CN113870882A - Howling suppression method, system, storage medium and earphone device - Google Patents
Howling suppression method, system, storage medium and earphone device Download PDFInfo
- Publication number
- CN113870882A CN113870882A CN202111149444.8A CN202111149444A CN113870882A CN 113870882 A CN113870882 A CN 113870882A CN 202111149444 A CN202111149444 A CN 202111149444A CN 113870882 A CN113870882 A CN 113870882A
- Authority
- CN
- China
- Prior art keywords
- sound signal
- pickup microphone
- howling
- picked
- howling suppression
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000001629 suppression Effects 0.000 title claims abstract description 75
- 238000000034 method Methods 0.000 title claims abstract description 52
- 230000005236 sound signal Effects 0.000 claims abstract description 201
- 238000003062 neural network model Methods 0.000 claims abstract description 38
- 238000001228 spectrum Methods 0.000 claims description 41
- 238000013528 artificial neural network Methods 0.000 claims description 18
- 230000008569 process Effects 0.000 claims description 18
- 238000000605 extraction Methods 0.000 claims description 12
- 238000001514 detection method Methods 0.000 claims description 11
- 238000001914 filtration Methods 0.000 claims description 7
- 230000005764 inhibitory process Effects 0.000 claims description 7
- 238000006243 chemical reaction Methods 0.000 claims description 5
- 230000008030 elimination Effects 0.000 claims description 2
- 238000003379 elimination reaction Methods 0.000 claims description 2
- 230000003321 amplification Effects 0.000 description 6
- 238000003199 nucleic acid amplification method Methods 0.000 description 6
- 230000009471 action Effects 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 3
- 208000032041 Hearing impaired Diseases 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 239000013598 vector Substances 0.000 description 2
- 230000000903 blocking effect Effects 0.000 description 1
- 210000005069 ears Anatomy 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000008054 signal transmission Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
- G10L21/0208—Noise filtering
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/27—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique
- G10L25/30—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique using neural networks
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04R—LOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; DEAF-AID SETS; PUBLIC ADDRESS SYSTEMS
- H04R5/00—Stereophonic arrangements
- H04R5/033—Headphones for stereophonic communication
Landscapes
- Engineering & Computer Science (AREA)
- Signal Processing (AREA)
- Acoustics & Sound (AREA)
- Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Quality & Reliability (AREA)
- Circuit For Audible Band Transducer (AREA)
Abstract
The application discloses a howling suppression method, a system, a storage medium and earphone equipment, wherein the howling suppression method comprises the following steps: extracting characteristic information in a sound signal picked up by a pickup microphone; inputting the characteristic information into a preset howling suppression neural network model so as to output path information of transmitting a sound signal of a loudspeaker to a pickup microphone through the howling suppression neural network model; and calculating a target sound signal picked up by the sound pickup microphone in the sound signals of the loudspeaker according to the path information, and eliminating the target sound signal from the sound signals picked up by the sound pickup microphone. The method and the device are not influenced by linear signals or nonlinear signals, and can block the conditions for generating the howling and inhibit the howling.
Description
Technical Field
The present invention relates to the field of sound control, and in particular, to a howling suppression method, system, storage medium, and earphone device.
Background
Under market excitation, the auxiliary hearing function of a TWS (True Wireless Stereo) earphone, namely, gain amplification of a missing frequency band of an hearing-impaired patient, is becoming more and more important. The distance between the speaker and the sound-collecting microphone in the TWS headset is short, and therefore, the sound signal S collected by the sound-collecting microphoneMICExcept that it comprises a clean sound signal SIAlso includes a part of loudspeakerAmplified sound signal SL1The sound signal S picked up by the pickup microphoneMICAmplified by the internal algorithm of earphone and sent to loudspeaker for output, and sound signal S is obtained by multiple circulationL1Infinite superposition and amplification can generate howling, which can bring secondary damage to hearing. Therefore, the sound signal S needs to be converted into a sound signalMICSound signal S corresponding to the loudspeaker in (1)L1Howling is eliminated, thereby suppressing.
In the prior art, the signal S is output from the loudspeaker, typically by means of an adaptive filterL0Estimating a sound signal S picked up by a pickup microphone of a loudspeakerL1However, the adaptive filter can only estimate a signal which changes linearly, and when the TWS earphone realizes the auxiliary hearing function, the TWS earphone performs gain amplification on the relevant frequency band, and the gain amplification causes the increase of vibration molecules around the speaker, thereby causing the speaker to output the signal SL0Non-linear distortion occurs, and accordingly, the loudspeaker output signal SL0The feedback path from the TWS earphone shell to the pickup microphone after being reflected by all parts in the TWS earphone shell cannot be accurately estimated, so that the self-adaptive filter cannot accurately estimate the sound signal S picked up by the pickup microphone by the loudspeakerL1That is, the sound signal S picked up by the pickup microphone cannot be eliminatedMICOfL1Therefore, the existing howling suppression scheme has poor effect and brings bad experience to users.
Therefore, how to provide a solution to the above technical problem is a problem that needs to be solved by those skilled in the art.
Disclosure of Invention
The present application aims to provide a howling suppression method, a howling suppression system, a storage medium, and an earphone device, which are not affected by a linear signal or a nonlinear signal, and can block a condition of howling generation and suppress howling.
In order to solve the above technical problem, the present application provides a howling suppression method, including:
extracting characteristic information in a sound signal picked up by a pickup microphone;
inputting the characteristic information into a preset howling suppression neural network model so as to output path information of transmitting a sound signal of a loudspeaker to the pickup microphone through the howling suppression neural network model;
and calculating a target sound signal picked up by the pickup microphone in the sound signals of the loudspeaker according to the path information, and eliminating the target sound signal from the sound signals picked up by the pickup microphone.
Optionally, the sound signal of the speaker is obtained through a feedback microphone.
Optionally, before extracting the feature information in the sound signal picked up by the sound pickup microphone, the howling suppression method further includes:
howling detection is carried out on the sound signals picked up by the pickup microphone;
correspondingly, the process of extracting the characteristic information in the sound signal picked up by the pickup microphone includes:
when it is determined that howling occurs, feature information in a sound signal picked up by a pickup microphone is extracted.
Optionally, the process of performing howling detection on the sound signal picked up by the pickup microphone includes:
and when the amplitude of the time domain of the sound signal picked up by the pickup microphone is larger than a first preset value, or when the amplitude of the frequency domain of the sound signal picked up by the pickup microphone is larger than a second preset value, judging that howling occurs.
Optionally, the howling suppression method further includes:
determining a plurality of groups of training parameters, wherein each group of training parameters comprises a pure voice signal, a sound signal picked up by the pickup microphone, a sound signal of the loudspeaker and analog path information;
extracting the characteristics of each group of training parameters to obtain characteristic information;
and training a neural network according to the characteristic information of each group of training parameters to obtain the howling inhibition neural network model.
Optionally, the process of extracting the features of the sound signal of the speaker to obtain the feature information includes:
and carrying out feature extraction on the sound signal of the loudspeaker through Fbank to obtain feature information.
Optionally, the process of extracting the feature information in the sound signal picked up by the pickup microphone includes:
converting an energy spectrum of a sound signal picked up by the pickup microphone into a power spectrum;
performing Mel filtering on the power spectrum to obtain a target sound spectrum;
calculating a log spectrum of the target sound spectrum;
and performing DCT conversion on the log spectrum to obtain cepstrum coefficients as characteristic information.
In order to solve the above technical problem, the present application further provides a howling suppression system, including:
the extraction module is used for extracting characteristic information in the sound signals picked up by the pickup microphone;
the processing module is used for inputting the characteristic information into a preset howling suppression neural network model so as to output path information of a sound signal of a loudspeaker transmitted to the pickup microphone through the howling suppression neural network model;
and the elimination module is used for calculating a target sound signal picked up by the pickup microphone in the sound signals of the loudspeaker according to the path information and eliminating the target sound signal from the sound signals picked up by the pickup microphone.
In order to solve the above technical problem, the present application further provides a storage medium, where computer-executable instructions are stored, and when the computer-executable instructions are loaded and executed by a processor, the steps of the howling suppression method are implemented as described in any one of the above.
To solve the above technical problem, the present application further provides a headphone apparatus including the storage medium as described above.
The howling suppression method includes the steps that firstly, a howling suppression neural network model is built based on a neural network, characteristic information in sound signals picked up by a pickup microphone is input into the howling suppression neural network model, the neural network is not influenced by linear signals or nonlinear signals, therefore, the howling suppression model can accurately estimate path information of the sound signals of a loudspeaker transmitted to the pickup microphone, target sound signals picked up by the pickup microphone in the sound signals of the loudspeaker are calculated according to the path information, the target sound signals are removed from the sound signals picked up by the pickup microphone, and conditions generated by howling can be blocked, and accordingly howling is suppressed. The application also provides a howling suppression system, a storage medium and an earphone device, which have the same beneficial effects as the howling suppression method.
Drawings
In order to more clearly illustrate the embodiments of the present application, the drawings needed for the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and that other drawings can be obtained by those skilled in the art without inventive effort.
Fig. 1 is a flowchart illustrating steps of a howling suppression method according to the present application;
fig. 2 is a schematic structural diagram of a howling suppression system provided in the present application.
Detailed Description
The core of the application is to provide a howling suppression method, a howling suppression system, a storage medium and an earphone device, which are not affected by linear signals or nonlinear signals, and can block the conditions generated by howling and suppress the howling.
In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, 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, and it is obvious that the described embodiments are some embodiments of the present application, but not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
Referring to fig. 1, fig. 1 is a flowchart illustrating a howling suppression method according to the present application, where the howling suppression method includes:
s101: extracting characteristic information in a sound signal picked up by a pickup microphone;
specifically, this embodiment is applied to an earphone device, and the earphone device includes a pickup microphone, an auxiliary microphone and a speaker, where the pickup microphone is used to collect a sound signal in an environment where the pickup microphone is located, and since the pickup microphone and the speaker in the earphone device are closer to each other, the sound signal collected by the pickup microphone may include a sound signal of the speaker, a noise signal, and the like in addition to a pure human sound signal.
Generally, sound signals picked up by the pickup microphone are subjected to noise reduction, amplification and other processing in the earphone device, then the sound signals are sent to the loudspeaker and output by the loudspeaker, and sound signals of the loudspeaker are picked up by the microphone.
It can be understood that different sound signals picked up by the microphone have different characteristic information, and in order to make the sound signals sent to the speaker by the microphone include only pure speech signals, the characteristic information in the sound signals picked up by the microphone needs to be accurately acquired, and the characteristic information can determine an adjustment parameter for suppressing howling.
S102: inputting the characteristic information into a preset howling suppression neural network model so as to output path information of transmitting a sound signal of a loudspeaker to a pickup microphone through the howling suppression neural network model;
specifically, taking an example that the sound signal picked up by the microphone only includes a clean speech signal and a picked-up speaker output signal, let the sound signal picked up by the microphone be X, the clean speech signal be S, and the sound signal picked up by the pickup microphone, that is, the target sound signal, be N, where X is S + N, the purpose of this embodiment is to remove N. The target sound signal N of the speaker picked up by the pickup microphone is a signal that reaches the pickup microphone after the sound signal Y of the speaker is shielded and reflected by the casing and the elements in the casing of the earphone device, and then both the inner wall of the casing and the elements in the casing can be regarded as a filter, so that the target sound signal N of the speaker obtained by filtering the complete sound signal Y of the speaker for multiple times through the casing and the elements in the casing finally reaches the pickup microphone and is picked up by the pickup microphone. Since the complete sound signal Y of the speaker is different, and the transmission path to the sound pickup microphone may also be different when the complete sound signal Y is transmitted in the housing, that is, different Y may be filtered by different filters, and N ═ F × Y, F is the path information of the sound signal of the speaker transmitted from the speaker to the microphone in the housing of the earphone device, so to obtain accurate N, it is necessary to accurately estimate F.
Specifically, in the field of hearing aids, after gain amplification is performed on some frequency bands missing from hearing-impaired patients, nonlinear distortion occurs in signals output by a speaker, the scheme in the prior art can only estimate transmission paths of linear signals, and transmission paths of parts where the nonlinear distortion occurs cannot be accurately estimated. In order to solve the problem of signal nonlinear distortion, in this embodiment, a processing model is first constructed based on a Neural Network, that is, a howling suppression Neural Network model in this step, specifically, an RNN (Recurrent Neural Network) may be selected to construct the howling suppression Neural Network model. The neural network may not be affected by signal linearity or nonlinearity, and after the characteristic information of the sound signal picked up by the microphone is input into the howling suppression neural network model, the howling suppression neural network model may automatically output path information of the sound signal of the speaker transmitted to the sound pickup microphone in the casing of the earphone device.
S103: and calculating a target sound signal picked up by the sound pickup microphone in the sound signals of the loudspeaker according to the path information, and eliminating the target sound signal from the sound signals picked up by the sound pickup microphone.
Specifically, in consideration of the fact that when the target sound signal is estimated, the sound signal actually output by the speaker needs to be used as a reference, before the step is executed, the method further includes an operation of acquiring the sound signal of the speaker, calculating the target sound signal N picked up by the pickup microphone according to the acquired sound signal Y of the speaker and the path information F, and eliminating the target sound signal Y from the sound signal X picked up by the pickup microphone, so that the sound signal input to the speaker by the pickup microphone only includes the pure voice signal S, thereby blocking a condition of generating howling, further suppressing the howling, avoiding secondary damage of the howling to the hearing of the user, and improving the competitiveness of the earphone device.
It can be seen that, in this embodiment, a howling suppression neural network model is first constructed based on a neural network, feature information in a sound signal picked up by a pickup microphone is input into the howling suppression neural network model, and the neural network is not affected by a linear signal or a nonlinear signal, so that the howling suppression model can accurately estimate path information of the sound signal transmitted from a speaker to the pickup microphone, calculate a target sound signal picked up by the pickup microphone in the sound signal of the speaker according to the path information, and remove the target sound signal from the sound signal picked up by the pickup microphone, that is, block a condition generated by howling, thereby suppressing howling.
On the basis of the above-described embodiment:
as an alternative embodiment, the sound signal of the speaker is the sound signal of the speaker acquired by the feedback microphone.
Referring to the above, the earphone device further includes a feedback microphone, in this embodiment, the feedback microphone is used to acquire the sound signal of the speaker, so that factors such as signal transmission delay can be removed, and a more accurate sound signal of the speaker is acquired, thereby further ensuring accurate estimation of the target sound signal.
As an alternative embodiment, before extracting feature information in a sound signal picked up by a pickup microphone, the howling suppression method further includes:
performing howling detection on a sound signal picked up by a pickup microphone;
correspondingly, the process of extracting the characteristic information in the sound signal picked up by the pickup microphone includes:
when it is determined that howling occurs, feature information in a sound signal picked up by a pickup microphone is extracted.
In this embodiment, after the sound signal picked up by the pickup microphone is acquired, howling detection is performed on the sound signal, that is, whether howling occurs is determined, and then subsequent operations for acquiring feature information are performed after the howling occurs, so as to reduce the data processing amount. Since the sound signal with howling is reflected in the frequency domain or the time domain, whether howling occurs can be judged according to the amplitude value of the time domain or the amplitude value of the frequency domain. Specifically, when the amplitude of the time domain of the sound signal picked up by the pickup microphone is greater than a first preset value, or when the amplitude of the frequency domain of the sound signal picked up by the pickup microphone is greater than a second preset value, it is determined that howling occurs; correspondingly, when the amplitude of the time domain of the sound signal picked up by the pickup microphone is smaller than or equal to a first preset value, or when the amplitude of the frequency domain of the sound signal picked up by the pickup microphone is smaller than or equal to a second preset value, it is determined that howling does not occur.
Of course, in addition to the above scheme, howling detection may also be implemented by other schemes, and this embodiment is not limited in detail herein.
As an alternative embodiment, the howling suppression method further includes:
determining a plurality of groups of training parameters, wherein each group of training parameters comprises a pure voice signal, a sound signal picked up by a pickup microphone, a sound signal of a loudspeaker and analog path information;
extracting the characteristics of each group of training parameters to obtain characteristic information;
and training the neural network according to the characteristic information of each group of training parameters to obtain a howling inhibition neural network model.
Specifically, a plurality of groups of training parameters are obtained, the training parameters include known pure voice signals, sound signals picked up by a pickup microphone, sound signals of a loudspeaker and analog path information, corresponding characteristic information of the sound signals is extracted, and a neural network is trained according to the characteristic information of each group of training parameters, so that a howling inhibition neural network model is obtained. Specifically, the feature parameter of feature extraction is the sum of the energy of frequency domain points, firstly, feature extraction is performed on a sound signal picked up by a pickup microphone, feature analysis is performed on a pure sound signal output by a pure sound source, feature extraction is performed on a sound signal of a loudspeaker after gain control, and then the extraction result, the simulation path and four inputs are quantized into a matrix to perform input and output amplitude values so as to train an RNN (neural network) to obtain a howling inhibition neural network model. In practical application, when the characteristic information of the sound signal is input into the howling suppression neural network model, the suppression neural network model can automatically generate path information.
As an alternative embodiment, the process of extracting the features of the sound signal of the speaker to obtain the feature information includes:
feature information is obtained by performing feature extraction on a sound signal of a speaker through an Fbank (filter bank).
Specifically, a feedback microphone is used for picking up a sound signal of a loudspeaker, and feature information in the sound signal is extracted through Fbank, wherein the number of feature values in the feature information is 22.
As an alternative embodiment, the process of extracting the characteristic information in the sound signal picked up by the pickup microphone includes:
converting an energy spectrum of a sound signal picked up by a pickup microphone into a power spectrum;
performing Mel filtering on the power spectrum to obtain a target sound spectrum;
calculating a logarithmic spectrum of the target sound spectrum;
and performing DCT conversion on the log spectrum to obtain cepstrum coefficients as characteristic information.
Specifically, for convenience of analysis, an energy spectrum of a sound signal picked up by the pickup microphone is converted into a power spectrum by a first relation, where the first relation is:
Then Mel filtering is carried out on the power spectrum by referring to a second relational expression, the power spectrum is converted into a sound spectrum which accords with the hearing habits of human ears, and then logarithm is taken so as to convert a unit into dB. The second relation is:
where m denotes the mth filter coefficient, and there are 26 filter banks for feature extraction, and there are 26 features in total.
And finally, performing DCT (Discrete Cosine Transform) on the reciprocal spectrum by referring to a third relational expression, reserving 2 nd to 13 th coefficients of the DCT, and calculating a first difference, a second difference and an energy value of the 13-dimensional feature vector to obtain a cepstrum coefficient of the 39-dimensional MFCC feature. The third relation is:
where C denotes a DCT transform system and N denotes a frame.
It can be understood that after the sound signals picked up by the sound pickup microphone and the sound signals picked up by the feedback microphone are subjected to the feature extraction, the number of feature values in the obtained feature information is 78, and after the feature information is input into the howling suppression neural network model, the path information of the sound signals of the speaker transmitted from the speaker to the sound pickup microphone in the earphone device can be obtained, and the path information can be represented as a filter coefficient of 27 th order, so the path information specifically can include 27 vectors.
Referring to fig. 2, fig. 2 is a schematic structural diagram of a howling suppression system provided in the present application, where the howling suppression system includes:
an extraction module 11, configured to extract feature information in a sound signal picked up by a pickup microphone;
the processing module 12 is configured to input the feature information into a preset howling suppression neural network model, so that the howling suppression neural network model outputs path information that a sound signal of the speaker is transmitted to the pickup microphone;
and the eliminating module 13 is used for calculating a target sound signal picked up by the pickup microphone in the sound signals of the loudspeaker according to the path information, and eliminating the target sound signal from the sound signals picked up by the pickup microphone.
It can be seen that, in this embodiment, a howling suppression neural network model is first constructed based on a neural network, feature information in a sound signal picked up by a pickup microphone is input into the howling suppression neural network model, and the neural network is not affected by a linear signal or a nonlinear signal, so that the howling suppression model can accurately estimate path information of the sound signal transmitted from a speaker to the pickup microphone, calculate a target sound signal picked up by the pickup microphone in the sound signal of the speaker according to the path information, and remove the target sound signal from the sound signal picked up by the pickup microphone, that is, block a condition generated by howling, thereby suppressing howling.
As an alternative embodiment, the sound signal of the speaker is the sound signal of the speaker acquired by the feedback microphone.
As an alternative embodiment, the howling suppression system further includes:
the detection module is used for carrying out howling detection on the sound signals picked up by the pickup microphone;
correspondingly, the process of extracting the characteristic information in the sound signal picked up by the pickup microphone includes:
the extracting module 11 is specifically configured to, when it is determined that howling occurs, extract feature information in a sound signal picked up by the sound pickup microphone.
As an alternative embodiment, the process of performing howling detection on the sound signal picked up by the pickup microphone includes:
when the amplitude of the time domain of the sound signal picked up by the pickup microphone is larger than a first preset value, or when the amplitude of the frequency domain of the sound signal picked up by the pickup microphone is larger than a second preset value, the occurrence of howling is judged;
and when the amplitude of the time domain of the sound signal picked up by the pickup microphone is smaller than or equal to a first preset value, or when the amplitude of the frequency domain of the sound signal picked up by the pickup microphone is smaller than or equal to a second preset value, judging that howling does not occur.
As an alternative embodiment, the howling suppression system further includes:
the training module is used for determining a plurality of groups of training parameters, wherein each group of training parameters comprises a pure voice signal, a sound signal picked up by a pickup microphone, a sound signal of a loudspeaker and analog path information; extracting the characteristics of each group of training parameters to obtain characteristic information; and training the neural network according to the characteristic information of each group of training parameters to obtain a howling inhibition neural network model.
As an alternative embodiment, the process of extracting the features of the sound signal of the speaker to obtain the feature information includes:
and extracting the characteristics of the sound signals of the loudspeaker through Fbank to obtain characteristic information.
As an alternative embodiment, the process of extracting the characteristic information in the sound signal picked up by the pickup microphone includes:
converting an energy spectrum of a sound signal picked up by a pickup microphone into a power spectrum;
performing Mel filtering on the power spectrum to obtain a target sound spectrum;
calculating a logarithmic spectrum of the target sound spectrum;
and performing DCT conversion on the log spectrum to obtain cepstrum coefficients as characteristic information.
The present application also provides a storage medium, which may include: Read-Only Memory (ROM), Random Access Memory (RAM), and various other media that can store program codes. The storage medium having stored thereon computer-executable instructions that, when loaded and executed by a processor, perform the steps of: extracting characteristic information in a sound signal picked up by a pickup microphone; inputting the characteristic information into a preset howling suppression neural network model so as to output path information of transmitting a sound signal of a loudspeaker to a pickup microphone through the howling suppression neural network model; and calculating a target sound signal picked up by the sound pickup microphone in the sound signals of the loudspeaker according to the path information, and eliminating the target sound signal from the sound signals picked up by the sound pickup microphone.
In the embodiment, a howling suppression neural network model is firstly constructed based on a neural network, characteristic information in a sound signal picked up by a pickup microphone is input into the howling suppression neural network model, and the neural network is not influenced by a linear signal or a nonlinear signal, so that the howling suppression model can accurately estimate path information of the sound signal transmitted from a loudspeaker to the pickup microphone, a target sound signal picked up by the pickup microphone in the sound signal of the loudspeaker is calculated according to the path information, and the target sound signal is removed from the sound signal picked up by the pickup microphone, so that the condition generated by the howling can be blocked, and the howling can be suppressed.
As an alternative embodiment, when the computer executable instructions stored in the storage medium are loaded and executed by the processor, the following steps can be specifically realized: performing howling detection on a sound signal picked up by a pickup microphone; when it is determined that howling occurs, feature information in a sound signal picked up by a pickup microphone is extracted.
As an alternative embodiment, when the computer executable instructions stored in the storage medium are loaded and executed by the processor, the following steps can be specifically realized: when the amplitude of the time domain of the sound signal picked up by the pickup microphone is larger than a first preset value, or when the amplitude of the frequency domain of the sound signal picked up by the pickup microphone is larger than a second preset value, the occurrence of howling is judged; and when the amplitude of the time domain of the sound signal picked up by the pickup microphone is smaller than or equal to a first preset value, or when the amplitude of the frequency domain of the sound signal picked up by the pickup microphone is smaller than or equal to a second preset value, judging that howling does not occur.
As an alternative embodiment, when the computer executable instructions stored in the storage medium are loaded and executed by the processor, the following steps can be specifically realized: determining a plurality of groups of training parameters, wherein each group of training parameters comprises a pure voice signal, a sound signal picked up by a pickup microphone, a sound signal of a loudspeaker and analog path information; extracting the characteristics of each group of training parameters to obtain characteristic information; and training the neural network according to the characteristic information of each group of training parameters to obtain a howling inhibition neural network model.
As an alternative embodiment, when the computer executable instructions stored in the storage medium are loaded and executed by the processor, the following steps can be specifically realized: and extracting the characteristics of the sound signals of the loudspeaker through Fbank to obtain characteristic information.
As an alternative embodiment, when the computer executable instructions stored in the storage medium are loaded and executed by the processor, the following steps can be specifically realized: converting an energy spectrum of a sound signal picked up by a pickup microphone into a power spectrum; performing Mel filtering on the power spectrum to obtain a target sound spectrum; calculating a logarithmic spectrum of the target sound spectrum; and performing DCT conversion on the log spectrum to obtain cepstrum coefficients as characteristic information.
In another aspect, the present application further provides a headphone apparatus including the storage medium described in any one of the above embodiments.
Please refer to the above embodiments for the introduction of the earphone device provided in the present application, which is not described herein again.
The earphone device provided by the application has the same beneficial effects as the howling suppression method.
It is further noted that, in the present specification, relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. A howling suppression method, comprising:
extracting characteristic information in a sound signal picked up by a pickup microphone;
inputting the characteristic information into a preset howling suppression neural network model so as to output path information of transmitting a sound signal of a loudspeaker to the pickup microphone through the howling suppression neural network model;
and calculating a target sound signal picked up by the pickup microphone in the sound signals of the loudspeaker according to the path information, and eliminating the target sound signal from the sound signals picked up by the pickup microphone.
2. The howling suppression method according to claim 1, wherein the sound signal of the speaker is obtained by a feedback microphone.
3. The howling suppression method according to claim 1, wherein before extracting the feature information in the sound signal picked up by the pickup microphone, the howling suppression method further comprises:
howling detection is carried out on the sound signals picked up by the pickup microphone;
correspondingly, the process of extracting the characteristic information in the sound signal picked up by the pickup microphone includes:
when it is determined that howling occurs, feature information in a sound signal picked up by a pickup microphone is extracted.
4. The howling suppression method according to claim 3, wherein the process of performing howling detection on the sound signal picked up by the pickup microphone includes:
and when the amplitude of the time domain of the sound signal picked up by the pickup microphone is larger than a first preset value, or when the amplitude of the frequency domain of the sound signal picked up by the pickup microphone is larger than a second preset value, judging that howling occurs.
5. The howling suppression method according to claim 1, wherein the howling suppression method further comprises:
determining a plurality of groups of training parameters, wherein each group of training parameters comprises a pure voice signal, a sound signal picked up by the pickup microphone, a sound signal of the loudspeaker and analog path information;
extracting the characteristics of each group of training parameters to obtain characteristic information;
and training a neural network according to the characteristic information of each group of training parameters to obtain the howling inhibition neural network model.
6. The howling suppression method according to claim 5, wherein the process of extracting features of the sound signal of the speaker to obtain feature information comprises:
and carrying out feature extraction on the sound signal of the loudspeaker through Fbank to obtain feature information.
7. The howling suppression method according to any one of claims 1 to 6, wherein the process of extracting the feature information in the sound signal picked up by the pickup microphone includes:
converting an energy spectrum of a sound signal picked up by the pickup microphone into a power spectrum;
performing Mel filtering on the power spectrum to obtain a target sound spectrum;
calculating a log spectrum of the target sound spectrum;
and performing DCT conversion on the log spectrum to obtain cepstrum coefficients as characteristic information.
8. A howling suppression system, comprising:
the extraction module is used for extracting characteristic information in the sound signals picked up by the pickup microphone;
the processing module is used for inputting the characteristic information into a preset howling suppression neural network model so as to output path information of a sound signal of a loudspeaker transmitted to the pickup microphone through the howling suppression neural network model;
and the elimination module is used for calculating a target sound signal picked up by the pickup microphone in the sound signals of the loudspeaker according to the path information and eliminating the target sound signal from the sound signals picked up by the pickup microphone.
9. A storage medium having stored thereon computer-executable instructions which, when loaded and executed by a processor, carry out the steps of the howling suppression method according to any one of claims 1 to 7.
10. An earphone device, characterized in that it comprises a storage medium according to claim 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111149444.8A CN113870882A (en) | 2021-09-29 | 2021-09-29 | Howling suppression method, system, storage medium and earphone device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111149444.8A CN113870882A (en) | 2021-09-29 | 2021-09-29 | Howling suppression method, system, storage medium and earphone device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113870882A true CN113870882A (en) | 2021-12-31 |
Family
ID=78992460
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111149444.8A Pending CN113870882A (en) | 2021-09-29 | 2021-09-29 | Howling suppression method, system, storage medium and earphone device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113870882A (en) |
-
2021
- 2021-09-29 CN CN202111149444.8A patent/CN113870882A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112767963B (en) | Voice enhancement method, device and system and computer readable storage medium | |
| MX2014011556A (en) | Apparatus and method for improving the perceived quality of sound reproduction by combining active noise cancellation and perceptual noise compensation. | |
| CN110611871A (en) | Howling suppression method and system for digital hearing aid and special DSP | |
| CN112863535A (en) | Residual echo and noise elimination method and device | |
| CN115884032B (en) | Smart call noise reduction method and system for feedback earphone | |
| CN110556125A (en) | Feature extraction method and device based on voice signal and computer storage medium | |
| CN111988708A (en) | Single-microphone-based howling suppression method and device | |
| CN111798864A (en) | Echo cancellation method and device, electronic equipment and storage medium | |
| CN103929704B (en) | The method and system that a kind of adaptive acoustic feedback based on transform domain is eliminated | |
| JP2024517721A (en) | Audio optimization for noisy environments | |
| CN112802490B (en) | Beam forming method and device based on microphone array | |
| KR101557779B1 (en) | Method and apparatus for noise reduction in a communication device having two microphones | |
| CN113870882A (en) | Howling suppression method, system, storage medium and earphone device | |
| CN115620737A (en) | Voice signal processing device, method, electronic equipment and sound amplification system | |
| CN113593518A (en) | Howling suppression method and device, in-ear earphone and storage medium | |
| CN114420153A (en) | Sound quality adjusting method, device, equipment and storage medium | |
| CN114664322A (en) | Single-microphone hearing-aid noise reduction method based on Bluetooth headset chip and Bluetooth headset | |
| CN115314823A (en) | Hearing aid method, system and equipment based on digital sounding chip | |
| CN113611319A (en) | Wind noise suppression method, device, equipment and system based on voice component | |
| JP2022025908A (en) | Filter processing method, filter processing device, and filter processing program | |
| CN116887160B (en) | Digital hearing aid howling suppression method and system based on neural network | |
| Fabry et al. | Perceptual filter optimization for active acoustic equalization | |
| EP3837621B1 (en) | Dual-microphone methods for reverberation mitigation | |
| CN115835092B (en) | Audio amplification feedback suppression method, system, computer and storage medium | |
| KR102408689B1 (en) | Musical noise eliminating device using neural network voice activity detection |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |