Broadband adaptive filtering and reverse sound wave synthesis method of intelligent noise reduction equipment

Abstract

The invention relates to the technical field of sound processing, in particular to a broadband adaptive filtering and reverse sound wave synthesis method for intelligent noise reduction equipment, which comprises the following steps of: performing local mean decomposition on an original audio signal to obtain a plurality of product function components and a monotonic residual signal, analyzing each product function component based on the original audio signal to obtain an equipment characteristic index of each product function component, and analyzing an amplitude change of each product function component to obtain a human voice interference index of each product function component; determining the interference weight of each product function component according to the equipment characteristic index and the human voice interference index of each product function component; and obtaining an interference signal corresponding to the noise according to the plurality of product function components, the monotonic residual signal and the interference weight of each product function component. According to the invention, environmental noise and human voice interference during equipment operation can be accurately identified, so that the noise reduction effect of industrial audio is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of sound processing, and in particular to a wideband adaptive filtering and reverse sound wave synthesis method of an intelligent noise reduction device. BACKGROUND

[0002] In voice communication, voice recognition, human-computer interaction and other voice input scenarios, background environmental noise has a significant impact on the intelligibility and recognizability of voice signals. Especially in open spaces or complex noise environments, the noise frequency band is wide, the components are complex, and the changes are rapid. The use of an intelligent noise reduction device can effectively suppress noise interference in voice.

[0003] The intelligent noise reduction device usually has an active noise reduction (ANC) function and the ability to automatically identify noise characteristics and dynamically adjust noise reduction strategies. The ANC function mainly neutralizes noise by synthesizing reverse sound waves to achieve the effect of noise reduction.

[0004] Current active noise reduction technology is mainly applied to closed scenes and is mainly aimed at low-frequency narrowband noise, which is difficult to apply to open or wide-frequency multi-source noise environments. At the same time, traditional adaptive filtering technology is mostly fixed-bandwidth or single-frequency band filtering, and lacks the ability to dynamically adapt to complex wideband noise environments, especially in industrial acoustic state monitoring processes, it is difficult to analyze and eliminate noise that overlaps with the operating frequency band of some devices, which can easily affect the sound quality of the target audio and cannot obtain a relatively ideal noise reduction result.

[0005] That is, the active noise reduction scheme provided by the prior art has poor noise reduction effect on industrial audio. SUMMARY

[0006] In order to solve the technical problem that the active noise reduction scheme provided by the prior art has poor noise reduction effect on industrial audio, the purpose of the present application is to provide a wideband adaptive filtering and reverse sound wave synthesis method of an intelligent noise reduction device, and the technical solution adopted is as follows: In a first aspect, an embodiment of the present application provides a wideband adaptive filtering and reverse sound wave synthesis method of an intelligent noise reduction device, which comprises: locally mean-decomposing an original audio signal to obtain a plurality of product function components and a monotonic residual signal, the original audio signal being derived from an industrial production environment; analyzing each product function component based on the original audio signal to obtain a device characteristic index of each product function component, the device characteristic index being used to represent the probability of the product function component corresponding to the device operating sound in the industrial production environment; An amplitude variation of each product function component is analyzed to obtain a human voice interference index of each product function component, the human voice interference index being used to represent a probability of the product function component corresponding to human voice; An interference weight of each product function component is determined according to the device feature index and the human voice interference index of each product function component; An interference signal is obtained according to the plurality of product function components, the monotonic residual signal, and the interference weight of each product function component, the interference signal being used to generate a noise reduction signal, the noise reduction signal and the interference signal having the same amplitude but opposite phases, the noise reduction signal being used to neutralize noise in the original audio signal.

[0007] In an embodiment, the original audio signal is subjected to local mean decomposition to obtain a plurality of product function components and a monotonic residual signal, including: The original audio signal is split based on a preset frame length to obtain a plurality of signal frames; The plurality of signal frames are subjected to windowing processing respectively to obtain a plurality of window signals; The plurality of window signals are subjected to Fourier transform respectively to obtain a plurality of signal transform information; Signal energy differences between adjacent window signals are analyzed based on the plurality of signal transform information to obtain a decomposition interference index of each window signal, wherein the decomposition interference index is used to represent a probability of the window signal interfering with the local mean decomposition operation; The original audio signal is subjected to local mean decomposition based on the decomposition interference index of each window signal to obtain a plurality of product function components and a monotonic residual signal.

[0008] In an embodiment, the signal transform information includes a plurality of signal energies of a plurality of frequencies in the corresponding window signal; The signal energy differences between adjacent window signals are analyzed based on the plurality of signal transform information to obtain a decomposition interference index of each window signal, including: Based on the plurality of signal transform information, an energy front difference value and an energy rear difference value corresponding to each frequency in each window signal are obtained, wherein the energy front difference value is a signal energy difference between the corresponding frequency in the previous window signal and the corresponding frequency in the corresponding window signal, and the energy rear difference value is a signal energy difference between the corresponding frequency in the corresponding window signal and the corresponding frequency in the next window signal; A sum value of the energy front difference value and the energy rear difference value corresponding to each frequency in each window signal is calculated to obtain an energy sum value corresponding to each frequency in each window signal; A product of the signal energy of each frequency in each window signal and the energy sum value corresponding to the frequency is calculated to obtain an energy product of each frequency in each window signal; calculating a sum of a plurality of energy products of a plurality of frequencies in each window signal to obtain an energy reference value corresponding to each window signal; calculating a ratio of the energy reference value corresponding to each window signal and an energy entropy value corresponding to each window signal to obtain a decomposition interference index of each window signal, wherein the energy entropy value is used to represent an information entropy of a plurality of signal energies of a plurality of frequencies in the corresponding window signal.

[0009] In one embodiment, in the corresponding plurality of decomposition operations of the local mean decomposition, the extreme value data of each local extreme point in each decomposition operation is a product of an original extreme value of the local extreme point and a correction coefficient corresponding to the local extreme point, the correction coefficient is a ratio of a first interference index and a second interference index of the corresponding local extreme point, the first interference index is a decomposition interference index of a window signal to which the corresponding local extreme point belongs, and the second interference index is a sum of a decomposition interference index of a window signal to which a previous local extreme point of the corresponding local extreme point belongs and a decomposition interference index of a window signal to which a next local extreme point of the corresponding local extreme point belongs.

[0010] In one embodiment, the analysis of each product function component based on the original audio signal to obtain a device feature index of each product function component comprises: performing a fast Fourier transform on each product function component to obtain a plurality of sinusoidal wave components corresponding to each product function component; determining a sinusoidal wave acoustic index and a frequency difference index of each sinusoidal wave component corresponding to each product function component based on the original audio signal, the sinusoidal wave acoustic index being a frequency acoustic index of a frequency component with a minimum frequency difference between the corresponding frequency component and the original audio signal, the frequency acoustic index being used to represent a probability that the corresponding frequency component indicates a device running sound in the corresponding industrial production environment, and the frequency difference index being used to indicate a minimum frequency difference between the corresponding sinusoidal wave component and the plurality of frequency components included in the original audio signal; determining a component feature index of each sinusoidal wave component corresponding to each product function component based on the sinusoidal wave acoustic index, the frequency difference index and the amplitude of each sinusoidal wave component corresponding to each product function component; calculating a sum of a plurality of component feature indexes of a plurality of sinusoidal wave components corresponding to each product function component to obtain a device feature index of each product function component.

[0011] In one embodiment, the frequency acoustic index of each frequency component included in the original audio signal is obtained by: splitting the original audio signal based on a preset frame length to obtain a plurality of signal frames; The plurality of signal frames are respectively windowed to obtain a plurality of window signals; The amplitude variation of each frequency component in the plurality of window signals is analyzed to obtain an amplitude fluctuation index corresponding to each frequency component; The phase variation of each frequency component in the plurality of window signals is analyzed to obtain a phase variation index corresponding to each frequency component; The frequency acoustic index of each frequency component is determined according to the amplitude fluctuation index and the phase variation index corresponding to each frequency component.

[0012] In one embodiment, the amplitude variation of each frequency component in the plurality of window signals is analyzed to obtain an amplitude fluctuation index corresponding to each frequency component, comprising: The absolute value of the amplitude difference between adjacent two window signals of each frequency component is calculated to obtain a plurality of window amplitude differences corresponding to each frequency component; The variance of the plurality of window amplitude differences corresponding to each frequency component is calculated to obtain the amplitude fluctuation index corresponding to each frequency component.

[0013] In one embodiment, the phase variation of each frequency component in the plurality of window signals is analyzed to obtain a phase variation index corresponding to each frequency component, comprising: The absolute value of the phase difference between adjacent two window signals of each frequency component is calculated to obtain a phase difference sequence corresponding to each frequency component; In the phase difference sequence corresponding to each frequency component, the sequence mean value and a plurality of sequence peak values corresponding to each frequency component are determined, and the difference between the plurality of sequence peak values and the sequence mean value is calculated respectively to obtain a plurality of sequence jump values corresponding to each frequency component; The mean value of the plurality of sequence jump values corresponding to each frequency component is calculated to obtain the phase jump index corresponding to each frequency component; The phase difference sequence corresponding to each frequency component is curve-fitted to obtain a phase difference fitting curve corresponding to each frequency component; The absolute value of the derivative of each sequence point in the phase difference sequence corresponding to each frequency component in the corresponding phase difference fitting curve is calculated to obtain a plurality of sequence point derivative values corresponding to each frequency component; The mean value of the plurality of sequence point derivative values corresponding to each frequency component is determined as the phase drift index corresponding to each frequency component; The phase variation index corresponding to each frequency component is obtained according to the phase drift index and the phase jump index corresponding to each frequency component.

[0014] In one embodiment, the analyzing the amplitude variation of each product function component to obtain a vocal interference index of each product function component comprises: performing signal splitting on each product function component to obtain a plurality of signal segments corresponding to each product function component; performing clustering on the plurality of signal segments corresponding to each product function component to obtain a plurality of clusters corresponding to each product function component; sorting the plurality of clusters corresponding to each product function component in descending order of the amplitude mean value of the cluster to obtain a cluster sequence corresponding to each product function component; calculating the difference between the amplitude mean values of two adjacent clusters in the cluster sequence corresponding to each product function component to obtain a plurality of cluster amplitude differences corresponding to each product function component, wherein the cluster amplitude difference is the difference between the amplitude mean values of the corresponding cluster and the previous cluster; determining the cluster corresponding to the largest cluster amplitude difference as the critical cluster corresponding to each product function component, and determining the plurality of clusters after the critical cluster in the corresponding cluster sequence as a plurality of strong signal clusters corresponding to each product function component; analyzing the amplitude variation between the plurality of strong signal clusters corresponding to each product function component to obtain a vocal interference index of each product function component.

[0015] In one embodiment, the analyzing the amplitude variation between the plurality of strong signal clusters corresponding to each product function component to obtain a vocal interference index of each product function component comprises: analyzing the cluster difference between each strong signal cluster and its adjacent next strong signal cluster in the plurality of strong signal clusters corresponding to each product function component to obtain a plurality of cluster difference values corresponding to each product function component; calculating the difference between the cluster amplitude mean values of each strong signal cluster and its adjacent next strong signal cluster in the plurality of strong signal clusters corresponding to each product function component to obtain a plurality of target amplitude differences corresponding to each product function component; calculating the ratio of each target amplitude difference and the corresponding cluster difference value corresponding to each product function component to obtain a plurality of cluster interference indexes corresponding to each product function component; calculating the mean value of the plurality of cluster interference indexes corresponding to each product function component to obtain a vocal interference index of each product function component.

[0016] In a second aspect, another embodiment of the present application provides a wideband adaptive filtering and reverse sound wave synthesis system of an intelligent noise reduction device, the system comprising: a signal decomposition module, configured to perform local mean decomposition on an original audio signal to obtain a plurality of product function components and a monotonic residual signal, wherein the original audio signal is from an industrial production environment; a device feature analysis module, configured to analyze each product function component based on the original audio signal to obtain a device feature index of each product function component, wherein the device feature index is used to represent a probability of the product function component corresponding to a device running sound in the industrial production environment; a human voice analysis module, configured to analyze an amplitude variation of each product function component to obtain a human voice interference index of each product function component, wherein the human voice interference index is used to represent a probability of the product function component corresponding to a human voice; an interference analysis module, configured to determine an interference weight of each product function component according to the device feature index and the human voice interference index of each product function component; an interference determination module, configured to obtain an interference signal according to the plurality of product function components, the monotonic residual signal and the interference weight of each product function component, wherein the interference signal is used to generate a noise reduction signal, the noise reduction signal and the interference signal have the same amplitude but opposite phases, and the noise reduction signal is used to neutralize noise in the original audio signal.

[0017] In a third aspect, a further embodiment of the present application provides an electronic device, including a processor, a memory, and a computer program stored in the memory and executable on the processor, and when the computer program is executed by the processor, the steps of the method in the first aspect are implemented.

[0018] In a fourth aspect, a further embodiment of the present application provides a computer readable storage medium, and the computer readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method in the first aspect are implemented.

[0019] The present application has the following advantages: The application accurately decomposes the original audio signal into multiple product function components and a monotonic residual signal by local mean decomposition on the original audio signal to adapt to the non-stationary, wideband, nonlinear characteristics of the original audio signal from the industrial production environment, and then obtains the equipment characteristic index and the human voice interference index of each product function component to determine the probability of each product function component corresponding to the equipment running sound and the probability of corresponding human voice, that is, to determine the probability of each product function component belonging to the key sound that needs to be retained and the probability of each product function component belonging to the noise that needs to be filtered out, so as to accurately evaluate the interference weight of each product function component from multiple aspects and determine a more accurate interference signal, that is, to accurately identify the environmental noise and human voice interference that overlap with the equipment running sound in the industrial production environment, thereby improving the noise reduction effect of industrial audio. BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art and the advantages thereof, a brief introduction will be given to the drawings needed in the embodiments or the prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0021] Figure 1 A schematic flow chart of a wideband adaptive filtering and reverse sound wave synthesis method of an intelligent noise reduction device provided by an embodiment of the present application; Figure 2 A structural schematic diagram of a wideband adaptive filtering and reverse sound wave synthesis system of an intelligent noise reduction device provided by an embodiment of the present application; Figure 3 A structural schematic diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0022] In order to further illustrate the technical means and effects adopted by the present application to achieve the predetermined object, the following describes the wideband adaptive filtering and reverse sound wave synthesis method of an intelligent noise reduction device according to the present application, its specific implementation, structure, features and effects in detail in combination with the drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0023] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0024] The application provides a specific scheme of a wide-band adaptive filtering and reverse sound wave synthesizing method of an intelligent noise reduction device.

[0025] The application provides a wide-band adaptive filtering and reverse sound wave synthesizing method of an intelligent noise reduction device. Figure 1 Fig. 1 is a schematic flowchart of a wide-band adaptive filtering and reverse sound wave synthesizing method of an intelligent noise reduction device according to an embodiment of the application, which comprises the following steps: Step S1, locally average-decomposing an original audio signal to obtain a plurality of product function components and a monotonic residual signal.

[0026] The original audio signal is from an industrial production environment.

[0027] In the industrial production environment, the original audio signal collected is complex, including not only persistent device running sound (such as motor sound, fan sound, mechanical vibration sound) and environmental noise, but also sudden device starting sound, impact sound, operation alarm sound and human voice. The frequency distribution span of the above-mentioned sound is usually 20 Hz to 16 kHz, and the spectrum characteristics change in real time.

[0028] In the application, in order to effectively collect the original audio signal containing device running sound, environmental noise and sudden noise, the following arrangement can be made: A multi-channel microphone array (usually ≥4 channels) is arranged near the fixed key components of the device (such as motor bearing, gear box, pump, etc.). The array structure of the multi-channel microphone array can be adjusted to be linear, annular or planar according to the space arrangement of the production site. Each microphone collects the sound signal of the site at a sampling rate of 44.1 kHz to 192 kHz, with a quantization accuracy of ≥16 bit and a dynamic range of ≥96 dB. All microphones are synchronously sampled with time stamping to ensure the completeness of the multi-channel audio phase information.

[0029] The above-mentioned multi-channel microphone array can be used to collect the above-mentioned original audio signal.

[0030] In the above-mentioned step, the original audio signal is decomposed to separate the environmental noise and the device running sound into different components, so as to adaptively suppress the noise in different components based on the different performances of the noise in different components, while retaining the characteristics of the device running sound as much as possible.

[0031] The reason for choosing the local mean decomposition processing is that the original audio signal in the industrial production environment has the characteristics of non-stationary, wide frequency, non-linear, and is mixed with background noise (fan, pipeline sound), human voice, sudden impact sound, equipment working inherent frequency vibration sound and other complex components. The local mean decomposition processing can utilize the local characteristics of the original audio signal itself for audio decomposition, better adapt to the characteristics of the original audio signal of non-stationary, wide frequency, non-linear, and ensure the accuracy of the plurality of product function components and the monotonic residual signal obtained by decomposition, and further ensure the accuracy of the subsequent determined noise signal.

[0032] Further, the local mean decomposition of the original audio signal to obtain a plurality of product function components and a monotonic residual signal comprises: Splitting the original audio signal based on a preset frame length to obtain a plurality of signal frames; Windowing processing the plurality of signal frames respectively to obtain a plurality of window signals; Performing Fourier transform on the plurality of window signals respectively to obtain a plurality of signal transform information; Based on the plurality of signal transform information, analyzing the signal energy difference between adjacent window signals to obtain a decomposition interference index of each window signal, wherein the decomposition interference index is used to represent the probability of window signal interference with the local mean decomposition operation; Based on the decomposition interference index of each window signal, performing local mean decomposition on the original audio signal to obtain a plurality of product function components and a monotonic residual signal.

[0033] Exemplarily, the frame length can be 50 milliseconds (ms), and the window function used in the windowing processing can be a Hamming window function.

[0034] After Fourier transform of each window signal, the obtained signal transform information will include a plurality of sinusoidal waves with different amplitudes, frequencies and phases corresponding to the window signal.

[0035] Each window signal includes a plurality of sinusoidal waves, and the frequencies of the plurality of sinusoidal waves are different, so it can also be understood that each window signal includes a plurality of frequencies, and the amplitude and phase of each frequency are the amplitude and phase of the corresponding sinusoidal wave.

[0036] When there is a sudden impact sound in the industrial field (such as hammering, valve explosion, accidental operation), these strong mutation signals will cause the audio extreme points to increase sharply in a short time, and the energy is high. At this time, if the local mean decomposition is directly performed, there will be a high risk of fitting distortion, which will cause PF decomposition error, false modal or modal aliasing, and affect the determination of the subsequent noise signal.

[0037] Based on this, in the above process, the signal energy difference between adjacent window signals is analyzed to obtain the decomposition interference index of each window signal, that is, the probability of the existence of the burst impact sound in each signal segment of the original audio signal is quantitatively represented, so as to guide the subsequent local mean decomposition operation, suppress the adverse effect of the strong mutation signal generated by the burst impact sound on the local mean decomposition operation, and make the plurality of product function components and a monotonic residual signal obtained after decomposition more accurate and reliable.

[0038] Further, the signal transformation information includes a plurality of signal energies of a plurality of frequencies in the corresponding window signal. The signal energy difference between adjacent window signals is analyzed based on the plurality of signal transformation information to obtain the decomposition interference index of each window signal, including: Based on the plurality of signal transformation information, the energy front difference value and the energy rear difference value corresponding to each frequency in each window signal are obtained, wherein the energy front difference value is the signal energy difference between the corresponding frequency in the previous window signal and the corresponding frequency in the corresponding window signal, and the energy rear difference value is the signal energy difference between the corresponding frequency in the corresponding window signal and the corresponding frequency in the next window signal. The sum value of the energy front difference value and the energy rear difference value corresponding to each frequency in each window signal is calculated to obtain the energy sum value corresponding to each frequency in each window signal. The product of the signal energy of each frequency in each window signal and the energy sum value is calculated to obtain the energy product of each frequency in each window signal. The sum value of the plurality of energy products of the plurality of frequencies in each window signal is calculated to obtain the energy reference value corresponding to each window signal. The ratio of the energy reference value corresponding to each window signal and the energy entropy value corresponding to each window signal is calculated to obtain the decomposition interference index of each window signal, wherein the energy entropy value is used to represent the information entropy of the plurality of signal energies of the plurality of frequencies in the corresponding window signal.

[0039] The signal energy difference between the corresponding frequency in the previous window signal and the corresponding frequency in the corresponding window signal is specifically the difference value between the signal energy of the corresponding frequency in the previous window signal and the signal energy of the corresponding frequency in the corresponding window signal, wherein the signal energy can be understood as the square of the amplitude of the corresponding frequency.

[0040] Similarly, the signal energy difference between the corresponding frequency in the corresponding window signal and the corresponding frequency in the next window signal is specifically the difference value between the signal energy of the corresponding frequency in the corresponding window signal and the signal energy of the corresponding frequency in the next window signal.

[0041] The energy reference value corresponding to the window signal is used to represent the signal energy fluctuation degree (usually represented as signal energy rapid attenuation) in the corresponding window signal. The greater the energy reference value, the more severe the signal energy fluctuation degree in the corresponding window signal, that is, the higher the probability (represented by the decomposition interference index) of containing the burst impact sound in the corresponding window signal.

[0042] Similarly, the greater the energy entropy value corresponding to the window signal, the more information entropy of the plurality of signal energies of the plurality of frequencies in the corresponding window signal, that is, the more audio signals of different frequencies contained in the corresponding window signal, and the higher the possibility of containing other noises in the corresponding window signal except the sound of the device running, and the lower the probability of containing the burst impact sound in the corresponding window signal.

[0043] In the application, in order to avoid the influence of extreme values, the energy front difference value and the energy rear difference value corresponding to each frequency in each window signal can be normalized (such as using a sigmoid function for normalization processing) to normalize the energy front difference value and the energy rear difference value corresponding to each frequency in each window signal to the interval (0, 1).

[0044] Exemplarily, the information entropy of the plurality of signal energies of the plurality of frequencies in the window signal can be obtained by the following process: The sum of the plurality of signal energies of the plurality of frequencies in the window signal is calculated to obtain the total energy value corresponding to the window signal; The ratio of the plurality of signal energies of the plurality of frequencies in the window signal to the total energy value corresponding thereto is calculated respectively to obtain the signal energy proportion of the plurality of frequencies in the window signal, wherein the signal energy proportion of the plurality of frequencies in the window signal can be represented as , (j = 1,..., num), num is the total number of the plurality of frequencies in the window signal; Correspondingly, the information entropy of the plurality of signal energies of the plurality of frequencies in the window signal can be represented as: In one example, the decomposition interference index of a window signal can be represented as: wherein, represents the decomposition interference index of the window signal, n represents the total number of the plurality of frequencies in the window signal, represents the normalized value of the signal energy of the sinusoidal wave of the i-th frequency in the window signal (the normalization process can be completed by using a Z-Score standardization algorithm), and represents the energy front difference value and the energy rear difference value corresponding to the sinusoidal wave of the i-th frequency in the window signal, represents the energy entropy value corresponding to the window signal.

[0045] Further, in the multiple decomposition operations corresponding to the local mean decomposition, the extreme value data of each local extreme point in each decomposition operation is the product of the original extreme value of the local extreme point and the correction coefficient corresponding to the local extreme point, the correction coefficient being the ratio of the first interference index of the corresponding local extreme point and the second interference index of the corresponding local extreme point, the first interference index being the decomposition interference index of the window signal to which the corresponding local extreme point belongs, and the second interference index being the sum of the decomposition interference index of the window signal to which the previous local extreme point of the corresponding local extreme point belongs and the decomposition interference index of the window signal to which the next local extreme point of the corresponding local extreme point belongs.

[0046] Here, the original extreme value of the local extreme point can be understood as the original amplitude of the local extreme point.

[0047] Specifically, the multiple decomposition operations corresponding to the local mean decomposition are as follows: Step one, extracting a plurality of local extreme points (including local maximum values and local minimum values) from the original audio signal.

[0048] Step two, in the above plurality of local extreme points, calculating the product of the amplitude of each local extreme point and the correction coefficient corresponding to the local extreme point to obtain the corrected amplitude of each local extreme point.

[0049] Step three, in the above plurality of local extreme points, taking the average of the corrected amplitudes of adjacent local extreme points to obtain the local mean.

[0050] Step four, then in the above plurality of local extreme points, calculating half of the absolute value of the amplitude difference between adjacent local extreme points to obtain the envelope estimate. Step five, using the moving average method to smooth the local mean and the envelope estimate respectively to obtain the smoothed local mean function and the local envelope estimate function . Step six, subtracting the local mean function from the original audio signal to obtain the zero mean signal . Step seven, demodulating , specifically by dividing by to obtain the envelope normalized signal . Step eight, repeating steps one to seven above until the local envelope estimate function , at this time the envelope normalized signal obtained is a pure frequency modulation signal, which can be defined as It is to be noted that the local extreme points used in each decomposition operation are extracted from the zero-mean signal obtained in the previous decomposition operation, and the local extreme points used in the first decomposition operation are extracted from the original audio signal, and the audio signal used to subtract the local mean function in each decomposition operation is the zero-mean signal obtained in the previous decomposition operation.

[0051] Step nine, multiplying all the local envelope estimation functions obtained through the above multiple decomposition operations to obtain an envelope signal Then, the first product function component (i.e., the PF component) can be expressed as Step ten, subtracting the envelope signal from the original audio signal to obtain a residual signal, and repeating steps one to nine (it is to be noted that in the repeated execution of the above steps, the original audio signal in step one is changed to the residual signal obtained in the last time, and similarly, the original audio signal in step six is changed to the residual signal obtained in the last time minus the zero-mean function obtained by subtracting the local mean functions), until the residual signal is a monotonic function (i.e., a monotonic residual signal), at which time the original audio signal is decomposed into a plurality of PF components and a monotonic residual signal, i.e., the decomposition is completed.

[0052] In the above process, the decomposition interference index based on each window signal is used to calculate the correction coefficient of each local extreme point, so that the amplitude of each local extreme point participating in the mean and envelope estimation calculation is adaptively corrected according to the probability of the local extreme point belonging to the window signal containing the sudden impact sound, so as to suppress the extreme amplitude influence of the local extreme point caused by the sudden impact sound, avoid the fitting distortion risk of the local mean estimation operation, and make the obtained plurality of product function components and a monotonic residual signal more accurate.

[0053] Step S2, analyzing each product function component based on the original audio signal to obtain a device characteristic index of each product function component.

[0054] The device characteristic index is used to represent the probability of the product function component corresponding to the equipment running sound in the industrial production environment.

[0055] Further, the analysis of each product function component based on the original audio signal to obtain a device characteristic index of each product function component comprises: performing a fast Fourier transform on each product function component to obtain a plurality of sinusoidal wave components corresponding to each product function component; determine a sinusoidal acoustic index and a frequency difference index of each sinusoidal component corresponding to each product function component based on the original audio signal, the sinusoidal acoustic index being a frequency acoustic index of a frequency component in the plurality of frequency components included in the original audio signal, which has a minimum frequency difference with the corresponding sinusoidal component, the frequency acoustic index being used to indicate a probability that the corresponding frequency component indicates the equipment running sound in the corresponding industrial production environment, the frequency difference index being used to indicate a minimum frequency difference between the corresponding sinusoidal component and the plurality of frequency components included in the original audio signal; determine a component feature index of each sinusoidal component corresponding to each product function component based on the sinusoidal acoustic index, the frequency difference index and the amplitude of each sinusoidal component corresponding to each product function component; calculate a sum of the component feature indexes of the plurality of sinusoidal components corresponding to each product function component to obtain an equipment feature index of each product function component.

[0056] In the above process, the component feature index of each sinusoidal component corresponding to each product function component is calculated based on the sinusoidal acoustic index, the frequency difference index and the amplitude of each sinusoidal component corresponding to each product function component, that is, by the amplitude of each sinusoidal wave, the probability that the most similar frequency of the sinusoidal wave in the original audio signal indicates the equipment running sound, and the difference degree (or the similarity degree) of the most similar frequency of the sinusoidal wave in the original audio signal, the probability that each sinusoidal component corresponding to each product function component indicates the equipment running sound is comprehensively evaluated, and then the plurality of probabilities of the plurality of sinusoidal components of each product function component are summarized to obtain a more accurate equipment feature index.

[0057] Exemplarily, the equipment feature index of one product function component may be represented as: wherein M represents the total number of the plurality of sinusoidal components included in the product function component, represents the amplitude of the mth sinusoidal component in the product function component (to avoid the influence of extreme values, the amplitude can be normalized), represents the sinusoidal acoustic index of the mth sinusoidal component in the product function component, represents the frequency difference index of the mth sinusoidal component in the product function component.

[0058] Further, the step of obtaining the frequency acoustic index of each frequency component included in the original audio signal comprises: split the original audio signal based on a preset frame length to obtain a plurality of signal frames; The plurality of signal frames are respectively subjected to windowing processing to obtain a plurality of window signals; The amplitude variation of each frequency component in the plurality of window signals is analyzed to obtain an amplitude fluctuation index corresponding to each frequency component; The phase variation of each frequency component in the plurality of window signals is analyzed to obtain a phase variation index corresponding to each frequency component; The frequency acoustic index of each frequency component is determined according to the amplitude fluctuation index and the phase variation index corresponding to each frequency component.

[0059] In this process, the amplitude variation and the phase variation of each frequency component in the plurality of window signals are analyzed to evaluate the amplitude fluctuation and the phase variation of each frequency component, and the frequency acoustic index of each frequency component in the original audio signal is determined accordingly, which can more accurately evaluate the probability of each frequency component indicating the device running sound.

[0060] Wherein, the greater the amplitude fluctuation index, the more dramatic the amplitude variation of the corresponding frequency component between the plurality of window signals, the less matched the sound characteristics generated by the stable device running process, and the lower the probability of the corresponding frequency component indicating the device running sound (specifically, the normal running condition).

[0061] And the greater the phase variation index, the more significant the phase variation of the corresponding frequency between the plurality of window signals, the more matched the phase trajectory anomaly (such as phase jump and phase drift) in the abnormal device running process, and the higher the probability of the corresponding frequency component indicating the device running sound (specifically, the abnormal running condition, such as bearing wear, gear crack, etc.).

[0062] Exemplarily, the frequency acoustic index of the frequency component can be represented as: Wherein, is used to represent the phase variation index of the frequency component, is used to represent the amplitude fluctuation index of the frequency component.

[0063] Specifically, the analysis of the amplitude variation of each frequency component in the plurality of window signals to obtain the amplitude fluctuation index corresponding to each frequency component comprises: The absolute value of the amplitude difference between adjacent two window signals of each frequency component is calculated to obtain a plurality of window amplitude differences corresponding to each frequency component; The variance of the plurality of window amplitude differences corresponding to each frequency component is calculated to obtain the amplitude fluctuation index corresponding to each frequency component.

[0064] ​In the flow, the variance of the amplitude difference between adjacent window signals at each frequency component is selected to quantify the amplitude fluctuation of each frequency component in the plurality of window signals, instead of the variance of the amplitude of each frequency in the plurality of window signals, so that the determined amplitude fluctuation index is more accurate.

[0065] Specifically, the analysis of the phase change of each frequency component in the plurality of window signals obtains a phase change index corresponding to each frequency component, including: The absolute value of the phase difference between the two adjacent window signals of each frequency component is calculated to obtain a phase difference sequence corresponding to each frequency component. In the phase difference sequence corresponding to each frequency component, a sequence mean value and a plurality of sequence peak values corresponding to each frequency component are determined, and the difference between the plurality of sequence peak values and the sequence mean value is calculated to obtain a plurality of sequence jump values corresponding to each frequency component. The mean value of the plurality of sequence jump values corresponding to each frequency component is calculated to obtain a phase jump index corresponding to each frequency component. The phase difference sequence corresponding to each frequency component is curve-fitted to obtain a phase difference fitting curve corresponding to each frequency component. The absolute value of the derivative of each sequence point in the phase difference sequence corresponding to each frequency component in the corresponding phase difference fitting curve is calculated to obtain a plurality of sequence point derivative values corresponding to each frequency component. The mean value of the plurality of sequence point derivative values corresponding to each frequency component is determined as a phase drift index corresponding to each frequency component. According to the phase drift index and the phase jump index corresponding to each frequency component, a phase change index corresponding to each frequency component is obtained.

[0066] The sequence mean value is the average value of the plurality of sequence elements included in the corresponding phase difference sequence, the sequence peak value is the value of the peak element in the plurality of sequence elements included in the corresponding phase difference sequence, and the peak element is the sequence element in the plurality of sequence elements included in the corresponding phase difference sequence, which is greater than the previous sequence element and greater than the next sequence element.

[0067] The phase jump index is used to represent the degree of deviation of the phase peak value in the corresponding frequency component from the phase mean value, and the greater the phase jump index, the more significant the phase jump feature of the corresponding frequency component, and the more matched the phase trajectory anomaly in the abnormal operation process of the device.

[0068] The phase drift index is used to represent the degree of phase value deviation in the corresponding frequency component, and the greater the phase drift index, the more significant the phase drift feature of the corresponding frequency component, and the more matched the phase trajectory anomaly in the abnormal operation process of the device.

[0069] In the application, the phase value of the frequency component in the window signal may be expressed as: wherein, represents the phase angle value of the frequency component in the window signal, and cos(.) represents the cosine function.

[0070] In one example, the product of the phase drift index and the phase jump index corresponding to each frequency component can be calculated to obtain the phase change index corresponding to each frequency component, and in this case, the curve fitting process can be completed based on the least square method.

[0071] Step S3, analyzing the amplitude change of each product function component to obtain the vocal interference index of each product function component.

[0072] wherein, the vocal interference index is used to represent the probability of the product function component corresponding to the vocal.

[0073] Further, the analyzing the amplitude change of each product function component to obtain the vocal interference index of each product function component comprises: performing signal splitting on each product function component to obtain a plurality of signal segments corresponding to each product function component; performing clustering on the plurality of signal segments corresponding to each product function component to obtain a plurality of clusters corresponding to each product function component; taking the amplitude mean value of the cluster as a comparison index, sorting the plurality of clusters corresponding to each product function component in descending order to obtain a cluster sequence corresponding to each product function component; in the cluster sequence corresponding to each product function component, calculating the difference between the amplitude mean values of adjacent two clusters to obtain a plurality of cluster amplitude differences corresponding to each product function component, wherein the cluster amplitude difference is the difference between the amplitude mean values of the corresponding cluster and the previous cluster; in the plurality of cluster amplitude differences corresponding to each product function component, determining the cluster corresponding to the largest cluster amplitude difference as the critical cluster corresponding to each product function component, and determining the plurality of clusters after the critical cluster in the corresponding cluster sequence as a plurality of strong signal clusters corresponding to each product function component; analyzing the amplitude change between the plurality of strong signal clusters corresponding to each product function component to obtain the vocal interference index of each product function component.

[0074] Exemplarily, each product function component can be signal split based on a preset segment length to obtain a plurality of signal segments corresponding to each product function component, wherein the segment length can be 50 milliseconds.

[0075] The clustering process can be completed by using a K-Means algorithm, and the K value of the K-Means algorithm can be obtained according to an elbow method. The mean of the plurality of signal amplitudes in the corresponding signal segment and the starting time of the signal segment (i.e., the time indicated by the first signal point of the signal segment) can be used as clustering indicators. That is, the mean amplitudes of different signal segments in the same cluster (i.e., the mean of the plurality of signal amplitudes in the signal segment) are close, and the starting times are also close.

[0076] In the above process, the amplitude mean value of the cluster is selected as the comparison indicator, the plurality of clusters corresponding to each product function component are sorted in descending order to obtain a cluster sequence corresponding to each product function component, and the cluster with the largest amplitude mean value difference between clusters is selected as the critical cluster, so as to determine a plurality of strong signal clusters corresponding to the strong signal from the plurality of clusters, thereby avoiding the interference of noise corresponding to low signal energy as much as possible, and making the determined vocal interference index more accurate.

[0077] Further, the amplitude variation between the plurality of strong signal clusters corresponding to each product function component is analyzed to obtain a vocal interference index of each product function component, including: In the plurality of strong signal clusters corresponding to each product function component, the cluster difference degree between each strong signal cluster and the adjacent next strong signal cluster is analyzed to obtain a plurality of cluster difference values corresponding to each product function component. In the plurality of strong signal clusters corresponding to each product function component, the difference between the cluster amplitude mean values of each strong signal cluster and the adjacent next strong signal cluster is calculated to obtain a plurality of target amplitude differences corresponding to each product function component. The ratio of each target amplitude difference and the corresponding cluster difference value corresponding to each product function component is calculated to obtain a plurality of cluster interference indexes corresponding to each product function component. The mean of the plurality of cluster interference indexes corresponding to each product function component is calculated to obtain a vocal interference index of each product function component.

[0078] The cluster amplitude mean value of the strong signal cluster is the average of the plurality of amplitude mean values of the plurality of signal segments included in the strong signal cluster.

[0079] In the above process, the cluster difference degree between each strong signal cluster and its adjacent next strong signal cluster and the difference between the cluster amplitude mean values are analyzed to accurately evaluate the probability of each strong signal cluster corresponding to the voice interference from the difference between the clusters as a whole and the difference in the amplitude dimension, so that the determined voice interference index is more accurate.

[0080] The greater the target amplitude difference is, the greater the signal amplitude (which can also be understood as signal energy) increase between the adjacent two strong signal clusters is, the more matched the situation of gradually increasing voice in the industrial production environment (due to the over-strong background sound leading to repeated and gradually increasing voice communication) is, and the greater the probability of the corresponding product function component indicating voice is.

[0081] The greater the cluster difference value is, the lower the similarity between the adjacent two strong signal clusters is, the lower the matching degree of the repeated voice in the industrial production environment is, and the smaller the probability of the corresponding product function component indicating voice is.

[0082] Exemplarily, the cluster difference degree between each strong signal cluster and its adjacent next strong signal cluster can be quantified by calculating the dynamic time warping distance between each strong signal cluster and its adjacent next strong signal cluster, so as to obtain a plurality of cluster difference values corresponding to each product function component. The target amplitude difference is specifically the difference between the cluster amplitude mean value of the next cluster adjacent to the corresponding cluster and the cluster amplitude mean value of the corresponding cluster.

[0083] In one example, the voice interference index of the product function component can be represented as: wherein U represents the total number of strong signal clusters included in the product function component, represents the target amplitude difference of the rth strong signal cluster included in the product function component (i.e. the difference between the cluster amplitude mean value of the r+1th strong signal cluster included in the product function component and the cluster amplitude mean value of the rth strong signal cluster), represents the cluster difference value of the rth strong signal cluster included in the product function component (i.e. the dynamic time warping distance between the rth strong signal cluster and the r+1th strong signal cluster included in the product function component).

[0084] Step S4, determining the interference weight of each product function component according to the device feature index and the voice interference index of each product function component.

[0085] Specifically, the ratio of the voice interference index and the device feature index of each product function component is calculated to obtain the initial weight of each product function component. The initial weight of each product function component is normalized to obtain an interference weight of each product function component.

[0086] In step S5, an interference signal is obtained according to the plurality of product function components, the monotonic residual signal and the interference weight of each product function component.

[0087] The interference signal is used to generate a noise reduction signal, the noise reduction signal has the same amplitude as the interference signal but has opposite phase, and the noise reduction signal is used to neutralize the noise in the original audio signal.

[0088] In one example, the plurality of product function components can be weighted and calculated to obtain a weighted average signal based on the interference weight of each product function component, and then the weighted average signal and the monotonic residual signal are added to obtain the interference signal.

[0089] It should be noted that the monotonic residual signal does not have the non-stationary, wideband, nonlinear and other characteristics of the equipment running sound, and thus can be directly regarded as environmental noise as a basic component signal of the interference signal.

[0090] In applications, the noise reduction signal and the original audio signal can be added to obtain a target audio signal after adaptive noise reduction processing. The target audio signal can be used to analyze the running state, fault risk, etc. of each device in the industrial production environment, and the use of the target audio signal is not limited by the present application.

[0091] In summary, the present application adapts to the non-stationary, wideband, nonlinear characteristics of the original audio signal from the industrial production environment by performing local mean decomposition on the original audio signal to obtain a plurality of product function components and a monotonic residual signal, and then obtains the equipment characteristic index and the vocal interference index of each product function component to determine the probability of each product function component corresponding to the equipment running sound and the probability of each product function component corresponding to the vocal sound, i.e., to determine the probability of each product function component belonging to the key sound that needs to be retained and the probability of each product function component belonging to the noise that needs to be filtered out, so as to accurately evaluate the interference weight of each product function component from multiple aspects, and to determine a more accurate interference signal, i.e., to accurately identify the environmental noise and vocal interference that overlap with the equipment running sound in the industrial production environment, thereby improving the noise reduction effect of the industrial audio.

[0092] The present application provides a wideband adaptive filtering and reverse sound wave synthesis system of an intelligent noise reduction device, please refer to Figure 2 which shows a structure schematic diagram of a wideband adaptive filtering and reverse sound wave synthesis system 200 of an intelligent noise reduction device according to an embodiment of the present application, and the system comprises: The signal decomposition module 201 is configured to perform local mean decomposition on an original audio signal to obtain a plurality of product function components and a monotonic residual signal, wherein the original audio signal is from an industrial production environment. The equipment feature analysis module 202 is configured to analyze each product function component based on the original audio signal to obtain an equipment feature index of each product function component, wherein the equipment feature index is used to represent a probability of the product function component corresponding to a device running sound in the industrial production environment. The human voice analysis module 203 is configured to analyze an amplitude variation of each product function component to obtain a human voice interference index of each product function component, wherein the human voice interference index is used to represent a probability of the product function component corresponding to human voice. The interference analysis module 204 is configured to determine an interference weight of each product function component according to the equipment feature index and the human voice interference index of each product function component. The interference determination module 205 is configured to obtain an interference signal according to the plurality of product function components, the monotonic residual signal and the interference weight of each product function component, wherein the interference signal is used to generate a noise reduction signal, the noise reduction signal and the interference signal have the same amplitude but opposite phases, and the noise reduction signal is used to neutralize noise in the original audio signal.

[0093] It should be noted that the system provided in the above embodiments is only used as an example for the division of the above functional modules. In actual applications, the above functions can be completed by different functional modules according to needs, that is, the internal structure of the computer device is divided into different functional modules to complete all or part of the above described functions. In addition, the wideband adaptive filtering and reverse sound wave synthesis system and the wideband adaptive filtering and reverse sound wave synthesis method of the intelligent noise reduction device provided in the above embodiments belong to the same concept, and the specific implementation process is detailed in the method embodiments, which will not be repeated here.

[0094] The embodiment of the present application also provides an electronic device. Please refer to Figure 3 The electronic device can include a processor 301, a memory 302, and a program 3021 stored in the memory 302 and executable on the processor 301.

[0095] When the program 3021 is executed by the processor 301, it can implement Figure 1 Any step in the corresponding method embodiment and achieve the same beneficial effects, which will not be repeated here.

[0096] Those skilled in the art can understand that all or part of the steps of the above-mentioned embodiment methods can be completed by program instructions related to hardware, and the program can be stored in a readable medium.

[0097] The embodiment of the present application also provides a readable storage medium, wherein the readable storage medium stores a computer program, and the computer program can realize the above method when executed by a processor. Figure 1 Any step in the corresponding method embodiment can be achieved, and the same technical effects can be achieved. To avoid repetition, it will not be repeated here.

[0098] The computer readable storage medium of the embodiment of the present application can adopt any combination of one or more computer readable media. The computer readable medium can be a computer readable signal medium or a computer readable storage medium. The computer readable storage medium may, for example, be an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or apparatus, or any combination thereof. More specific examples (non-exhaustive list) of the computer readable storage medium include: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this document, the computer readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, device or apparatus.

[0099] The computer readable signal medium can include a data signal propagating in a baseband or as part of a carrier wave propagating through a transmission medium, in which the computer readable program code is embodied. Such a propagating data signal can take many forms, including but not limited to electro-magnetic, optical, or any suitable combination thereof. The computer readable signal medium can also be any computer readable medium that is not a computer readable storage medium and that can transmit, propagate or transport program for use by or in connection with an instruction execution system, apparatus or device.

[0100] The program code contained on the storage medium can be transmitted in any suitable medium, including but not limited to wireless, wire, cable, optical fiber, RF, etc., or any suitable combination thereof.

[0101] Computer program code for carrying out operations of the present application can be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C++ or the like, and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider).

[0102] The embodiments of the present application further provide a computer program product, which, when running on a computer, enables the computer to perform the above related steps to realize the method for wideband adaptive filtering and inverse sound wave synthesis of the intelligent noise reduction device provided by the above embodiments.

[0103] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or can be advantageous.

[0104] Each of the embodiments in the specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the difference from other embodiments.

Claims

1. A method for broadband adaptive filtering and reverse sound wave synthesis of an intelligent noise reduction device, characterized in that: The method comprises: Performing local mean decomposition on an original audio signal to obtain a plurality of product function components and a monotonic residual signal, wherein the original audio signal originates from an industrial production environment; Analyzing each product function component based on the original audio signal to obtain a device characteristic index for each product function component, wherein the device characteristic index is used to indicate the probability that the product function component corresponds to the sound of equipment operation in an industrial production environment; Analyzing the amplitude change of each product function component to obtain a human voice interference index for each product function component, wherein the human voice interference index is used to indicate the probability that the product function component corresponds to a human voice; Determining an interference weight of each product function component according to a device characteristic index and a human voice interference index of each product function component; An interference signal is obtained based on multiple product function components, a monotonic residual signal, and an interference weight of each product function component. The interference signal is used to generate a noise reduction signal. The noise reduction signal and the interference signal have the same amplitude but opposite phase. The noise reduction signal is used to neutralize noise in the original audio signal.

2. The method for broadband adaptive filtering and inverse sound wave synthesis of an intelligent noise reduction device according to claim 1, characterized in that: The method of performing local mean decomposition on the original audio signal to obtain multiple product function components and a monotonic residual signal includes: Splitting the original audio signal based on a preset frame length to obtain multiple signal frames; Performing windowing processing on multiple signal frames respectively to obtain multiple window signals; Performing Fourier transform on multiple window signals respectively to obtain multiple signal transformation information; Analyzing the signal energy difference between adjacent window signals based on the plurality of signal transformation information to obtain a decomposition interference index for each window signal, wherein the decomposition interference index is used to indicate a probability that the window signal interferes with the local mean decomposition operation; Based on the decomposition interference index of each window signal, the original audio signal is subjected to local mean decomposition to obtain multiple product function components and a monotonic residual signal.

3. The method for wideband adaptive filtering and inverse sound wave synthesis of an intelligent noise reduction device according to claim 2, characterized in that: The signal transformation information includes a plurality of signal energies corresponding to a plurality of frequencies in the window signal; The analyzing the signal energy difference between adjacent window signals based on the plurality of signal transformation information to obtain a decomposition interference index of each window signal includes: Based on the multiple signal transformation information, obtaining a front energy difference value and a rear energy difference value corresponding to each frequency in each window signal, wherein the front energy difference value is a signal energy difference between a corresponding frequency in a previous window signal and a corresponding frequency in a corresponding window signal, and the rear energy difference value is a signal energy difference between a corresponding frequency in a corresponding window signal and a corresponding frequency in a subsequent window signal; Calculate the sum of the energy difference before and after each frequency in each window signal to obtain the energy sum corresponding to each frequency in each window signal; Calculating the product of the signal energy of each frequency in each window signal and the corresponding energy sum value to obtain the energy product of each frequency in each window signal; Calculating a sum of multiple energy products of multiple frequencies in each window signal to obtain an energy reference value corresponding to each window signal; The ratio of the energy reference value and the corresponding energy entropy value corresponding to each window signal is calculated to obtain the decomposition interference index of each window signal, wherein the energy entropy value is used to represent the information entropy of multiple signal energies of multiple frequencies in the corresponding window signal.

4. The method for broadband adaptive filtering and inverse sound wave synthesis of an intelligent noise reduction device according to claim 2, characterized in that: In the multiple decomposition operations corresponding to the local mean decomposition, the extreme value data of each local extreme point in each decomposition operation is the product of the original extreme value of the local extreme point and the correction coefficient corresponding to the local extreme point, and the correction coefficient is the ratio of the first interference index and the second interference index of the corresponding local extreme point. The first interference index is the decomposition interference index of the window signal to which the corresponding local extreme point belongs, and the second interference index is the sum of the decomposition interference index of the window signal to which the previous local extreme point of the corresponding local extreme point belongs and the decomposition interference index of the window signal to which the next local extreme point of the corresponding local extreme point belongs.

5. The method for wideband adaptive filtering and inverse sound wave synthesis of an intelligent noise reduction device according to claim 1, characterized in that: The analyzing each product function component based on the original audio signal to obtain a device characteristic index of each product function component includes: Performing a fast Fourier transform on each product function component to obtain multiple sinusoidal wave components corresponding to each product function component; Determining, based on the original audio signal, a sinusoidal wave acoustic index and a frequency difference index for each sinusoidal wave component corresponding to each product function component, wherein the sinusoidal wave acoustic index is a frequency acoustic index of a frequency component having the smallest frequency difference with the corresponding sinusoidal wave component among multiple frequency components included in the original audio signal, the frequency acoustic index being used to indicate a probability that the corresponding frequency component indicates a sound of equipment operation in a corresponding industrial production environment, and the frequency difference index being used to indicate a minimum frequency difference between the corresponding sinusoidal wave component and the multiple frequency components included in the original audio signal; Determining a component characteristic index of each sine wave component corresponding to each product function component based on a sine wave acoustic index, a frequency difference index, and an amplitude of each sine wave component corresponding to each product function component; The sum of multiple component characteristic indices of multiple sinusoidal wave components corresponding to each product function component is calculated to obtain a device characteristic index of each product function component.

6. The method for wideband adaptive filtering and inverse sound wave synthesis of an intelligent noise reduction device according to claim 5, characterized in that: The steps of obtaining the frequency acoustic index of each frequency component included in the original audio signal include: Splitting the original audio signal based on a preset frame length to obtain multiple signal frames; Performing windowing processing on multiple signal frames respectively to obtain multiple window signals; Analyze the amplitude change of each frequency component in multiple window signals to obtain the amplitude fluctuation index corresponding to each frequency component; Analyze the phase change of each frequency component in multiple window signals to obtain the phase change index corresponding to each frequency component; The frequency acoustic index of each frequency component is determined according to the amplitude fluctuation index and the phase change index corresponding to each frequency component.

7. The method for wideband adaptive filtering and inverse sound wave synthesis of an intelligent noise reduction device according to claim 6, characterized in that: The analyzing the amplitude change of each frequency component in the multiple window signals to obtain the amplitude fluctuation index corresponding to each frequency component includes: Calculating the absolute value of the amplitude difference of each frequency component between two adjacent window signals to obtain multiple window amplitude differences corresponding to each frequency component; The variance of the multiple window amplitude differences corresponding to each frequency component is calculated to obtain the amplitude fluctuation index corresponding to each frequency component.

8. The method for wideband adaptive filtering and inverse sound wave synthesis of an intelligent noise reduction device according to claim 6, characterized in that: The analyzing the phase change of each frequency component in the multiple window signals to obtain the phase change index corresponding to each frequency component includes: Calculate the absolute value of the phase difference between two adjacent window signals of each frequency component to obtain a phase difference sequence corresponding to each frequency component; In the phase difference sequence corresponding to each frequency component, a sequence mean and multiple sequence peaks corresponding to each frequency component are determined, and differences between the multiple sequence peaks and the sequence mean are respectively calculated to obtain multiple sequence jump values ​​corresponding to each frequency component; Calculating the average of multiple sequence jump values ​​corresponding to each frequency component to obtain a phase jump index corresponding to each frequency component; Perform curve fitting on the phase difference sequence corresponding to each frequency component to obtain a phase difference fitting curve corresponding to each frequency component; Calculating the absolute value of the derivative of each sequence point in the phase difference sequence corresponding to each frequency component in the corresponding phase difference fitting curve to obtain multiple sequence point derivative values ​​corresponding to each frequency component; The average of multiple sequence point derivative values ​​corresponding to each frequency component is determined as the phase drift index corresponding to each frequency component; According to the phase drift index and the phase jump index corresponding to each frequency component, the phase change index corresponding to each frequency component is obtained.

9. The method for wideband adaptive filtering and inverse sound wave synthesis of an intelligent noise reduction device according to claim 1, characterized in that: The analysis of the amplitude change of each product function component to obtain the human voice interference index of each product function component includes: Performing signal splitting on each product function component to obtain multiple signal segments corresponding to each product function component; Clustering multiple signal segments corresponding to each product function component to obtain multiple clusters corresponding to each product function component; Using the cluster amplitude mean as a comparison index, sort the multiple clusters corresponding to each product function component in descending order to obtain a cluster sequence corresponding to each product function component; In the cluster sequence corresponding to each product function component, the difference in the amplitude means between two adjacent clusters is calculated to obtain multiple cluster amplitude differences corresponding to each product function component, wherein the cluster amplitude difference is the difference in the amplitude mean between the corresponding cluster and its previous cluster; Among the multiple cluster amplitude differences corresponding to each product function component, the cluster corresponding to the largest cluster amplitude difference is determined as the critical cluster corresponding to each product function component, and the multiple clusters located after the critical cluster in the corresponding cluster sequence are determined as the multiple strong signal clusters corresponding to each product function component; The amplitude changes between multiple strong signal clusters corresponding to each product function component are analyzed to obtain the human voice interference index of each product function component.

10. The method for wideband adaptive filtering and inverse sound wave synthesis of an intelligent noise reduction device according to claim 9, characterized in that: The analyzing the amplitude changes between the multiple strong signal clusters corresponding to each product function component to obtain the human voice interference index of each product function component includes: In the multiple strong signal clusters corresponding to each product function component, the cluster difference degree between each strong signal cluster and its adjacent next strong signal cluster is analyzed to obtain the multiple cluster difference values ​​corresponding to each product function component; In the multiple strong signal clusters corresponding to each product function component, the difference between the cluster amplitude means of each strong signal cluster and its adjacent next strong signal cluster is calculated to obtain multiple target amplitude differences corresponding to each product function component; Calculating the ratio of each target amplitude difference corresponding to each product function component to the corresponding cluster difference value to obtain multiple cluster interference indices corresponding to each product function component; The mean of the multiple cluster interference indices corresponding to each product function component is calculated to obtain the human voice interference index of each product function component.

Images