation.

[0026] In summary, by using multi-microphone array acquisition, dynamic sampling rate adjustment, precise time-frequency decomposition, and weighted calculation, the environmental noise weight was quantitatively represented. Combined with a frequency band differentiated processing strategy, a scientific distribution of signal processing efficiency was formed, providing a precise data foundation for subsequent steps of speech separation and noise reduction parameter optimization. This effectively improved the accuracy of noise feature capture and the controllability of signal processing efficiency in complex environments.

[0027] In step S102, based on the preliminary distribution of environmental noise weights and signal processing efficiency, user speech and background voice are separated to obtain the user speech signal.

[0028] In one optional implementation, the step of separating user speech from background voice based on the preliminary distribution of environmental noise weights and signal processing efficiency to obtain the user speech signal includes: The input audio is acquired and decomposed into a first spectral signal through Fourier transform. Based on the preliminary distribution of environmental noise weights and signal processing efficiency, the target frequency band for noise feature extraction is determined. The noise features are extracted from the target frequency band of the first spectral signal using short-time Fourier transform to obtain the first noise weight; Based on the first noise weight and referring to the preliminary distribution of the signal processing efficiency, the filtering parameters are adjusted to filter the first spectral signal to obtain the first separated signal; Extract user speech features from the first separated signal, separate background human voice, and obtain a second separated signal containing signal amplitude compensation. If the speech clarity of the second separated signal is lower than a preset speech clarity threshold, the signal amplitude compensation amount is adjusted by dynamic range compression to obtain the user's speech signal.

[0029] It should be noted that when acquiring the input audio and decomposing it into the first spectral signal using Fourier transform, the input audio originates from the mixed sound (including user voice, background voice, and environmental noise) collected by the Bluetooth headset microphone array. A 2048-point Fast Fourier Transform (FFT) is used to convert the time-domain signal into the first spectral signal in the frequency domain. This point setting combines the computing power of the Bluetooth headset chip (at a sampling rate of 48kHz, the computational cost of a single 2048-point FFT is approximately 1.2 MIPS, lower than the chip's 3 MIPS single audio frame processing limit) with the frequency resolution requirements (a resolution of 23.4Hz can accurately distinguish sound signals in similar frequency bands). For example, in a multi-person conference scenario, it can effectively separate the 1kHz (core frequency band for user speech) and 1.05kHz (concentrated frequency band for background discussion) signals, avoiding separation distortion caused by frequency band overlap.

[0030] When extracting noise features from the target frequency band of the first spectral signal using Short Time Fourier Transform (STFT), based on the previously determined "preliminary distribution of environmental noise weight and signal processing efficiency," frequency bands with noise weight ≥ the average value of the entire frequency band and signal processing efficiency ≥ 0.7dB / MIPS are preferentially selected as target frequency bands (such as the 50-200Hz band in the subway scenario). Then, the mean and variance of the amplitude of each audio frame audio data within the target frequency band are calculated, and frequency components exceeding "mean + 2 times variance" are marked as noise features (this judgment standard has been calibrated through 100 sets of different noise scenarios, with a false judgment rate ≤ 3%). Finally, the first noise weight is obtained by combining the energy proportion of each noise feature through the energy weighting method. When calculating the weight, frequency bands with high signal processing efficiency are assigned a weight coefficient of 1.2 times to ensure that the noise features of the high-efficiency processing frequency band are given priority consideration. For example, the first noise weight of air conditioning noise (concentrated in 400-600Hz, signal processing efficiency 0.8dB / MIPS) in the office scenario can be calculated as 0.3.

[0031] When filtering the first spectral signal based on the first noise weight, a minimum mean square error (LMS) adaptive filter is used, with the filter update step size set to 0.01 (this value was determined through comparative testing: with a step size of 0.01, the filter convergence speed is 40% faster than with 0.005, and the steady-state error only increases by 0.5dB, balancing convergence efficiency and filtering accuracy). During the filtering process, the attenuation strategy is adjusted with reference to the preliminary distribution of signal processing efficiency—for frequency bands with high noise weight and high signal processing efficiency, the attenuation is calculated as "noise weight × 0.05" (e.g., a frequency band with a weight of 0.25 corresponds to an attenuation of 1.25dB), and for frequency bands with high noise weight but low signal processing efficiency, the attenuation is calculated as "noise weight × 0.03" (to avoid computational overload caused by excessive calculation). For example, in a restaurant environment, the 500-1.5kHz frequency band (first noise weight 0.28, signal processing efficiency 0.9dB / MIPS) where background conversations are concentrated belongs to the "high weight + high efficiency" frequency band. According to the rules, it is attenuated by 1.4dB to obtain the first separation signal that initially filters the background noise.

[0032] The user's speech features are extracted from the first separated signal, and background voices are separated to obtain a second separated signal containing signal amplitude compensation. If the speech intelligibility of the second separated signal is lower than a preset speech intelligibility threshold, the signal amplitude compensation is adjusted by dynamic range compression to obtain the user's speech signal. For example, if the amplitude of the user's speech at 800Hz (within the male fundamental frequency range) is detected to be 4dB lower than the target value (the target value is the average amplitude of user speech in this frequency band, determined by offline statistics), the signal amplitude compensation is set to +4dB.

[0033] The preset speech clarity threshold is set to MOS (Mean Opinion Score) of 3.5 (out of 1-5) through subjective listening tests. This threshold is based on industry call clarity standards (e.g., a MOS score of ≥3.5 for Bluetooth calls can meet daily communication needs). It is determined by averaging the scores of 50 sets of speech signals from different noise scenarios by 20 listeners. If the MOS score of the second separated signal is lower than this value (e.g., a score of 3.2 due to residual low-frequency noise from the coffee machine in a coffee shop scenario), the signal amplitude compensation is adjusted through dynamic range compression (DRC): the speech dynamic range is compressed from 60dB to 40dB (compression ratio of 2:1, which can increase the amplitude of weak signals by 3-5dB in actual tests without significant distortion). During the compression process, a compression factor of 0.8 is used for the user's speech fundamental frequency band (85-255Hz) to avoid changes in timbre caused by excessive compression of the fundamental frequency signal. For example, by boosting the amplitude of the 300Hz frequency band (close to the fundamental frequency band) signal when the user speaks softly by an additional 2dB, a user voice signal with a MOS score of 4.0 or higher can be obtained.

[0034] In summary, through precise spectral decomposition, adaptive filtering, feature extraction, and dynamic optimization, efficient separation of user speech and background voices was achieved. This preserved the clarity of the user speech while suppressing background interference to the greatest extent, providing a high-quality speech signal foundation for subsequent call status analysis.

[0035] In step S103, based on the user's voice signal, the user's call state features are extracted, state judgment and signal analysis are performed, and preliminary parameters for noise reduction depth adjustment and real-time load detection results are obtained.

[0036] In one optional implementation, the step of extracting user call state features based on the user's voice signal, performing state judgment and signal analysis, and obtaining preliminary parameters for noise reduction depth adjustment and real-time load detection results includes: Real-time load detection data is obtained through the built-in sensors in the headphones; Spectral data is obtained from the user's voice signal, and the spectral data is decomposed by short-time Fourier transform to extract the main frequency band energy distribution and obtain the first energy distribution feature. Based on the first energy distribution characteristics, the degree of overlap between the main frequency band energy and the background human voice frequency is calculated to obtain the first overlap characteristic; If the first overlap feature is greater than the preset call state classification threshold, then the first energy distribution feature and the first overlap feature are classified to obtain the first call state classification result. Based on the first call status classification result and combined with the real-time load detection data, the classification threshold is dynamically optimized, and the call status is reclassified to obtain the current call status. Based on the current call status and the first overlap feature, the dynamic gain requirements of the main frequency band energy concentration and noise reduction are analyzed to obtain preliminary parameters for noise reduction depth adjustment, and the real-time load detection results are recorded simultaneously.

[0037] It should be noted that when acquiring real-time load detection data through the built-in sensors in the earphone, the sensors include a processor load monitoring module, a memory usage sensor, and a Bluetooth link bandwidth detector. The sampling interval is set to 100ms, collecting processor load (in %), memory usage (in %), and Bluetooth bandwidth utilization (in %). For example, in a subway call scenario, real-time load data of 68% processor load, 52% memory usage, and 75% bandwidth utilization can be obtained.

[0038] After acquiring spectral data from the user's speech signal, the "time-frequency" energy distribution is obtained through short-time Fourier transform (STFT, 20ms window, Hanning window). The energy proportion of the main frequency band (the core frequency band of user speech, 300-3kHz, which is statistically determined based on the extraction of effective speech segments using Mel-frequency cepstral coefficients (MFCC) from a large number of daily conversations of users of different genders and ages, and can cover the main speech energy distribution range of more than 95% of users' clear conversations) is extracted as the first energy distribution feature. For example, if the user's speech energy proportion is detected to be 65% in the 300-1kHz frequency band and 35% in the 1-3kHz frequency band, then the specific data for this feature is formed.

[0039] When calculating the first overlap feature based on the first energy distribution characteristic, the following steps are first taken: Preliminary noise analysis (in typical scenarios such as offices, subways, and restaurants, a mixture of background noise and human voice signals is collected; the energy weighting method mentioned earlier is used to calculate the background noise weight distribution, determining the main frequency range of background human voice in different scenarios; the background human voice frequency range is dynamically updated in real-time: initially using the default frequency band 500-2kHz, then every 10 seconds the spectral energy distribution of the most recent 10 frames of audio is detected; if the energy change rate of a certain frequency band exceeds 15%, that frequency band is added to the target frequency band; and the background human voice frequency range for each scenario needs to be verified through 100 sets of sample tests to ensure coverage of more than 80% of the background human voice energy distribution in that scenario) determines the frequency range of background human voice (e.g., 500-1.8kHz in an office scenario). Then, the percentage of energy in the overlapping frequency band (500-1.8kHz) of this range relative to the total energy of the main frequency band is calculated. For example, if the overlapping frequency band energy accounts for 22%, then the first overlap feature is 22%.

[0040] The preset call status classification threshold was calibrated experimentally (100 different interference scenarios were selected, including offices, subways, and restaurants; 20 samples of user calls mixed with background noise were collected for each scenario, for a total of 2000 samples. For each sample, "interferenced call" or "clear call" was manually labeled, and the first overlap feature was calculated. Using the labeling result as the tag, it was determined that when the first overlap feature was 20%, the ratio of "interferenced call" to "clear call" samples was close to 1:1, and the classification threshold was set at 20%). If the first overlap feature (22%) was greater than this threshold, The Support Vector Machine (SVM) algorithm is used to classify the first energy distribution feature and the first overlap feature. The input to the SVM algorithm is the first energy distribution feature (energy percentage in the 300-1kHz frequency band and energy percentage in the 1-3kHz frequency band) and the first overlap feature. The training process uses the aforementioned 2000 labeled samples, employing a radial basis function (RBF) as the kernel function. Cross-validation is used to determine the penalty parameter C as 10 and the gamma parameter as 0.1. After training, the classification accuracy is ≥92%, and the output is a first classification result of "interferenced call" or "clear call". For example, in the above scenario, a preliminary judgment of "interferenced call" would be obtained.

[0041] When dynamically optimizing the classification threshold based on the initial classification result and real-time load detection data, if the processor load exceeds 60% (e.g., 68%), the classification threshold is increased to 25% to avoid overcomputation under high load. This adjustment is based on testing 100 high-load scenarios (covering typical scenarios such as subways, shopping malls, and office areas). In these scenarios, when the classification threshold is increased to 25%, the computational load of the classification algorithm can be reduced by about 30%, while the classification accuracy only decreases by about 2%, achieving a good balance between computational efficiency and classification accuracy. During reclassification, because 22% < 25%, the current call status is ultimately "clear call".

[0042] When calculating the preliminary parameters for noise reduction depth based on the current call status and the first overlap feature, the basic dynamic gain requirement for "clear calls" is set to 10dB, and for "interference calls" it is set to 15dB. This basic gain value was determined in a laboratory environment through subjective listening tests and objective indicator (such as signal-to-noise ratio and Mean Opinion Score, MOS) analysis of multiple sets of voice samples with different call qualities. For "clear calls," applying a 10dB gain can improve the MOS score from 3.5 to 4.0 or higher; for "interference calls," applying a 15dB gain can improve the MOS score from approximately 2.8 to 3.8 or higher, meeting the needs for clear calls in daily life. During parameter calculation, the percentage value of the first overlap feature is first converted to a decimal, added to 1, and then multiplied by the basic dynamic gain for the corresponding call status to obtain the preliminary parameters for noise reduction depth adjustment. If real-time load detection data shows a processor load >70%, the gain value is reduced by 10% based on the above calculation results. For example, when there is "clear call" with an overlap of 22% and processor load of 68% (not exceeding 70%), first convert 22% to 0.22, add it to 1 to get 1.22, then multiply it by the 10dB base gain to finally get the preliminary parameter of 12.2dB, and simultaneously record the real-time load detection results (processor 68%, memory 52%, bandwidth 75%).

[0043] In summary, by employing multi-dimensional load monitoring, precise voice feature extraction, and dynamic classification algorithms, real-time judgment of call status and scientific calculation of noise reduction parameters were achieved. This ensured the accuracy of status analysis and avoided resource waste through load linkage optimization, providing a reliable decision-making basis for subsequent resource scheduling.

[0044] In step S104, based on the preliminary parameters of the noise reduction depth adjustment and the real-time load detection results, the resource allocation order and switching frequency are adjusted, and the resource allocation priority and frequency switching delay parameters are determined.

[0045] In one optional implementation, the step of adjusting the resource allocation order and switching frequency based on the preliminary parameters adjusted according to the noise reduction depth and the real-time load detection results, and determining the resource allocation priority and frequency switching delay parameters, includes: Obtain user call duration data; Based on the call duration data, combined with the preliminary parameters of the noise reduction depth adjustment and the real-time load detection results, power consumption requirements are assessed to obtain power consumption constraints. If the power consumption constraint exceeds the preset power consumption constraint threshold, the resource allocation is prioritized to obtain the resource allocation priority. Based on the resource allocation priority, the bandwidth utilization rate is analyzed to obtain bandwidth utilization data, and the link switching frequency is dynamically adjusted to obtain the frequency switching delay parameter.

[0046] It should be noted that when acquiring the user's call duration data, the system clock built into the Bluetooth headset records the call start time and current time, accurate to the second. Simultaneously, it combines this with the battery management module to obtain data on consumed and remaining battery power, forming complete call duration-power consumption correlation data. For example, if the user has been on a call for 18 minutes, has consumed 28% of the battery, and has 72% remaining...

[0047] When assessing power consumption requirements based on call duration data and real-time load detection results, a load-power mapping model must first be constructed. Specifically, the core power consumption modules of the Bluetooth headset (including the noise-canceling chip, Bluetooth communication module, microphone array, and audio processing unit) must be disassembled. The basic power consumption values ​​of each module in idle state (no call, no noise cancellation operation) are recorded, and the load correlation dimensions corresponding to each module are clarified (e.g., processor load corresponds to the computational intensity of the audio processing unit and noise-canceling chip, and Bluetooth link bandwidth utilization corresponds to the data transmission intensity of the Bluetooth module). When clarifying the load correlation dimensions corresponding to each module, a quantitative correlation between load and module power consumption must be established in conjunction with specific calculation methods. For the correlation between processor load and the computational intensity of the audio processing unit and noise-canceling chip, it is calculated through the "percentage of computational instructions": the number of Fourier transform, filtering, and other computational instructions processed per second by the audio processing unit, and the number of noise weight calculation and clock synchronization adjustment instructions per second by the noise-canceling chip are counted. The percentage of the sum of these two to the total computational capacity of the processor per second is the processor load. Regarding the relationship between Bluetooth link bandwidth utilization and Bluetooth module data transmission strength, it is calculated based on the "actual transmission rate ratio": the total rate of real-time transmission of user voice signals (e.g., 64kbps) and noise reduction parameter configuration data (e.g., 2kbps) by the Bluetooth module is divided by the maximum transmission rate of the Bluetooth link (e.g., 2Mbps), and then multiplied by 100%, i.e., (8+2) / 2000×100%=0.5%, corresponding to a Bluetooth link bandwidth utilization of 0.5%, thus achieving a quantitative explanation of the load correlation dimension. Subsequently, in a controlled laboratory environment (e.g., 25℃, no electromagnetic interference), the load parameters were gradually adjusted by controlling a single variable (e.g., adjusting the processor load from 10% to 100% in 10% increments, while keeping other parameters such as Bluetooth bandwidth utilization and microphone sampling rate stable). Real-time power consumption under each load gradient was recorded using a high-precision power monitoring device. The load-power consumption correlation coefficient was determined through data statistical analysis (e.g., obtaining a coefficient that the power consumption increases by 8mA for every 10% increase in processor load), and a complete load-power consumption mapping model was established. Combining this model with the real-time load detection results (e.g., the current processor load is 65%), the power consumption required for the remaining call was calculated and predicted (e.g., 22mA). Finally, the power consumption constraint was formed by the ratio of the total power consumption requirement to the remaining battery capacity.

[0048] The preset power consumption constraint threshold is usually set to 40% of the remaining battery capacity (reserve emergency power). If the assessed power consumption demand (e.g., 32mA·h) exceeds this threshold (e.g., if the threshold is 40mA·h when the remaining battery capacity is 100mA·h, it will not trigger; if the remaining capacity is 50mA·h, the threshold is 20mA·h, it will trigger), the resource allocation is prioritized according to its core importance: Level 1 (user voice signal processing), Level 2 (noise reduction parameter calculation), Level 3 (Bluetooth link maintenance), Level 4 (non-essential functions such as LED indicators), thus obtaining the resource allocation priority.

[0049] When analyzing bandwidth utilization based on resource allocation priorities, the Bluetooth bandwidth usage of each priority task is calculated (e.g., Level 1 50%, Level 2 20%, Level 3 15%), and the sum is the bandwidth utilization data. When bandwidth utilization exceeds 80%, a preset warning value of 80% is set based on the bandwidth threshold for stable Bluetooth module transmission (to avoid data packet loss). When bandwidth utilization exceeds this value, the link switching frequency is reduced from the default 100 times / second to 80 times / second, and the frequency switching latency parameter is calculated as "1 divided by the switching frequency", i.e., 12.5ms. For example, in a multi-person conference scenario, when bandwidth utilization reaches 85%, link overhead is reduced by decreasing the switching frequency.

[0050] In summary, by evaluating the power consumption in conjunction with call duration and load, and by combining dynamic resource sorting and bandwidth optimization, precise control of resource allocation was achieved. This balanced power consumption constraints while meeting the requirements for noise reduction depth, thus providing resource assurance for the stable operation of the device.

[0051] In step S105, based on the resource allocation priority and the frequency switching delay parameter, and combined with the user call status characteristics, the clock synchronization of the noise reduction chip is adjusted and optimized to obtain the adjusted chip clock frequency and power consumption allocation rules.

[0052] In one optional implementation, the step of adjusting and optimizing the clock synchronization of the noise reduction chip based on the resource allocation priority and the frequency switching delay parameter, combined with the user call state characteristics, to obtain the adjusted chip clock frequency and power consumption allocation rules, includes: Based on the resource allocation priority and combined with the main frequency band energy distribution in the user call status characteristics, calculate the energy distribution data and perform dynamic frequency adjustment to generate frequency adjustment parameters and determine the chip clock synchronization scheme. Based on the chip clock synchronization scheme and the frequency adjustment parameters, the frequency switching delay is optimized to obtain the adjusted chip clock frequency. Based on the adjusted chip clock frequency, power allocation data is calculated, and signal quality is adjusted in conjunction with the noise reduction chip hardware operating parameters and audio signal processing requirements to determine power allocation rules.

[0053] It should be noted that when calculating energy distribution data based on resource allocation priority (user voice signal processing task is the highest) and the main frequency band energy distribution in the user call status characteristics, the energy proportion of each sub-band of the main frequency band is used as the basis (e.g., 300-1kHz accounts for 60%, 1-3kHz accounts for 40%). The sub-bands of the main frequency band (300-3kHz, covering more than 95% of the main voice energy for clear user calls) are divided according to voice signal characteristics: 300-1kHz is the low-frequency sub-band (containing the user's voice fundamental tone and main formants), and 1-3kHz is the mid-high frequency sub-band (containing key harmonic components for voice intelligibility), based on the energy proportion of each sub-band (e.g., 300-1kHz accounts for 60%, 1-3kHz accounts for 40%). The dynamic frequency adjustment parameters dynamically query a pre-calibrated "Noise Scene - Optimal Operating Frequency" lookup table (LUT) based on the complexity of the current noise environment and real-time load, selecting the chip operating frequency that meets the requirements of real-time processing and optimal power consumption. The "Noise Scene - Optimal Operating Frequency" lookup table (LUT) is determined by laboratory testing. The chip clock synchronization solution uses PLL phase-locked loop technology to synchronize the noise reduction chip clock with the Bluetooth master clock (32MHz), with the synchronization error controlled within ±10ppm, ensuring consistent data processing timing.

[0054] When optimizing frequency switching delay based on chip clock synchronization scheme and frequency adjustment parameters, the default switching delay is 5ms. If the synchronization error is less than 15ppm and the frequency adjustment parameters are in the 120-140MHz range, the delay can be compressed to 3ms. This setting is based on laboratory test data: within this synchronization error and frequency range, the chip's internal timing deviation is small, the data transmission bit error rate is less than 0.001%, and signal integrity can still be guaranteed after delay compression. At the same time, the chip circuit response speed within this frequency range is adapted to the 3ms delay, and there will be no signal processing interruption due to the shortened delay, thereby reducing signal distortion during frequency switching. For example, with a synchronization error of 8ppm under 130MHz adjustment parameters, the switching delay is optimized to 3ms, ultimately resulting in a chip clock frequency of 130MHz±5MHz (allowing ±5MHz dynamic fluctuation to adapt to load changes).

[0055] When calculating power consumption allocation data based on the adjusted chip clock frequency, a frequency-power linear model needs to be constructed first. Specifically, the model parameters are determined through laboratory hardware testing: At different clock frequencies of the noise reduction chip (gradually increasing from 80MHz to 150MHz, with each 10MHz increment representing a test gradient), while keeping other chip operating parameters stable, the real-time operating current corresponding to each frequency is recorded using a high-precision current monitoring device. After multiple tests and data fitting, a linear relationship is derived: "100MHz corresponds to 20mA, and power consumption increases by 2mA for every 10MHz increase." This establishes the frequency-power linear model. When applying this model subsequently, if the adjusted chip clock frequency is 130MHz, the corresponding power consumption calculated through the model is 26mA (100MHz corresponds to 20mA). The 130MHz frequency increases by 30MHz compared to 100MHz, resulting in an additional power consumption of 3×2mA=6mA, or 20mA+6mA=26mA. Combined with optimization methods for the noise reduction chip (such as using a 128th-order filter to enhance suppression of low-frequency noise below 300Hz), this 26mA of power consumption is allocated according to resource priority (voice processing module > noise reduction module > link maintenance > reserved). Specifically, the voice processing module accounts for 50% (13mA), the noise reduction module for 30% (7.8mA), link maintenance for 15% (3.9mA), and the reserved power supply for 5% (1.3mA), forming a clear power allocation rule. For example, in noisy street scenarios, this rule prioritizes power supply to the noise reduction module, ensuring the effective suppression of low-frequency noise by the 128th-order filter and maintaining the stable operation of the core noise reduction algorithm.

[0056] In summary, by prioritizing resources through frequency adjustment, high-precision clock synchronization, and refined power consumption allocation, the optimal balance between noise reduction chip performance and energy consumption is achieved. This not only meets the processing needs of different call scenarios but also avoids resource waste through a dynamic adjustment mechanism, providing hardware-level support for the long-term operation of the device.

[0057] In step S106, based on the adjusted chip clock frequency and the power consumption allocation rule, the power consumption curve and real-time status feedback data are obtained through current monitoring, temperature threshold control and noise analysis.

[0058] In one optional implementation, the step of obtaining the power consumption curve and real-time status feedback data based on the adjusted chip clock frequency and power allocation rules, through current carrying capacity monitoring, temperature threshold control, and noise analysis, includes: Acquire chip temperature sensor data and overload current monitoring data; Based on the adjusted chip clock frequency and the power consumption allocation rule, the overlap between the chip temperature sensor data and the background human voice frequency is analyzed to generate the initial frequency adjustment amplitude and obtain the chip temperature management status. The chip temperature sensor data is compared with a preset chip safety temperature threshold. If the temperature exceeds the safety temperature threshold, the initial frequency adjustment amplitude is reduced to obtain the temperature-optimized frequency adjustment amplitude. Based on the frequency adjustment amplitude optimized by the temperature, noise suppression processing is performed on the overlap of the background human voice frequencies to optimize the noise reduction depth and obtain real-time status feedback data. The overload current monitoring data is compared with the preset chip safe current threshold. If the monitored current exceeds the safe current threshold, the frequency adjustment range after temperature optimization is further reduced to obtain the final frequency adjustment range. Based on the final frequency adjustment amplitude, the power consumption per unit time is calculated in conjunction with the power consumption allocation rules. Power consumption data is recorded in time series to generate a power consumption curve. The power consumption curve and the real-time status feedback data are then integrated as the result.

[0059] It should be noted that when acquiring chip temperature sensor data and overload current monitoring data, temperature data is collected through the thermistor built into the noise reduction chip (sampling interval 500ms), in °C; current data is collected through a precision sampling resistor connected in series in the power supply circuit (sampling interval 100ms), in mA (Bluetooth headset operating current is typically in the milliampere range). For example, after 30 minutes of continuous talk time, monitoring data of chip temperature 42°C and operating current 60mA can be obtained.

[0060] When analyzing the overlap between chip temperature data and background voice frequencies based on the adjusted chip clock frequency (e.g., 130MHz) and power allocation rules, the overlap ratio between the background voice and the user's main voice frequency band is first calculated (e.g., 45%). Combined with the power consumption level (26mA) at 130MHz, an initial frequency adjustment range is generated. Specifically, a base frequency value is first set (generally the current chip clock frequency), and then dynamically adjusted according to the overlap ratio between the background voice and the user's main voice frequency band: a higher overlap ratio indicates a greater need for processing, so the corresponding frequency value is increased from the base frequency. Simultaneously, the current power consumption level is considered to ensure it is within a reasonable range, guaranteeing that the power consumption after increasing the frequency does not exceed a preset threshold. Finally, the initial frequency adjustment range is derived. The specific adjustment relationship needs to be tested beforehand to assess the balance between processing capability and power consumption limits under different adjustment relationships. For example, data experiments have shown that increasing the frequency by 2MHz for every 10% overlap ratio can effectively achieve processing differentiation while maintaining an acceptable power consumption level. If the chip temperature is within a safe range (e.g., 42°C < 50°C threshold), the initial value of 130MHz is maintained, and the chip temperature management status is marked as "normal".

[0061] The preset safe temperature threshold for the chip is calibrated according to the chip model, typically 50°C (normal scenarios) or 55°C (high-temperature environments). When the chip temperature exceeds the safe threshold, a temperature control strategy is activated. This strategy, based on the chip's thermal model and real-time power consumption, uses a proportional-integral (PI) controller to dynamically reduce the operating frequency, stabilizing the temperature within a safe range. Similarly, when the operating current exceeds the threshold, an overcurrent protection strategy is activated, prioritizing reducing the complexity of processing tasks or the operating frequency to ensure the current returns to normal levels.

[0062] The power consumption per unit time is calculated based on the final frequency adjustment amplitude and power allocation rules, generating a power consumption curve. For example, if the final frequency adjustment amplitude after adjustment based on temperature conditions and current limits is 104MHz, the power consumption per unit time is calculated using the power allocation rules. This power consumption is recorded in a time series of 1 minute / data point, with the horizontal axis representing call duration and the vertical axis representing cumulative power consumption, generating a power consumption curve. This curve is then integrated with real-time status feedback data (temperature 49°C, current 95mA, signal-to-noise ratio 25dB) to form a complete status monitoring result.

[0063] In summary, by implementing dual-dimensional hardware protection based on temperature and current, dynamic frequency adjustment, and noise suppression optimization, a balance between safe operation and noise reduction effect is achieved. The generated power curve and status feedback data provide real-time basis for precise control of the final noise reduction output, ensuring the stability and security of calls in complex environments.

[0064] In step S107, a status analysis is performed based on the power consumption curve and real-time status feedback. If an overload interruption is detected and the temperature rise rate exceeds the preset temperature change threshold, the protection response speed and noise reduction depth are adjusted to obtain the noise reduction output of the Bluetooth headset.

[0065] In one optional implementation, the step of performing state analysis based on the power consumption curve and real-time status feedback, and adjusting the protection response speed and noise reduction depth to obtain the noise reduction output of the Bluetooth headset if an overload interruption is detected and the temperature rise rate exceeds a preset temperature change threshold, includes: Based on the power consumption curve and the real-time status feedback data, collect device status parameters and obtain overload interruption trigger signal and temperature rise rate data. If an overload interruption is detected and the rate of temperature rise exceeds the preset temperature change threshold, the protection response speed is adjusted by using a dynamic threshold to obtain the adjusted response speed parameters. Based on the adjusted response speed parameters, the noise reduction depth is adjusted to obtain the optimized noise reduction depth parameters. By optimizing the noise reduction depth parameters and data stream stability processing, the signal processing is optimized to obtain the final noise reduction output of the Bluetooth headset.

[0066] It should be noted that when collecting device status parameters based on power consumption curves and real-time status feedback data, the focus is on extracting overload interruption trigger signals (generated when the current exceeds the safety threshold of 1.0A for 100ms) and temperature rise rate data (calculated through 5 consecutive temperature samples, in °C / s). For example, in a sudden strong noise scenario in a subway, overload interruption trigger signals and a temperature rise rate of 1.2°C / s can be collected.

[0067] The preset temperature change threshold is set to 0.8°C / s through reliability testing. This threshold is based on high-temperature environment testing calibration (test conditions: ambient temperature 35±2°C, noise source 75dB steady-state traffic noise, sampling period 60 seconds). A protection response is triggered when the temperature rise rate exceeds the threshold three consecutive times. If an overload interruption is detected simultaneously and the temperature rise rate (1.2°C / s) exceeds the threshold, a dynamic threshold judgment algorithm is activated—adjusting the protection response speed based on the slope of the power consumption curve (e.g., 5mA·h / min, representing a rapid increase in power consumption), compressing the default 200ms response time to 100ms, resulting in the adjusted response speed parameter. For example, during long outdoor calls in summer, this mechanism can quickly respond to overheating risks.

[0068] When adjusting the noise reduction depth based on the adjusted response speed parameter (100ms), a "response speed-noise reduction depth" linkage model needs to be constructed first. Specifically, the model association rules are determined through laboratory scenario testing: under different call environments (such as quiet indoors, noisy streets, and subway commuting), the protection response speed is gradually adjusted (from 200ms to 50ms, with each 50ms increment representing a gradient). Simultaneously, the optimal value of noise reduction depth for different response speeds is monitored using audio analysis equipment. After multiple tests and verifications, the correlation relationship of "for every 50ms increase in response speed, the noise reduction depth can increase by 2dB" is obtained, and the noise reduction depth is set accordingly. The maximum noise reduction depth threshold is 30dB to avoid excessive noise reduction causing speech distortion. This establishes a complete "response speed-noise reduction depth" linkage model. When applying this model, if the adjusted response speed parameter is 100ms, compared with the default response speed (assuming the default response speed is 150ms, 150ms to 100ms corresponds to a 50ms increase in response speed), the noise reduction depth can be increased by 2dB according to the model rules. If the original noise reduction depth is 25dB, and 25dB + 2dB = 27dB does not exceed the maximum threshold of 30dB, then the optimized noise reduction depth parameter can be obtained as 27dB.

[0069] When processing data stream stability using the optimized noise reduction depth parameter (27dB), a data packet compression algorithm (compression ratio 1.5:1) is employed to reduce data transmission volume, lowering the processing latency from 50ms to 30ms. Simultaneously, signal processing is optimized: for the core audio band of 300-3kHz, the current environmental noise type (such as steady-state noise or sudden noise) is first identified through spectrum analysis, and then the filtering parameters are dynamically adjusted—if steady-state noise (such as air conditioning noise) is detected, the filtering intensity of the low-frequency band (300-1kHz) is automatically increased while the attenuation of the high-frequency band (1-3kHz) is reduced; if sudden noise (such as crowd noise) is detected, the real-time monitoring frequency across the entire frequency band is temporarily enhanced, and the filtering coefficients are updated every 10ms to ensure that voice details are preserved while suppressing noise. The final output Bluetooth headset noise-reduced signal achieves a signal-to-noise ratio ≥28dB and a speech clarity MOS score ≥4.2, enabling clear calls even in noisy shopping mall environments.

[0070] In summary, by monitoring overload and temperature rise, dynamically accelerating the protection response, and precisely controlling the noise reduction depth, the system achieves coordinated protection of equipment safety and call quality. The final output noise-reduced signal not only meets the need for clear calls in complex environments, but also ensures the long-term stable operation of the equipment through a multi-layered protection mechanism.

[0071] Reference Figure 2 The second embodiment of the present invention provides a Bluetooth headset active noise cancellation device for calls, comprising: The environmental feature acquisition module is used to obtain the preliminary distribution of environmental noise weights and signal processing efficiency; The speech separation processing module is used to separate user speech from background human voice based on the preliminary distribution of environmental noise weights and signal processing efficiency to obtain the user speech signal. The call status analysis module is used to extract user call status features based on the user's voice signal, perform status judgment and signal analysis, and obtain preliminary parameters for noise reduction depth adjustment and real-time load detection results. The resource scheduling module is used to adjust the resource allocation order and switching frequency based on the preliminary parameters of the noise reduction depth adjustment and the real-time load detection results, and to determine the resource allocation priority and frequency switching delay parameters. The chip optimization module is used to adjust and optimize the clock synchronization of the noise reduction chip based on the resource allocation priority and the frequency switching delay parameter, combined with the user call status characteristics, to obtain the adjusted chip clock frequency and power consumption allocation rules. The status feedback module is used to obtain the power consumption curve and real-time status feedback data based on the adjusted chip clock frequency and the power consumption allocation rules, through current monitoring, chip temperature threshold control and noise analysis. The noise reduction output module is used to perform state analysis based on the power consumption curve and real-time status feedback data. If an overload interruption is detected and the temperature rise rate exceeds the preset temperature change threshold, the protection response speed for overload interruption and abnormal temperature rise, as well as the noise reduction depth, are adjusted to obtain the noise reduction output of the Bluetooth headset.

[0072] It should be noted that the Bluetooth headset active noise cancellation device provided in this embodiment of the invention is used to execute all the process steps of the Bluetooth headset active noise cancellation method in the above embodiment. The working principle and beneficial effects of the two are one-to-one, so they will not be described again.

[0073] This invention also provides an electronic device. The electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a Bluetooth headset active noise cancellation program. When the processor executes the computer program, it implements the steps in the above-described embodiments of the Bluetooth headset active noise cancellation method, for example... Figure 1 The step S101 shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above-described device embodiments, such as the environmental feature acquisition module.

[0074] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0075] The electronic device may be a desktop computer, laptop, handheld computer, or smart tablet, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. It may include more or fewer components than described above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0076] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting all parts of the electronic device via various interfaces and lines.

[0077] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0078] Wherein, if the modules / units integrated in the electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0079] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0080] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for active noise cancellation during calls in a Bluetooth headset, characterized in that, include: Obtain the preliminary distribution of environmental noise weights and signal processing efficiency; Based on the preliminary distribution of environmental noise weights and signal processing efficiency, user speech and background voice are separated to obtain the user speech signal; Based on the user's voice signal, extract the user's call status features, perform status judgment and signal analysis, and obtain preliminary parameters for noise reduction depth adjustment and real-time load detection results. Based on the preliminary parameters of the noise reduction depth adjustment and the real-time load detection results, adjust the resource allocation order and switching frequency, and determine the resource allocation priority and frequency switching delay parameters. Based on the resource allocation priority and the frequency switching delay parameter, combined with the user call status characteristics, the clock synchronization of the noise reduction chip is adjusted and optimized to obtain the adjusted chip clock frequency and power consumption allocation rules. Based on the adjusted chip clock frequency and the power consumption allocation rule, the power consumption curve and real-time status feedback data are obtained through current carrying capacity monitoring, chip temperature threshold control and noise analysis. Based on the power consumption curve and real-time status feedback data, status analysis is performed. If an overload interruption is detected and the temperature rise rate exceeds the preset temperature change threshold, the overload interruption and abnormal temperature rise protection response speed and noise reduction depth are adjusted to obtain the noise reduction output of the Bluetooth headset.

2. The active noise cancellation method for Bluetooth headset calls according to claim 1, characterized in that, The preliminary distribution of environmental noise weights and signal processing efficiency is obtained, including: Ambient audio data is collected through the microphone array of a Bluetooth headset, and the ambient audio data is digitally processed using a preset sampling rate to obtain the original audio data. The original audio data is decomposed into time and frequency using short-time Fourier transform to generate the spectral distribution of the audio signal; If there are frequency components in the spectrum distribution that are higher than the preset spectrum component threshold, the environmental noise weight is obtained by weighted calculation. By jointly analyzing the environmental noise weights and the spectral distribution, and adjusting the signal processing parameters based on the original audio time-frequency characteristics, a preliminary distribution of signal processing efficiency is obtained.

3. The active noise cancellation method for Bluetooth headset calls according to claim 1, characterized in that, The step of separating user speech from background voice based on the preliminary distribution of environmental noise weights and signal processing efficiency to obtain the user speech signal includes: The input audio is acquired and decomposed into a first spectral signal through Fourier transform. Based on the preliminary distribution of environmental noise weights and signal processing efficiency, the target frequency band for noise feature extraction is determined. The noise features are extracted from the target frequency band of the first spectral signal using short-time Fourier transform to obtain the first noise weight; Based on the first noise weight and referring to the preliminary distribution of the signal processing efficiency, the filtering parameters are adjusted to filter the first spectral signal to obtain the first separated signal; Extract user speech features from the first separated signal, separate background human voice, and obtain a second separated signal containing signal amplitude compensation. If the speech clarity of the second separated signal is lower than a preset speech clarity threshold, the signal amplitude compensation amount is adjusted by dynamic range compression to obtain the user's speech signal.

4. The active noise cancellation method for Bluetooth headset calls according to claim 1, characterized in that, The step of extracting user call state features based on the user's voice signal, performing state judgment and signal analysis, and obtaining preliminary parameters for noise reduction depth adjustment and real-time load detection results includes: Real-time load detection data is obtained through the built-in sensors in the headphones; Spectral data is obtained from the user's voice signal, and the spectral data is decomposed by short-time Fourier transform to extract the main frequency band energy distribution and obtain the first energy distribution feature. Based on the first energy distribution characteristics, the degree of overlap between the main frequency band energy and the background human voice frequency is calculated to obtain the first overlap characteristic; If the first overlap feature is greater than the preset call state classification threshold, then the first energy distribution feature and the first overlap feature are classified to obtain the first call state classification result. Based on the first call status classification result and combined with the real-time load detection data, the classification threshold is dynamically optimized, and the call status is reclassified to obtain the current call status. Based on the current call status and the first overlap feature, the dynamic gain requirements of the main frequency band energy concentration and noise reduction are analyzed to obtain preliminary parameters for noise reduction depth adjustment, and the real-time load detection results are recorded simultaneously.

5. The active noise cancellation method for Bluetooth headsets according to claim 1, characterized in that, The process of adjusting the resource allocation order and switching frequency based on the preliminary parameters adjusted according to the noise reduction depth and the real-time load detection results, and determining the resource allocation priority and frequency switching delay parameters, includes: Obtain user call duration data; Based on the call duration data, combined with the preliminary parameters of the noise reduction depth adjustment and the real-time load detection results, power consumption requirements are assessed to obtain power consumption constraints. If the power consumption constraint exceeds the preset power consumption constraint threshold, the resource allocation is prioritized to obtain the resource allocation priority. Based on the resource allocation priority, the bandwidth utilization rate is analyzed to obtain bandwidth utilization data, and the link switching frequency is dynamically adjusted to obtain the frequency switching delay parameter.

6. The active noise cancellation method for Bluetooth headset calls according to claim 1, characterized in that, The step of adjusting and optimizing the clock synchronization of the noise reduction chip based on the resource allocation priority, the frequency switching delay parameter, and the user call state characteristics, to obtain the adjusted chip clock frequency and power consumption allocation rules, includes: Based on the resource allocation priority and combined with the main frequency band energy distribution in the user call status characteristics, calculate the energy distribution data and perform dynamic frequency adjustment to generate frequency adjustment parameters and determine the chip clock synchronization scheme. Based on the chip clock synchronization scheme and the frequency adjustment parameters, the frequency switching delay is optimized to obtain the adjusted chip clock frequency. Based on the adjusted chip clock frequency, power allocation data is calculated, and signal quality is adjusted in conjunction with the noise reduction chip hardware operating parameters and audio signal processing requirements to determine power allocation rules.

7. The active noise cancellation method for Bluetooth headsets according to claim 4, characterized in that, Based on the adjusted chip clock frequency and the power consumption allocation rules, and through current carrying capacity monitoring, temperature threshold control, and noise analysis, the power consumption curve and real-time status feedback data are obtained, including: Acquire chip temperature sensor data and overload current monitoring data; Based on the adjusted chip clock frequency and the power consumption allocation rule, the overlap between the frequency of the chip temperature sensor data and the background human voice frequency is analyzed to generate the initial frequency adjustment amplitude and obtain the chip temperature sensor data. The chip temperature sensor data is compared with a preset chip safety temperature threshold. If the temperature exceeds the safety temperature threshold, the initial frequency adjustment amplitude is reduced to obtain the temperature-optimized frequency adjustment amplitude. Based on the frequency adjustment amplitude optimized by the temperature, noise suppression processing is performed on the overlap of the background human voice frequencies to optimize the noise reduction depth and obtain real-time status feedback data. The overload current monitoring data is compared with the preset chip safe current threshold. If the monitored current exceeds the chip safe current threshold, the frequency adjustment range after temperature optimization is further reduced to obtain the final frequency adjustment range. Based on the final frequency adjustment amplitude, the power consumption per unit time is calculated in conjunction with the power consumption allocation rules. Power consumption data is recorded in time series to generate a power consumption curve. The power consumption curve and the real-time status feedback data are then integrated as the result.

8. The active noise cancellation method for Bluetooth headsets according to claim 1, characterized in that, The process involves analyzing the state based on the power consumption curve and real-time status feedback data. If an overload interruption is detected and the temperature rise rate exceeds a preset temperature change threshold, the protection response speed and noise reduction depth are adjusted to obtain the noise reduction output of the Bluetooth headset, including: Based on the power consumption curve and the real-time status feedback data, collect device status parameters and obtain overload interruption trigger signal and temperature rise rate data. If an overload interruption is detected and the rate of temperature rise exceeds the preset temperature change threshold, the protection response speed is adjusted by using a dynamic threshold to obtain the adjusted response speed parameters. Based on the adjusted response speed parameters, the noise reduction depth is adjusted to obtain the optimized noise reduction depth parameters. By optimizing the noise reduction depth parameters and data stream stability processing, the signal processing efficiency is improved, resulting in the final noise reduction output of the Bluetooth headset.

9. An active noise cancellation device for Bluetooth headset calls, characterized in that, include: The environmental feature acquisition module is used to obtain the preliminary distribution of environmental noise weights and signal processing efficiency; The speech separation processing module is used to separate user speech from background human voice based on the preliminary distribution of environmental noise weights and signal processing efficiency to obtain the user speech signal. The call status analysis module is used to extract user call status features based on the user's voice signal, perform status judgment and signal analysis, and obtain preliminary parameters for noise reduction depth adjustment and real-time load detection results. The resource scheduling module is used to adjust the resource allocation order and switching frequency based on the preliminary parameters of the noise reduction depth adjustment and the real-time load detection results, and to determine the resource allocation priority and frequency switching delay parameters. The chip optimization module is used to adjust and optimize the clock synchronization of the noise reduction chip based on the resource allocation priority and the frequency switching delay parameter, combined with the user call status characteristics, to obtain the adjusted chip clock frequency and power consumption allocation rules. The status feedback module is used to obtain the power consumption curve and real-time status feedback data based on the adjusted chip clock frequency and the power consumption allocation rules, through current monitoring, chip temperature threshold control and noise analysis. The noise reduction output module is used to perform state analysis based on the power consumption curve and real-time status feedback data. If an overload interruption is detected and the temperature rise rate exceeds the preset temperature change threshold, the protection response speed for overload interruption and abnormal temperature rise, as well as the noise reduction depth, are adjusted to obtain the noise reduction output of the Bluetooth headset.