Wireless earphone and intelligent voice control system for wireless earphone

Abstract

The invention relates to the technical field of earphones, in particular to a wireless earphone and an intelligent voice control system for the wireless earphone, which comprises a multi-microphone array module, a voiceprint processing module, a motion compensation module, a dynamic frequency control module, a wireless communication module and an earphone main control chip. Through the multi-microphone array module and the motion compensation module, noise is suppressed, sound field distortion is corrected, and voice signal quality is improved. And the voiceprint processing module generates a voiceprint confidence score, so that false triggering is avoided, and the security and accuracy of voiceprint recognition are enhanced. And the dynamic frequency control module adjusts the frequency of the processor in real time according to environmental conditions and user behaviors, and dynamically adjusts performance and power consumption. Efficient interaction between the voiceprint processing module and the dynamic frequency control module is realized by adopting an asynchronous communication mechanism, the mode is switched to a low-power-consumption mode when high-performance processing is not needed, and the low-power-consumption mode is optimized. All the modules work cooperatively to dynamically adapt to user behaviors and environmental conditions, and intelligence and real-time performance are enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of earphones, and in particular to a wireless earphone and an intelligent voice control system for the wireless earphone. Background Art

[0002] With the widespread use of wireless headphones, the performance requirements of headphones continue to increase, especially in voice recognition, environmental adaptability and battery life.

[0003] There is often a contradiction between the speech recognition accuracy and battery life of traditional wireless headphones, especially in noisy environments or when the wearer exercises frequently. How to improve speech recognition accuracy and extend the use time of the headphones without increasing the battery burden is an important challenge in existing technologies, so it is necessary to improve it. Summary of the invention

[0004] The purpose of the present invention is to address the deficiencies of the prior art and provide a wireless headset and an intelligent voice control system for wireless headsets, which suppresses noise and corrects sound field distortion through a multi-microphone array module and a motion compensation module to improve the quality of voice signals. The voiceprint confidence score is generated through a voiceprint processing module to avoid false triggering and enhance the security and accuracy of voiceprint recognition. The processor frequency is adjusted in real time according to environmental conditions and user behavior through a dynamic frequency control module to dynamically adjust performance and power consumption. An asynchronous communication mechanism is used to achieve efficient interaction between the voiceprint processing module and the dynamic frequency control module, and switch to a low power consumption mode when high-performance processing is not required to optimize the low power consumption mode. The modules work together to dynamically adapt to user behavior and environmental conditions to enhance intelligence and real-time performance.

[0005] To achieve the above-mentioned purpose, a wireless headset and an intelligent voice control system for a wireless headset of the present invention include a multi-microphone array module, a voiceprint processing module, a motion compensation module, a dynamic frequency control module, a wireless communication module and a headset main control chip, wherein the multi-microphone array module, the voiceprint processing module, the motion compensation module, the dynamic frequency control module and the wireless communication module are all connected to the headset main control chip;

[0006] A multi-microphone array module is used to collect audio signals and output beamforming voice signals to suppress environmental noise; a voiceprint processing module is used to detect valid voice segments in voice signals and generate voiceprint confidence scores for valid voice segments; a motion compensation module is connected to the nine-axis inertial sensor, and calculates the time compensation amount ΔT and the space compensation parameter θ through the motion compensation parameters collected by the nine-axis inertial sensor to correct the sound field distortion caused by the earphone wearing offset; a dynamic frequency control module is used to adjust the processor core frequency of the earphone main control chip in real time according to the environmental noise signal-to-noise ratio, the remaining battery power of the earphone and the motion compensation parameters to balance performance and power consumption; a wireless communication module is used to transmit data with external devices through the Bluetooth low energy protocol; the voiceprint processing module and the dynamic frequency control module interact through an asynchronous communication mechanism. When the voiceprint confidence score exceeds a preset threshold, the dynamic frequency control module is triggered to switch the processor core frequency of the earphone main control chip to a low power mode.

[0007] Beneficial effects of the present invention: The present invention suppresses noise and corrects sound field distortion through a multi-microphone array module and a motion compensation module, thereby improving the quality of voice signals. The voiceprint confidence score is generated through the voiceprint processing module to avoid false triggering and enhance the security and accuracy of voiceprint recognition. The processor frequency is adjusted in real time according to environmental conditions and user behavior through the dynamic frequency control module, and the performance and power consumption are dynamically adjusted. The asynchronous communication mechanism is used to achieve efficient interaction between the voiceprint processing module and the dynamic frequency control module, and switch to low power consumption mode when high-performance processing is not required, thereby optimizing the low power consumption mode. The modules work together to dynamically adapt to user behavior and environmental conditions, thereby enhancing intelligence and real-time performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 It is a schematic diagram of the functional modules of the present invention.

[0009] Figure 2 It is a schematic diagram of the structure of the composite sound filtering cotton of the present invention.

[0010] Reference numerals include:

[0011] 1. First sound-absorbing cotton; 2. Interlayer; 3. Second sound-absorbing cotton; 4. Sound-conducting gap. DETAILED DESCRIPTION

[0012] The present invention is described in detail below with reference to the accompanying drawings.

[0013] like Figure 1 to Figure 2As shown, a wireless headset and an intelligent voice control system for a wireless headset of the present invention include a multi-microphone array module, a voiceprint processing module, a motion compensation module, a dynamic frequency control module, a wireless communication module and a headset main control chip. The multi-microphone array module, the voiceprint processing module, the motion compensation module, the dynamic frequency control module and the wireless communication module are all connected to the headset main control chip, so that the headset main control chip receives data from each module and coordinates the operation of each module.

[0014] The multi-microphone array module is used to collect audio signals and output beamforming voice signals to suppress ambient noise. Using an array of multiple microphones, beamforming technology can be used to enhance voice signals from a specific direction (such as the direction of the user's mouth) while suppressing signals from other directions (such as ambient noise).

[0015] Improve the clarity of voice collection and reduce the interference of environmental noise, so that high-quality voice signals can be obtained even in noisy environments.

[0016] Preferably, the collection end of the multi-microphone array module is provided with a composite sound filter cotton, the composite sound filter cotton includes a first sound-absorbing cotton 1, an interlayer 2 and a second sound-absorbing cotton 3 arranged in sequence, the second sound-absorbing cotton 3 is attached to the collection end of the multi-microphone array module, the composite sound filter cotton is provided with a sound guide gap 4, the sound guide gap 4 is arranged through the first sound-absorbing cotton 1, the interlayer 2 and the second sound-absorbing cotton 3, and the spacing of the sound guide gap 4 of the interlayer 2 is greater than the spacing of the sound guide gap 4 of the first sound-absorbing cotton 1 and the second sound-absorbing cotton 3, and the spacing of the sound guide gap 4 of the first sound-absorbing cotton 1 is greater than the spacing of the sound guide gap 4 of the second sound-absorbing cotton 3.

[0017] The first sound-absorbing cotton 1 can be open-cell polyurethane foam, melamine resin foam or gradient density fiber cotton. The interlayer 2 can be aluminum honeycomb, plastic honeycomb, damping rubber, damping silica gel, polymer damping material. The second sound-absorbing cotton 3 can be high-density polyester fiber cotton, glass fiber cotton or acoustic damping felt.

[0018] The first sound-absorbing cotton 1 serves as the outermost barrier. It absorbs high-frequency sound waves (such as environmental noise and wind noise) through a porous fiber structure and uses the principle of converting sound energy into heat energy to attenuate the sound wave energy directly incident on the microphone. This reduces the direct interference of environmental noise (such as keyboard sound and friction sound) on the microphone. It also avoids the sound wave reflection on the microphone housing or the surface of the collection end causing signal distortion.

[0019] The interlayer 2 forms an acoustic impedance mutation interface through the air gap. When the sound wave passes through the first sound-absorbing cotton 1 and enters the cavity, due to the difference in acoustic impedance between air and solid materials, part of the sound wave is reflected back to the first sound-absorbing cotton 1 for further absorption, and the remaining sound wave is reflected multiple times in the cavity and gradually attenuated. The cavity resonance effect can absorb medium and low frequency noise (such as air conditioning sound, low frequency vibration of traffic) and reduce the interference of external mechanical vibration on the microphone.

[0020] The second sound-absorbing cotton 3 further absorbs the residual sound waves that penetrate the interlayer 2, especially the mid-high frequency components, through a denser fiber structure or gradient density design, while blocking the indirect noise from the external environment (such as wall reflection sound). Together with the first sound-absorbing cotton 1, it forms a coarse and fine filter double-layer structure to enhance the full-band noise reduction capability. It blocks the sound waves inside the microphone from radiating outwards to avoid the formation of a feedback loop (such as howling).

[0021] The sound-conducting gap 4 provides a controllable propagation path for the sound waves, causing the incident sound waves to be refracted, reflected and scattered multiple times in the gap, thereby extending the propagation path of the sound waves in the composite sound-filtering cotton and increasing the sound energy loss.

[0022] The larger sound-conducting gap 4 of the interlayer 2 forms an "acoustic energy trap". When the sound wave passes through the first sound-absorbing cotton 1, the sound wave diffuses due to the sudden change in the cavity volume, and the energy is dispersed to a larger space, reducing the sound pressure intensity; while the narrower sound-conducting gap 4 of the first sound-absorbing cotton 1 and the second sound-absorbing cotton 3 ensures that high-frequency sound waves are quickly absorbed to avoid penetration. The large-gap sound-conducting gap 4 absorbs low frequencies, and the narrow-gap sound-conducting gap 4 absorbs high frequencies, covering a wider frequency band. At the same time, it prevents the size of the sound-conducting gap 4 from resonating with a specific frequency (such as cavity resonance noise).

[0023] By making the spacing of the sound-conducting gap 4 of the first sound-absorbing cotton 1 larger than the spacing of the sound-conducting gap 4 of the second sound-absorbing cotton 3, an acoustic impedance gradient design is achieved. The wide sound-conducting gap 4 of the first sound-absorbing cotton 1 allows more sound waves to enter the interlayer 2, and the low acoustic impedance characteristics of the wide gap are utilized to fully diffuse the sound waves in the cavity; while the narrow sound-conducting gap 4 of the second sound-absorbing cotton 3 forms a high acoustic impedance barrier, forcing the residual sound waves to be reflected multiple times in a smaller space, thereby accelerating the attenuation of sound energy.

[0024] At the same time, frequency selective absorption is also achieved. The wide sound guide gap 4 has a stronger ability to capture low-frequency sound waves (longer wavelengths), while the narrow sound guide gap 4 performs secondary absorption on medium and high frequency sound waves (shorter wavelengths), forming complementary filtering.

[0025] Specifically, a wide gap allows low-frequency sound waves to enter the cavity and be absorbed by resonance, reducing low-frequency noise (such as environmental vibration and wind noise). A narrow gap limits the propagation path of high-frequency sound waves, preventing high-frequency signal leakage or the formation of standing waves. By using the difference in gap size, the attenuation efficiency of sound waves in different frequency bands is balanced to avoid signal distortion caused by excessive attenuation of a single frequency band.

[0026] The voiceprint processing module is used to detect valid voice segments in the voice signal and generate voiceprint confidence scores for the valid voice segments. This module identifies valid voice segments by analyzing the characteristics of the voice signal, and calculates the matching degree of these voice segments with the preset voiceprint database to generate voiceprint confidence scores.

[0027] It can accurately identify the user's voice and improve the accuracy of voice recognition; at the same time, the voiceprint confidence score can be used to determine the validity of the voice signal and provide a basis for subsequent processing.

[0028] The motion compensation module is connected to the nine-axis inertial sensor, and calculates the time compensation amount ΔT and the space compensation parameter θ through the motion compensation parameters collected by the nine-axis inertial sensor to correct the sound field distortion caused by the headphone wearing offset. Specifically, by connecting to the nine-axis inertial sensor, the acceleration, angular velocity and other parameters of the headphone wearer during movement are collected, and the time compensation amount ΔT and the space compensation parameter θ are calculated to correct the sound field distortion caused by the wearing offset.

[0029] Ensure that the earphones can still provide stable sound quality in different sports conditions and avoid sound quality degradation caused by wearing deviation.

[0030] The dynamic frequency control module is connected to the processor of the headphone main control chip through the clock tree, and is used to adjust the processor core frequency of the headphone main control chip in real time according to the environmental noise signal-to-noise ratio (SNR), the remaining battery power of the headphone, and the motion compensation parameters to balance performance and power consumption. This module adjusts the processor core frequency of the headphone main control chip in real time according to the environmental noise signal-to-noise ratio (SNR), the remaining battery power of the headphone, and the motion compensation parameters.

[0031] Minimize power consumption while ensuring headset performance. Increase processor frequency when noise is high or high-performance processing is required; reduce processor frequency to save power when battery is low or the environment is quiet.

[0032] Wireless communication module for data transmission with external devices via Bluetooth Low Energy (BLE) protocol. Bluetooth Low Energy (BLE) protocol is used for data transmission with external devices, which has the characteristics of low power consumption and stable connection.

[0033] Realize seamless connection between headphones and external devices (such as mobile phones, computers, etc.), while reducing energy consumption and extending the use time of headphones.

[0034] The voiceprint processing module and the dynamic frequency control module interact through an asynchronous communication mechanism. When the voiceprint confidence score exceeds the preset threshold, the dynamic frequency control module is triggered to switch the processor core frequency of the headphone main control chip to low power mode.

[0035] Specifically, the asynchronous communication mechanism is a communication method that decouples modules and allows modules to transfer information in a non-blocking manner. In this application, the interaction between the voiceprint processing module and the dynamic frequency control module is implemented through the asynchronous communication mechanism. The specific method is as follows:

[0036] T1. Data generation and packaging, the voiceprint processing module calculates the voiceprint confidence score by analyzing the valid voice segment of the voice signal. When the score exceeds the preset threshold, the score and related information are packaged as a message or event.

[0037] The package content includes: voiceprint confidence score, timestamp (used to mark the time when the event occurred), and other information (such as the current ambient noise signal-to-noise ratio, remaining battery power of the headset, etc.).

[0038] T2. Message passing: the encapsulated message is passed to the dynamic frequency control module through the message queue or event bus.

[0039] This transmission method is non-blocking, that is, the voiceprint processing module can continue to perform other tasks without waiting for the response of the dynamic frequency control module.

[0040] Specifically, if a message queue is used, the message will be stored in the queue and the dynamic frequency control module will read it on demand.

[0041] If an event-driven framework is used, messages are published in the form of events, and the dynamic frequency control module subscribes to relevant events and responds.

[0042] T3. Trigger condition judgment: After receiving the message, the dynamic frequency control module analyzes the voiceprint confidence score and determines whether the trigger condition is met (such as whether the score exceeds the preset threshold). If the condition is met, the corresponding operation is performed (such as switching the processor core frequency to low power mode).

[0043] T4. Action execution: The dynamic frequency control module adjusts the processor core frequency of the headset main control chip through the clock tree according to the trigger conditions.

[0044] For example, when the voiceprint confidence score is high, the processor frequency is reduced to save power consumption; when the score is low or the system load increases, the processor frequency is increased to ensure performance.

[0045] By switching the processor core frequency to low power mode in a timely manner, the energy consumption of the headset can be significantly reduced. This is crucial to extending the battery life of the headset and reducing the frequency of charging, especially when it is used for a long time or cannot be charged in time.

[0046] While ensuring that valid voice signals are accurately recognized and processed, a balance between performance and power consumption can be achieved by intelligently adjusting the processor frequency. This not only ensures the performance requirements of the headset when processing complex tasks, but also avoids unnecessary energy consumption.

[0047] By intelligently managing processor frequency, the risk of system crashes or performance degradation due to processor overheating or overloading can be reduced. This helps improve the stability and reliability of headset use.

[0048] In summary, the multi-microphone array module and motion compensation module can suppress noise and correct sound field distortion to improve the quality of voice signals. The voiceprint processing module generates voiceprint confidence scores to avoid false triggering and enhance the security and accuracy of voiceprint recognition. The dynamic frequency control module adjusts the processor frequency in real time according to environmental conditions and user behavior, and dynamically adjusts performance and power consumption. The asynchronous communication mechanism is used to achieve efficient interaction between the voiceprint processing module and the dynamic frequency control module, switching to low power mode when high-performance processing is not required, and optimizing low power mode. The modules work together to dynamically adapt to user behavior and environmental conditions, enhancing intelligence and real-time performance.

[0049] The nine-axis inertial sensor consists of the following three sub-sensors:

[0050] Three-axis accelerometer, measures linear acceleration.

[0051] Three-axis gyroscope, measuring angular velocity.

[0052] Three-axis magnetometer, measures magnetic field strength.

[0053] The data from these sub-sensors of the nine-axis inertial sensor are comprehensively processed through the sensor fusion algorithm, which can realize functions such as posture estimation, motion detection and spatial positioning, and provide important input data for motion compensation, temperature prediction and global adaptation of wireless headphones.

[0054] The voiceprint processing module of this embodiment includes an endpoint detection submodule.

[0055] The endpoint detection submodule is used to perform voice activation detection on the voice signal transmitted by the multi-microphone array module, detect the start and end points of the valid voice segment in the voice signal, and filter the non-voice segment;

[0056] The working steps of the endpoint detection submodule are:

[0057] A1. Read the voice signal transmitted by the multi-microphone array module and eliminate the echo in the voice signal through an adaptive filter, wherein the reference signal of the adaptive filter comes from the audio output signal of the wireless communication module.

[0058] Specifically, by reading the speech signal with echo cancellation, possible echoes are eliminated to prevent the sound, as produced by the earphone speaker, from being captured again by the microphone.

[0059] A2. Resample the speech signal, adjust the sampling frequency from 48kHz to 16kHz, and dynamically adjust the detection window length W based on the improved Teager energy operator. The detection window length W satisfies:

[0060] W=max(20ms,100ms-10×SNR+5×a rms ),

[0061] Detect the start and end points of valid speech segments in speech signals and filter out non-speech segments;

[0062] Among them, SNR is the current environmental noise signal-to-noise ratio, a rms It is the RMS value of acceleration output by the nine-axis inertial sensor.

[0063] The improved Teager energy operator is: Ψ[x(n)] = x 2 (n)-x(n-1)x(n+1)+0.2∣x(n)∣.

[0064] Specifically, by resampling and adjusting the detection window, and resampling the speech signal, the sampling frequency is adjusted from 48kHz to 16kHz. Reducing the sampling frequency can reduce the demand for computing resources while retaining the key information of the speech signal. Reducing the sampling frequency can reduce the demand for storage and computing resources. And adapt to the subsequent processing requirements, complete the preprocessing step.

[0065] At the same time, the detection window length W is dynamically adjusted based on the improved Teager energy operator. The calculation of the detection window length W takes into account the current signal-to-noise ratio (SNR) and the root mean square value of the acceleration output by the nine-axis inertial sensor (a rms ) to ensure that the start and end points of the speech segment can be accurately detected in different environments.

[0066] By eliminating echoes and dynamically adjusting the detection window length W, the endpoint detection submodule can more accurately identify valid speech segments and reduce false detections and missed detections.

[0067] By dynamically adjusting parameters using signal-to-noise ratio and acceleration information, the headphones can maintain stable performance under different noise levels and motion conditions.

[0068] Preferably, the method for dynamically adjusting the detection window is:

[0069] A21. Preprocess the input signal, read the voice signal and eliminate the echo through the adaptive filter; resample the voice signal and reduce the sampling frequency from 48kHz to 16kHz;

[0070] A22. Calculate signal characteristics,

[0071] Signal-to-noise ratio (SNR) is used to calculate the signal-to-noise ratio of the current speech signal, reflecting the strength of the speech signal relative to the background noise. The signal-to-noise ratio can be estimated by a variety of methods, such as based on spectrum analysis or energy ratio. rms ), obtain the user's acceleration data from the nine-axis inertial sensor, calculate its RMS value, and reflect the user's motion state.

[0072] A23. Dynamically adjust the window length and calculate the detection window length (W) according to the formula:

[0073] W=max(20ms,100ms-10×SNR+5×a rms );

[0074] The value of the window length W is dynamically adjusted as the signal-to-noise ratio and acceleration change.

[0075] A24. Sliding window detection, using the adjusted window length W to slide the detection signal, extract the starting point and end point of the valid speech segment. In the window, combined with the Improved Teager Energy Operator (ITEO) or other feature extraction methods, determine whether there is valid speech.

[0076] The method of dynamically adjusting the detection window has the following specific characteristics:

[0077] Strong adaptability, the window length is dynamically adjusted according to the signal-to-noise ratio and the RMS value of acceleration, and can adapt to speech signals under different environmental conditions. For example, in a noisy environment (low signal-to-noise ratio), the window length increases to reduce noise interference; when the user is moving (high acceleration), the window length decreases to capture rapidly changing speech signals.

[0078] Good robustness, combined with signal-to-noise ratio and acceleration information, improves the adaptability to complex scenes. The minimum window length limit (20ms) ensures the stability of detection.

[0079] High real-time performance, simple ITEO calculation, suitable for real-time processing of embedded devices (such as wireless headphones). The adjustment of window length only relies on simple mathematical operations, with low computational overhead.

[0080] The voiceprint processing module of this embodiment further includes a neural network verification submodule, which is used to extract voiceprint features from valid speech segments and generate voiceprint confidence scores;

[0081] The working method of the neural network verification submodule is:

[0082] A3. Perform time-frequency mask fusion on the speech signal of the input valid speech segment. The time-frequency mask fusion calculates the speech signal energy S(t,f) and the noise signal energy N(t,f) in the time-frequency domain according to the formula: Calculate the time-frequency mask value, use the time-frequency mask to perform weighted processing on the original speech signal, and obtain an enhanced speech signal;

[0083] A4. Voiceprint feature extraction and confidence score generation: extract voiceprint features from the enhanced speech signal and use a neural network model to generate the original voiceprint confidence score S;

[0084] A5. Temperature scaling calibration, real-time monitoring of processor core temperature T cpu , according to the formula: Calibrate confidence scores;

[0085] Among them, S′ is the calibrated voiceprint confidence score, S is the original voiceprint confidence score, T cpu is the processor core temperature, and ΔT is the time compensation amount.

[0086] The neural network verification submodule first performs time-frequency mask fusion processing on the speech signal of the input valid speech segment. The calculation of the time-frequency mask is based on the energy distribution of the speech signal and the noise signal in the time-frequency domain. Through the Sigmoid activation function, the input is mapped to the [0,1] interval to generate mask weights. These weights are used to dynamically assign weights to speech and noise. High mask values ​​(close to 1) enhance speech, and low mask values ​​(close to 0) suppress noise. In addition, the calculation of the time-frequency mask also takes into account the time compensation amount ΔT to correct the sound field distortion caused by wearing offset or movement.

[0087] Through time-frequency mask fusion, background noise and other interference components can be effectively suppressed to improve the clarity of speech signals. Clear speech signals help neural networks to more accurately extract voiceprint features, thereby improving the accuracy of voiceprint recognition.

[0088] After extracting the voiceprint features, the neural network verification submodule generates a voiceprint confidence score S. To further improve the accuracy of voiceprint recognition, the module also performs temperature scaling calibration based on the processor core temperature. The calibration formula takes into account the processor core temperature monitored in real time by the dynamic frequency control module, and adjusts the voiceprint confidence score to compensate for the performance degradation that may be caused by the increase in processor temperature.

[0089] Temperature scaling calibration helps to compensate for the impact of processor temperature increase on voiceprint recognition performance, thereby improving the accuracy of voiceprint recognition. By considering the impact of processor status (such as temperature) on voiceprint recognition, the overall stability of the system can be enhanced and false recognition or recognition failure caused by environmental factors can be reduced.

[0090] The motion compensation module of this embodiment includes a time compensation calculation unit, a space compensation calculation unit and a dynamic weight allocator.

[0091] The time compensation calculation unit is used to calculate the time compensation amount ΔT. The calculation formula of the time compensation amount ΔT is:

[0092]

[0093] Among them, ΔT is the time compensation, a(t) is the real-time acceleration signal, μ a is the average acceleration in the 100ms window, a max Maximum value of acceleration;

[0094] The space compensation calculation unit is used to calculate the space compensation parameter θ. The calculation formula is:

[0095]

[0096] Among them, Ω is the angular velocity signal from the gyroscope in the nine-axis inertial sensor; ∫Ωdt represents the integral of the angular velocity Ω over time t, which is used to calculate the change in the rotation angle; Represents the derivative of angular velocity Ω with respect to time t, which is used to calculate rotational acceleration; ∫∫Ωdt 2 It represents the double integral of angular velocity Ω over time t, and is used to calculate the change of rotational displacement. This formula is used to comprehensively consider the influence of rotation angle, rotational acceleration and rotational displacement.

[0097] Dynamic weight allocator, used to rms Adjust the weight ω of time and space compensation, the calculation formula is:

[0098] ω=0.2+0.1×tanh(0.5a rms ),

[0099] Among them, w is a dynamic weight, which is used to adjust the relative importance of time compensation and space compensation; a rms It is the RMS value of acceleration output by the nine-axis inertial sensor.

[0100] The time compensation calculation unit corrects the sound field distortion caused by user movement by calculating the time compensation ΔT. The calculation is based on the acceleration data output by the nine-axis inertial sensor. The mean value μ_a of the acceleration is calculated within a 100ms window, and the time compensation ΔT is obtained by integrating the variance formula. max The maximum value of acceleration is used for normalization processing to ensure that the value of ΔT is within a reasonable range. By calculating the time compensation amount, the sound field distortion caused by the user's head movement can be effectively corrected, thereby improving the accuracy of voiceprint recognition. The unit can dynamically respond to movements of different intensities, providing users with a more stable and accurate voiceprint recognition experience.

[0101] The spatial compensation calculation unit compensates for the change in the sound source position caused by user movement by calculating the spatial compensation parameter θ. The calculation is based on the angular velocity data Ω output by the nine-axis inertial sensor, and θ is obtained through operations such as integration, differentiation, and quadratic integration. Among them, each coefficient (0.25, 0.1, -0.05) is optimized to ensure that θ can accurately reflect the change in the sound source position. By calculating the spatial compensation parameters, the change in the sound source position caused by user movement can be effectively compensated, thereby improving the stability and accuracy of voiceprint recognition. This unit can reduce the voiceprint recognition error caused by movement, so that users can get a good voiceprint recognition experience even during exercise.

[0102] The dynamic weight distributor is based on the acceleration root mean square value a output by the nine-axis inertial sensor. rms Dynamically adjust the weight ω of time and space compensation. The calculation is based on the tanh function, and the coefficient and offset are adjusted to ensure that the value of ω is within a reasonable range and can reflect the compensation requirements under different accelerations. By dynamically adjusting the weight, it can be ensured that time and space compensation can achieve the best effect under different accelerations, thereby improving the accuracy and stability of voiceprint recognition. The distributor can be adaptively adjusted according to different motion states, making the system more flexible and intelligent.

[0103] In summary, the wireless headset of the present application realizes accurate perception and compensation of the user's motion state by integrating a motion compensation module, which not only improves the accuracy and stability of voiceprint recognition, but also enhances the flexibility of the system and user experience.

[0104] The specific working method of the motion compensation module is:

[0105] B1. Input signal, acceleration signal a(t) and angular velocity signal Ω from the nine-axis inertial sensor;

[0106] B2. Time compensation calculation, calculating the time compensation amount ΔT, which is used to dynamically adjust the time alignment of the speech signal;

[0107] B3. Spatial compensation calculation, calculating the spatial compensation parameter θ, which is used to dynamically adjust the spatial alignment of the speech signal;

[0108] B4. Dynamic weight allocation, adjust the weights ω of time compensation and space compensation according to the acceleration root mean square value arms;

[0109] B5. Output the result, combine the time compensation amount ΔT and the space compensation parameter θ, and generate the final motion compensation parameter to optimize the speech signal processing effect.

[0110] The advantages of the motion compensation module working method include:

[0111] Strong adaptability, time compensation and space compensation can be dynamically adjusted according to the user's motion state to adapt to different usage scenarios. Good robustness, comprehensive consideration of the effects of acceleration and angular velocity, and improved accuracy of motion compensation. High real-time performance, simple formulas and low computational overhead, suitable for real-time processing of embedded devices (such as wireless headphones).

[0112] The motion compensation module is suitable for the following scenarios:

[0113] Speech enhancement: In noisy environments, it dynamically compensates for the impact of user movement on speech signals and improves speech quality. Speech processing in sports scenarios: In sports scenarios such as running and fitness, it optimizes the time domain and spatial domain alignment of speech signals. Wireless headphones: When users wear headphones and move, they can ensure the clarity and stability of speech signals.

[0114] The dynamic frequency control module of this embodiment includes a basic frequency calculation unit, a burst response unit and an over-temperature protection unit.

[0115] Basic frequency calculation unit, used to perform the formula:

[0116]

[0117] Among them, SNR is the signal-to-noise ratio, which reflects the strength of the speech signal relative to the background noise, and ΔT is the time compensation amount;

[0118] The burst response unit is used when the voice signal strength changes When , the control frequency change rate satisfies:

[0119] The over-temperature protection unit is used to limit the maximum frequency according to the real-time temperature T. The calculation formula is:

[0120] Fmax=1.8-0.02×|T-25|GHz, where T is the real-time temperature.

[0121] The basic frequency calculation unit calculates the basic operating frequency (F) of the wireless headset using a given formula.base ). This formula takes into account the signal-to-noise ratio (SNR) and the time compensation (ΔT). The signal-to-noise ratio (SNR) reflects the ratio of signal to noise and is an important indicator of signal quality. The time compensation (ΔT) may be used to adjust the frequency to adapt to different environmental conditions or device status. The specific formula is:

[0122]

[0123] in, It represents the signal-to-noise ratio level of 20 decibels (dB), and ΔT is the time compensation amount.

[0124] By dynamically adjusting the basic operating frequency, wireless headphones can better adapt to different signal qualities and environmental conditions, thereby providing more stable and clearer audio transmission. This helps improve the user experience and reduce the degradation of sound quality caused by signal interference or environmental changes.

[0125] The burst response unit is used to monitor the rate of change of voice signal strength When the voice signal strength change rate exceeds the preset threshold (such as 30dB / s), the unit will control the frequency change rate To meet specific requirements (such as ). This helps wireless headphones quickly adjust their operating frequency to maintain stable audio transmission when the voice signal changes suddenly.

[0126] By quickly responding to changes in voice signal strength, the burst response unit can reduce sound quality distortion or interruption caused by sudden signal changes. This helps improve the stability and reliability of wireless headsets in complex communication environments, especially in applications such as voice calls or speech recognition.

[0127] The over-temperature protection unit is used to monitor the real-time temperature (T) of the wireless headset and limit the maximum operating frequency (F_max) according to the temperature. The specific formula is:

[0128]

[0129] This formula takes into account the impact of temperature on the performance of wireless headphones and prevents device damage or performance degradation due to overheating by limiting the maximum operating frequency. By implementing an over-temperature protection strategy, wireless headphones can maintain stable performance in high temperature environments and avoid sound quality degradation or device damage due to overheating. This helps extend the service life of wireless headphones while improving the user experience. In addition, this feature also helps improve the safety of wireless headphones and reduce potential risks caused by overheating.

[0130] Working method of dynamic frequency control module:

[0131] C1. Calculate the processor's base frequency F based on the signal-to-noise ratio SNR and the time compensation ΔT base ;

[0132] Specifically, obtain the current signal-to-noise ratio SNR and the time compensation amount ΔT; calculate The value of , amplifies the contribution of high signal-to-noise ratio to the fundamental frequency; weights the time compensation ΔT proportionally (the coefficient is 0.05) to obtain its impact on the fundamental frequency; adds the two items to obtain the fundamental frequency F base .

[0133] C2. When the voice signal strength changes When the threshold is exceeded, the processor frequency is quickly increased to cope with sudden demands;

[0134] Specifically, real-time monitoring of the voice signal strength change rate if Triggering the burst response mechanism; ensuring the processor frequency change rate At least 200MHz / ms, quickly increase the operating frequency; after the sudden demand ends, gradually return to the basic frequency F base .

[0135] C3. Over-temperature protection: according to the real-time temperature T, the maximum frequency Fmax of the processor is limited to prevent device damage caused by high temperature;

[0136] Specifically, obtain the current real-time temperature T; calculate (T-25) 2 , reflecting the degree to which the temperature deviates from 25°C; the maximum frequency Fmax is calculated according to the formula to ensure that the processor frequency is reduced in a high temperature environment; if the current operating frequency exceeds Fmax, the frequency is immediately reduced to below Fmax.

[0137] C4. Dynamically adjust the processor frequency, taking into account the basic frequency, burst response and over-temperature protection, and dynamically adjust the processor's operating frequency;

[0138] Specifically, the fundamental frequency F is calculated according to step 1 base ; According to step 2, determine whether a burst response needs to be triggered and adjust the frequency change rate. According to step 3, calculate the maximum frequency Fmax to ensure that the operating frequency does not exceed Fmax.

[0139] The processor's operating frequency is dynamically adjusted based on changes in the signal-to-noise ratio, the dynamic characteristics of the voice signal, the user's motion status, and real-time temperature. In low signal-to-noise ratio or static state, the frequency is reduced to save power. When the voice signal changes suddenly, the frequency is quickly increased to ensure performance. In high temperature environments, the maximum frequency is limited to protect the device.

[0140] This embodiment also includes a safety protection module, which is connected to the processor of the headset main control chip. Specifically, if the safety protection module is an independent hardware unit (such as a dedicated fault detection chip or circuit), it directly communicates with the processor through an interrupt pin or other hardware interface. If the safety protection module is a software module embedded in the main control chip, it interacts with the processor through a shared memory or an internal communication mechanism.

[0141] Working method of the security protection module:

[0142] D1. When the cumulative number of failures is C fail When ≥3, the system enters the fuse state, maintains basic functions and reduces system load, triggers the fuse timer, and the fuse duration τ satisfies:

[0143] minute;

[0144] D2. During the fuse period, the degraded mode controller performs the following operations:

[0145] Switch the sampling frequency of the multi-microphone array module from 48kHz to 8kHz;

[0146] Activate the 4-bit quantization mode of the neural network verification submodule;

[0147] D3. Return to normal. After the fuse duration ends, normal working state will be restored.

[0148] Among them, τ is the fuse duration, C fail is the cumulative number of failures.

[0149] The safety protection module is a key component in wireless headphones, designed to ensure stable operation of the device by monitoring the cumulative number of faults. When the cumulative number of faults reaches or exceeds the preset threshold (such as 3 times), the module triggers the fuse timer. The fuse duration (τ) is extended according to the increase in the cumulative number of faults, but there is an upper limit (such as 1440 minutes, i.e. 24 hours).

[0150] During the fuse period, the degradation mode controller performs a series of operations to reduce the power consumption and performance requirements of the device to prevent further failures. By triggering the fuse timer and implementing the degradation mode, the security protection module can effectively prevent the wireless headset from continuously failing in a short period of time and protect the device from potential damage. Reducing the performance and power consumption requirements of the device during the fuse period helps reduce wear and aging of the device, thereby extending the service life of the wireless headset. Although the degradation mode will reduce the performance of certain functions (such as sampling frequency and the accuracy of the neural network verification submodule), this temporary performance reduction can still provide certain functional guarantees and improve the overall user experience compared to the complete inability of the device to work. The fuse timer is a key component in the security protection module, which is used to start timing when the cumulative number of failures reaches a preset threshold. The timing length is dynamically adjusted according to the increase in the cumulative number of failures, but there is a fixed upper limit.

[0151] By dynamically adjusting the fuse duration according to the cumulative number of faults, the fuse timer can provide a more flexible and effective device protection strategy. Setting an upper limit on the fuse duration can prevent the fuse mechanism from being triggered frequently due to minor faults, thereby avoiding excessive impact on the user experience. During the fuse period, the degradation mode controller is responsible for performing a series of operations to reduce the performance and power consumption requirements of the wireless headset. These operations include switching the sampling frequency of the multi-microphone array module from 48kHz to 8kHz, and activating the 4-bit quantization mode of the neural network verification submodule.

[0152] By reducing the sampling frequency and activating the quantization mode, the degraded mode controller can significantly reduce the power consumption of wireless headphones, thereby extending battery life. Reducing performance and power requirements helps to reduce the processing burden of wireless headphones and reduce overheating and wear problems caused by continuous operation. Although the degraded mode will reduce the performance of some advanced functions, it still provides basic sound collection and verification functions to ensure that the wireless headphones can still maintain a certain degree of usability during the fuse period.

[0153] The failures in the above methods mainly refer to abnormal conditions related to hardware, software, system level or user behavior that occur during the operation of the device. These failures may include microphone failure (for example, a microphone cannot normally collect audio signals, resulting in incomplete voice input), signal processing failure, neural network reasoning errors, communication interruption, etc. By counting the cumulative number of failures and triggering the fuse mechanism, the system can effectively respond to frequent failures, protect the device and maintain basic functions.

[0154] This embodiment also includes a temperature compensation module, which is connected to the processor of the headset main control chip. If the temperature compensation module is an independent hardware unit (such as a dedicated temperature sensor or thermal management chip), the real-time temperature data is transmitted to the processor through communication protocols such as I2C and SPI. If the temperature compensation module is a software module embedded in the main control chip, it interacts with the processor through shared memory or internal communication mechanism.

[0155] The temperature compensation module is used to correct the noise reduction coefficient α and predict temperature changes.

[0156] The calculation formula of the modified noise reduction coefficient α is:

[0157] α T =α×(1-0.005(T-25)+0.001ΔT), where α T is the corrected noise reduction coefficient, α is the original noise reduction coefficient, T is the real-time temperature, and ΔT is the time compensation;

[0158] Specifically, -0.005(T-25) is the linear effect of temperature deviation from 25°C, which is used to adjust the noise reduction coefficient. When the temperature increases (T>25), the noise reduction coefficient will decrease to reduce power consumption and heat accumulation. When the temperature decreases (T<25), the noise reduction coefficient will increase to improve the noise reduction effect.

[0159] +0.001ΔT is the linear effect of the time compensation on the noise reduction coefficient. A larger time compensation (user vigorous exercise) will cause the noise reduction coefficient to increase to adapt to the rapidly changing ambient noise.

[0160] Corrected noise reduction coefficient α T The effects of temperature and time compensation are taken into account to dynamically adjust the strength of the noise reduction algorithm. In an active noise reduction (ANC) system, the noise reduction coefficient determines the strength of the noise reduction algorithm. By dynamically adjusting the noise reduction coefficient, the noise reduction effect can be optimized under different temperatures and motion conditions, while avoiding increased power consumption or device overheating due to excessive noise reduction intensity.

[0161] Solve differential equations to predict temperature changes:

[0162] in, Temperature change rate, F processor operating frequency, RMS value of acceleration, T real-time temperature;

[0163] 0.1F 2 It is a quadratic function of the processor operating frequency and is used to reflect the effect of frequency on temperature. When the processor frequency is higher, more heat is generated and the temperature rises faster.

[0164] It is a quadratic function of the RMS value of acceleration and is used to reflect the impact of user movement on temperature. When the user moves vigorously, the device may heat up due to the external environment or internal vibration.

[0165] -0.2(T-25) is the negative feedback term for the temperature deviation from 25°C, which is used to simulate the tendency of the device to return to the ambient temperature. When the temperature is above 25°C, the temperature will gradually decrease; when the temperature is below 25°C, the temperature will gradually increase. The differential equation describes the change of device temperature over time and is used to predict future temperature change trends. In embedded devices (such as wireless headphones), temperature changes may occur due to processor load, user movement, or environmental conditions. By predicting temperature changes, measures can be taken in advance (such as reducing the frequency or adjusting the noise reduction coefficient) to prevent the device from overheating or performance degradation.

[0166] Working method of temperature compensation module:

[0167] E1. Adjust the noise reduction algorithm and use the corrected noise reduction coefficient α T Updated the parameters of the active noise cancellation (ANC) algorithm; reduced the intensity of noise cancellation in high temperatures or during intense exercise to reduce power consumption and heat accumulation;

[0168] E2. Control the processor frequency. If it is predicted that the temperature will rise rapidly, the processor's operating frequency F is reduced to reduce heat generation.

[0169] E3. Trigger the protection mechanism. If it is predicted that the temperature may exceed the safety threshold, the fuse mechanism is triggered or the degradation mode is entered.

[0170] By correcting the noise reduction coefficient to take temperature changes into account, the temperature compensation module ensures that the wireless headset maintains a stable noise reduction effect under different temperature conditions. This helps improve the user experience, especially in environments with large temperature changes.

[0171] By predicting the temperature change trend, the temperature compensation module can adjust the working parameters of the wireless headset (such as noise reduction coefficient, operating frequency, etc.) in advance to adapt to the upcoming temperature changes. This helps to reduce performance fluctuations caused by temperature changes and improve the stability and reliability of the device.

[0172] By optimizing noise reduction performance and predicting temperature changes, the temperature compensation module helps reduce wear and aging of wireless headphones due to overheating or performance fluctuations. This helps extend the life of the equipment and reduce maintenance costs.

[0173] In summary, the temperature compensation module can improve the overall performance and user experience of wireless headphones by optimizing noise reduction performance and adapting to temperature changes. Whether in a cold or hot environment, users can enjoy a stable and clear audio experience.

[0174] This embodiment also includes a global adaptation engine, which is connected to the processor of the headset main control chip. If the global adaptation engine is an independent hardware unit (such as a dedicated coprocessor or a state monitoring chip), the state information is transmitted to the processor through communication protocols such as I2C and SPI. If the global adaptation engine is a software module embedded in the main control chip, it interacts with the processor through shared memory or internal communication mechanism.

[0175] The global adaptation engine is used to construct the three-dimensional state space {SNR, a rms ,T},

[0176] When SNR<15dB, and a rms When >1g, do the following:

[0177] Activate to force the processor to operate at a frequency of 1.5 GHz to provide higher computing power.

[0178] Increase the weight ω, ω = 0.3;

[0179] Switch the neural network sparsity of the neural network validation submodule to 60%.

[0180] The global adaptation engine is a high-level management component in wireless headsets that is responsible for dynamically adjusting multiple parameters of the headset based on environmental conditions and device status to ensure optimal performance and user experience. The engine achieves this function primarily by constructing a three-dimensional state space consisting of signal-to-noise ratio (SNR), root mean square acceleration (RMS), and the like. rms , representing the vibration or motion state of the equipment) and temperature (T).

[0181] When certain conditions are met (i.e. SNR is less than 15dB and a rms If the global adaptation engine exceeds 1g, it will perform a series of preset operations, including:

[0182] Activating Force increases the processor's operating frequency to 1.5GHz to provide higher computing power. This is to improve the headset's signal processing capabilities in situations where the signal-to-noise ratio is low and the ambient noise is high.

[0183] Increase the weight ω, ω = 0.3. This is to better stabilize the performance of the headphones during movement or vibration and reduce the degradation of sound quality caused by external interference.

[0184] Switch the neural network sparsity of the neural network verification submodule to 60%. Sparsity refers to the proportion of neurons in a neural network with zero or near-zero weights. By adjusting the sparsity, the computational complexity and power consumption of the neural network can be reduced while maintaining a certain performance level.

[0185] The global adaptation engine can dynamically adjust headphone parameters based on real-time conditions such as ambient noise levels, device vibration status and temperature, ensuring optimal sound quality and performance in various environments.

[0186] In a low signal-to-noise ratio environment where the SNR is less than 15dB, by forcibly setting the operating frequency and increasing the compensation weight, the global adaptation engine can significantly improve the signal processing capability and stability of the headphones, reduce noise interference, and improve sound clarity.

[0187] Optimize performance in motion: When the device is in motion (i.e. a rms The global adaptation engine can reduce power consumption and computational complexity while maintaining sound quality, thereby improving the endurance and response speed of the headset by adjusting the compensation weights and the sparsity of the neural network.

[0188] By dynamically adjusting the headphone parameters to adapt to different environments and device states, the global adaptation engine can significantly improve the user's auditory experience, allowing users to enjoy stable and clear sound quality whether in a quiet environment or in a noisy or moving state. At the same time, by optimizing power consumption and performance, it can also extend the life of the headphones and reduce maintenance costs.

[0189] This embodiment provides an intelligent voice control system for a wireless headset, including the wireless headset described above, an external device and a voice control instruction library.

[0190] The external device is connected to the wireless headset via the wireless communication module;

[0191] The voice control command library is stored in an external device and is used to respond to the voiceprint confidence score to execute playback control, noise reduction mode switching, or low-latency transmission commands.

[0192] Workflow of external devices:

[0193] F1. Data reception: the external device receives data from the wireless headset through a wireless communication module (such as Bluetooth, Wi-Fi).

[0194] The data includes the voiceprint confidence score S.

[0195] Voice commands (such as "play music", "turn on noise reduction").

[0196] F2. Command matching: the external device determines the credibility of the voice command based on the voiceprint confidence score S.

[0197] If S reaches or exceeds a certain threshold (such as (S>0.8)), the voice command is considered to be credible.

[0198] If S is lower than a certain threshold (such as (S<0.8)), the voice command is considered unreliable and is ignored.

[0199] F3. Command execution: If the voice command is credible, the corresponding command is selected from the voice control command library and executed.

[0200] For example:

[0201] If the voice command is "play music", the playback control command is executed.

[0202] If the voice command is "turn on noise reduction", it switches to active noise reduction mode.

[0203] F4. User feedback: the external device provides feedback of the command execution results to the user through the user interface or voice prompts.

[0204] For example:

[0205] "Music Played" is displayed on the screen.

[0206] The user is informed through voice prompts that "noise reduction mode is turned on".

[0207] The intelligent voice control system is mainly composed of wireless headphones and external devices. The wireless headphones establish a connection with the external devices through the built-in wireless communication module to realize the interaction of data transmission and control instructions.

[0208] The system integrates the collaborative work of wireless headphones and external devices, improving the overall convenience and intelligence level of use.

[0209] Through external devices, the system can easily expand functions to meet the diverse needs of users. The external device is wirelessly connected to the wireless headset through a wireless communication module (such as Bluetooth, Wi-Fi, etc.). This connection method allows the external device to send control instructions to the wireless headset and receive data from the wireless headset (such as voiceprint recognition results). Specifically, the external device is a smartphone or a tablet computer.

[0210] The voice control command library is stored in the external device and contains a variety of control commands, such as playback control (play, pause, previous song, next song, etc.), noise reduction mode switching (turn on / off noise reduction), and low-latency transmission commands (used to optimize the gaming or video call experience). These commands are associated with the voiceprint confidence score received by the wireless headset, and the corresponding operations are performed according to the results of voiceprint recognition.

[0211] Through the voice control command library, the system can realize intelligent playback control, noise reduction mode switching and low-latency transmission according to the user's voice commands, which improves the user experience. Users can customize the voice control command library according to their needs to achieve personalized intelligent control. The wireless headset recognizes the user's voice through the built-in voice recognition technology and calculates the voiceprint confidence score. This score reflects the reliability of the voice recognition result. The external device determines whether the corresponding control command needs to be executed based on the received voiceprint confidence score. Through the judgment of the voiceprint confidence score, the system can avoid misoperation caused by misrecognition and improve the accuracy of control. In scenarios where identity authentication is required (such as unlocking specific functions), the voiceprint confidence score can be used as an important basis for judging the user's identity and enhancing the security of the system.

[0212] In summary, the intelligent voice control system of the present application realizes functions such as intelligent playback control, noise reduction mode switching and low-latency transmission by integrating wireless headphones with external devices, realizing wireless connection, building a voice control command library, and responding using voiceprint confidence scores, thereby improving user experience and system security.

[0213] The voice control instruction library of this embodiment is used to trigger noise reduction mode switching or low-latency transmission instructions by matching streaming voice text through a dynamic sliding window when it is detected that the user's voiceprint confidence score exceeds a threshold.

[0214] The voice control command library is a preset command set stored in an external device. The command library not only contains a series of predefined voice control commands, such as noise reduction mode switching and low-latency transmission commands, but also integrates a dynamic sliding window matching algorithm. When the wireless headset detects the user's voice input through the built-in voice recognition technology and calculates the corresponding voiceprint confidence score, if the score exceeds the preset threshold, it indicates that the user's voice is accurately recognized and has a high degree of credibility, and the system will proceed to the next step. At this time, the external device will use the dynamic sliding window matching algorithm to perform real-time analysis of the streaming voice text. The algorithm can dynamically adjust the window size to match voice clips of different lengths, thereby accurately capturing the control command keywords issued by the user. Once the match is successful, the corresponding noise reduction mode switching or low-latency transmission command is triggered.

[0215] Through the threshold judgment of the voiceprint confidence score, the system can screen out high-confidence voice input, avoid misoperation caused by misrecognition, and thus improve the accuracy of command execution. The application of the dynamic sliding window matching algorithm enables the system to capture the control commands issued by the user in real time and accurately, greatly shortening the response time. Users do not need to perform complex operations or memorize specific command formats. They can easily control the noise reduction mode and low-latency transmission function of wireless headphones through natural voice communication, which greatly improves the convenience and comfort of use. This function reflects the high intelligence of the intelligent voice control system, which can automatically adjust the working status of the device according to the user's voice commands, meeting the personalized needs of users in different scenarios. The design of the voice control command library has good scalability, and commands can be easily added or modified to meet possible new functional requirements in the future. At the same time, the wireless connection method of the system with external devices also ensures good compatibility and can be widely used in various smart devices and platforms.

[0216] Specific working methods:

[0217] G1. Voice signal acquisition and processing. The wireless headset acquires voice signals through a multi-microphone array module. The voiceprint processing module extracts voiceprint features and generates a voiceprint confidence score S.

[0218] G2. Voiceprint confidence score judgment: if S>Threshold, the voice command is considered credible and proceed to the next step; otherwise, the voice command is ignored;

[0219] G3. Dynamic sliding window matching, converting voice signals into streaming voice text; using dynamic sliding windows to perform real-time matching of streaming voice text; matching rules can be based on keywords (such as "noise reduction", "low latency") or semantic understanding.

[0220] G4. Trigger command execution. If a predefined control command is matched, the corresponding operation is triggered. For example: "Turn on noise reduction" → switch to active noise reduction mode. "Switch to low latency mode" → enable low latency transmission protocol.

[0221] Specific method of dynamic sliding window matching:

[0222] G31. Sliding window design, window length: usually set to a few seconds (such as 2-5 seconds) to cover the complete voice command. Sliding step, the time interval between each slide (such as 0.5 seconds), to ensure real-time performance.

[0223] G32. Text matching algorithm, keyword matching: Extract keywords (such as "noise reduction", "low latency") within the sliding window. Use simple string matching algorithms (such as regular expressions) to identify instructions.

[0224] Semantic understanding: Use natural language processing (NLP) technology to parse the semantics of streaming speech text. For example, identify the user's intention based on deep learning models (such as Transformer).

[0225] G33. Streaming speech-to-text generation, using speech recognition engines (such as Google Speech-to-Text, iFlytek) to convert speech signals into text in real time.

[0226] Specific method to trigger the execution of instructions:

[0227] G41. Noise reduction mode switch, Active Noise Reduction (ANC) mode: Send instructions through external devices to adjust the noise reduction parameters of the headset. Transparency mode: Allows ambient sound to enter the headset to enhance user perception.

[0228] G42. Low latency transmission command, enable low latency audio transmission protocol (such as aptX Low Latency or LDAC), and ensure that audio data has low latency (such as <40ms) during transmission.

[0229] In summary, the voice control command library function in the intelligent voice control system of the present application, by combining voiceprint confidence score judgment, dynamic sliding window matching algorithm and predefined voice control commands, realizes accurate recognition and rapid response to user voice commands, greatly improving the user experience and system intelligence level.

[0230] The above contents are only preferred embodiments of the present invention. For ordinary technicians in this field, according to the concept of the present invention, there will be changes in the specific implementation methods and application scopes. The content of this specification should not be understood as limiting the present invention.

Claims

1. A wireless headset, characterized in that: It includes a multi-microphone array module, a voiceprint processing module, a motion compensation module, a dynamic frequency control module, a wireless communication module and an earphone main control chip, and the multi-microphone array module, the voiceprint processing module, the motion compensation module, the dynamic frequency control module and the wireless communication module are all connected to the earphone main control chip; Multi-microphone array module, used to collect audio signals and output beam-formed voice signals to suppress ambient noise; A voiceprint processing module, used to detect valid speech segments in speech signals and generate voiceprint confidence scores for the valid speech segments; The motion compensation module is connected to the nine-axis inertial sensor, and calculates the time compensation amount ΔT and the space compensation parameter θ through the motion compensation parameters collected by the nine-axis inertial sensor to correct the sound field distortion caused by the earphone wearing offset; Dynamic frequency control module, used to adjust the processor core frequency of the headset main control chip in real time according to the ambient noise signal-to-noise ratio, the remaining battery power of the headset and motion compensation parameters, balancing performance and power consumption; Wireless communication module, used for data transmission with external devices via Bluetooth low energy protocol; Asynchronous communication mechanism, the voiceprint processing module and the dynamic frequency control module interact through the asynchronous communication mechanism. When the voiceprint confidence score exceeds a preset threshold, the dynamic frequency control module is triggered to switch the processor core frequency of the headset main control chip to a low power consumption mode.

2. The wireless headset according to claim 1, characterized in that: The voiceprint processing module includes an endpoint detection submodule; The endpoint detection submodule is used to perform voice activation detection on the voice signal transmitted by the multi-microphone array module, detect the start and end points of the valid voice segment in the voice signal, and filter the non-voice segment; The working steps of the endpoint detection submodule are: A1. Read the voice signal transmitted by the multi-microphone array module, and eliminate the echo in the voice signal through an adaptive filter, wherein the reference signal of the adaptive filter is derived from the audio output signal of the wireless communication module; A2. The speech signal is resampled, the sampling frequency is adjusted from 48 kHz to 16 kHz, and the detection window length W is dynamically adjusted based on the improved Teager energy operator.

3. The wireless headset according to claim 2, characterized in that: The voiceprint processing module further includes a neural network verification submodule, wherein the neural network verification submodule is used to extract voiceprint features from valid speech segments and generate a voiceprint confidence score; The working method of the neural network verification submodule is: A3. Perform time-frequency mask fusion on the speech signal of the input valid speech segment, calculate the speech signal energy S(t,f) and the noise signal energy N(t,f) in the time-frequency domain, and use the time-frequency mask to perform weighted processing on the original speech signal to obtain an enhanced speech signal; A4. Voiceprint feature extraction and confidence score generation: extract voiceprint features from the enhanced speech signal and use a neural network model to generate the original voiceprint confidence score S; A5. Temperature scaling calibration, real-time monitoring of processor core temperature T cpu , according to the formula: Calibrate confidence scores; Among them, S′ is the calibrated voiceprint confidence score, S is the original voiceprint confidence score, T cpu is the processor core temperature, and ΔT is the time compensation amount.

4. The wireless headset according to claim 3, characterized in that: The motion compensation module includes a time compensation calculation unit, a space compensation calculation unit and a dynamic weight allocator. A time compensation calculation unit, used for calculating a time compensation amount ΔT; A space compensation calculation unit, used for calculating a space compensation parameter θ; Dynamic weight allocator, used to rms Adjust the weights ω of time and space compensation; The specific working method of the motion compensation module is: B1. Input signal, acceleration signal a(t) and angular velocity signal Ω from the nine-axis inertial sensor; B2. Time compensation calculation, calculating the time compensation amount ΔT, which is used to dynamically adjust the time alignment of the speech signal; B3. Spatial compensation calculation, calculating the spatial compensation parameter θ, which is used to dynamically adjust the spatial alignment of the speech signal; B4. Dynamic weight allocation, adjust the weights ω of time compensation and space compensation according to the acceleration root mean square value arms; B5. Output the result, combine the time compensation amount ΔT and the space compensation parameter θ, and generate the final motion compensation parameter to optimize the speech signal processing effect.

5. The wireless headset according to claim 4, characterized in that: The dynamic frequency control module includes a basic frequency calculation unit, a burst response unit and an over-temperature protection unit. Basic frequency calculation unit, used to perform the formula: Among them, SNR is the signal-to-noise ratio, which reflects the strength of the speech signal relative to the background noise, and ΔT is the time compensation amount; The burst response unit is used when the voice signal strength changes When the frequency change rate is controlled The over-temperature protection unit is used to limit the maximum frequency according to the real-time temperature T. The calculation formula is: F max =1.8-0.02×|T-25|GHz, where T is the real-time temperature; Working method of dynamic frequency control module: C1. Calculate the processor's base frequency F based on the signal-to-noise ratio SNR and the time compensation ΔT base ; C2. When the voice signal strength changes When the threshold is exceeded, the processor frequency is increased to cope with the sudden demand; C3. Over-temperature protection, according to the real-time temperature T, limit the maximum frequency F of the processor max ; C4. Dynamically adjust the processor frequency, comprehensively consider the basic frequency, burst response and over-temperature protection, and dynamically adjust the processor's operating frequency.

6. The wireless headset according to claim 3, characterized in that: It also includes a safety protection module, which is connected to the processor of the headset main control chip. The working method of the safety protection module is as follows: D1. When the cumulative number of failures is C fail When ≥3, the system enters the fuse state, maintains basic functions and reduces system load, triggers the fuse timer, and the fuse duration τ satisfies: D2. During the fuse period, the degraded mode controller performs the following operations: Switch the sampling frequency of the multi-microphone array module from 48kHz to 8kHz; Activate the 4-bit quantization mode of the neural network verification submodule; D3, back to normal, after the fuse time ends, the normal working state is restored; Among them, τ is the fuse duration, C fail is the cumulative number of failures.

7. The wireless headset according to claim 6, characterized in that: It also includes a temperature compensation module, which is connected to the processor of the headset main control chip. The temperature compensation module is used to correct the noise reduction coefficient α and predict temperature changes. The calculation formula of the modified noise reduction coefficient α is: α T =α×(1-0.005(T-25)+0.001ΔT), where α T is the corrected noise reduction coefficient, α is the original noise reduction coefficient, T is the real-time temperature, and ΔT is the time compensation; Solve differential equations to predict temperature changes: in, is the temperature change rate, F is the operating frequency of the processor, is the RMS value of acceleration, T is the real-time temperature; Working method of temperature compensation module: E1. Adjust the noise reduction algorithm and use the corrected noise reduction coefficient α T Updated the parameters of the active noise reduction algorithm; reduced the intensity of noise reduction in high temperatures or during intense exercise to reduce power consumption and heat accumulation; E2. Control the processor frequency. If it is predicted that the temperature will rise rapidly, the processor operating frequency F is reduced to reduce heat generation; E3. Trigger the protection mechanism. If it is predicted that the temperature may exceed the safety threshold, the fuse mechanism is triggered or the degradation mode is entered.

8. The wireless headset according to claim 5, characterized in that: It also includes a global adaptation engine, which is connected to the processor of the headset main control chip and is used to construct a three-dimensional state space {SNR, a rms ,T}, When SNR<15dB, and a rms When >1g, do the following: Activate to force the processor to operate at a frequency of 1.5 GHz to provide higher computing power. Increase the weight ω, ω = 0.3; Switch the neural network sparsity of the neural network validation submodule to 60%.

9. An intelligent voice control system for a wireless headset, comprising the wireless headset according to any one of claims 1 to 8, characterized in that: Also includes external device and voice control command library, An external device connected to the wireless headset via the wireless communication module; The voice control instruction library is stored in the external device and is used to respond to the voiceprint confidence score to execute playback control, noise reduction mode switching or low-latency transmission instructions.

10. The intelligent voice control system according to claim 9, characterized in that: The voice control instruction library is used to trigger noise reduction mode switching or low-latency transmission instructions by matching streaming voice text through a dynamic sliding window when it is detected that the user's voiceprint confidence score exceeds a threshold.

Images