CN109119093A - Voice de-noising method, device, storage medium and mobile terminal - Google Patents
Voice de-noising method, device, storage medium and mobile terminal Download PDFInfo
- Publication number
- CN109119093A CN109119093A CN201811273582.5A CN201811273582A CN109119093A CN 109119093 A CN109119093 A CN 109119093A CN 201811273582 A CN201811273582 A CN 201811273582A CN 109119093 A CN109119093 A CN 109119093A
- Authority
- CN
- China
- Prior art keywords
- noise reduction
- amplitude spectrum
- noise
- voice
- signal
- 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
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Processing of the speech or voice signal 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
- G10L21/0316—Speech enhancement, e.g. noise reduction or echo cancellation by changing the amplitude
- G10L21/0324—Details of processing therefor
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Processing of the speech or voice signal 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
- G10L21/0216—Noise filtering characterised by the method used for estimating noise
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; 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
Abstract
The embodiment of the present application discloses voice de-noising method, device, storage medium and mobile terminal.Wherein, method includes: the voice signal for receiving and being acquired based on earphone single microphone, extract the initial amplitude spectrum and phase spectrum of the voice signal, initial amplitude spectrum is transmitted in noise reduction model trained in advance, the first noise reduction amplitude spectrum is obtained, masking processing is carried out to the first noise reduction amplitude spectrum, generates the second noise reduction amplitude spectrum, the second noise reduction amplitude spectrum and the phase spectrum are recombinated, the first reducing noise of voice is generated.By using above scheme, keep the phase spectrum of voice signal constant, noise reduction process is carried out based on amplitude spectrum of the noise reduction model to voice signal, and the first noise reduction amplitude spectrum obtained to processing carries out masking processing, eliminate the problem of dtmf distortion DTMF in noise reduction process, it solves the problems, such as that the pronunciation signal noise acquired under noisy environment is big, realizes the quick high accuracy noise reduction of the voice signal acquired to earphone.
Description
Technical field
The invention relates to voice processing technology field more particularly to a kind of voice de-noising method, device, storage Jie
Matter and mobile terminal.
Background technique
With the fast development of the mobile terminals such as mobile phone, earphone becomes the important component of the mobile terminals such as mobile phone, can
Realize audio broadcastings, voice instant messaging by earphone, the functions such as dial and receive calls, by more and more users receive with
It uses.
During making a phone call, voice signal is acquired by ear microphone, voice signal is sent to mobile terminal,
Voice signal is sent to counterpart telephone by uplink.But earphone picks up sound, noise reduction generally by single microphone
Ability is poor, when user is when the noisy public place such as subway, bus is made a phone call, passes through what ear microphone acquired
Pronunciation signal noise is big, and the noise reduction of high quality can not be carried out to the voice signal of acquisition, and speech quality is poor.
Summary of the invention
The embodiment of the present application provides voice de-noising method, device, storage medium and mobile terminal, acquires to ear microphone
Voice signal carry out high quality noise reduction, improve speech quality.
In a first aspect, the embodiment of the present application provides a kind of voice de-noising method, comprising:
Receive the voice signal acquired based on earphone single microphone；
The initial amplitude spectrum and phase spectrum for extracting the voice signal, are transmitted to training in advance for initial amplitude spectrum
In noise reduction model, the first noise reduction amplitude spectrum is obtained；
Masking processing is carried out to the first noise reduction amplitude spectrum, generates the second noise reduction amplitude spectrum；
The second noise reduction amplitude spectrum and the phase spectrum are recombinated, the first reducing noise of voice is generated.
Second aspect, the embodiment of the present application provide a kind of voice noise reduction device, comprising:
Speech reception module, for receiving the voice signal based on the acquisition of earphone single microphone；
First noise reduction amplitude spectrum generation module, the initial amplitude for extracting the voice signal is composed and phase spectrum, by institute
It states initial amplitude spectrum to be transmitted in noise reduction model trained in advance, obtains the first noise reduction amplitude spectrum；
Second noise reduction amplitude spectrum generation module generates second for carrying out masking processing to the first noise reduction amplitude spectrum
Noise reduction amplitude spectrum；
First reducing noise of voice generation module, it is raw for being recombinated to the second noise reduction amplitude spectrum and the phase spectrum
At the first reducing noise of voice.
The third aspect, the embodiment of the present application provide a kind of computer readable storage medium, are stored thereon with computer journey
Sequence realizes the voice de-noising method as described in the embodiment of the present application when the program is executed by processor.
Fourth aspect, the embodiment of the present application provide a kind of mobile terminal, including memory, processor and are stored in storage
On device and the computer program that can run on a processor, the processor realize such as the application when executing the computer program
Voice de-noising method described in embodiment.
The voice de-noising method provided in the embodiment of the present application is believed by receiving the voice acquired based on earphone single microphone
Number, the initial amplitude spectrum and phase spectrum of the voice signal are extracted, initial amplitude spectrum is transmitted to noise reduction trained in advance
In model, the first noise reduction amplitude spectrum is obtained, masking processing is carried out to the first noise reduction amplitude spectrum, generates the second noise reduction amplitude
Spectrum, recombinates the second noise reduction amplitude spectrum and the phase spectrum, generates the first reducing noise of voice.By using above-mentioned side
Case keeps the phase spectrum of voice signal constant, carries out noise reduction process based on amplitude spectrum of the noise reduction model to voice signal, and to place
It manages the first obtained noise reduction amplitude spectrum and carries out masking processing, eliminate the problem of dtmf distortion DTMF in noise reduction process, solve in noisy environment
The big problem of the pronunciation signal noise of lower acquisition realizes the quick high accuracy noise reduction of the voice signal acquired to earphone.
Detailed description of the invention
Fig. 1 is a kind of flow diagram of voice de-noising method provided by the embodiments of the present application；
Fig. 2 is the flow diagram of another voice de-noising method provided by the embodiments of the present application；
Fig. 3 is the flow diagram of another voice de-noising method provided by the embodiments of the present application；
Fig. 4 is a kind of structural schematic diagram of voice noise reduction device provided by the embodiments of the present application；
Fig. 5 is a kind of structural schematic diagram of mobile terminal provided by the embodiments of the present application；
Fig. 6 is the structural schematic diagram of another mobile terminal provided by the embodiments of the present application.
Specific embodiment
Further illustrate the technical solution of the application below with reference to the accompanying drawings and specific embodiments.It is understood that
It is that specific embodiment described herein is used only for explaining the application, rather than the restriction to the application.It further needs exist for illustrating
, part relevant to the application is illustrated only for ease of description, in attached drawing rather than entire infrastructure.
It should be mentioned that some exemplary embodiments are described as before exemplary embodiment is discussed in greater detail
The processing or method described as flow chart.Although each step is described as the processing of sequence by flow chart, many of these
Step can be implemented concurrently, concomitantly or simultaneously.In addition, the sequence of each step can be rearranged.When its operation
The processing can be terminated when completion, it is also possible to have the additional step being not included in attached drawing.The processing can be with
Corresponding to method, function, regulation, subroutine, subprogram etc..
Fig. 1 is a kind of flow diagram of voice de-noising method provided by the embodiments of the present application, and this method can be by voice
Denoising device executes, and wherein the device can be implemented by software and/or hardware, and can generally integrate in the terminal.Such as Fig. 1 institute
Show, this method comprises:
The voice signal that step 101, reception are acquired based on earphone single microphone.
Step 102, the initial amplitude spectrum and phase spectrum for extracting the voice signal, initial amplitude spectrum are transmitted to pre-
First in trained noise reduction model, the first noise reduction amplitude spectrum is obtained.
Step 103 carries out masking processing to the first noise reduction amplitude spectrum, generates the second noise reduction amplitude spectrum.
Step 104 recombinates the second noise reduction amplitude spectrum and the phase spectrum, generates the first reducing noise of voice.
Illustratively, the mobile terminal in the embodiment of the present application may include that mobile phone and tablet computer etc. have call function
Smart machine.Earphone in the present embodiment can send out the voice signal of acquisition with mobile terminal wired connection or wireless connection
It send to mobile terminal.It should be noted that acquiring voice signal based on the single microphone being arranged in earphone, Mike in earphone is reduced
Wind quantity reduces hardware cost.
After ear microphone acquires analog signal, the voice signal of acquisition is carried out by analog-to-digital conversion based on analog-digital converter,
Audio digital signals are obtained, audio digital signals are transmitted to mobile terminal.Optionally, earphone carries out the analog signal of acquisition
Segmentation, will treated each section of voice signal real-time Transmission to mobile terminal.For example, can will be 0 per signal strength twice
Time point between analog signal be divided into a signal segment.Optionally, carrying out segmentation to analog signal includes: with microphone
Acquisition voice signal Startup time is initial time, is denoted as sart point in time t0；It searches in analog signal after sart point in time t0
First time occur signal strength be 0 signal time point, by after sart point in time t0 first time occur signal it is strong
The time point for the signal that degree is 0 is as end time point t1；By the simulation letter between sart point in time t0 and end time point t1
Number it is divided into a signal segment；The signal for being 0 with the presence or absence of signal strength is continued to test, if so, will be after end time point t1
First time occur signal strength be 0 signal time point, by after end time point t1 first time occur signal it is strong
Degree for 0 signal time point as next signal segment at the beginning of point t0, and execution above-mentioned steps are recycled, in real time to voice
The analog signal of acquisition equipment acquisition is segmented.Optionally, it can also be and analog signal carried out according to fixed interval
Segmentation.Illustratively, fixed interval T is triggered with sound acquisition equipment constantly for initial time, the segmentation of analog signal
It can be 0-T, T-2T and 2T-3T etc., wherein T can be 500ms or 1s.
The voice signal that mobile terminal real-time reception earphone is sent carries out Short Time Fourier Transform to voice signal, generates
The initial amplitude of voice signal is composed and phase spectrum.Wherein, initial amplitude spectrum is that the range value of voice signal is bent with the variation of frequency
Line, phase spectrum are the phase curves varying with frequency of voice signal.In the present embodiment, initial amplitude is composed based on noise reduction model
It is handled, to realize to the function of voice signal noise reduction, while phase spectrum is constant, ensure that the accurate of voice signal after noise reduction
Degree.
Optionally, the noise reduction model is deep neural network model (the Deep Neural with decrease of noise functions
Networks, DNN) or production confrontation network model (Generative Adversarial Net, GAN).For having
The DNN network model of decrease of noise functions is generated by great amount of samples training.Wherein sample includes clean speech and based on clean speech
Noise signal generates plus voice of making an uproar is added, DNN network model can be the training by way of supervised learning and obtain, specifically
Voice of making an uproar will be added to be input in DNN noise reduction model to be trained, based on DNN noise reduction model to be trained, obtain processed drop
It makes an uproar voice, which is compared with clean speech, when reducing noise of voice and clean speech difference, according to reducing noise of voice
And the deviation of clean speech reversely adjusts the weight parameter in DNN noise reduction model, and iteration executes above-mentioned training process, until DNN
The reducing noise of voice and clean speech of the output of noise reduction model reach default similarity.The DNN noise reduction model that training is completed has language
Sound decrease of noise functions, after the initial amplitude spectrum of the received voice signal of mobile terminal is input to the DNN noise reduction model of training completion,
It can obtain the first noise reduction amplitude spectrum of the voice signal.
It for the GAN network model with decrease of noise functions, including generates sub-network and differentiates sub-network, generate sub-network and use
In carrying out noise reduction process to the amplitude spectrum of input, differentiate sub-network for differentiating to the amplitude spectrum of input, it is determined whether packet
Noisy acoustical signal.Such as first differentiation submodel is trained based on training sample, it is improved by adjusting network parameter and differentiates son
The discrimination precision of model, after the completion of differentiating submodel training, the fixed network parameter for differentiating submodel, to generate submodel into
Row training, adjusts the network parameter for generating submodel, so that including the differentiation of noise signal in generation submodel output amplitude spectrum
Probability decline.Above-mentioned training process is recycled, when the output result for differentiating submodel and generation submodel meets default error, really
Determine production confrontation network model training to complete.After the completion of the training of GAN network model, initial amplitude spectrum is input to GAN net
The amplitude spectrum for generating submodel input is determined as the first noise reduction amplitude spectrum by the generation submodel of network model.Optionally, to first
Before beginning amplitude spectrum is handled, it can also be and initial amplitude spectrum is input in the differentiation submodel of GAN network model, according to
Whether differentiate that the output result of submodel determines in initial amplitude spectrum includes noise signal, input if so, initial amplitude is composed
To the generation submodel of GAN network model, if not, it is determined that the voice signal of earphone acquisition is clean signal, without being dropped
It makes an uproar processing.In some embodiments, it can also be the differentiation submodel that the first noise reduction amplitude spectrum is input to GAN network model
In, when being greater than preset value according to the probability in the output result for differentiating submodel in the first noise reduction amplitude spectrum including noise signal,
It is input to the generation submodel of GAN network model to the first noise reduction amplitude spectrum circulation, carries out secondary noise reduction, until output result is full
Sufficient preset requirement.By the detection to the first noise reduction amplitude spectrum, noise reduction precision is improved, voice is believed after further improving processing
Number clarity.
In the present embodiment, after determining the first noise reduction amplitude spectrum, masking processing is carried out to the first noise reduction amplitude spectrum, is used
In providing the quality of voice signal, the distorted signals caused by noise reduction process is avoided.Masking processing is for removing the first noise reduction width
Distorted signal in degree spectrum compensates.Optionally, distorted signals is judged whether there is according to the first noise reduction amplitude spectrum, if it is not,
Step 103 is then omitted, the first noise reduction amplitude spectrum is recombinated with phase spectrum, obtains the first reducing noise of voice；If so, to described
One noise reduction amplitude spectrum carries out masking processing, obtains the second noise reduction amplitude spectrum, recombinates to the second noise reduction amplitude spectrum and phase spectrum,
And it carries out inverse Fourier transform in short-term and generates the first reducing noise of voice.
After obtaining the first reducing noise of voice, the first reducing noise of voice is transmitted.Illustratively, mobile terminal processing is logical
First reducing noise of voice is then transmitted to the uplink of call, is transferred to counterpart mobile terminal by speech phase.Illustratively, mobile
First reducing noise of voice is sent to instant communication server, is transmitted to counterpart mobile terminal by terminal processes instant communication state.It is real
Now to the quick high accuracy noise reduction of the voice signal of earphone acquisition, speech quality is improved.
The voice de-noising method provided in the embodiment of the present application is believed by receiving the voice acquired based on earphone single microphone
Number, the initial amplitude spectrum and phase spectrum of the voice signal are extracted, initial amplitude spectrum is transmitted to noise reduction trained in advance
In model, the first noise reduction amplitude spectrum is obtained, masking processing is carried out to the first noise reduction amplitude spectrum, generates the second noise reduction amplitude
Spectrum, recombinates the second noise reduction amplitude spectrum and the phase spectrum, generates the first reducing noise of voice.By using above-mentioned side
Case keeps the phase spectrum of voice signal constant, carries out noise reduction process based on amplitude spectrum of the noise reduction model to voice signal, and to place
It manages the first obtained noise reduction amplitude spectrum and carries out masking processing, eliminate the problem of dtmf distortion DTMF in noise reduction process, solve in noisy environment
The big problem of the pronunciation signal noise of lower acquisition realizes the quick high accuracy noise reduction of the voice signal acquired to earphone.
Fig. 2 is the flow diagram of another voice de-noising method provided by the embodiments of the present application, referring to fig. 2, this implementation
The method of example includes the following steps:
The voice signal that step 201, reception are acquired based on earphone single microphone.
Step 202, the initial amplitude spectrum and phase spectrum for extracting the voice signal, initial amplitude spectrum are transmitted to pre-
First in trained noise reduction model, the first noise reduction amplitude spectrum is obtained.
The range value of each frequency point of current demand signal frame in the first noise reduction amplitude spectrum and described second drop in step 203
The range value of the correspondence frequency point of a upper signal frame in amplitude spectrum of making an uproar is smoothed, and generates the second noise reduction width of current demand signal frame
Degree spectrum.
Step 204 recombinates the second noise reduction amplitude spectrum and the phase spectrum, generates the first reducing noise of voice.
In the present embodiment, during carrying out masking processing to the first noise reduction amplitude spectrum, as unit of signal frame
Reason, for arbitrary signal frame, based on the second noise reduction amplitude spectrum in preceding signal frame to the first noise reduction amplitude spectrum of current demand signal frame
It is smoothed, specifically, to any frequency point in current demand signal frame, by the range value of the frequency point and a upper signal frame the
The range value of the frequency point is smoothed in two noise reduction amplitude spectrums, obtains the frequency point in current demand signal frame the second noise reduction amplitude spectrum
Range value.Wherein, the masking factor being smoothed can be fixed preset value, can also be according to real-time change
Range value determines.
Optionally, masking processing is carried out to the first noise reduction amplitude spectrum and meets following formula:
Wherein, λ (m, k) is masking factor, which meets following formula:
And
Wherein,For the second noise reduction amplitude spectrum,For the first noise reduction amplitude spectrum, m is the frame number of voice signal, and k is frequency point,
σ is standard deviation.
The voice de-noising method provided in the embodiment of the present application is kept away by carrying out masking processing to the first noise reduction amplitude spectrum
Exempt from distorted signals caused by noise reduction process, improves noise-reduced speech signal quality.
Fig. 3 is the flow diagram of another voice de-noising method provided by the embodiments of the present application, and the present embodiment is above-mentioned
One optinal plan of embodiment, correspondingly, as shown in figure 3, the method for the present embodiment includes the following steps:
The voice signal that step 301, reception are acquired based on earphone single microphone.
Step 302, the initial amplitude spectrum and phase spectrum for extracting the voice signal, initial amplitude spectrum are transmitted to pre-
First in trained noise reduction model, the first noise reduction amplitude spectrum is obtained.
Step 303 carries out masking processing to the first noise reduction amplitude spectrum, generates the second noise reduction amplitude spectrum.
Step 304 recombinates the second noise reduction amplitude spectrum and the phase spectrum, generates the first reducing noise of voice, will
First reducing noise of voice is sent to the uplink of call, is transferred to counterpart mobile terminal.
Step 305, the call voice signal for receiving downlink transmission carry out noise to the call voice signal and estimate
Meter.
Step 306, when determining that the call voice signal includes noise signal according to noise estimation, by the call voice
The amplitude spectrum of signal is input in noise reduction model trained in advance, obtains third noise reduction amplitude spectrum.
Step 307 generates the second reducing noise of voice according to third noise reduction amplitude spectrum, and second reducing noise of voice is broadcast
It puts.
In the present embodiment, step 301 after carrying out noise reduction to voice signal, is passed to step 304 for acquiring voice signal
Counterpart mobile terminal is transported to, realizes the transmission of voice signal in communication process；Step 305 is moved to step 307 for receiving other side
The voice signal that dynamic terminal is sent, when the voice signal is there are when noise signal, after carrying out noise reduction to received voice signal, base
Loudspeaker in earphone or mobile terminal plays out, and realizes the reception of voice signal in communication process.
Wherein, in step 305, the call voice signal for receiving downlink transmission, makes an uproar to call voice signal
Sound estimation, when call voice signal be clean speech signal, when not including noise signal, which is broadcast
It puts.When call voice signal includes noise signal, Short Time Fourier Transform is carried out to call voice signal, obtains call voice
The amplitude spectrum and phase spectrum of signal will be carried out at noise reduction based on amplitude spectrum of the noise reduction model trained in advance to call voice signal
Reason, obtains third noise reduction amplitude spectrum, the phase spectrum of third noise reduction amplitude spectrum and call voice signal is recombinated, and carry out short
When inverse Fourier transform, obtain the second reducing noise of voice, by the second reducing noise of voice be sent to the loudspeaker of mobile terminal or earphone into
Row plays.By carrying out noise estimation to call voice signal, the call voice signal comprising noise signal be directed to
Property noise reduction process, avoid carrying out clean speech signal invalid noise reduction process, improve Speech processing efficiency, avoid leading to
Letter delay, improves speech quality.
In the present embodiment, when the generation submodel of application GAN network model carries out noise reduction process to voice signal, base
Noise estimation is carried out to call voice signal in the differentiation submodel of application GAN network model.It is described that the call voice is believed
Number carry out noise estimation, comprising: by the call voice signal be input to production confrontation network model differentiation submodel in,
Determine whether the call voice signal includes noise signal according to the output result for differentiating submodel, wherein described to sentence
Small pin for the case model is used to carry out noise estimation to input voice, carries out noise to voice signal respectively based on same GAN network model
Estimation and noise reduction process not only increase signal processing efficiency, also save network it is not necessary that different network models is respectively set
The memory source that model occupies.
In some embodiments, described that noise estimation is carried out to the call voice signal, comprising: by the call voice
Signal is compared with preset template signal, calculates the difference power of the call voice signal and preset template signal, root
Determine whether the call voice signal includes noise signal according to the difference power.Wherein, it is more to can be storage in mobile terminal
Template Information similar in the voice content of the template signal of a different phonetic content, selection and call voice signal is to call voice
Signal carries out noise evaluation.Illustratively, call voice signal and the difference power of preset template signal are bigger, show language of conversing
Sound signal includes that noise signal is more, and call voice signal and the difference power of preset template signal are smaller, shows call voice
Signal includes that noise signal is fewer.Power threshold is set, when call voice signal and the difference power of preset template signal are greater than
Or when being equal to power threshold, determine that the call voice signal includes noise signal, when call voice signal and preset template
When the difference power of signal is less than power threshold, determine that call voice signal does not include noise signal, i.e., call voice signal is dry
Net voice signal.Wherein, power threshold can be is arranged according to user demand, if it is desired to which high quality is conversed, is then arranged lesser
Power threshold.It can be after receiving the voice signal acquired based on earphone single microphone, the voice signal of acquisition carried out
Noise evaluation directly sends the upper of call for the voice signal of acquisition when determining in voice signal does not include noise signal
Row access, is transferred to counterpart mobile terminal.
It should be noted that step 301 is not limited in the present embodiment to step 304 and step 305 to the execution of step 307
Sequentially, in other embodiments, it can also be and first carry out step 305 to step 307, then execute step 301 to step 304.
The voice de-noising method provided in the embodiment of the present application, by believing the voice of earphone acquisition under through state
Number and counterpart mobile terminal send call voice signal judged, when comprising noise signal, be based on preset noise reduction
Model carries out noise reduction process, solves the problems, such as that speech quality is poor when conversing under noisy environment, improves noise reduction precision and lead to
Talk about quality.
Fig. 4 is a kind of structural block diagram of voice noise reduction device provided by the embodiments of the present application, the device can by software and/or
Hardware realization is typically integrated in mobile terminal, can by execute mobile terminal voice de-noising method come to desktop layouts into
Edlin.As shown in figure 4, the device includes: speech reception module 401, the first noise reduction amplitude spectrum generation module 402, second drop
It makes an uproar amplitude spectrum generation module 403 and the first reducing noise of voice generation module 404.
Speech reception module 401, for receiving the voice signal based on the acquisition of earphone single microphone；
First noise reduction amplitude spectrum generation module 402, the initial amplitude for extracting the voice signal is composed and phase spectrum, will
The initial amplitude spectrum is transmitted in noise reduction model trained in advance, obtains the first noise reduction amplitude spectrum；
Second noise reduction amplitude spectrum generation module 403 generates for carrying out masking processing to the first noise reduction amplitude spectrum
Two noise reduction amplitude spectrums；
First reducing noise of voice generation module 404, for being recombinated to the second noise reduction amplitude spectrum and the phase spectrum,
Generate the first reducing noise of voice.
The voice noise reduction device provided in the embodiment of the present application keeps the phase spectrum of voice signal constant, is based on noise reduction mould
Type carries out noise reduction process to the amplitude spectrum of voice signal, and the first noise reduction amplitude spectrum obtained to processing carries out masking processing, disappears
It except the problem of dtmf distortion DTMF in noise reduction process, solves the problems, such as that the pronunciation signal noise acquired under noisy environment is big, realizes to ear
The quick high accuracy noise reduction of the voice signal of machine acquisition.
On the basis of the above embodiments, the noise reduction model is that deep neural network model or production fight network
Model.
On the basis of the above embodiments, the second noise reduction amplitude spectrum generation module 403 is used for:
By the range value of each frequency point of current demand signal frame in the first noise reduction amplitude spectrum and the second noise reduction amplitude spectrum
In the range value of correspondence frequency point of a upper signal frame be smoothed, generate the second noise reduction amplitude spectrum of current demand signal frame.
On the basis of the above embodiments, the second noise reduction amplitude spectrum generation module 403 meets following formula:
Wherein, masking factor λ (m, k) meets following formula:
And
Wherein,For the second noise reduction amplitude spectrum,For the first noise reduction amplitude spectrum, m is the frame number of voice signal, and k is frequency point,
σ is standard deviation.
On the basis of the above embodiments, the initial amplitude spectrum of the voice signal and phase spectrum are based on believing the voice
Number carry out Short Time Fourier Transform generation；
The reducing noise of voice based on to after recombination the second noise reduction amplitude spectrum and the phase spectrum carry out in short-term Fourier it is inverse
Transformation generates.
On the basis of the above embodiments, further includes:
Noise estimation module, for receive downlink transmission call voice signal, to the call voice signal into
The estimation of row noise；
Third noise reduction amplitude spectrum generation module, for determining that the call voice signal includes noise letter according to noise estimation
Number when, the amplitude spectrum of the call voice signal is input in advance trained noise reduction model, obtains third noise reduction amplitude spectrum；
Second reducing noise of voice generation module, for generating the second reducing noise of voice according to third noise reduction amplitude spectrum, and will be described
Second reducing noise of voice plays out.
On the basis of the above embodiments, noise estimation module is used for:
The call voice signal is compared with preset template signal, calculate the call voice signal and is preset
Template signal difference power, determine whether the call voice signal includes noise signal according to the difference power.
On the basis of the above embodiments, noise estimation module is used for:
The call voice signal is input in the differentiation submodel of production confrontation network model, according to the differentiation
The output result of submodel determines whether the call voice signal includes noise signal, wherein the differentiation submodel is used for
Noise estimation is carried out to input voice.
The embodiment of the present application also provides a kind of storage medium comprising computer executable instructions, and the computer is executable
Instruction is used to execute voice de-noising method when being executed by computer processor, this method comprises:
Receive the voice signal acquired based on earphone single microphone；
The initial amplitude spectrum and phase spectrum for extracting the voice signal, are transmitted to training in advance for initial amplitude spectrum
In noise reduction model, the first noise reduction amplitude spectrum is obtained；
Masking processing is carried out to the first noise reduction amplitude spectrum, generates the second noise reduction amplitude spectrum；
The second noise reduction amplitude spectrum and the phase spectrum are recombinated, the first reducing noise of voice is generated.
Storage medium --- any various types of memory devices or storage equipment.Term " storage medium " is intended to wrap
It includes: install medium, such as CD-ROM, floppy disk or magnetic tape equipment；Computer system memory or random access memory, such as
DRAM, DDRRAM, SRAM, EDORAM, blue Bath (Rambus) RAM etc.；Nonvolatile memory, such as flash memory, magnetic medium (example
Such as hard disk or optical storage)；Register or the memory component of other similar types etc..Storage medium can further include other types
Memory or combinations thereof.In addition, storage medium can be located at program in the first computer system being wherein performed, or
It can be located in different second computer systems, second computer system is connected to the first meter by network (such as internet)
Calculation machine system.Second computer system can provide program instruction to the first computer for executing.Term " storage medium " can
To include two or more that may reside in different location (such as in the different computer systems by network connection)
Storage medium.Storage medium can store the program instruction that can be performed by one or more processors and (such as be implemented as counting
Calculation machine program).
Certainly, a kind of storage medium comprising computer executable instructions, computer provided by the embodiment of the present application
The voice de-noising operation that executable instruction is not limited to the described above, can also be performed voice provided by the application any embodiment
Relevant operation in noise-reduction method.
The embodiment of the present application provides a kind of mobile terminal, and language provided by the embodiments of the present application can be integrated in the mobile terminal
Sound denoising device.Fig. 5 is a kind of structural schematic diagram of mobile terminal provided by the embodiments of the present application.Mobile terminal 500 can wrap
It includes: memory 501, processor 502 and the computer program that is stored on memory 501 and can be run in processor 502, it is described
Processor 502 realizes the voice de-noising method as described in the embodiment of the present application when executing the computer program.
Mobile terminal provided by the embodiments of the present application keeps the phase spectrum of voice signal constant, based on noise reduction model to language
The amplitude spectrum of sound signal carries out noise reduction process, and the first noise reduction amplitude spectrum obtained to processing carries out masking processing, eliminates noise reduction
Problem of dtmf distortion DTMF in the process solves the problems, such as that the pronunciation signal noise acquired under noisy environment is big, realizes and acquires to earphone
Voice signal quick high accuracy noise reduction.
Fig. 6 is the structural schematic diagram of another mobile terminal provided by the embodiments of the present application.The mobile terminal may include:
Shell (not shown), memory 601, central processing unit (central processing unit, CPU) 602 (are also known as located
Manage device, hereinafter referred to as CPU), circuit board (not shown) and power circuit (not shown).The circuit board is placed in institute
State the space interior that shell surrounds；The CPU602 and the memory 601 are arranged on the circuit board；The power supply electricity
Road, for each circuit or the device power supply for the mobile terminal；The memory 601, for storing executable program generation
Code；The CPU602 is run and the executable journey by reading the executable program code stored in the memory 601
The corresponding computer program of sequence code, to perform the steps of
Receive the voice signal acquired based on earphone single microphone；
The initial amplitude spectrum and phase spectrum for extracting the voice signal, are transmitted to training in advance for initial amplitude spectrum
In noise reduction model, the first noise reduction amplitude spectrum is obtained；
Masking processing is carried out to the first noise reduction amplitude spectrum, generates the second noise reduction amplitude spectrum；
The second noise reduction amplitude spectrum and the phase spectrum are recombinated, the first reducing noise of voice is generated.
The mobile terminal further include: Peripheral Interface 603, RF (Radio Frequency, radio frequency) circuit 605, audio-frequency electric
Road 606, loudspeaker 611, power management chip 608, input/output (I/O) subsystem 609, other input/control devicess 610,
Touch screen 612, other input/control devicess 610 and outside port 604, these components pass through one or more communication bus
Or signal wire 607 communicates.
It should be understood that illustrating the example that mobile terminal 600 is only mobile terminal, and mobile terminal 600
It can have than shown in the drawings more or less component, can combine two or more components, or can be with
It is configured with different components.Various parts shown in the drawings can include one or more signal processings and/or dedicated
It is realized in the combination of hardware, software or hardware and software including integrated circuit.
Just the mobile terminal provided in this embodiment for operating to voice de-noising is described in detail below, the movement
Terminal takes the mobile phone as an example.
Memory 601, the memory 601 can be accessed by CPU602, Peripheral Interface 603 etc., and the memory 601 can
It can also include nonvolatile memory to include high-speed random access memory, such as one or more disk memory,
Flush memory device or other volatile solid-state parts.
The peripheral hardware that outputs and inputs of equipment can be connected to CPU602 and deposited by Peripheral Interface 603, the Peripheral Interface 603
Reservoir 601.
I/O subsystem 609, the I/O subsystem 609 can be by the input/output peripherals in equipment, such as touch screen 612
With other input/control devicess 610, it is connected to Peripheral Interface 603.I/O subsystem 609 may include 6091 He of display controller
For controlling one or more input controllers 6092 of other input/control devicess 610.Wherein, one or more input controls
Device 6092 processed receives electric signal from other input/control devicess 610 or sends electric signal to other input/control devicess 610,
Other input/control devicess 610 may include physical button (push button, rocker buttons etc.), dial, slide switch, behaviour
Vertical pole clicks idler wheel.It is worth noting that input controller 6092 can with it is following any one connect: keyboard, infrared port,
The indicating equipment of USB interface and such as mouse.
Touch screen 612, the touch screen 612 are the input interface and output interface between customer mobile terminal and user,
Visual output is shown to user, visual output may include figure, text, icon, video etc..
Display controller 6091 in I/O subsystem 609 receives electric signal from touch screen 612 or sends out to touch screen 612
Electric signals.Touch screen 612 detects the contact on touch screen, and the contact that display controller 6091 will test is converted to and is shown
The interaction of user interface object on touch screen 612, i.e. realization human-computer interaction, the user interface being shown on touch screen 612
Object can be the icon of running game, the icon for being networked to corresponding network etc..It is worth noting that equipment can also include light
Mouse, light mouse are the extensions for the touch sensitive surface for not showing the touch sensitive surface visually exported, or formed by touch screen.
RF circuit 605 is mainly used for establishing the communication of mobile phone Yu wireless network (i.e. network side), realizes mobile phone and wireless network
The data receiver of network and transmission.Such as transmitting-receiving short message, Email etc..Specifically, RF circuit 605 receives and sends RF letter
Number, RF signal is also referred to as electromagnetic signal, and RF circuit 605 converts electrical signals to electromagnetic signal or electromagnetic signal is converted to telecommunications
Number, and communicated by the electromagnetic signal with communication network and other equipment.RF circuit 605 may include for executing
The known circuit of these functions comprising but it is not limited to antenna system, RF transceiver, one or more amplifiers, tuner, one
A or multiple oscillators, digital signal processor, CODEC (COder-DECoder, coder) chipset, user identifier mould
Block (Subscriber Identity Module, SIM) etc..
Voicefrequency circuit 606 is mainly used for receiving audio data from Peripheral Interface 603, which is converted to telecommunications
Number, and the electric signal is sent to loudspeaker 611.
Loudspeaker 611 is reduced to sound for mobile phone to be passed through RF circuit 605 from the received voice signal of wireless network
And the sound is played to user.
Power management chip 608, the hardware for being connected by CPU602, I/O subsystem and Peripheral Interface are powered
And power management.
The application, which can be performed, in voice noise reduction device, storage medium and the mobile terminal provided in above-described embodiment arbitrarily implements
Voice de-noising method provided by example has and executes the corresponding functional module of this method and beneficial effect.Not in above-described embodiment
In detailed description technical detail, reference can be made to voice de-noising method provided by the application any embodiment.
Note that above are only the preferred embodiment and institute's application technology principle of the application.It will be appreciated by those skilled in the art that
The application is not limited to specific embodiment described here, be able to carry out for a person skilled in the art it is various it is apparent variation,
The protection scope readjusted and substituted without departing from the application.Therefore, although being carried out by above embodiments to the application
It is described in further detail, but the application is not limited only to above embodiments, in the case where not departing from the application design, also
It may include more other equivalent embodiments, and scope of the present application is determined by the scope of the appended claims.
Claims (10)
1. a kind of voice de-noising method characterized by comprising
Receive the voice signal acquired based on earphone single microphone；
The initial amplitude spectrum and phase spectrum for extracting the voice signal, are transmitted to noise reduction trained in advance for initial amplitude spectrum
In model, the first noise reduction amplitude spectrum is obtained；
Masking processing is carried out to the first noise reduction amplitude spectrum, generates the second noise reduction amplitude spectrum；
The second noise reduction amplitude spectrum and the phase spectrum are recombinated, the first reducing noise of voice is generated.
2. the method according to claim 1, wherein the noise reduction model is deep neural network model or life
An accepted way of doing sth fights network model.
3. being given birth to the method according to claim 1, wherein carrying out masking processing to the first noise reduction amplitude spectrum
At the second noise reduction amplitude spectrum, comprising:
On in the range value of each frequency point of current demand signal frame in the first noise reduction amplitude spectrum and the second noise reduction amplitude spectrum
The range value of the correspondence frequency point of one signal frame is smoothed, and generates the second noise reduction amplitude spectrum of current demand signal frame.
4. according to the method described in claim 3, it is characterized in that, by the first noise reduction amplitude spectrum current demand signal frame it is each
The range value of frequency point is smoothed with the range value of the corresponding frequency point of a upper signal frame in the second noise reduction amplitude spectrum, full
The following formula of foot:
Wherein, masking factor λ (m, k) meets following formula:
And
Wherein,For the second noise reduction amplitude spectrum,For the first noise reduction amplitude spectrum, m is the frame number of voice signal, and k is frequency point, and σ is
Standard deviation.
5. the method according to claim 1, wherein the initial amplitude spectrum and phase spectrum of the voice signal are based on
Short Time Fourier Transform generation is carried out to the voice signal；
The reducing noise of voice based on to after recombination the second noise reduction amplitude spectrum and the phase spectrum carry out inverse Fourier transform in short-term
It generates.
6. the method according to claim 1, wherein the method also includes:
The call voice signal for receiving downlink transmission carries out noise estimation to the call voice signal；
When determining that the call voice signal includes noise signal according to noise estimation, by the amplitude spectrum of the call voice signal
It is input in noise reduction model trained in advance, obtains third noise reduction amplitude spectrum；
The second reducing noise of voice is generated according to third noise reduction amplitude spectrum, and second reducing noise of voice is played out.
7. according to the method described in claim 6, it is characterized in that, it is described to the call voice signal carry out noise estimation,
Include:
The call voice signal is compared with preset template signal, calculates the call voice signal and preset mould
The difference power of partitioned signal determines whether the call voice signal includes noise signal according to the difference power；Alternatively,
The call voice signal is input in the differentiation submodel of production confrontation network model, according to the differentiation submodule
The output result of type determines whether the call voice signal includes noise signal, wherein the differentiation submodel is used for defeated
Enter voice and carries out noise estimation.
8. a kind of voice noise reduction device characterized by comprising
Speech reception module, for receiving the voice signal based on the acquisition of earphone single microphone；
First noise reduction amplitude spectrum generation module, the initial amplitude for extracting the voice signal is composed and phase spectrum, will be described first
Beginning amplitude spectrum is transmitted in noise reduction model trained in advance, obtains the first noise reduction amplitude spectrum；
Second noise reduction amplitude spectrum generation module generates the second noise reduction for carrying out masking processing to the first noise reduction amplitude spectrum
Amplitude spectrum；
First reducing noise of voice generation module generates for recombinating to the second noise reduction amplitude spectrum and the phase spectrum
One reducing noise of voice.
9. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the program is held by processor
The voice de-noising method as described in any in claim 1-7 is realized when row.
10. a kind of mobile terminal, which is characterized in that including memory, processor and storage are on a memory and can be in processor
The computer program of operation, the processor realize language as claimed in claim 1 when executing the computer program
Sound noise-reduction method.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811273582.5A CN109119093A (en) | 2018-10-30 | 2018-10-30 | Voice de-noising method, device, storage medium and mobile terminal |
| PCT/CN2019/107579 WO2020088154A1 (en) | 2018-10-30 | 2019-09-24 | Method for voice audio noise reduction, storage medium and mobile terminal |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811273582.5A CN109119093A (en) | 2018-10-30 | 2018-10-30 | Voice de-noising method, device, storage medium and mobile terminal |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109119093A true CN109119093A (en) | 2019-01-01 |
Family
ID=64854717
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811273582.5A Pending CN109119093A (en) | 2018-10-30 | 2018-10-30 | Voice de-noising method, device, storage medium and mobile terminal |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN109119093A (en) |
| WO (1) | WO2020088154A1 (en) |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110751960A (en) * | 2019-10-16 | 2020-02-04 | 北京网众共创科技有限公司 | Method and device for determining noise data |
| CN110797041A (en) * | 2019-10-21 | 2020-02-14 | 珠海市杰理科技股份有限公司 | Voice noise reduction processing method and device, computer equipment and storage medium |
| CN110795892A (en) * | 2019-10-23 | 2020-02-14 | 北京邮电大学 | Channel simulation method and device based on generation countermeasure network |
| WO2020088154A1 (en) * | 2018-10-30 | 2020-05-07 | Oppo广东移动通信有限公司 | Method for voice audio noise reduction, storage medium and mobile terminal |
| CN111128230A (en) * | 2019-12-31 | 2020-05-08 | 广州市百果园信息技术有限公司 | Voice signal reconstruction method, device, equipment and storage medium |
| CN111341304A (en) * | 2020-02-28 | 2020-06-26 | 广州国音智能科技有限公司 | Method, device and equipment for training speech characteristics of speaker based on GAN |
| CN111370017A (en) * | 2020-03-18 | 2020-07-03 | 苏宁云计算有限公司 | Voice enhancement method, device and system |
| CN111477237A (en) * | 2019-01-04 | 2020-07-31 | 北京京东尚科信息技术有限公司 | Audio noise reduction method and device and electronic equipment |
| CN111933171A (en) * | 2020-09-21 | 2020-11-13 | 北京达佳互联信息技术有限公司 | Noise reduction method and device, electronic equipment and storage medium |
| CN112053421A (en) * | 2020-10-14 | 2020-12-08 | 腾讯科技（深圳）有限公司 | Signal noise reduction processing method, device, equipment and storage medium |
| CN112382284A (en) * | 2020-10-22 | 2021-02-19 | 珠海格力电器股份有限公司 | Pickup noise reduction method and device and server |
| WO2021042538A1 (en) * | 2019-09-02 | 2021-03-11 | 平安科技（深圳）有限公司 | Method and device for audio processing, and computer storage medium |
| CN113066476A (en) * | 2019-12-13 | 2021-07-02 | 科大讯飞股份有限公司 | Synthetic speech processing method and related device |
| CN113098901A (en) * | 2021-04-29 | 2021-07-09 | 心动互动娱乐有限公司 | High-concurrency real-time voice call method, system, device and medium |
| CN113539290A (en) * | 2020-04-22 | 2021-10-22 | 华为技术有限公司 | Voice noise reduction method and device |
| CN113823312A (en) * | 2021-02-19 | 2021-12-21 | 北京沃东天骏信息技术有限公司 | Speech enhancement model generation method and device and speech enhancement method and device |
| CN114007157A (en) * | 2021-10-28 | 2022-02-01 | 中北大学 | Intelligent noise reduction communication earphone |
| CN114333882A (en) * | 2022-03-09 | 2022-04-12 | 深圳市友杰智新科技有限公司 | Voice noise reduction method, device and equipment based on amplitude spectrum and storage medium |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112700786B (en) * | 2020-12-29 | 2024-03-12 | 西安讯飞超脑信息科技有限公司 | Speech enhancement method, device, electronic equipment and storage medium |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104200811A (en) * | 2014-08-08 | 2014-12-10 | 华迪计算机集团有限公司 | Self-adaption spectral subtraction and noise elimination processing method and device for voice signals |
| CN104810024A (en) * | 2014-01-28 | 2015-07-29 | 上海力声特医学科技有限公司 | Double-path microphone speech noise reduction treatment method and system |
| CN105405439A (en) * | 2015-11-04 | 2016-03-16 | 科大讯飞股份有限公司 | Voice playing method and device |
| CN105427859A (en) * | 2016-01-07 | 2016-03-23 | 深圳市音加密科技有限公司 | Front voice enhancement method for identifying speaker |
| US9607627B2 (en) * | 2015-02-05 | 2017-03-28 | Adobe Systems Incorporated | Sound enhancement through deverberation |
| CN107077860A (en) * | 2014-10-21 | 2017-08-18 | 三菱电机株式会社 | Method for will there is audio signal of making an uproar to be converted to enhancing audio signal |
| CN107452389A (en) * | 2017-07-20 | 2017-12-08 | 大象声科（深圳）科技有限公司 | A kind of general monophonic real-time noise-reducing method |
| CN107945811A (en) * | 2017-10-23 | 2018-04-20 | 北京大学 | A kind of production towards bandspreading resists network training method and audio coding, coding/decoding method |
| CN108231086A (en) * | 2017-12-24 | 2018-06-29 | 航天恒星科技有限公司 | A kind of deep learning voice enhancer and method based on FPGA |
| CN108597496A (en) * | 2018-05-07 | 2018-09-28 | 广州势必可赢网络科技有限公司 | A kind of speech production method and device for fighting network based on production |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2012155301A (en) * | 2011-01-21 | 2012-08-16 | Wrk Solution Co Ltd | State recognition type speech recognition method |
| CN107293289B (en) * | 2017-06-13 | 2020-05-29 | 南京医科大学 | Speech generation method for generating confrontation network based on deep convolution |
| CN108346433A (en) * | 2017-12-28 | 2018-07-31 | 北京搜狗科技发展有限公司 | A kind of audio-frequency processing method, device, equipment and readable storage medium storing program for executing |
| CN109119093A (en) * | 2018-10-30 | 2019-01-01 | Oppo广东移动通信有限公司 | Voice de-noising method, device, storage medium and mobile terminal |
-
2018
- 2018-10-30 CN CN201811273582.5A patent/CN109119093A/en active Pending
-
2019
- 2019-09-24 WO PCT/CN2019/107579 patent/WO2020088154A1/en active Application Filing
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104810024A (en) * | 2014-01-28 | 2015-07-29 | 上海力声特医学科技有限公司 | Double-path microphone speech noise reduction treatment method and system |
| CN104200811A (en) * | 2014-08-08 | 2014-12-10 | 华迪计算机集团有限公司 | Self-adaption spectral subtraction and noise elimination processing method and device for voice signals |
| CN107077860A (en) * | 2014-10-21 | 2017-08-18 | 三菱电机株式会社 | Method for will there is audio signal of making an uproar to be converted to enhancing audio signal |
| US9607627B2 (en) * | 2015-02-05 | 2017-03-28 | Adobe Systems Incorporated | Sound enhancement through deverberation |
| CN105405439A (en) * | 2015-11-04 | 2016-03-16 | 科大讯飞股份有限公司 | Voice playing method and device |
| CN105427859A (en) * | 2016-01-07 | 2016-03-23 | 深圳市音加密科技有限公司 | Front voice enhancement method for identifying speaker |
| CN107452389A (en) * | 2017-07-20 | 2017-12-08 | 大象声科（深圳）科技有限公司 | A kind of general monophonic real-time noise-reducing method |
| CN107945811A (en) * | 2017-10-23 | 2018-04-20 | 北京大学 | A kind of production towards bandspreading resists network training method and audio coding, coding/decoding method |
| CN108231086A (en) * | 2017-12-24 | 2018-06-29 | 航天恒星科技有限公司 | A kind of deep learning voice enhancer and method based on FPGA |
| CN108597496A (en) * | 2018-05-07 | 2018-09-28 | 广州势必可赢网络科技有限公司 | A kind of speech production method and device for fighting network based on production |
Non-Patent Citations (3)
| Title |
|---|
| 徐勇: "基于深层神经网络的语音增强方法研究", 《中国博士学位论文全文数据库信息科技辑（月刊）》 * |
| 王海武: "基于生成对抗网络的语音增强方法研究", 《中国优秀硕士学位论文全文数据库信息科技辑（月刊）》 * |
| 韩伟等: "基于感知掩蔽深度神经网络的单通道语音增强方法", 《自动化学报》 * |
Cited By (28)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020088154A1 (en) * | 2018-10-30 | 2020-05-07 | Oppo广东移动通信有限公司 | Method for voice audio noise reduction, storage medium and mobile terminal |
| CN111477237B (en) * | 2019-01-04 | 2022-01-07 | 北京京东尚科信息技术有限公司 | Audio noise reduction method and device and electronic equipment |
| CN111477237A (en) * | 2019-01-04 | 2020-07-31 | 北京京东尚科信息技术有限公司 | Audio noise reduction method and device and electronic equipment |
| WO2021042538A1 (en) * | 2019-09-02 | 2021-03-11 | 平安科技（深圳）有限公司 | Method and device for audio processing, and computer storage medium |
| CN110751960B (en) * | 2019-10-16 | 2022-04-26 | 北京网众共创科技有限公司 | Method and device for determining noise data |
| CN110751960A (en) * | 2019-10-16 | 2020-02-04 | 北京网众共创科技有限公司 | Method and device for determining noise data |
| CN110797041A (en) * | 2019-10-21 | 2020-02-14 | 珠海市杰理科技股份有限公司 | Voice noise reduction processing method and device, computer equipment and storage medium |
| CN110795892B (en) * | 2019-10-23 | 2021-10-01 | 北京邮电大学 | Channel simulation method and device based on generation countermeasure network |
| CN110795892A (en) * | 2019-10-23 | 2020-02-14 | 北京邮电大学 | Channel simulation method and device based on generation countermeasure network |
| CN113066476A (en) * | 2019-12-13 | 2021-07-02 | 科大讯飞股份有限公司 | Synthetic speech processing method and related device |
| CN111128230B (en) * | 2019-12-31 | 2022-03-04 | 广州市百果园信息技术有限公司 | Voice signal reconstruction method, device, equipment and storage medium |
| CN111128230A (en) * | 2019-12-31 | 2020-05-08 | 广州市百果园信息技术有限公司 | Voice signal reconstruction method, device, equipment and storage medium |
| CN111341304A (en) * | 2020-02-28 | 2020-06-26 | 广州国音智能科技有限公司 | Method, device and equipment for training speech characteristics of speaker based on GAN |
| CN111370017A (en) * | 2020-03-18 | 2020-07-03 | 苏宁云计算有限公司 | Voice enhancement method, device and system |
| CN111370017B (en) * | 2020-03-18 | 2023-04-14 | 苏宁云计算有限公司 | Voice enhancement method, device and system |
| CN113539290B (en) * | 2020-04-22 | 2024-04-12 | 华为技术有限公司 | Voice noise reduction method and device |
| CN113539290A (en) * | 2020-04-22 | 2021-10-22 | 华为技术有限公司 | Voice noise reduction method and device |
| CN111933171B (en) * | 2020-09-21 | 2021-01-22 | 北京达佳互联信息技术有限公司 | Noise reduction method and device, electronic equipment and storage medium |
| CN111933171A (en) * | 2020-09-21 | 2020-11-13 | 北京达佳互联信息技术有限公司 | Noise reduction method and device, electronic equipment and storage medium |
| CN112053421A (en) * | 2020-10-14 | 2020-12-08 | 腾讯科技（深圳）有限公司 | Signal noise reduction processing method, device, equipment and storage medium |
| CN112053421B (en) * | 2020-10-14 | 2023-06-23 | 腾讯科技（深圳）有限公司 | Signal noise reduction processing method, device, equipment and storage medium |
| CN112382284A (en) * | 2020-10-22 | 2021-02-19 | 珠海格力电器股份有限公司 | Pickup noise reduction method and device and server |
| CN113823312A (en) * | 2021-02-19 | 2021-12-21 | 北京沃东天骏信息技术有限公司 | Speech enhancement model generation method and device and speech enhancement method and device |
| CN113823312B (en) * | 2021-02-19 | 2023-11-07 | 北京沃东天骏信息技术有限公司 | Speech enhancement model generation method and device, and speech enhancement method and device |
| CN113098901A (en) * | 2021-04-29 | 2021-07-09 | 心动互动娱乐有限公司 | High-concurrency real-time voice call method, system, device and medium |
| CN114007157A (en) * | 2021-10-28 | 2022-02-01 | 中北大学 | Intelligent noise reduction communication earphone |
| CN114333882B (en) * | 2022-03-09 | 2022-08-19 | 深圳市友杰智新科技有限公司 | Voice noise reduction method, device and equipment based on amplitude spectrum and storage medium |
| CN114333882A (en) * | 2022-03-09 | 2022-04-12 | 深圳市友杰智新科技有限公司 | Voice noise reduction method, device and equipment based on amplitude spectrum and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2020088154A1 (en) | 2020-05-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109119093A (en) | Voice de-noising method, device, storage medium and mobile terminal | |
| JP6393730B2 (en) | Voice identification method and apparatus | |
| CN109119090A (en) | Method of speech processing, device, storage medium and electronic equipment | |
| JP6099556B2 (en) | Voice identification method and apparatus | |
| CN110457457B (en) | Training method of dialogue generation model, dialogue generation method and device | |
| CN109903773B (en) | Audio processing method, device and storage medium | |
| WO2021139327A1 (en) | Audio signal processing method, model training method, and related apparatus | |
| CN111554321B (en) | Noise reduction model training method and device, electronic equipment and storage medium | |
| CN107995370B (en) | Call control method, device, storage medium and mobile terminal | |
| CN110364156A (en) | Voice interactive method, system, terminal and readable storage medium storing program for executing | |
| CN108449493A (en) | Voice communication data processing method, device, storage medium and mobile terminal | |
| CN110505332A (en) | A kind of noise-reduction method, device, mobile terminal and storage medium | |
| CN106791067A (en) | A kind of In Call adjusting method, device and mobile terminal | |
| CN108449506A (en) | Voice communication data processing method, device, storage medium and mobile terminal | |
| CN108172237A (en) | Voice communication data processing method, device, storage medium and mobile terminal | |
| CN110931035B (en) | Audio processing method, device, equipment and storage medium | |
| CN108449497A (en) | Voice communication data processing method, device, storage medium and mobile terminal | |
| CN108449495A (en) | Voice communication data processing method, device, storage medium and mobile terminal | |
| CN109119097A (en) | Fundamental tone detecting method, device, storage medium and mobile terminal | |
| CN108418982A (en) | Voice communication data processing method, device, storage medium and mobile terminal | |
| CN113643706B (en) | Speech recognition method, device, electronic equipment and storage medium | |
| CN113808566B (en) | Vibration noise processing method and device, electronic equipment and storage medium | |
| CN108960213A (en) | Method for tracking target, device, storage medium and terminal | |
| CN115273909A (en) | Voice activity detection method, device, equipment and computer readable storage medium | |
| CN117059068A (en) | Speech processing method, device, storage medium and computer equipment |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190101 |
|
| RJ01 | Rejection of invention patent application after publication |