A howling frequency point screening method, device and system of audio data

Abstract

The application discloses a howling frequency point screening method, device and system of audio data, through receiving the first audio data after noise processing and obtaining the first frequency spectrum data by performing frequency spectrum transformation on the first audio data, determining the real part data and imaginary part data corresponding to each point in the first frequency spectrum data and calculating the power value corresponding to each point, determining the first value through the comparison result of the power value and the first standard value, and determining whether the frequency screening is needed for the first frequency spectrum data according to the comparison result of the first value and the first threshold value to obtain the corresponding first howling frequency point. The application performs a screening on the first frequency spectrum data before performing the frequency screening on the first frequency spectrum data, improves the accuracy of the subsequent frequency screening to obtain the howling frequency point, and also makes the suppression precision of the notch filter for the audio data howling signal be effectively improved according to the parameter update of the howling frequency point obtained by screening.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of audio data transmission, in particular to a howling frequency point screening method, device and system of audio data. BACKGROUND

[0002] In the current audio system, if the gain between the loudspeaker(s) and the microphone(s) is greater than 1 and in phase, that is, the audio signal of the audio system is affected by positive feedback, after self-excited oscillation, the audio signal becomes larger and larger, and the howling sound produced is annoying and needs to be eliminated. How to eliminate such feedback howling sound has become a big problem in the industry.

[0003] The prior art uses frequency shifting technology and phase adjustment technology to process the howling feedback suppressor, but the above two technologies have the problems of low search accuracy of howling frequency points and slow calculation speed, and at the same time, since there is no bottom noise removal, the problem of low suppression accuracy occurs when the howling is suppressed at the end. SUMMARY

[0004] The present application provides a howling frequency point screening method, device and system of audio data to achieve the technical effect of accurate screening of howling frequency points and effective removal of bottom noise.

[0005] To solve the above technical problems, the present application provides a howling frequency point screening method of audio data, comprising the following steps:

[0006] Receiving first audio data, performing frequency spectrum transformation on the first audio data to obtain first frequency spectrum data and determining the real part data and the imaginary part data corresponding to each point in the first frequency spectrum data; wherein the first audio data is obtained by noise processing of first to-be-processed data;

[0007] According to the real part data and the imaginary part data corresponding to each point in the first frequency spectrum data, the power value corresponding to each point in the first frequency spectrum data is calculated and obtained, and the first value is determined according to the comparison result of the power value and the preset first standard value;

[0008] Comparing the first value with the preset first threshold value, if the first value is greater than or equal to the first threshold value, frequency screening is performed on the first frequency spectrum data to obtain a first howling frequency point;

[0009] If the first value is less than the first threshold value, return to obtain second frequency spectrum data.

[0010] The siren frequency point screening method provided by the application, by receiving the first audio data processed by noise and performing spectrum transformation on the first audio data to obtain first spectrum data, provides data basis for subsequent frequency screening of the first spectrum data, and also provides parameter basis for subsequent determination of real part data and imaginary part data. The comparison result of the power value corresponding to each point of the first spectrum data obtained by calculation of the real part data and the imaginary part data and the first standard value is determined, the first value is compared with the first threshold value, and whether frequency screening is needed is determined according to the comparison result. Before the corresponding first siren frequency point is obtained by frequency screening of the first spectrum data, the first spectrum data is screened once, avoiding excessive consumption of computing resources caused by subsequent frequency screening of a large amount of spectrum data. If it is determined that there is no siren frequency point in the first spectrum data, the spectrum data can be ignored and the next group of spectrum data is directly screened once, improving the efficiency of siren frequency point screening and also improving the accuracy of subsequent frequency screening of the first spectrum data obtained by one-time screening to obtain siren frequency points.

[0011] The siren frequency point obtained by the above siren frequency point screening method is used as a parameter to update the notch filter, improving the accuracy and precision of the notch processing of the first audio data by the notch filter, so that the audio data processed by the notch filter is greatly suppressed, so that the audio data without siren is obtained, improving the user experience.

[0012] Meanwhile, the first audio data obtained by the siren frequency point screening method provided by the application is obtained by noise processing of the first to-be-processed data, and the floor noise in the first to-be-processed data is removed to further improve the accuracy of siren frequency point screening. In addition, the sound quality of the first audio data after noise removal is also improved, so that the audio data obtained by the method further improves the user experience.

[0013] As a preferred example, the first spectrum data is obtained by performing spectrum transformation on the first audio data, and the real part data and the imaginary part data corresponding to each point in the first spectrum data are determined, specifically:

[0014] The first audio data is calculated by the fast Fourier transform algorithm, so that the signal type of the first audio data is transformed from a time domain signal to a frequency domain signal, and the first spectrum data is obtained, and a preset window function is called to perform windowing processing on the first spectrum data.

[0015] The first spectrum data after windowing processing is processed by real and imaginary separation to obtain the real part data and the imaginary part data corresponding to each point in the first spectrum data.

[0016] The first audio data is calculated by a fast Fourier transform algorithm to obtain first spectrum data, which facilitates subsequent calculation of real part data and imaginary part data, and provides a data basis for subsequent frequency screening. Meanwhile, the first spectrum data is windowed by a preset window function, which reduces energy consumption in data transmission, smoothes discontinuous changes at the truncated end of the spectrum data, reduces noise interference, and also locks the first spectrum data, facilitating subsequent calculation operations.

[0017] As a preferred example, the first number value is determined according to a comparison result of the power value and a preset first standard value, specifically:

[0018] The power value corresponding to each point of the first spectrum data is compared with a preset first standard value respectively, and the number of times that the power value is greater than or equal to the first standard value is accumulated.

[0019] After the comparison, the accumulated number of times is taken as the first number value.

[0020] By comparing the preset first standard value with the power value of each point obtained by calculation, the possible existing howling frequency points in the spectrum data are determined, and a screening is performed first between subsequent frequency screening, which reduces the calculation pressure of subsequent frequency screening. Meanwhile, the number of times that the power value is greater than or equal to the first standard value is accumulated to obtain a first number value, which is the number of possible existing howling frequency points in the first spectrum data. The first number value is provided for comparison with a first threshold value to determine whether frequency screening needs to be performed on the first spectrum data, which further reduces the calculation pressure of subsequent frequency screening.

[0021] As a preferred example, the first spectrum data is subjected to frequency screening to obtain a first howling frequency point, specifically including:

[0022] The first spectrum data is subjected to frequency screening to obtain a first howling frequency point value, and then the first howling frequency point value is normalized by a normalization algorithm to obtain the first howling frequency point.

[0023] The howling frequency point value is normalized by the normalization algorithm, which improves the accuracy of the obtained howling frequency point corresponding to the howling frequency point, and the first howling frequency point is sent to a notch filter, so that the notch filter updates corresponding parameters according to the howling frequency point, improving the suppression accuracy of the notch filter for the howling signal of the first audio data.

[0024] As a preferred example, the first audio data is obtained by noise processing of first to-be-processed data, specifically including:

[0025] The first data to be processed is subjected to gain judgment calculation and processing, and the first gain data is output. The first gain data is subjected to gain smoothness calculation and the first gain is output. The first gain and the first data to be processed are mixed to obtain the first audio data.

[0026] The first audio data is obtained by noise processing of the first data to be processed through the above steps. This not only improves the accuracy of subsequent howling suppression based on the first audio data, but also improves the howling signal suppression accuracy of the notch filter adjusted by the howling frequency points obtained by the howling frequency point screening method. Furthermore, it improves the sound quality of the audio data processed by the adjusted notch filter, thereby enhancing the user experience.

[0027] Meanwhile, the noise removal steps described above not only remove noise from the first data to be processed, but also improve the sound quality of the subsequently obtained audio data, and further enhance the user experience.

[0028] As a preferred example, the step of performing gain determination calculation on the first data to be processed and outputting the first gain data specifically involves:

[0029] The absolute value of the first data to be processed is calculated to obtain the first absolute value data, and the first absolute value data is compared with the preset first threshold attribute.

[0030] If the first absolute value data is greater than the first threshold attribute, then the first gain data is set to 1 and output;

[0031] If the first absolute value data is less than the first threshold attribute, then the first gain data is set to 0 and output.

[0032] The first gain data obtained by the above method provides a data basis for subsequent gain smoothness calculation, which facilitates the subsequent gain smoothness calculation.

[0033] As a preferred example, the step of calculating the gain smoothness of the first gain data and outputting the first gain specifically involves:

[0034] The first release time, the first attack time, and the first duration are calculated using a preset sampling rate. Simultaneously, the gain smoothness is calculated based on the first release time, the first attack time, the first duration, and the first gain data to obtain and output the first gain.

[0035] The first gain data is calculated and processed by the calculation method to obtain the first gain mixed with the first audio data, which is provided as a data basis to a subsequent howling frequency point screening step, thereby improving the accuracy of the howling frequency point screening and the sound quality of the first audio data.

[0036] Correspondingly, the application further provides a howling frequency point screening device for audio data, which comprises a spectrum transformation module, a power comparison module and a frequency screening module.

[0037] The spectrum transformation module is configured to receive first audio data, perform spectrum transformation on the first audio data to obtain first spectrum data, and determine real part data and imaginary part data corresponding to each point in the first spectrum data.

[0038] The power comparison module is configured to calculate a power value corresponding to each point in the first spectrum data according to the real part data and the imaginary part data corresponding to each point in the first spectrum data, and determine a first value according to a comparison result of the power value and a preset first standard value.

[0039] The frequency screening module is configured to compare the first value with a preset first threshold value, if the first value is greater than or equal to the first threshold value, perform frequency screening on the first spectrum data to obtain a first howling frequency point, and if the first value is less than the first threshold value, return to obtain second spectrum data.

[0040] As a preferred example, the howling frequency point screening device further comprises a gain judgment module and a smoothness calculation module.

[0041] The gain judgment module is configured to perform gain judgment calculation processing on the first processing data and output first gain data.

[0042] The smoothness calculation module is configured to perform gain smoothness calculation on the first gain data, output a first gain, mix the first gain with the first processing data, and obtain first audio data.

[0043] Correspondingly, the application further provides a howling frequency point screening system for audio data, which comprises a memory, a processor and a computer program stored in the memory and running on the processor, and the processor implements the howling frequency point screening method for audio data according to any one of the above when processing the computer program. BRIEF DESCRIPTION OF DRAWINGS

[0044] Figure 1A flowchart of an embodiment of a howling frequency point screening method for audio data provided by the present application;

[0045] Figure 2 A flowchart of an embodiment of step 101 provided by the present application;

[0046] Figure 3 A flowchart of an embodiment of step 102 and step 103 provided by the present application;

[0047] Figure 4 A flowchart of an embodiment of a noise processing method for to-be-processed data provided by the present application;

[0048] Figure 5 A structural diagram of an embodiment of a howling frequency point screening device for audio data provided by the present application;

[0049] Figure 6 A structural diagram of an embodiment of a howling signal elimination device for audio data provided by the present application. DETAILED DESCRIPTION

[0050] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0051] Embodiment one

[0052] Please refer to Figure 1 A flowchart of an embodiment of a howling frequency point screening method for audio data provided by the present application, comprising steps 101 to 103, and each step is specifically as follows:

[0053] Step 101: receiving first audio data, performing frequency spectrum transformation on the first audio data to obtain first frequency spectrum data and determining real part data and imaginary part data corresponding to each point in the first frequency spectrum data.

[0054] The howling frequency point screening method provided by the present embodiment, by receiving the first audio data processed by the noise processing and performing frequency spectrum transformation on the first audio data to obtain the first frequency spectrum data, provides data basis for subsequent frequency screening on the first frequency spectrum data, and also provides parameter basis for subsequent determination of the real part data and the imaginary part data. The real part data and the imaginary part data determined according to the first frequency spectrum data are data basis for subsequent determination of the power value corresponding to each point of the first frequency spectrum data.

[0055] Meanwhile, the first audio data obtained by the howling frequency point screening method provided in the embodiment is obtained by noise processing of the first to-be-processed data, and the floor noise in the first to-be-processed data is removed, thereby further improving the accuracy of howling frequency point screening. In addition, the sound quality of the first audio data after noise removal is also improved, and therefore the audio data obtained by the method further improves the experience of users.

[0056] Step 102: Obtain a power value corresponding to each point in the first frequency spectrum data according to the real part data and the imaginary part data corresponding to each point in the first frequency spectrum data, and determine a first value according to a comparison result of the power value and a preset first standard value.

[0057] In the embodiment, the power value corresponding to each point is obtained according to the real part data and the imaginary part data corresponding to each point in the first frequency spectrum data, and the power value is compared with the preset first standard value, and a first value is determined according to a comparison result, thereby providing a data basis for a subsequent comparison step with a first threshold value.

[0058] In the embodiment, the calculation method of the power value corresponding to each point in the first frequency spectrum data according to the real part data and the imaginary part data is as follows: first, calculate the square of the real part data and the square of the imaginary part data, and calculate the sum of the two, then take the square root of the sum, and take the value obtained by the square root calculation as the power value.

[0059] Meanwhile, the preset first standard value is not further limited in the embodiment, and a user can adjust the specific value of the first standard value according to needs. The preset first standard value should be as close as possible to the power value corresponding to the howling frequency point, so that the first howling frequency point exists in the first frequency spectrum data that needs to continue frequency screening after one screening based on the first value, thereby improving the screening accuracy of the first howling frequency point.

[0060] Step 103: Compare the first value with a preset first threshold value. If the first value is greater than or equal to the first threshold value, frequency screening is performed on the first frequency spectrum data to obtain a first howling frequency point. If the first value is less than the first threshold value, return to obtain second frequency spectrum data.

[0061] The method determines whether the first frequency spectrum data needs to be frequency screened, thereby reducing the calculation resources required for frequency screening. The first frequency spectrum data is screened once before frequency screening, thereby avoiding excessive consumption of calculation resources caused by subsequent frequency screening on a large amount of frequency spectrum data. If it is determined that there is no howling frequency point in the first frequency spectrum data, the frequency spectrum data can be ignored, and the next set of frequency spectrum data is directly screened once, thereby improving the efficiency of howling frequency point screening and the accuracy of subsequent frequency screening on the first frequency spectrum data obtained by the once screening.

[0062] In summary, the howling frequency point obtained by the above howling frequency point screening method is used as a parameter to adaptively update the notch filter, thereby improving the accuracy and precision of the notch filter in notch processing of the first audio data, achieving adaptive adjustment of the notch filter parameters for the howling signal, and greatly suppressing the audio data processed by the notch filter, thereby obtaining audio data without howling, improving the experience of the user, and further improving the screening accuracy of the howling frequency point because the method does not cause sound distortion or loss of a wide frequency band.

[0063] The embodiment does not further limit the specific type of the notch filter, but the notch filter of the embodiment is preferably a high-Q notch filter realized by a double second-order IIR, and the high Q value is an ultra-narrow frequency band.

[0064] As another example of the embodiment, refer to Figure 2 , Figure 2 The flowchart of one embodiment of step 101 provided by the application includes steps 201 to 202, and each step is as follows:

[0065] Step 201: The first audio data is calculated by a fast Fourier transform algorithm, so that the signal type of the first audio data is transformed from a time domain signal to a frequency domain signal, and the first frequency spectrum data is obtained.

[0066] In the embodiment, the fast Fourier transform algorithm is used to calculate the first audio data to obtain the first frequency spectrum data, which facilitates the calculation of real part data and imaginary part data and provides a data basis for subsequent frequency screening.

[0067] In the embodiment, the fast Fourier transform algorithm is a 4096-point fast Fourier transform (FFT) of the first to-be-processed data, which transforms the signal type of the first to-be-processed data from a time domain signal to a frequency domain signal, and then obtains the first frequency spectrum data.

[0068] Step 202: calling a preset window function to perform windowing processing on the first spectrum data, and performing real-imaginary separation processing on the first spectrum data after the windowing processing, to obtain the real part data and the imaginary part data corresponding to each point in the first spectrum data.

[0069] The embodiment performs windowing processing on the first spectrum data through a preset window function, reduces the consumption of energy in data transmission, smoothes the discontinuous change at the truncated place of the spectrum data, reduces the interference of noise, and can also lock the first spectrum data, thereby facilitating subsequent calculation operations.

[0070] As another example of the embodiment, refer to Figure 3 , Figure 3 The flowchart of an embodiment of steps 102 and 103 provided by the application includes steps 301 to 302. Step 301 is a specific implementation method of part of the technical features in step 102, and step 302 is a specific implementation method of part of the technical features in step 103.

[0071] Step 301: comparing the power value corresponding to each point in the first spectrum data with a preset first standard value respectively, and accumulating the number of times that the power value is greater than or equal to the first standard value, and after the comparison, taking the accumulated number of times as the first number value.

[0072] The above execution method is a specific implementation method of determining the first number value according to the comparison result of the power value and the preset first standard value in step 102. The power value of each point obtained by calculation is compared with the preset first standard value, to determine the possible howling frequency point in the spectrum data, to perform a screening first in the subsequent frequency screening, thereby reducing the calculation pressure of the subsequent frequency screening, and the first number value obtained by accumulating the number of times that the power value is greater than or equal to the first standard value is the number of possible howling frequency points in the first spectrum data, which is provided to the subsequent comparison with the first threshold value to determine whether the frequency screening needs to be performed on the first spectrum data, thereby further reducing the calculation pressure of the subsequent frequency screening.

[0073] Step 302: comparing the first number value with a preset first threshold value, if the first number value is greater than or equal to the first threshold value, performing frequency screening on the first spectrum data to obtain a first howling frequency point value, and then performing normalization processing on the first howling frequency point value through a normalization algorithm to obtain the first howling frequency point.

[0074] The execution method is a specific implementation method of frequency screening on the first frequency spectrum data to obtain the first howling frequency point in step 103. The normalized algorithm is used to normalize the howling frequency point value, so as to improve the accuracy of the obtained howling frequency point, and the first howling frequency point is sent to the notch filter, so that the notch filter updates the corresponding parameters according to the howling frequency point, and improves the suppression accuracy of the notch filter for the howling signal of the first audio data.

[0075] In the present embodiment, the specific value of the first threshold is preferably 3. The first threshold is set to 3 in combination with the FFT processing time of 4096 points in the above fast Fourier transform. The first threshold is set to 3 to balance the time of receiving the first audio data and obtaining the real part data and the imaginary part data by processing the first audio data, that is, to balance the comprehensive time of receiving the first audio data and performing 4096-point FFT processing on the first audio data and performing windowing processing on the first frequency spectrum data, so that the data receiving and processing can start at the same time and end at the same time, that is, the above steps are synchronously operated, the zero-delay howling frequency point screening work is realized, and the operation efficiency of the howling frequency point screening device is improved.

[0076] As another example of the present embodiment, refer to Figure 4 , Figure 4 The flowchart of an embodiment of a noise processing method for to-be-processed data provided by the present application comprises steps 401 to 402, and each step is as follows:

[0077] Step 401: Perform gain judgment calculation processing on the first to-be-processed data to output first gain data.

[0078] In the present embodiment, the specific implementation process of performing gain judgment calculation processing on the first to-be-processed data is preferably to perform absolute value calculation on the first to-be-processed data S n to obtain first absolute value data |S n |, compare the first absolute value data with a preset first threshold attribute T Lin .

[0079] If the first absolute value data |S n | is greater than the first threshold attribute T Lin , the first gain data G c is set to 1 and output.

[0080] If the first absolute value data |S n | is less than the first threshold attribute T Lin , the first gain data Gc is set to 0 and output.

[0081] wherein the preset first threshold attribute T Lin is obtained by preset first attribute T dB is obtained by calculation according to the first attribute T dB is obtained by calculation according to the first attribute T Lin The calculation formula of the first threshold attribute T

[0082]

[0083] wherein the first threshold attribute is a threshold attribute converted into a linear domain.

[0084] The first gain data obtained by the above method provides a data basis for subsequent gain smoothness calculation, facilitating subsequent gain smoothness calculation.

[0085] Step 402: Perform gain smoothness calculation on the first gain data, output a first gain, and mix the first gain with the first to-be-processed data to obtain first audio data.

[0086] In the embodiment, the specific implementation process of performing gain smoothness calculation on the first gain data and outputting the first gain is preferably that a first release time, a first attack time and a first duration time are obtained by preset sampling rate calculation, and gain smoothness calculation is performed according to the first release time, the first attack time, the first duration time and the first gain data to obtain and output the first gain.

[0087] wherein the calculation formula of the first release time is:

[0088] The F s is a sampling rate, and T r is a user-preset release time.

[0089] The calculation formula of the first attack time is:

[0090] The F s is a sampling rate, and T a is a user-preset attack time.

[0091] The first gain is obtained by calculation and processing of the first gain data by the above calculation method, and the first audio data is obtained by mixing the first gain with the first to-be-processed data, which is provided as a data basis to a subsequent howling frequency point screening step, realizes high-precision and high-sensitivity noise processing, improves the accuracy of the howling frequency point screening, and also improves the sound quality of the first audio data.

[0092] In order to better illustrate the working principle and step flow of the audio data howling frequency point screening method, device and system of the present application, reference can be made to the relevant description in the foregoing, but not limited thereto.

[0093] Correspondingly, reference can be made to Figure 5 Figure 5 The structure diagram of an embodiment of the audio data howling frequency point screening device provided by the present application is shown in Figure 5 The howling frequency point screening device comprises a gain judgment module 501, a smoothness calculation module 502, a spectrum transformation module 503, a power comparison module 504 and a frequency screening module 505.

[0094] The gain judgment module 501 is configured to perform gain judgment calculation processing on the first to-be-processed data and output first gain data, specifically: performing absolute value calculation on the first to-be-processed data to obtain first absolute value data, and comparing the first absolute value data with a preset first threshold attribute.

[0095] If the first absolute value data is greater than the first threshold attribute, the first gain data is set to 1 and output; if the first absolute value data is less than the first threshold attribute, the first gain data is set to 0 and output.

[0096] The smoothness calculation module 502 is configured to perform gain smoothness calculation on the first gain data and output first gain, specifically: calculating a first release time, a first attack time and a first duration through a preset sampling rate, and simultaneously performing gain smoothness calculation according to the first release time, the first attack time, the first duration and the first gain data to obtain the first gain and output. After obtaining the first gain, the first gain and the first to-be-processed data are mixed to obtain first audio data.

[0097] The spectrum transformation module 503 is configured to receive first audio data, perform spectrum transformation on the first audio data to obtain first spectrum data and determine the real part data and the imaginary part data corresponding to each point in the first spectrum data; wherein the first audio data is obtained by noise processing on the first to-be-processed data. Specifically, it comprises:

[0098] The first audio data is calculated through a fast Fourier transform algorithm, so that the signal type of the first audio data is transformed from a time domain signal to a frequency domain signal, the first spectrum data is obtained, a preset window function is called to perform windowing processing on the first spectrum data, and the real part data and the imaginary part data corresponding to each point in the first spectrum data are obtained through real-imaginary separation processing on the windowing-processed first spectrum data.

[0099] The power comparison module 504 is configured to receive the first spectrum data, perform power comparison on the first spectrum data to obtain first power data, and determine the real part data and the imaginary part data corresponding to each point in the first spectrum data.​​​​​​​​​​​​​​​​​The power comparison module 504 is configured to calculate a power value corresponding to each point in the first frequency spectrum data according to the real part data and the imaginary part data corresponding to each point in the first frequency spectrum data, and determine a first value according to a comparison result of the power value and a preset first standard value.

[0100] The power value corresponding to each point in the first frequency spectrum data is compared with the preset first standard value respectively, and the number of times that the power value is greater than or equal to the first standard value is accumulated, and after the comparison ends, the accumulated number of times is taken as the first value.

[0101] The frequency screening module 505 is configured to compare the first value with a preset first threshold value, if the first value is greater than or equal to the first threshold value, frequency screening is performed on the first frequency spectrum data to obtain a first howling frequency point, specifically, a first howling frequency point value is obtained by performing frequency screening on the first frequency spectrum data, and the first howling frequency point value is normalized by a normalization algorithm to obtain the first howling frequency point. The first value is compared with the preset first threshold value, if the first value is less than the first threshold value, second frequency spectrum data is obtained.

[0102] Embodiment two

[0103] Reference is made to Figure 6 , Figure 6 Fig. 1 is a structural schematic diagram of an embodiment of a howling signal elimination device for audio data provided by the application, the howling signal elimination device comprising an audio data input module 601, an analog-to-digital conversion module 602, a digital signal processing module 603, a digital-to-analog conversion module 604 and an audio data output module 605.

[0104] The audio data input module 601 is configured to receive audio data and transmit the audio data to the analog-to-digital conversion module 602.

[0105] The analog-to-digital conversion module 602 is configured to convert the signal type of the received audio data from an analog signal to a digital signal, and send the audio data converted into a digital signal to the digital signal processing module 603.

[0106] After receiving the audio data converted into a digital signal, the digital signal processing module 603 performs howling frequency point screening on the audio data by using the howling frequency point screening device for audio data in Embodiment One, and sends the screened howling frequency point to a notch filter to update the parameters of the notch filter, so that the notch filter performs howling signal elimination processing on the audio data, and sends the audio data after howling signal elimination to the digital-to-analog conversion module 604.

[0107] The digital-to-analog conversion module 604 is configured to convert the signal type of the received audio data after the howling signal elimination processing from a digital signal to an analog signal, and send the converted audio data to the audio data output module 605.

[0108] The audio data output module 605 is configured to output the received audio data sent by the digital-to-analog conversion module 604.

[0109] To sum up, the present application provides a howling frequency point screening method, device and system of audio data, by receiving the first audio data after noise processing and performing spectrum transformation on the first audio data to obtain first spectrum data, determining the real part data and imaginary part data corresponding to each point in the first spectrum data and determining the power value corresponding to each point according to the real part data and imaginary part data, determining the first value according to the comparison result of the power value and the first standard value, and determining whether the first spectrum data needs to be frequency screened to obtain the corresponding first howling frequency point according to the comparison result of the first value and the first threshold value. The present application performs a screening on the first spectrum data before frequency screening, improves the accuracy of the subsequent frequency screening to obtain the howling frequency point, saves a large amount of computing resources consumed by the frequency screening of the spectrum data, and effectively improves the suppression accuracy of the notch filter for the howling signal of the audio data according to the howling frequency point obtained by screening.

[0110] The above specific embodiments further illustrate the purpose, technical solutions and advantages of the present application. It should be understood that the above description is only a specific embodiment of the present application and is not intended to limit the protection scope of the present application. It should be particularly pointed out that any modification, equivalent replacement, improvement, etc. made by those skilled in the art within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A howling frequency point screening method of audio data, characterized by, The method comprises the following steps: receiving first audio data, performing spectrum transformation on the first audio data to obtain first spectrum data, and determining real part data and imaginary part data corresponding to each point in the first spectrum data; wherein the first audio data is obtained by performing noise processing on first to-be-processed data; calculating a power value corresponding to each point in the first spectrum data according to the real part data and the imaginary part data corresponding to each point in the first spectrum data, and determining a first value according to a comparison result of the power value and a preset first standard value; comparing the first value with a preset first threshold value, if the first value is greater than or equal to the first threshold value, performing frequency screening on the first spectrum data to obtain a first howling frequency point; if the first value is less than the first threshold value, returning to obtain second spectrum data; wherein the first audio data is obtained by performing noise processing on the first to-be-processed data, and specifically comprises: performing gain judgment calculation processing on the first to-be-processed data to output first gain data, performing gain smoothness calculation on the first gain data to output a first gain, mixing the first gain and the first to-be-processed data to obtain the first audio data, wherein the gain smoothness calculation on the first gain data to output the first gain specifically comprises: calculating a first release time, a first attack time and a first duration by a preset sampling rate, and simultaneously performing gain smoothness calculation according to the first release time, the first attack time, the first duration and the first gain data to obtain and output the first gain.

2. The howling frequency point screening method of audio data according to claim 1, wherein, The spectrum transformation on the first audio data to obtain the first spectrum data and the determination of the real part data and the imaginary part data corresponding to each point in the first spectrum data specifically comprise: performing calculation on the first audio data by a fast Fourier transform algorithm, so that the signal type of the first audio data is transformed from a time domain signal to a frequency domain signal to obtain the first spectrum data, and simultaneously performing windowing processing on the first spectrum data by calling a preset window function; performing real-imaginary separation processing on the first spectrum data after the windowing processing to obtain the real part data and the imaginary part data corresponding to each point in the first spectrum data.

3. The howling frequency point screening method of audio data according to claim 1, wherein, The determination of the first value according to the comparison result of the power value and the preset first standard value specifically comprises: comparing the power value corresponding to each point in the first spectrum data with the preset first standard value respectively, and accumulating the number of times that the power value is greater than or equal to the first standard value; after the comparison, taking the accumulated number of times as the first value.

4. The howling frequency filtering method of audio data as recited in claim 1, wherein, The frequency screening on the first spectrum data to obtain the first howling frequency point specifically comprises: performing frequency screening on the first spectrum data to obtain a first howling frequency point value, and performing normalization processing on the first howling frequency point value by a normalization algorithm to obtain the first howling frequency point.

5. The method of claim 1, wherein the step of filtering the audio data is performed by a filter bank. 5 The gain judgment calculation processing on the first to-be-processed data to output the first gain data specifically comprises: The first to be processed data is subjected to absolute value calculation to obtain first absolute value data, and the first absolute value data is compared with a preset first threshold attribute; If the first absolute value data is greater than the first threshold attribute, the first gain data is set to 1 and outputted; If the first absolute value data is less than the first threshold attribute, the first gain data is set to 0 and outputted.

6. An apparatus for whistling frequency point screening of audio data, characterized by comprising: The howling frequency point screening device comprises a frequency spectrum conversion module, a power comparison module and a frequency screening module; The frequency spectrum conversion module is configured to receive first audio data, perform frequency spectrum conversion on the first audio data to obtain first frequency spectrum data, and determine real part data and imaginary part data corresponding to each point in the first frequency spectrum data; wherein the first audio data is obtained by noise processing on first to be processed data; The power comparison module is configured to calculate power values corresponding to each point in the first frequency spectrum data according to the real part data and the imaginary part data corresponding to each point in the first frequency spectrum data, and determine a first value according to a comparison result of the power values and a preset first standard value; The frequency screening module is configured to compare the first value with a preset first threshold value, if the first value is greater than or equal to the first threshold value, perform frequency screening on the first frequency spectrum data to obtain first howling frequency points, and if the first value is less than the first threshold value, return to obtain second frequency spectrum data; The first audio data is obtained by noise processing on first to be processed data, specifically comprising: performing gain judgment calculation processing on the first to be processed data to output first gain data, performing gain smoothness calculation on the first gain data to output first gain, mixing the first gain and the first to be processed data to obtain first audio data, wherein the gain smoothness calculation on the first gain data to output first gain specifically comprises: calculating a first release time, a first attack time and a first duration by a preset sampling rate, and simultaneously performing gain smoothness calculation according to the first release time, the first attack time, the first duration and the first gain data to obtain and output the first gain.

7. A howling frequency point screening system of audio data, characterized by, The howling frequency point screening system comprises a memory, a processor and a computer program stored on the memory and running on the processor, and the processor implements the howling frequency point screening method of the audio data according to any one of claims 1-5 when processing the computer program.

Images