CN103440870A - Method and device for voice frequency noise reduction - Google Patents
Method and device for voice frequency noise reduction Download PDFInfo
- Publication number
- CN103440870A CN103440870A CN201310359719XA CN201310359719A CN103440870A CN 103440870 A CN103440870 A CN 103440870A CN 201310359719X A CN201310359719X A CN 201310359719XA CN 201310359719 A CN201310359719 A CN 201310359719A CN 103440870 A CN103440870 A CN 103440870A
- Authority
- CN
- China
- Prior art keywords
- noise
- audio signal
- segment
- frequency spectrum
- noise reduction
- 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
Images
Abstract
The invention provides a method and device for voice frequency noise reduction. The method particularly comprises the steps that a voice frequency signal segment with the minimum energy is searched in a voice frequency signal needing noise reduction and serves as a noise segment; spectral analysis is conducted on the noise segment so that the characteristics of a frequency domain can be obtained and serve as noise characteristics; noise reduction processing is conducted on the voice frequency signal needing noise reduction according to the noise characteristics. The method and device for voice frequency noise reduction can improve the convenience of voice frequency noise reduction.
Description
Technical field
The application relates to the Audio Signal Processing technical field, particularly relates to a kind of audio defeat method and device.
Background technology
The audio defeat technology, refer to and utilize signal to process and the method for pattern-recognition, and from contain noisy audio frequency, by noise remove, the audio frequency that makes to remove after noise has higher signal to noise ratio (S/N ratio) and preferable quality.Audio defeat is one of Audio Signal Processing field gordian technique that need to solve.
A kind of existing typical audio defeat method, for after choosing the noise segment data, learn corresponding noise characteristic according to selected noise segment data the people, and according to described noise characteristic, corresponding sound signal carried out to noise reduction process.Choose the noise segment in sound signal owing to usually utilizing at present artificial pre-mode of listening, often need artificial multi-pass operations could realize choosing of noise segment, therefore, the convenience of existing audio defeat method is poor, and does not meet the succinct interaction characteristic of mobile device.
Summary of the invention
The application's technical matters to be solved is to provide a kind of audio defeat method and device, can improve the convenience of audio defeat.
In order to address the above problem, the application discloses a kind of audio defeat method, comprising:
The audio signal segment of search energy minimum in treating noise reduction audio signal, using it as noise segment;
Described noise segment is carried out to spectrum analysis and obtain its frequency domain character, as noise characteristic;
Treat that to described noise reduction audio signal carries out noise reduction process according to described noise characteristic.
Preferably, described audio signal segment of searching for the energy minimum in treating noise reduction audio signal, the step using it as noise segment comprises:
All or part of treat noise reduction audio signal in the audio signal segment of search energy minimum, as noise segment.
Preferably, described audio signal segment of searching for the energy minimum in part is treated noise reduction audio signal, the step as noise segment comprises:
Travel through the described initial part for the treatment of noise reduction audio signal, and the energy of each audio signal segment in more described initial part; Wherein, the time domain of described each audio signal segment is equal in length;
Choose an audio signal segment of energy minimum from all audio signal segment, as noise segment.
Preferably, described noise segment is carried out to spectrum analysis and obtain its frequency domain character, the step as noise characteristic comprises:
Described noise segment is divided into to a plurality of equal-sized first windows on time domain;
Described each first window is carried out to the time-frequency conversion process;
According to the frequency domain data of described each first window, determine the frequency spectrum maximal value of all first windows on each frequency of described noise segment;
Frequency spectrum maximal value on described each frequency is recorded as to noise characteristic corresponding to described noise segment.
Preferably, the described noise characteristic of described foundation treats that to described noise reduction audio signal carries out the step of noise reduction process, comprising:
Treat that by described noise reduction audio signal and described noise characteristic carry out Frequency spectrum ratio;
Treat that to described noise reduction audio signal carries out the spectrum gain processing according to the frequency spectrum comparative result.
Preferably, describedly according to the frequency spectrum comparative result, to described, treat that noise reduction audio signal carries out the step of spectrum gain processing and comprise:
Determine and lower parameter according to described frequency spectrum comparative result, make described downward parameter reduce along with the increase of described frequency spectrum comparative result;
Treat that to described noise reduction audio signal carries out frequency spectrum and lower to process according to described downward parameter.
Preferably, the described frequency spectrum comparative result of described foundation is determined the downward parameter, and the step that makes described downward parameter reduce along with the increase of described frequency spectrum comparative result comprises:
According to size order, the frequency spectrum comparative result is divided into to several stages, wherein between the corresponding size order of lowering parameter of stage and stage, the size order of frequency spectrum comparative result is contrary.
On the other hand, disclosed herein as well is a kind of audio defeat device, comprising:
The noise search module, for the audio signal segment treating noise reduction audio signal search energy minimum, using it as noise segment;
The noise characteristic study module, obtain its frequency domain character for described noise segment is carried out to spectrum analysis, as noise characteristic; And
The noise reduction process module, for treating that to described noise reduction audio signal carries out noise reduction process according to described noise characteristic.
Preferably, described noise search module, specifically for all or part of treat noise reduction audio signal in the audio signal segment of search energy minimum, as noise segment.
Preferably, described noise search module comprises:
The traversal comparison sub-module, for traveling through the described initial part for the treatment of noise reduction audio signal, and the energy of each audio signal segment in more described initial part; Wherein, the time domain of described each audio signal segment is equal in length; And
Choose submodule, for choose an audio signal segment of energy minimum from all audio signal segment, as noise segment.
Preferably, described noise characteristic study module comprises:
Window is divided submodule, for described noise segment being divided on time domain to a plurality of equal-sized first windows;
Time-frequency conversion process submodule, for carrying out the time-frequency conversion process to described each first window;
Maximal value is determined submodule, for the frequency domain data according to described each first window, determines the frequency spectrum maximal value of all first windows on each frequency of described noise segment; And
Record sub module, be recorded as noise characteristic corresponding to described noise segment for the frequency spectrum maximal value by described each frequency.
Preferably, described noise reduction process module comprises:
The frequency spectrum comparison sub-module, for treating that by described noise reduction audio signal and described noise characteristic carry out Frequency spectrum ratio; And
Spectrum gain is processed submodule, for according to the frequency spectrum comparative result, to described, treating that noise reduction audio signal carries out the spectrum gain processing.
Preferably, described spectrum gain processing submodule comprises:
The parameter determining unit, for according to described frequency spectrum comparative result, determining and lower parameter, make described downward parameter reduce along with the increase of described frequency spectrum comparative result; And
Frequency spectrum is lowered processing unit, for according to described downward parameter, to described, treating that noise reduction audio signal carries out frequency spectrum and lower and process.
Preferably, described parameter determining unit comprises:
Stage is divided subelement, and for according to size order, the frequency spectrum comparative result being divided into to several stages, wherein between the corresponding size order of lowering parameter of stage and stage, the size order of frequency spectrum comparative result is contrary.
Compared with prior art, the application has the following advantages:
At first, choose the noise segment in sound signal with respect to the artificial pre-mode of listening of prior art utilization, the audio signal segment of the application's automatic search energy minimum in treating noise reduction audio signal, using it as noise segment, described automatic search is without artificial participation, therefore can save the artificial multi-pass operations that prior art spends, thereby can improve the convenience of audio defeat.
Secondly, the application, when being applied to mobile device, only needs the user to operate and trigger corresponding audio defeat flow process by one-touch, therefore, the application can be on mobile device the neighbourhood noise in one-touch filtering sound signal, can meet the succinct interaction characteristic of mobile device.
The accompanying drawing explanation
Fig. 1 is the process flow diagram of a kind of audio defeat embodiment of the method for the application;
Fig. 2 is the process flow diagram of the audio defeat embodiment of the method for a kind of movement-based equipment of the application;
Fig. 3 is the example of a kind of Second Window of the application;
Fig. 4 is the structural drawing of a kind of audio defeat device of the application embodiment.
Embodiment
For above-mentioned purpose, the feature and advantage that make the application can become apparent more, below in conjunction with the drawings and specific embodiments, the application is described in further detail.
With reference to Fig. 1, show the process flow diagram of a kind of audio defeat embodiment of the method for the application, specifically can comprise:
The embodiment of the present application can be applied on the equipment such as personal computer (PC, Personal Computer), mobile device, for carry out the audio defeat processing on various device.
The embodiment of the present application can be carried out noise reduction process to the sound signal of various forms, as Advanced Audio Coding (AAC, Advanced Audio Coding), dynamic image expert compression standard audio frequency aspect 3(MP3, Moving Picture Experts Group Audio Layer III) and WAVE form etc.; Wherein, the sound signal of noise reduction process can be the sound signal in simple audio file, can be also the sound signal in video file.In a word, the embodiment of the present application is not limited form and the document form of concrete sound signal.
Which kind of in addition, no matter be applied on equipment, with operating the audio defeat flow process that triggers the application by one-touch per family.
Existing audio defeat method is to utilize artificial pre-mode of listening to choose noise segment in sound signal, needs artificial multi-pass operations, and convenience is poor, and does not meet the succinct interaction characteristic of mobile device.
And the embodiment of the present application is found after deliberation: owing to not comprising main flow acoustical signal (as people's one's voice in speech etc.), the energy of pure neighbourhood noise is usually smaller, and also, energy is less, and the degree of approximation of corresponding signal audio signal segment and pure neighbourhood noise is higher; Therefore can think, in sound signal, the audio signal segment of energy minimum is exactly pure neighbourhood noise audio signal segment (following general designation noise segment), and this is also the characteristic of pure environmental noise power minimum.
Correspondingly, the characteristic of the pure environmental noise power minimum of described foundation, in treating noise reduction audio signal, the step 101 of search noise segment specifically can comprise: the audio signal segment of search energy minimum in treating noise reduction audio signal, using it as noise segment.
In specific implementation, can all or part of treat noise reduction audio signal in the audio signal segment of search energy minimum, as noise segment.Wherein, in all treating noise reduction audio signal, the search noise segment can effectively guarantee the search precision of noise segment, but has the shortcoming that operand is large; In part is treated noise reduction audio signal, the search noise segment has advantages of that operand is little, but can not effectively guarantee the search precision of noise segment.
Find after deliberation, at the initial part of sound signal, the likelihood ratio that the main flow acoustical signal occurs is lower, and the likelihood ratio that pure neighbourhood noise occurs is higher.
Therefore, in order effectively to guarantee the search precision of noise segment, in a kind of preferred embodiment of application, described audio signal segment of searching for the energy minimum in part is treated noise reduction audio signal, the step as noise segment may further include:
Sub-step S111, travel through the described initial part for the treatment of noise reduction audio signal, and the energy of each audio signal segment in more described initial part; Wherein, the time domain of described each audio signal segment is equal in length;
In specific implementation, those skilled in the art can select described front N second for the treatment of noise reduction audio signal as described initial part according to actual conditions, and wherein, N is natural number, for example, and N=10 etc.In addition, those skilled in the art also can determine the time length of field of described each audio signal segment according to actual conditions, and for example, in a kind of application example of the application, the time length of field of each audio signal segment is 2 seconds etc.In a word, the length of described initial part and wherein each signal segment the time length of field not as the application restric-tion of the embodiment of the present application.
Sub-step S112, choose an audio signal segment of energy minimum from all audio signal segment, as noise segment.
In the embodiment of the present application, it is also the process of noise characteristic study that described noise segment is carried out to the process that spectrum analysis obtains its frequency domain character, it specifically can comprise: by Fast Fourier Transform (FFT) (FFT, Fast Fourier Transformation) etc. the time-frequency conversion process is converted to frequency domain data by the noise segment data, and records its frequecy characteristic.
In a preferred embodiment of the present application, describedly described noise segment is carried out to spectrum analysis obtain its frequency domain character, as the step 102 of noise characteristic, specifically can comprise:
Sub-step S121, described noise segment is divided into to a plurality of equal-sized first windows on time domain;
The equal and opposite in direction here refers to that the byte number of each first window equates.For example, the noise segment of 2S can be divided into and take several first windows that 1024 bytes are base unit, type is INT or FLOAT.Be appreciated that above-mentioned is that the embodiment of the present application is not limited number, byte length and the type of concrete first window as example.
Sub-step S122, described each first window is carried out to the time-frequency conversion process;
Formula (1) is a kind of example of carrying out the time-frequency conversion process by FFT of the application:
Wherein, i=1,2...m, m means the number of first window in noise segment, and xi (n) means the time domain data of i first window, and N means count (one is exemplified as 1024) of FFT, k=0,1 ... N-1,
Sub-step S123, according to the frequency domain data of described each first window, determine the frequency spectrum maximal value of all first windows on each frequency of described noise segment;
Formula (2) is a kind of peaked example of the frequency spectrum of all first windows on each frequency of determining described noise segment of the application:
R(k)=max(X1(k),X2(k),....Xm(k)) （2）
Sub-step S124, the frequency spectrum maximal value on described each frequency is recorded as to noise characteristic corresponding to described noise segment.
Due to the noise segment audio signal segment that is energy minimum in sound signal, frequency spectrum maximal value in noise segment on each frequency more can represent pure neighbourhood noise, therefore the frequency spectrum maximal value in above-mentioned sub-step S21-S24 study noise segment on each frequency, as noise characteristic, can access and have more representational noise characteristic.
Be appreciated that, above-mentioned sub-step S21-S24 is the preferred embodiment as the application's noise characteristic learning method, in fact, other noise characteristic learning method (for example, it not the frequency spectrum maximal value recorded on each frequency, but record other frequecy characteristic) be also feasible, the application is not limited concrete noise characteristic learning method.
In specific implementation, can adopt various algorithms to carry out noise reduction process.For example, can be according to described noise characteristic structure low-pass filter, and adopt the low-pass filter of constructing to treat that to described noise reduction audio signal carries out noise reduction process etc.Intermediate frequency in low-pass filter elimination sound signal simply and high frequency (middle pitch and high pitch) composition, however the main flow acoustical signal in a lot of sound signal may comprise low frequency, intermediate frequency and radio-frequency component simultaneously; So the characteristic of low-pass filter easily causes obtaining inaccurate filter effect.In like manner, the characteristic of Hi-pass filter and bandpass filter also easily causes obtaining inaccurate filter effect.
In a preferred embodiment of the present application, the described noise characteristic of described foundation treats that to described noise reduction audio signal carries out noise reduction process step 103, specifically can comprise:
Sub-step S131, by described, treat that noise reduction audio signal and described noise characteristic carry out Frequency spectrum ratio;
In specific implementation, can treat that noise reduction audio signal is divided into a plurality of equal-sized Second Windows on time domain by described, described each Second Window is carried out to the time-frequency conversion process, and frequency domain data and the described noise characteristic of described each Second Window carried out to Frequency spectrum ratio.
Suppose describedly to treat that noise reduction audio signal includes the data of 4096 bytes, can treat that noise reduction audio signal is divided into 4 Second Windows by described so, wherein each Second Window includes the data of 1024 bytes.
It should be noted that, first window and Second Window in the embodiment of the present application are time-domain window, and the two statement is not both for its residing processing links more clearly is described, in fact, the two is the time-domain window that essence is identical.
Sub-step S132, according to the frequency spectrum comparative result, to described, treat that noise reduction audio signal carries out the spectrum gain processing.
In a preferred embodiment of the present application, describedly according to the frequency spectrum comparative result, to described, treat that noise reduction audio signal carries out the sub-step S132 of spectrum gain processing, specifically can comprise:
Sub-step S1321, the described frequency spectrum comparative result of foundation are determined the downward parameter, make described downward parameter reduce along with the increase of described frequency spectrum comparative result;
In this preferred embodiment, energy that can be based on pure neighbourhood noise is smaller characteristic usually, according to the frequency spectrum comparative result, carries out the spectrum gain processing, and corresponding principle is specifically as follows:
Treat that for described for noise reduction audio signal, its frequency spectrum more approaches the frequency spectrum of noise segment, the degree that in the spectrum gain processing procedure, its frequency spectrum is lowered is larger; Because the energy of pure neighbourhood noise is usually smaller, therefore in such cases, can think described and treat that noise reduction audio signal may only comprise pure neighbourhood noise, the likelihood ratio of main flow acoustical signal that also comprises other is lower, so noise reduction intensity is more intense;
Otherwise its frequency spectrum is more away from the frequency spectrum of noise segment, the degree that in the spectrum gain processing procedure, its frequency spectrum is lowered is less; Because the energy of pure neighbourhood noise is usually smaller, therefore in such cases, can think described and treat noise reduction audio signal except comprising pure neighbourhood noise, also may comprise other main flow acoustical signal, thus the noise reduction strength ratio a little less than.
Therefore, when implementing sub-step S1321 according to the definite downward of described frequency spectrum comparative result parameter, can determine that lowering parameter reduces its increase along with the frequency spectrum comparative result according to actual conditions.For example, can determine that described downward parameter and described frequency spectrum comparative result are inversely proportional to according to actual conditions.And for example, also can determine stage type downward parameter according to actual conditions, also, according to size order, the frequency spectrum comparative result is divided into to several stages, wherein the size order of corresponding downward of stage parameter is contrary with the size order of frequency spectrum comparative result; For example, counting of FFT, it is 1024 o'clock, the number of frequency spectrum comparative result is 1024, suppose, according to order from small to large, these 1024 frequency spectrum comparative results are divided into to 4 stages: stage 1, stage 2, stage 3, stage 4,4 corresponding size orders of lowering parameter of stage are contrary with the size order of frequency spectrum comparative result, are also the corresponding corresponding corresponding corresponding parameter of lowering of parameter>stage 4 of lowering of parameter>stage 3 of lowering of parameter>stage 2 of lowering of stage 1.
For example, in a preferred embodiment of the present application, the frequency spectrum comparative result can be divided into to 2 stages, particularly, when the described frequency spectrum until noise reduction audio signal is less than or equal to the frequency spectrum of described noise characteristic, described downward parameter is gain; When the described frequency spectrum until noise reduction audio signal is greater than the frequency spectrum of described noise characteristic, described downward parameter is gain* ((X (k) – R (k))/X (k); Wherein, 0<gain<1, X (K) means the described frequency spectrum for the treatment of noise reduction audio signal, R (k) means the frequency spectrum of described noise characteristic.
Sub-step S1322, according to described downward parameter, to described, treat that noise reduction audio signal carries out frequency spectrum and lower to process.
In a kind of application example of the application, described frequency spectrum is lowered to process and is specifically as follows, at first the frequency spectrum of described noise characteristic and described downward parameter are carried out to product calculation, then, treat that by described the frequency spectrum of noise reduction audio signal and product result of calculation carries out additive operation, corresponding additive operation result is exactly that frequency spectrum is lowered result.
It should be noted that, in the embodiment of the present application, spectrum gain is processed or frequency spectrum lowers that to process the result obtained be frequency domain data, and while also needing to be undertaken frequently by corresponding frequency domain data, conversion process is to obtain the sound signal after final noise reduction process.
Due to the spectrum gain of this preferred embodiment, processing is to carry out frequency spectrum downward processing according to lowering parameter, and described downward parameter reduces along with the increase of described frequency spectrum comparative result, therefore this preferred embodiment can carry out the noise reduction audio signal for the treatment of that approaches the frequency spectrum of noise segment the frequency spectrum of stronger degree and lower processing, and, will carry out the frequency spectrum of weak degree and lower and process away from the noise reduction audio signal for the treatment of of the frequency spectrum of noise segment; Therefore, with respect to the frequency content in the elimination sound signals simply such as low-pass filter, Hi-pass filter and bandpass filter, this preferred embodiment can be based on pure neighbourhood noise energy smaller characteristic usually, obtain filter effect comparatively accurately, and can access the filtering result that simultaneously comprises low frequency, intermediate frequency and radio-frequency component.
In a word, choose the noise segment in sound signal with respect to the artificial pre-mode of listening of prior art utilization, the audio signal segment of the application's automatic search energy minimum in treating noise reduction audio signal, using it as noise segment, described automatic search is without artificial participation, therefore can save the artificial multi-pass operations that prior art spends, thereby can improve the convenience of audio defeat.
In addition, the application, when being applied to mobile device, only needs the user to operate and trigger corresponding audio defeat flow process by one-touch, therefore, the application can be on mobile device the neighbourhood noise in one-touch filtering sound signal, can meet the succinct interaction characteristic of mobile device.
For making those skilled in the art understand better the application, below provide the example that the application is applied to mobile device.
The application scenarios of this example:
The user finds that the acoustical signal of main flow in this video is covered by environmental noise or disturbs after having taken one section video with mobile phone in a relatively noisy environment, wants to obtain the low noise video that can at once hear on mobile phone; So operate the audio defeat flow process that triggers the application by one-touch, corresponding audio defeat flow process specifically can comprise:
The frequency spectrum unified representation of supposing each Second Window after spectrum gain is processed is Y (k),, when the described frequency spectrum until noise reduction audio signal is greater than the frequency spectrum of described noise characteristic, can obtain Y (k) by formula (3):
Y(k)=X(k)-R(x)*gain*((X(k)–R(k))/X(k) （3）
When the described frequency spectrum until noise reduction audio signal is less than or equal to the frequency spectrum of described noise characteristic, can obtain Y (k) by formula (4):
Y(k)=X(k)-R(x)*gain （4）
The conversion process when frequency spectrum of step 209, each Second Window after spectrum gain is processed carries out frequency;
Show the example of a kind of Second Window of the application with reference to Fig. 3, can find out, after step 205 obtains Second Window, can be by the data buffer storage of all Second Windows, carry out step 206-noise reduction process of step 209 according to the data of each Second Window successively, and export successively the noise reduction process result of each Second Window in step 210.
With preceding method, embodiment is corresponding, and the application also provides a kind of audio defeat device, with reference to the structural drawing shown in Fig. 4, specifically can comprise:
Noise characteristic study module 402, obtain its frequency domain character for described noise segment is carried out to spectrum analysis, as noise characteristic; And
Noise reduction process module 403, for treating that to described noise reduction audio signal carries out noise reduction process according to described noise characteristic.
In a preferred embodiment of the present application, described noise search module 401, can specifically for all or part of treat noise reduction audio signal in the audio signal segment of search energy minimum, as noise segment.
In another preferred embodiment of the present application, described noise search module 401 specifically can comprise:
The traversal comparison sub-module, for traveling through the described initial part for the treatment of noise reduction audio signal, and the energy of each audio signal segment in more described initial part; Wherein, the time domain of described each audio signal segment is equal in length; And
Choose submodule, for choose an audio signal segment of energy minimum from all audio signal segment, as noise segment.
In another preferred embodiment of the application, described noise characteristic study module 402 specifically can comprise:
Window is divided submodule, for described noise segment being divided on time domain to a plurality of equal-sized first windows;
Time-frequency conversion process submodule, for carrying out the time-frequency conversion process to described each first window;
Maximal value is determined submodule, for the frequency domain data according to described each first window, determines the frequency spectrum maximal value of all first windows on each frequency of described noise segment; And
Record sub module, be recorded as noise characteristic corresponding to described noise segment for the frequency spectrum maximal value by described each frequency.
In a preferred embodiment of the present application, described noise reduction process module 403 specifically can comprise:
The frequency spectrum comparison sub-module, for treating that by described noise reduction audio signal and described noise characteristic carry out Frequency spectrum ratio; And
Spectrum gain is processed submodule, for according to the frequency spectrum comparative result, to described, treating that noise reduction audio signal carries out the spectrum gain processing.
In the embodiment of the present application, preferably, described spectrum gain is processed submodule and specifically can be comprised:
The parameter determining unit, for according to described frequency spectrum comparative result, determining and lower parameter, make described downward parameter reduce along with the increase of described frequency spectrum comparative result; And
Frequency spectrum is lowered processing unit, for according to described downward parameter, to described, treating that noise reduction audio signal carries out frequency spectrum and lower and process.
In the embodiment of the present application, preferably, described parameter determining unit may further include: the stage is divided subelement, for according to size order, the frequency spectrum comparative result being divided into to several stages, wherein between the corresponding size order of lowering parameter of stage and stage, the size order of frequency spectrum comparative result is contrary.
Each embodiment in this instructions all adopts the mode of going forward one by one to describe, and what each embodiment stressed is and the difference of other embodiment that between each embodiment, identical similar part is mutually referring to getting final product.For device embodiment, because it is substantially similar to embodiment of the method, so description is fairly simple, relevant part gets final product referring to the part explanation of embodiment of the method.
Above a kind of audio defeat method and the device that the application is provided, be described in detail, applied specific case herein the application's principle and embodiment are set forth, the explanation of above embodiment is just for helping to understand the application's method and core concept thereof; Simultaneously, for one of ordinary skill in the art, the thought according to the application, all will change in specific embodiments and applications, and in sum, this description should not be construed as the restriction to the application.
Claims (14)
1. an audio defeat method, is characterized in that, comprising:
The audio signal segment of search energy minimum in treating noise reduction audio signal, using it as noise segment;
Described noise segment is carried out to spectrum analysis and obtain its frequency domain character, as noise characteristic;
Treat that to described noise reduction audio signal carries out noise reduction process according to described noise characteristic.
2. the method for claim 1, is characterized in that, described audio signal segment of searching for the energy minimum in treating noise reduction audio signal, and the step using it as noise segment comprises:
All or part of treat noise reduction audio signal in the audio signal segment of search energy minimum, as noise segment.
3. method as claimed in claim 2, is characterized in that, described audio signal segment of searching for the energy minimum in part is treated noise reduction audio signal, and the step as noise segment comprises:
Travel through the described initial part for the treatment of noise reduction audio signal, and the energy of each audio signal segment in more described initial part; Wherein, the time domain of described each audio signal segment is equal in length;
Choose an audio signal segment of energy minimum from all audio signal segment, as noise segment.
4. as claim 1 or 2 or 3 described methods, it is characterized in that, described noise segment is carried out to spectrum analysis and obtain its frequency domain character, the step as noise characteristic comprises:
Described noise segment is divided into to a plurality of equal-sized first windows on time domain;
Described each first window is carried out to the time-frequency conversion process;
According to the frequency domain data of described each first window, determine the frequency spectrum maximal value of all first windows on each frequency of described noise segment;
Frequency spectrum maximal value on described each frequency is recorded as to noise characteristic corresponding to described noise segment.
5. as claim 1 or 2 or 3 described methods, it is characterized in that, the described noise characteristic of described foundation treats that to described noise reduction audio signal carries out the step of noise reduction process, comprising:
Treat that by described noise reduction audio signal and described noise characteristic carry out Frequency spectrum ratio;
Treat that to described noise reduction audio signal carries out the spectrum gain processing according to the frequency spectrum comparative result.
6. method as claimed in claim 5, is characterized in that, describedly according to the frequency spectrum comparative result, to described, treats that noise reduction audio signal carries out the step of spectrum gain processing and comprise:
Determine and lower parameter according to described frequency spectrum comparative result, make described downward parameter reduce along with the increase of described frequency spectrum comparative result;
Treat that to described noise reduction audio signal carries out frequency spectrum and lower to process according to described downward parameter.
7. method as claimed in claim 6, is characterized in that, the described frequency spectrum comparative result of described foundation is determined the downward parameter, and the step that makes described downward parameter reduce along with the increase of described frequency spectrum comparative result comprises:
According to size order, the frequency spectrum comparative result is divided into to several stages, wherein between the corresponding size order of lowering parameter of stage and stage, the size order of frequency spectrum comparative result is contrary.
8. an audio defeat device, is characterized in that, comprising:
The noise search module, for the audio signal segment treating noise reduction audio signal search energy minimum, using it as noise segment;
The noise characteristic study module, obtain its frequency domain character for described noise segment is carried out to spectrum analysis, as noise characteristic; And
The noise reduction process module, for treating that to described noise reduction audio signal carries out noise reduction process according to described noise characteristic.
9. device as claimed in claim 8, is characterized in that, described noise search module, specifically for all or part of treat noise reduction audio signal in the audio signal segment of search energy minimum, as noise segment.
10. device as claimed in claim 9, is characterized in that, described noise search module comprises:
The traversal comparison sub-module, for traveling through the described initial part for the treatment of noise reduction audio signal, and the energy of each audio signal segment in more described initial part; Wherein, the time domain of described each audio signal segment is equal in length; And
Choose submodule, for choose an audio signal segment of energy minimum from all audio signal segment, as noise segment.
11. as claim 8 or 9 or 10 described devices, it is characterized in that, described noise characteristic study module comprises:
Window is divided submodule, for described noise segment being divided on time domain to a plurality of equal-sized first windows;
Time-frequency conversion process submodule, for carrying out the time-frequency conversion process to described each first window;
Maximal value is determined submodule, for the frequency domain data according to described each first window, determines the frequency spectrum maximal value of all first windows on each frequency of described noise segment; And
Record sub module, be recorded as noise characteristic corresponding to described noise segment for the frequency spectrum maximal value by described each frequency.
12. as claim 8 or 9 or 10 described devices, it is characterized in that, described noise reduction process module comprises:
The frequency spectrum comparison sub-module, for treating that by described noise reduction audio signal and described noise characteristic carry out Frequency spectrum ratio; And
Spectrum gain is processed submodule, for according to the frequency spectrum comparative result, to described, treating that noise reduction audio signal carries out the spectrum gain processing.
13. device as claimed in claim 12, is characterized in that, described spectrum gain is processed submodule and is comprised:
The parameter determining unit, for according to described frequency spectrum comparative result, determining and lower parameter, make described downward parameter reduce along with the increase of described frequency spectrum comparative result; And
Frequency spectrum is lowered processing unit, for according to described downward parameter, to described, treating that noise reduction audio signal carries out frequency spectrum and lower and process.
14. device as claimed in claim 13, is characterized in that, described parameter determining unit comprises:
Stage is divided subelement, and for according to size order, the frequency spectrum comparative result being divided into to several stages, wherein between the corresponding size order of lowering parameter of stage and stage, the size order of frequency spectrum comparative result is contrary.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310359719XA CN103440870A (en) | 2013-08-16 | 2013-08-16 | Method and device for voice frequency noise reduction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310359719XA CN103440870A (en) | 2013-08-16 | 2013-08-16 | Method and device for voice frequency noise reduction |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103440870A true CN103440870A (en) | 2013-12-11 |
Family
ID=49694561
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310359719XA Pending CN103440870A (en) | 2013-08-16 | 2013-08-16 | Method and device for voice frequency noise reduction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103440870A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105791530A (en) * | 2014-12-26 | 2016-07-20 | 联芯科技有限公司 | Output volume adjusting method and device |
| CN105929261A (en) * | 2016-04-06 | 2016-09-07 | 北京瑞森新谱科技有限公司 | Efficient audio signal-to-noise ratio measuring method |
| CN106384597A (en) * | 2016-08-31 | 2017-02-08 | 广州市百果园网络科技有限公司 | Audio frequency data processing method and device |
| CN107886964A (en) * | 2017-09-25 | 2018-04-06 | 惠州市德赛西威汽车电子股份有限公司 | A kind of audio-frequency processing method and its system |
| CN112383879A (en) * | 2020-11-17 | 2021-02-19 | 蔡如青 | Mobile phone application alarm system and method |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1397929A (en) * | 2002-07-12 | 2003-02-19 | 清华大学 | Speech intensifying-characteristic weighing-logrithmic spectrum addition method for anti-noise speech recognization |
| CN1787079A (en) * | 2005-12-19 | 2006-06-14 | 北京中星微电子有限公司 | Apparatus and method for detecting moise |
| CN101111888A (en) * | 2005-02-02 | 2008-01-23 | 富士通株式会社 | Signal processing method and device thereof |
| CN101477800A (en) * | 2008-12-31 | 2009-07-08 | 瑞声声学科技(深圳)有限公司 | Voice enhancing process |
| US20100067710A1 (en) * | 2008-09-15 | 2010-03-18 | Hendriks Richard C | Noise spectrum tracking in noisy acoustical signals |
-
2013
- 2013-08-16 CN CN201310359719XA patent/CN103440870A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1397929A (en) * | 2002-07-12 | 2003-02-19 | 清华大学 | Speech intensifying-characteristic weighing-logrithmic spectrum addition method for anti-noise speech recognization |
| CN101111888A (en) * | 2005-02-02 | 2008-01-23 | 富士通株式会社 | Signal processing method and device thereof |
| CN1787079A (en) * | 2005-12-19 | 2006-06-14 | 北京中星微电子有限公司 | Apparatus and method for detecting moise |
| US20100067710A1 (en) * | 2008-09-15 | 2010-03-18 | Hendriks Richard C | Noise spectrum tracking in noisy acoustical signals |
| CN101477800A (en) * | 2008-12-31 | 2009-07-08 | 瑞声声学科技(深圳)有限公司 | Voice enhancing process |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105791530A (en) * | 2014-12-26 | 2016-07-20 | 联芯科技有限公司 | Output volume adjusting method and device |
| CN105791530B (en) * | 2014-12-26 | 2019-04-16 | 联芯科技有限公司 | Output volume adjusting method and apparatus |
| CN105929261A (en) * | 2016-04-06 | 2016-09-07 | 北京瑞森新谱科技有限公司 | Efficient audio signal-to-noise ratio measuring method |
| CN106384597A (en) * | 2016-08-31 | 2017-02-08 | 广州市百果园网络科技有限公司 | Audio frequency data processing method and device |
| CN107886964A (en) * | 2017-09-25 | 2018-04-06 | 惠州市德赛西威汽车电子股份有限公司 | A kind of audio-frequency processing method and its system |
| CN112383879A (en) * | 2020-11-17 | 2021-02-19 | 蔡如青 | Mobile phone application alarm system and method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109065030B (en) | Convolutional neural network-based environmental sound identification method and system | |
| CN110197670B (en) | Audio noise reduction method and device and electronic equipment | |
| CN107393542B (en) | Bird species identification method based on two-channel neural network | |
| US20170256271A1 (en) | Systems and methods facilitating selective removal of content from a mixed audio recording | |
| CN109379311B (en) | Ultra-short wave specific signal identification method based on convolutional neural network | |
| CN103440870A (en) | Method and device for voice frequency noise reduction | |
| Gabor | Communication theory and cybernetics | |
| CN103456312B (en) | A kind of single-channel voice blind separating method based on Computational auditory scene analysis | |
| CN105788592A (en) | Audio classification method and apparatus thereof | |
| CN101366078A (en) | Neural network classifier for separating audio sources from a monophonic audio signal | |
| CN104053120A (en) | Method and device for processing stereo audio frequency | |
| CN108461081B (en) | Voice control method, device, equipment and storage medium | |
| CN113611324B (en) | Method and device for suppressing environmental noise in live broadcast, electronic equipment and storage medium | |
| Fan et al. | Image denoising based on wavelet thresholding and Wiener filtering in the wavelet domain | |
| CN110767248B (en) | Anti-modulation interference audio fingerprint extraction method | |
| CN112183582A (en) | Multi-feature fusion underwater target identification method | |
| CN108806725A (en) | Speech differentiation method, apparatus, computer equipment and storage medium | |
| CN105227763B (en) | A kind of instrumental audio real time method for segmenting realized on Intelligent mobile equipment | |
| CN116913258B (en) | Speech signal recognition method, device, electronic equipment and computer readable medium | |
| Won et al. | Automatic music tagging with harmonic cnn | |
| CN104715756A (en) | Audio data processing method and device | |
| al Balushi et al. | Wavelet based human voice identification system | |
| CN112863517A (en) | Speech recognition method based on perceptual spectrum convergence rate | |
| CN113763976A (en) | Method and device for reducing noise of audio signal, readable medium and electronic equipment | |
| Kumari et al. | Audio signal classification based on optimal wavelet and support vector machine |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20131211 |
|
| RJ01 | Rejection of invention patent application after publication |