Noise reduction regulation and control method for earphone and noise reduction earphone

Abstract

The invention belongs to the technical field of earphone noise reduction, and provides a noise reduction regulation and control method for an earphone and a noise reduction earphone. The method comprises the following steps: acquiring a first environment sound signal of an environment area, and extracting human voice features from the first environment sound signal to construct a human voice feature template library; in response to a noise reduction regulation and control instruction, collecting a second environment sound signal of the environment area, converting the time domain signal into a frequency domain signal through Fourier transform, and extracting a frequency spectrum feature from the frequency domain signal; performing matching calculation on the spectrum features and a human voice feature template library to determine environment voice signals, human voice signals and noise signals; an active noise reduction algorithm is adopted to generate offset sound waves opposite to the noise signals in phase, the human voice signals are amplified, and the offset sound waves and the amplified human voice signals are output through a loudspeaker of the earphone. According to the invention, effective noise filtering and clear transparent transmission of specific human voice are realized, smooth communication can be realized without taking off the earphone, and the use experience is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of earphone noise reduction, in particular to a noise reduction control method for earphones and a noise reduction earphone. BACKGROUND

[0002] Currently, the mainstream earphone noise reduction technology is mainly realized by combining active noise reduction (ANC) and passive noise reduction. The active noise reduction technology uses a microphone to collect environmental noise, and then generates a sound wave with an opposite phase to the noise through a noise reduction circuit to cancel it out, thereby achieving the purpose of reducing environmental noise. Passive noise reduction relies on the physical structure of the earphone, such as sealed earplugs and ear covers, to block the transmission of external sounds.

[0003] However, the existing noise reduction technology has obvious limitations in actual application. When the user is in a scene that requires voice communication interaction, such as receiving instructions, alarms, and voice communication in high-noise environments such as construction sites, manufacturing workshops, mining areas, and hydropower stations, the traditional noise reduction mode will filter out most of the sounds in the environment, including useful sounds such as human voices, making it difficult for the user to normally receive instructions, alarms, and voice communication. The user must remove the earphone to hear the speaking voice, which not only complicates the operation but also may block the voice, easily causing safety production accidents and safety production hazards, and causing conflicts between occupational health and safety production.

[0004] Although some earphone products are equipped with an environmental sound transmission function, which attempts to pick up external sounds through a microphone and transmit them into the ear canal to solve the above problems, this transmission function often transmits all environmental sounds indiscriminately, including both human voices that need to be heard and various unwanted noises such as daily work noises (such as drilling machines, electric saws, and pile driving), equipment operation noises, and voices of undesired personnel. This makes it difficult for the user to obtain a clear and comfortable communication experience after turning on the transmission function, and it is difficult to meet the user's needs in complex environments.

[0005] Therefore, an earphone noise reduction control method that accurately identifies and transmits specific human voices while efficiently filtering other environmental noises is an important problem that needs to be solved in the current earphone noise reduction technology field, and it is of great significance for improving user experience and ensuring safety production. SUMMARY

[0006] To solve at least one of the above technical problems, the present application provides a noise reduction control method for earphones, a noise reduction earphone, an electronic device, a computer storage medium, and a computer program product.

[0007] In a first aspect of the present application, a noise reduction control method for earphones is provided, comprising the following method steps: Before receiving the noise reduction regulation instruction, a first environmental sound signal of an environment region where the earphone is located is collected by at least one microphone arranged on the earphone at a preset time interval, a human voice feature is extracted from the first environmental sound signal by using an attention mechanism, and a human voice feature template library is constructed. In response to the noise reduction regulation instruction, a second environmental sound signal of the environment region is collected, high-frequency interference and low-frequency bottom noise in the second environmental sound signal are removed to obtain a time domain signal, the time domain signal is converted into a frequency domain signal by Fourier transform, and a frequency spectrum feature is extracted from the frequency domain signal. The frequency spectrum feature is matched with the human voice feature template library for matching calculation, so as to determine a signal in the environmental sound signal that meets a human voice frequency range and mark the signal as a human voice signal, and mark a signal in other frequency ranges as a noise signal. An active noise reduction algorithm is used to generate a cancellation sound wave opposite in phase to the noise signal, and the human voice signal is amplified, and the cancellation sound wave and the amplified human voice signal are output through a loudspeaker of the earphone.

[0008] In a second aspect of the present application, a noise reduction earphone is provided, which comprises a control unit, a storage unit, a loudspeaker and at least one microphone. Before receiving the noise reduction regulation instruction, a first environmental sound signal of an environment region where the earphone is located is collected by at least one microphone arranged on the earphone at a preset time interval, a human voice feature is extracted from the first environmental sound signal by using an attention mechanism, and a human voice feature template library is constructed. In response to the noise reduction regulation instruction, a second environmental sound signal of the environment region is collected, high-frequency interference and low-frequency bottom noise in the second environmental sound signal are removed to obtain a time domain signal, the time domain signal is converted into a frequency domain signal by Fourier transform, and a frequency spectrum feature is extracted from the frequency domain signal. The frequency spectrum feature is matched with the human voice feature template library for matching calculation, so as to determine a signal in the environmental sound signal that meets a human voice frequency range and mark the signal as a human voice signal, and mark a signal in other frequency ranges as a noise signal. An active noise reduction algorithm is used to generate a cancellation sound wave opposite in phase to the noise signal, and the human voice signal is amplified, and the cancellation sound wave and the amplified human voice signal are output through a loudspeaker of the earphone.

[0009] In a third aspect, the present application provides an electronic device comprising at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, which, when executed by the processor, implements the method according to any one of the preceding aspects.

[0010] In a fourth aspect, the present application provides a computer storage medium storing a computer program executable by a processor to implement the method according to any one of the preceding aspects.

[0011] In a fifth aspect, the present application provides a computer program product comprising a computer program executable by a processor to implement the method according to any one of the preceding aspects.

[0012] The present application collects environmental sound in a preset time interval and constructs a human voice template library with an attention mechanism, extracts spectral features for matching by combining Fourier transform, and accurately distinguishes human voice from noise; active noise reduction is used to generate counteracting sound waves, while amplifying specific human voice corresponding to the human voice template library. The present application realizes effective noise filtering and clear transmission of specific human voice, and enables smooth communication without taking off the earphone, thereby improving the use experience in a complex environment and ensuring production safety. BRIEF DESCRIPTION OF DRAWINGS

[0013] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained without creative labor.

[0014] Figure 1 is a flowchart of a noise reduction regulation method for earphones disclosed by the embodiments of the present application; Figure 2 is a schematic diagram of constructing a human voice feature template library based on residual human voice signals disclosed by the embodiments of the present application; Figure 3 is an electrical structure schematic diagram of a noise reduction earphone disclosed by the embodiments of the present application. DETAILED DESCRIPTION

[0015] The embodiments of the present application will be described below by specific embodiments. Those skilled in the art can easily understand other advantages and effects of the present application from the content disclosed in the present specification. Obviously, the described embodiments are part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0016] In addition, the technical features involved in the different embodiments of the present application described below can be combined with each other as long as they do not conflict with each other.

[0017] As shown in the Figure 1 The noise reduction regulation method for the earphone comprises the following method steps: S10, before receiving the noise reduction regulation instruction, collecting the first environmental sound signal of the environment area where the earphone is located at a preset time interval through at least one microphone arranged on the earphone, extracting the human voice feature from the first environmental sound signal by using the attention mechanism, and constructing a human voice feature template library; wherein the preset time interval is determined based on the environmental sound feature representing the personnel flow extracted from the first environmental sound signal.

[0018] As shown in the Figure 3 In addition to being equipped with a conventional built-in loudspeaker and its electric control components, power supply components, the noise reduction earphone in the present application is also configured with a microphone for collecting the environmental sound signal of the environment area where it is located. The microphone is usually arranged on the noise reduction earphone shell, but the present application does not make specific limitation on its installation position and layout quantity. At the same time, the user can also turn on or off the noise reduction and transparent function of the specific human voice of the noise reduction earphone through the switch button on the player or the noise reduction earphone.

[0019] The user wears the above-mentioned noise reduction earphone into or remains in a specific environment area, and when no noise reduction regulation instruction is received, the first environmental sound signal of the environment area where the earphone is located is collected through at least one microphone arranged on the earphone. At the same time, the human voice feature is extracted by using the attention mechanism, and then the human voice feature template library is constructed. The attention mechanism can focus on the key information related to human voice and reduce the interference of other noise, thereby improving the accuracy of human voice feature extraction.

[0020] In addition, the collection of the first environmental sound signal before receiving the noise reduction regulation instruction is carried out at a preset time interval. Such setting can realize the continuous update of the human voice feature template library to adapt to the personnel flow change in the environment area. Specifically, the above-mentioned preset time interval is determined based on the environmental sound feature representing the personnel flow extracted from the first environmental sound signal. Such setting can make the time interval setting more targeted and adapt to the sound change characteristics under different personnel flow conditions, and then construct a human voice feature template library that is more suitable for the current environment.

[0021] S20, in response to the noise reduction regulation instruction, collecting the second environmental sound signal of the environment area, removing the high-frequency interference and low-frequency bottom noise in the second environmental sound signal to obtain a time domain signal, converting the time domain signal into a frequency domain signal by Fourier transform, and extracting the frequency spectrum feature from the frequency domain signal.

[0022] When the user needs to turn on the specific voice transparent function, the specific voice transparent function of the noise reduction earphone can be turned on through the switch button on the player or the noise reduction earphone, and then the noise reduction control instruction is automatically triggered.

[0023] In response to the noise reduction control instruction, the sound signal of the environment area is collected again, that is, the second environment sound signal is obtained, which is the sound signal that needs to be processed by noise reduction control. The high-frequency interference and low-frequency bottom noise are removed to obtain a more pure time-domain signal, so as to avoid the influence of irrelevant noise on subsequent analysis. Then, the time-domain signal obtained after the foregoing processing is converted into a frequency-domain signal through Fourier transform, so that the sound is converted from a form changing with time into a form distributed according to frequency, facilitating the analysis of the characteristics of the sound from the frequency dimension. Then, the frequency spectrum characteristics are extracted from the obtained frequency-domain signal for distinguishing the voice and other noises. The frequency spectrum characteristics include but are not limited to the fundamental frequency, the to-be-matched formant, the to-be-matched spectral envelope line, and the to-be-matched sound intensity change rate.

[0024] S30, the frequency spectrum characteristics are matched and calculated with the voice feature template library to determine the signals in the environment sound signal that conform to the voice frequency range and mark them as voice signals, and mark the signals of other frequency ranges as noise signals.

[0025] The frequency spectrum characteristics extracted in step S20 are matched and calculated with the voice feature template library constructed in step S10. With the accurate voice feature information in the template library, it can be accurately judged which part of the second environment sound signal belongs to the voice frequency range corresponding to the voice feature templates. The signals conforming to the voice frequency range are determined by matching and marked as voice signals, and the rest are marked as noise signals.

[0026] S40, an active noise reduction algorithm is used to generate a cancellation sound wave opposite in phase to the noise signal, and the voice signal is amplified, and the cancellation sound wave and the amplified voice signal are output through the speaker of the earphone.

[0027] For the marked noise signal, an active noise reduction algorithm (Active Noise Cancellation, ANC) is used to generate a cancellation sound wave opposite in phase, which can be output through the speaker to cancel the noise in the environment and achieve the effect of reducing noise. At the same time, the voice signal is amplified and then output through the speaker, which can ensure that the user can clearly hear the specific voice. Finally, the cancellation sound wave and the amplified voice signal are output at the same time, which realizes the goal of filtering other noises while clearly transmitting the specific voice corresponding to the voice feature template, and meets the user's demand for enjoying the noise reduction effect and smoothly communicating in a complex environment.

[0028] The scheme collects environmental sound through a preset time interval and constructs a human voice template library with an attention mechanism, combines Fourier transform to extract spectral features for matching, and accurately distinguishes human voice from noise; active noise reduction is used to generate counteracting sound waves, while amplifying specific human voice corresponding to the human voice template library. The application realizes effective noise filtering and clear transmission of specific human voice, and enables smooth communication without taking off the earphone, thereby improving the use experience in complex environments and ensuring production safety.

[0029] It should be noted that the scheme in the application can be packaged as a human voice transmission function of a noise reduction earphone in software form, which is used only for the transmission of the human voice of other personnel matched with the human voice feature template library. The human voice feature template library is periodically updated according to the above-mentioned time interval. In addition, a conventional human voice transmission function independent of the human voice feature template library can also be designed for the user, i.e., all human voice signals are output to the user. This function can also be turned on or off by the user through a switch button on the player (such as a mobile phone, tablet computer, etc.) or the noise reduction earphone. Details are not repeated.

[0030] As an example, the human voice feature is extracted from the first environmental sound signal by using the attention mechanism to construct a human voice feature template library, comprising: The attention mechanism is used to process multiple human voice signals in the first environmental sound signal, identify the virtual human voice signal played by the electronic device, and track the distance of the remaining human voice signals to obtain a tracking result indicating the human voice signal gradually moving away from the environment area; The virtual human voice signal played by the electronic device and the human voice signal gradually moving away from the environment area are excluded, and the human voice feature template library is constructed based on the remaining human voice signals.

[0031] Firstly, the attention mechanism focuses on the analysis of the spectral features, rhythm and rhyme of the sound signal, for example, the virtual human voice played by the electronic device often has differences in tone consistency, background noise characteristics, etc. from real human voice. The attention mechanism focuses on these differences to accurately distinguish real human voice from virtual human voice played by mobile phones, televisions, radios and other electronic devices. These human voice signals belonging to virtual human voice signals are excluded to avoid virtual human voice entering the human voice feature template library and prevent irrelevant human voice features from being mixed in the template library, which affects the recognition accuracy of real communication human voice.

[0032] Then, distance detection tracking is performed on the remaining vocal signals that do not belong to the virtual vocal signal, and a tracking result indicates that the vocal signals gradually move away from the environment region. The distance detection tracking combines the intensity variation, propagation delay and other parameters of the sound signal to determine the distance relationship between the vocal source and the user. When it is detected that the intensity of a certain vocal signal continuously weakens and the variation trend meets the characteristics of gradually moving away (for example, the hawking sound outside the window, the singing vocal of a passerby, etc.), the vocal signal is marked as a vocal signal gradually moving away. Such vocal signals are those with strong mobility in the environment and low possibility of subsequent communication with the user, and therefore, the vocal feature template library is not constructed based on them.

[0033] Therefore, the vocal feature template library constructed by the embodiment has more pure and targeted features, can significantly improve the accuracy of subsequent recognition and matching of target vocals in the second environmental sound signal, and further optimize the effect of the entire earphone noise reduction regulation method, so that the user can hear the vocals that need to be communicated more clearly.

[0034] As an example, the preset time interval is determined based on an environmental sound feature extracted from the first environmental sound signal, including: An environmental sound feature representing personnel mobility is extracted from the first environmental sound signal. The environmental sound feature includes the switching frequency of the vocal signal, the appearance rate of the new vocal feature, and the mutation frequency of the vocal signal intensity. A personnel mobility quantitative index is established based on the environmental sound feature, and a time interval correction value is obtained based on a preset correlation between the personnel mobility quantitative index and the time interval correction value. The personnel mobility quantitative index is positively correlated with the time interval correction value. The preset time interval is obtained by subtracting the time interval correction value from the base time interval.

[0035] The switching frequency of the vocal signal refers to the number of times different vocal signals appear alternately per unit time, reflecting the change speed of the conversation object in the environment. The appearance rate of the new vocal feature is the number of first-appeared vocal features per unit time, reflecting the situation of new personnel entering the environment. The mutation frequency of the vocal signal intensity is the number of times the vocal signal intensity changes significantly per unit time, indirectly reflecting the movement state of the personnel. The above environmental sound features are used to describe the activity level of personnel mobility in the environment in multiple dimensions.

[0036] The personnel mobility quantitative index is obtained by weighted calculation of the vocal switching frequency, the new vocal rate and the intensity mutation frequency, and these scattered features are integrated into a comprehensive personnel mobility quantitative index, so as to accurately quantify the degree of personnel mobility. The calculation formula is, for example: wherein, is the personnel mobility quantitative index, a standardized value of the switching frequency of the human voice signal (i.e., a ratio of the actual switching frequency to a maximum frequency threshold), a standardized value of the appearance rate of the new human voice feature (i.e., a ratio of the actual rate to a maximum rate threshold), a standardized value of the mutation frequency of the human voice signal intensity (i.e., a ratio of the actual frequency to a maximum frequency threshold); 、 、 is a weight coefficient, and the default values are 0.4, 0.3, and 0.3, respectively (which can be dynamically adjusted according to the actual scene).

[0037] The preset correlation relationship is a pre-set corresponding rule of the personnel mobility quantitative indicator and the correction value, for example, the correction value is increased by a certain value when the quantitative indicator is improved by one level, so as to ensure that the correction value increases when the personnel mobility is more active. The following table shows an example:

[0038] The basic time interval is a pre-set initial collection interval, for example, 5 minutes. When the personnel mobility is high (the quantitative indicator is large), the time interval correction value is large, and the preset time interval obtained by subtracting the two is short, so that the environmental sound signal can be collected more frequently, and the human voice feature template library can be updated quickly to adapt to the environment with frequent personnel changes; when the personnel mobility is low (the quantitative indicator is small), the correction value is small, and the preset time interval is long, which ensures the effectiveness of the template library while reducing unnecessary collection and calculation, reduces the earphone energy consumption, and balances the timeliness of the template library and resource consumption.

[0039] As an example, as shown in Figure 2 , the human voice feature template library is constructed based on the remaining human voice signals, including: a core feature parameter group is extracted from each of the remaining human voice signals, including a fundamental frequency, a formant, a spectral envelope line, and a sound intensity change rate, a first group of human voice feature templates is constructed based on the core feature parameter group; a variation feature parameter group is obtained by simulating a feature variation range of each core feature parameter in the core feature parameter group in a preset scene, a second group of human voice feature templates is constructed based on the variation feature parameter group; wherein the preset scene refers to a scene that causes short-term changes in sound features; a human voice feature template library is constructed based on the first group of human voice feature templates and the second group of human voice feature templates.

[0040] The fundamental frequency in the core feature parameter group reflects the pitch of the sound, the formant represents the resonance characteristics of the vocal tract, the spectral envelope represents the frequency energy distribution profile of the sound, and the sound intensity change rate represents the dynamic change of the sound intensity. The above core feature parameters represent the basic acoustic characteristics of the human voice from different dimensions. Based on the above core feature parameter group of each human voice signal, a corresponding human voice feature template is constructed, that is, a first group of human voice feature templates is obtained. These human voice feature templates are equivalent to the standard sound image of each human voice. The human voice feature template contains the typical value range and probability distribution of each core feature parameter, such as the normal fluctuation interval of the fundamental frequency, the main distribution frequency of the formant, etc. It can be understood that the human voice feature templates in the subsequent second group of human voice feature templates are also the typical value range and probability distribution of the variation feature parameters corresponding to each core feature parameter, and details are not repeated.

[0041] At the same time, the present application also sets up a plurality of preset scenes, which focus on the situation of causing short-term changes of sound, such as the change of throat state after drinking water (or some special drinks) may cause slight shift of fundamental frequency, the formant of the sound in the moment after coughing will appear temporary change, the sound intensity change rate will be significantly improved when emotional excitement, etc. By simulating the possible fluctuation range of each core parameter in these scenes (such as the fundamental frequency changes within ±5% interval, a formant frequency shifts 100-200Hz, etc.), a variation feature parameter group is generated. The specific processing process is generally as follows: (1) The preset scenes are classified and defined, and the typical scenes causing short-term changes of sound features are determined, including the change of throat state after drinking water, the adjustment of vocal cords after temporary coughing, the change of sound state caused by emotional excitement, the change of oral cavity resonance after eating, etc. For each type of scene, a scene and feature change association model (preferably based on CNN) is established in advance through sample collection - collect human voice signals of different people in the above scenes, extract the core parameters such as fundamental frequency, formant, spectral envelope and sound intensity change rate, compare the parameter differences before and after the scene, and statistics the change amplitude distribution of the feature parameters in each type of scene.

[0042] (2) Set the scene variation threshold for each core feature parameter. For example, for the drinking water scene: the fundamental frequency is allowed to fluctuate within ±8% of the original value, the first formant frequency shift is controlled within 50-150Hz, the energy peak decay rate of the spectral envelope does not change more than 20%, and the fluctuation amplitude of the sound intensity change rate is limited to ±15% of the original value. For the scene after coughing: the short-term fluctuation range of the fundamental frequency is expanded to ±12%, the second formant frequency shift can reach 80-200Hz, and the energy proportion of the spectral envelope in the medium and high frequency band (2-4kHz) is allowed to change ±25%. For the emotional fluctuation scene: the variation range of the sound intensity change rate is relaxed to ±30%, and the dynamic change rate (change per second) of the fundamental frequency can be increased to 1.5 times of the original value.

[0043] (3) Based on the original core feature parameter set and the above-mentioned variation threshold, a variation feature parameter set corresponding to each preset scene is generated. A random disturbance algorithm is used to adjust the original core feature parameter set within the set threshold, and multiple sets of samples conforming to the scene variation rule are generated for each core parameter. For example, for the fundamental frequency parameter, 5 different offset variation values are generated within a range of ±8%; for the formant parameter, 3-4 sets of variation frequency values are generated according to the scene-specific frequency offset interval. Then a plurality of variation feature parameter sets are obtained.

[0044] Based on the variation parameter set corresponding to each remaining human voice signal, a second group of human voice feature templates is constructed. This is equivalent to adding dynamic changing sound images to each human voice, which compensates for the limitation of the first group of templates that can only cope with stable states, and can cover the characteristic fluctuations of the sound caused by short-term physiological or state changes.

[0045] Finally, the first group of human voice feature templates and the second group of human voice feature templates are integrated to construct a final human voice feature template library.

[0046] As an example, the spectral features include a to-be-matched fundamental frequency, a to-be-matched formant, a to-be-matched spectral envelope line, and a to-be-matched sound intensity change rate; the spectral features are matched with the human voice feature template library to determine the signals in the environmental sound signal that conform to the human voice frequency range, including: The feature similarity between the spectral features and each human voice feature template in the human voice feature template library is calculated, and the feature similarity is calculated by weighted Euclidean distance, wherein the weight coefficients of the to-be-matched fundamental frequency and the to-be-matched formant are higher than those of the to-be-matched spectral envelope line and the to-be-matched sound intensity change rate; When the feature similarity is greater than or equal to a similarity threshold, it is determined that the corresponding signal conforms to the human voice frequency range, otherwise it is determined that the corresponding signal does not conform to the human voice frequency range; wherein the similarity threshold is dynamically adjusted according to the environmental noise interference intensity.

[0047] The spectral features obtained in the foregoing include a to-be-matched fundamental frequency, a to-be-matched formant, a to-be-matched spectral envelope line, and a to-be-matched sound intensity change rate. These features correspond one-to-one to the core feature parameter set (fundamental frequency, formant, spectral envelope line, and sound intensity change rate) extracted when the human voice feature template library is constructed.

[0048] The feature similarity between the spectral features and each human voice feature template in the human voice feature template library is calculated by weighted Euclidean distance. Among them, the fundamental frequency determines the pitch of the sound, and the formant reflects the resonance characteristics of the sound channel. Both are core identifiers for distinguishing different human voices, and have a greater impact on the accuracy of human voice recognition. Therefore, the weight coefficients of the to-be-matched fundamental frequency and the to-be-matched formant are set to be higher than those of the to-be-matched spectral envelope line and the to-be-matched sound intensity change rate, that is, by assigning higher weights to improve the accuracy of matching.

[0049] When the feature similarity is greater than or equal to the similarity threshold, it means that the spectral features of the signal have a high degree of match with the features in the human voice feature template library, and it is determined to be in line with the human voice frequency range; otherwise, it is determined to be not in line with the human voice frequency range.

[0050] In addition, a similarity threshold is dynamically adjusted based on the intensity of ambient noise interference. Specifically, in environments with high noise interference, the human voice signal may be obscured by the noise, resulting in a decrease in feature similarity. In this case, appropriately lowering the similarity threshold can prevent the misclassification of noise-affected human voices as non-human voices. In environments with low noise interference, human voice features are clearer, and the threshold can be appropriately raised to reduce the possibility of noise being misclassified as human voices. This dynamic adjustment mechanism enables matching decisions to adapt to different ambient noise conditions, further improving the robustness of human voice recognition.

[0051] like Figure 3 As shown, an embodiment of the present invention further provides a noise-canceling headset, comprising a control unit 1001, a storage unit 1002, a speaker 1003, and at least one microphone 1004; the control unit 1001 calls the computer program code stored in the storage unit 1002 to implement the following steps: Before receiving the noise reduction control instruction, at least one microphone provided on the headset collects a first ambient sound signal of the surrounding area at a preset time interval, and uses an attention mechanism to extract human voice features from the first ambient sound signal to construct a human voice feature template library; wherein the preset time interval is determined based on the ambient sound features representing the mobility of people extracted from the first ambient sound signal; In response to the noise reduction control instruction, collecting a second ambient sound signal of the environmental area, removing high-frequency interference and low-frequency background noise in the second ambient sound signal to obtain a time domain signal, converting the time domain signal into a frequency domain signal through Fourier transform, and extracting spectral features from the frequency domain signal; Matching the spectrum features with the human voice feature template library to determine the signal in the ambient sound signal that meets the human voice frequency range and mark it as a human voice signal, and mark the signal in other frequency ranges as a noise signal; An active noise reduction algorithm is used to generate a cancelling sound wave with a phase opposite to that of the noise signal, and the human voice signal is amplified, and the cancelling sound wave and the amplified human voice signal are output through the speaker of the headset.

[0052] It is understandable that noise-canceling headphones usually also include power supply components, shell structures and other components, which will not be detailed here.

[0053] As an example, the human voice features are extracted from the first environmental sound signal by using an attention mechanism to construct a human voice feature template library, including: The multiple human voice signals in the first environmental sound signal are processed by using an attention mechanism to identify the virtual human voice signals played by the electronic device; and the distance detection tracking is performed on the remaining human voice signals to obtain a tracking result indicating the human voice signals gradually moving away from the environment area; The virtual human voice signals played by the electronic device and the human voice signals gradually moving away from the environment area are excluded, and the human voice feature template library is constructed based on the remaining human voice signals.

[0054] As an example, the preset time interval is determined based on the environmental sound features extracted from the first environmental sound signal, including: The environmental sound features representing the personnel mobility are extracted from the first environmental sound signal, including the switching frequency of the human voice signals, the appearance rate of the new human voice features, and the mutation frequency of the human voice signal intensity; The personnel mobility quantitative index is established based on the environmental sound features, and the time interval correction value is obtained based on the preset correlation relationship between the personnel mobility quantitative index and the time interval correction value; wherein the personnel mobility quantitative index is positively correlated with the time interval correction value; The preset time interval is obtained based on the subtraction operation between the basic time interval and the time interval correction value.

[0055] As an example, the human voice feature template library is constructed based on the remaining human voice signals, including: The core feature parameter group is extracted from the remaining human voice signals, including the fundamental frequency, the formant, the spectral envelope line, and the sound intensity change rate, and the first group of human voice feature templates is constructed based on the core feature parameter group; The feature variation range of each core feature parameter in the core feature parameter group under a preset scene is simulated to obtain a variation feature parameter group, and the second group of human voice feature templates is constructed based on the variation feature parameter group; wherein the preset scene refers to a scene causing short-term changes in sound features; The human voice feature template library is constructed based on the first group of human voice feature templates and the second group of human voice feature templates.

[0056] As an example, the spectral features include the to-be-matched fundamental frequency, the to-be-matched formant, the to-be-matched spectral envelope line, and the to-be-matched sound intensity change rate; then the spectral features are matched with the human voice feature template library to determine the signals in the environmental sound signal that meet the human voice frequency range, including: The feature similarity is calculated by weighted Euclidean distance, wherein the weight coefficient of the to-be-matched fundamental frequency and the to-be-matched formant is higher than that of the to-be-matched spectral envelope and the to-be-matched sound intensity change rate. When the feature similarity is greater than or equal to a similarity threshold, it is determined that the corresponding signal meets the human voice frequency range, otherwise it is determined that the corresponding signal does not meet the human voice frequency range; wherein the similarity threshold is dynamically adjusted according to the environmental noise interference intensity.

[0057] The embodiment of the present application also provides an electronic device, comprising at least one processor, a memory and a computer program stored in the memory and executable on the at least one processor, wherein the computer program is executed by the processor to implement the method according to any one of the preceding method.

[0058] The embodiment of the present application also provides a computer storage medium storing a computer program executable by a processor to implement the method according to any one of the preceding method.

[0059] The embodiment of the present application also provides a computer program product comprising a computer program executable by a processor to implement the method according to any one of the preceding method.

[0060] Those skilled in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware, computer software or a combination of both, and in order to clearly illustrate the interchangeability of hardware and software, the components and steps of the examples have been described in general in the above description. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0061] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited to this, any skilled person in the art can easily think of various equivalent modifications or replacements within the technical range disclosed by the present application, and these modifications or replacements should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A noise reduction regulating method for a headphone, characterized by, The method comprises the following steps: Before receiving the noise reduction regulation instruction, a first environmental sound signal of an environment region is collected at a preset time interval through at least one microphone arranged on the earphone, a human voice feature is extracted from the first environmental sound signal by using an attention mechanism, and a human voice feature template library is constructed; wherein the preset time interval is determined based on an environmental sound feature representing personnel mobility extracted from the first environmental sound signal; In response to the noise reduction regulation instruction, a second environmental sound signal of the environment region is collected, high-frequency interference and low-frequency background noise in the second environmental sound signal are removed to obtain a time domain signal, the time domain signal is converted into a frequency domain signal through Fourier transform, and a frequency spectrum feature is extracted from the frequency domain signal; The frequency spectrum feature is matched with the human voice feature template library for calculation to determine a signal in the environmental sound signal that meets a human voice frequency range and mark it as a human voice signal, and mark a signal of other frequency ranges as a noise signal; An active noise reduction algorithm is used to generate a cancellation sound wave opposite in phase to the noise signal, and the human voice signal is amplified, and the cancellation sound wave and the amplified human voice signal are output through a loudspeaker of the earphone.

2. The noise reduction regulating method for earphones of claim 1, wherein: The human voice feature is extracted from the first environmental sound signal by using the attention mechanism to construct the human voice feature template library, comprising: The first environmental sound signal is processed by using the attention mechanism to identify a virtual human voice signal played by an electronic device, and distance detection and tracking are performed on the remaining human voice signals to obtain a tracking result indicating a human voice signal gradually moving away from the environment region; The virtual human voice signal played by the electronic device and the human voice signal gradually moving away from the environment region are excluded, and the human voice feature template library is constructed based on the remaining human voice signals.

3. The noise reduction regulating method for headphones of claim 2, wherein: The preset time interval is determined based on the environmental sound feature extracted from the first environmental sound signal, comprising: The environmental sound feature representing personnel mobility is extracted from the first environmental sound signal, and the environmental sound feature includes a switching frequency of the human voice signal, an appearance rate of a new human voice feature, and a mutation frequency of the human voice signal intensity; A personnel mobility quantitative index is established based on the environmental sound feature, and a time interval correction value is obtained based on a preset correlation between the personnel mobility quantitative index and the time interval correction value; wherein the personnel mobility quantitative index is positively correlated with the time interval correction value; The preset time interval is obtained based on a subtraction operation between a basic time interval and the time interval correction value.

4. The noise reduction regulating method for earphones of claim 3, wherein: The human voice feature template library is constructed based on the remaining human voice signals, comprising: A core feature parameter group is extracted from each of the remaining human voice signals, including a fundamental frequency, a formant, a spectral envelope line, and a sound intensity change rate, a first group of human voice feature templates is constructed based on the core feature parameter group; A variation feature parameter group is obtained by simulating a feature variation range of each core feature parameter in the core feature parameter group in a preset scenario, and a second group of human voice feature templates is constructed based on the variation feature parameter group; wherein the preset scenario refers to a scenario causing short-term changes in sound features. Construct a human voice feature template library based on the first set of human voice feature templates and the second set of human voice feature templates.

5. The noise reduction regulating method for earphones of claim 4, wherein: The spectral features include a to-be-matched fundamental frequency, a to-be-matched formant, a to-be-matched spectral envelope, and a to-be-matched sound intensity change rate; the spectral features are matched with the human voice feature template library to determine signals in the environmental sound signal that conform to a human voice frequency range, including: A feature similarity between the spectral features and each human voice feature template in the human voice feature template library is calculated, and the feature similarity is calculated by a weighted Euclidean distance, wherein a weight coefficient of the to-be-matched fundamental frequency and the to-be-matched formant is higher than that of the to-be-matched spectral envelope and the to-be-matched sound intensity change rate; When the feature similarity is greater than or equal to a similarity threshold, it is determined that the corresponding signal conforms to the human voice frequency range, otherwise, it is determined that the corresponding signal does not conform to the human voice frequency range; wherein the similarity threshold is dynamically adjusted according to the environmental noise interference intensity.

6. A noise reducing earphone, characterized by The control unit calls computer program code stored in the storage unit to implement the following steps: Before receiving a noise reduction regulation instruction, at least one microphone arranged on the earphone collects a first environmental sound signal of an environment region at a preset time interval, and a human voice feature is extracted from the first environmental sound signal using an attention mechanism to construct a human voice feature template library; wherein the preset time interval is determined based on an environmental sound feature representing personnel mobility extracted from the first environmental sound signal; In response to a noise reduction regulation instruction, a second environmental sound signal of the environment region is collected, high-frequency interference and low-frequency noise in the second environmental sound signal are removed to obtain a time-domain signal, the time-domain signal is converted into a frequency-domain signal by Fourier transform, and a spectral feature is extracted from the frequency-domain signal; The spectral feature is matched with the human voice feature template library to determine signals in the environmental sound signal that conform to a human voice frequency range, and the signals are marked as human voice signals, and other frequency range signals are marked as noise signals; An active noise reduction algorithm is used to generate a cancellation sound wave opposite in phase to the noise signal, and the human voice signal is amplified, and the cancellation sound wave and the amplified human voice signal are output through the loudspeaker of the earphone.

7. A noise reducing earphone according to claim 6, characterised in that: The human voice feature is extracted from the first environmental sound signal using an attention mechanism to construct a human voice feature template library, including: The first environmental sound signal is processed using an attention mechanism to identify virtual human voice signals played by electronic devices therefrom; and distance detection tracking is performed on the remaining human voice signals to obtain a tracking result indicating human voice signals gradually moving away from the environment region; The virtual human voice signals played by electronic devices and the human voice signals gradually moving away from the environment region are excluded, and a human voice feature template library is constructed based on the remaining human voice signals.

8. An electronic device, the electronic device comprising: At least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, characterized in that the computer program, when executed by the processor, implements the method of any of claims 1-5.

9. A computer storage medium, characterized in that: The computer storage medium stores a computer program executable by a processor to implement the method of any of claims 1-5.

10. A computer program product, characterised in that: The computer program product comprises a computer program executable by a processor to implement the method of any of claims 1-5.

Images