CN106782587A - Sound mask device and sound mask method - Google Patents
Sound mask device and sound mask method Download PDFInfo
- Publication number
- CN106782587A CN106782587A CN201611029084.7A CN201611029084A CN106782587A CN 106782587 A CN106782587 A CN 106782587A CN 201611029084 A CN201611029084 A CN 201611029084A CN 106782587 A CN106782587 A CN 106782587A
- Authority
- CN
- China
- Prior art keywords
- frame
- main channel
- power
- channel signal
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 33
- 230000000873 masking effect Effects 0.000 claims abstract description 31
- 238000001514 detection method Methods 0.000 claims abstract description 18
- 238000009432 framing Methods 0.000 claims abstract description 12
- 238000001914 filtration Methods 0.000 claims description 112
- 230000008569 process Effects 0.000 claims description 13
- 238000005070 sampling Methods 0.000 claims description 13
- 230000008859 change Effects 0.000 claims description 8
- 230000014509 gene expression Effects 0.000 claims description 7
- 230000001143 conditioned effect Effects 0.000 claims description 6
- 230000000052 comparative effect Effects 0.000 claims description 4
- 230000007423 decrease Effects 0.000 claims description 4
- 230000004044 response Effects 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 2
- 230000007274 generation of a signal involved in cell-cell signaling Effects 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
- G10L21/0208—Noise filtering
Landscapes
- Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Quality & Reliability (AREA)
- Signal Processing (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Noise Elimination (AREA)
- Time-Division Multiplex Systems (AREA)
Abstract
The invention discloses a kind of sound mask device and sound mask method, the sound mask device includes：Receive the main channel of voice signal；Receive the subaisle of background signal；Framing unit；The bandpass filter being connected with the framing unit；The speech detection unit being connected with the bandpass filter；The time varing filter being connected with the speech detection unit；The masking unit being connected with the framing unit；The sum unit being connected with time varing filter and masking unit；The present invention shelters that result is good, and flexibility is high.
Description
Technical field
The present invention relates to a kind of sound mask device and sound mask method.
Background technology
Background signal such as music and dialogue (explanation) are usually had at video display recording, live match or live radio station
Situation about occurring simultaneously.Now, in order to protrude dialogue (explanation) sound, then background signal level must be reduced, enables dialogue
Shelter background signal；At the end of dialogue (explanation), background signal is then restored to original size.Sound mask device is exactly real
A kind of dynamic processor of existing above-mentioned functions, it is mainly shelters main channel signal automatically by additional signal, once additional letter
Number disappear, main channel signal amplitude restores automatically.
The Chinese utility model patent of Publication No. CN205004028U provides a kind of for directly preventing other people from understanding
Or the effective single channel sound mask device of admission conversation content, its basic inventive concept is：Deposited using speech selection switch selection
Voice interference signal in reservoir, and convert speech into voice signal using sound capturing unit；Signal transacting is divided into two
Unit, the first signal processing unit voice signal is split after random combine, generate the first masking signal；Secondary signal treatment is single
Voice interference signal random combine is generated the second masking signal by unit；Finally, audio player is by the first masking signal generation the
One shelters voice output, and the second masking signal generation second is sheltered into voice output.This kind of sound mask mode of the prior art is deposited
In following problem：1. the two-way masking signal for generating will not automatically change size with voice signal size variation, and flexibility is not
It is high；2. any treatment is not made to input speech signal, noise may have an impact to sheltering result.
The content of the invention
The present invention is directed to the proposition of problem above, and develops a kind of sound mask device and sound mask method, this kind of sound
Shelter device and sound mask method can avoid it is weaker in voice signal, and noise signal it is stronger when, use fixed judgement electricity
It is flat to cause the inaccurate technical problem of the court verdict of voice and noise, and can realize when a certain frame in main channel is voice
During frame, it should not have lossy whole to pass through, but then be filtered as far as possible when the frame is noise frame, so as to by noise
Influence is preferably minimized；This kind of sound mask device and sound mask method can be realized judging that a certain frame in main channel is language simultaneously
During sound frame, the subaisle signal power in respective frame quickly reduces with phonetic speech power but remains to be identified, when judging that voice terminates
When, the subaisle signal power in respective frame slowly increases, until the frame end.
Technological means of the invention is as follows：
A kind of sound mask device, including：
Receive the main channel of voice signal；
Receive the subaisle of background signal；
Framing unit, each frame main channel signal is obtained for the signal on the main channel to be carried out into sub-frame processing, and
Signal on the subaisle is carried out into sub-frame processing and obtains each frame subaisle signal；
The bandpass filter being connected with the framing unit；The bandpass filter is used to believe each frame main channel
Number carry out bandpass filtering；
The speech detection unit being connected with the bandpass filter；After the speech detection unit is used for bandpass filtering
Each frame main channel signal judged for speech frame or noise frame；
The time varing filter being connected with the speech detection unit；The time varing filter be used for carry out speech frame or
Each frame main channel signal after noise frame judges carries out time-varying bandpass filtering treatment；
The masking unit being connected with the framing unit；When the time varing filter is carried out to each frame main channel signal
While becoming bandpass filtering treatment, the masking unit is used to that each frame subaisle signal to be carried out sheltering treatment；When a frame master is logical
When road signal is speech frame, the filtering bandwidth of the time varing filter gradually increases, while masking unit adjustment and the language
Certain corresponding frame subaisle signal of sound frame is gradually reduced, when a frame main channel signal is noise frame, the time varing filter
Filtering bandwidth be gradually reduced, while the masking unit adjusts certain frame subaisle signal corresponding with the noise frame gradually increasing
Greatly；
The sum unit being connected with time varing filter and masking unit；The sum unit is used for will be by time-varying band logical
A frame main channel signal after filtering process is with certain the frame subaisle corresponding with the frame main channel signal by sheltering treatment
Signal correspondence superposition obtains a frame output signal；
Further,
If it is in_x (n) to carry out a frame main channel signal before bandpass filtering in the corresponding sampled value of n-th sampled point,
Then the corresponding sampled value of n-th sampled point is by the output result after bandpass filter Wherein, IIR_A (k) is that k-th filter factor, IIR_B (r) are r-th
Filter factor, u (n-k) represent that this frame main channel signal the n-th-k corresponding output result of sampled point after bandpass filtering (works as n
During >=k) or the corresponding output result of previous frame main channel signal k-th sampled point reciprocal after bandpass filtering (work as n<During k),
In_x (n-r) represent before this frame main channel signal bandpass filtering the n-th-r corresponding sampled value of sampled point (as n >=r) or on
The corresponding sampled value of one frame main channel signal, r-th sampled point reciprocal before bandpass filtering (works as n<During r)；
Further, the speech detection unit includes：
Power computation module, for calculating the power by each frame main channel signal after bandpass filter；
The threshold value acquisition module being connected with the power computation module；The threshold value acquisition module is used for according to power meter
The power by each frame main channel signal after bandpass filter that calculates of module is calculated to obtain phonetic speech power set, noise
The power demarcation threshold of power set and voice and noise；
The P being connected with threshold value acquisition modulenKnow module；The PnKnow module for each of noise power set
Probability distribution of the noise power in the power of each frame main channel signal is sued for peace, and when obtain and value is more than first
The maximum in each noise power is defined as P during preset valuen；Work as PnMeet pre-conditioned, and the phonetic speech power collection
It is described when meeting preset relation between each noise power average in each phonetic speech power average and the noise power set in conjunction
PnKnow that module updates Pn, otherwise keep PnIt is constant；
With threshold value acquisition module and PnKnow the determination module that module is connected；The determination module is used for basis and calculates
Power and P by the frame main channel signal after bandpass filternSize comparative result come determine the frame main channel believe
Number be speech frame or noise frame；
Further, the power computation module passes through formula Px=-10 × log10(Preal/327682) calculate process
The power of the frame main channel signal after bandpass filtering, in formula：PxRepresent by the frame main channel signal after bandpass filtering
Power,LEN represents that the corresponding sampling number of the frame length of main channel signal one, u (n) represent frame master
Channel signal carries out after bandpass filtering the corresponding sampled value of n-th sampled point；
Further, the threshold value acquisition module obtains phonetic speech power set, noise power set and voice and noise
The detailed process of power demarcation threshold be：
The frequency that the performance number of each frame main channel signal of the default frame number calculated to power computation module each occurs
Counted；
The probability distribution histogram of each frame main channel signal performance number of default frame number is set up, according to the probability distribution Nogata
Figure obtains phonetic speech power distributed areas and noise power distributed areas；
The power demarcation threshold of voice and noise is calculated using maximum variance between clusters；
Further, the phonetic speech power collection in setting phonetic speech power distributed areas is combined into [0, Pth- 1], noise power distribution
Noise power collection in region is combined into [Pth,Pm], wherein, Pth- 1 represents that the phonetic speech power in phonetic speech power distributed areas is minimum
Value, PthRepresent noise power maximum, the P in noise power distributed areasmRepresent the noise work(in noise power distributed areas
Rate minimum value；The threshold value acquisition module calculates the power demarcation threshold P of voice and noise using maximum variance between clustersthbest
Detailed process be：
By formulaCalculate each phonetic speech power average in phonetic speech power set, by formulaEach noise power average in noise power set is calculated, wherein, uvTable
Show each phonetic speech power average, the u in phonetic speech power setnRepresent each noise power average, p in noise power setiRepresent voice
The distribution probability of power i, pkValue is [0, P to distribution probability, the i of expression phonetic speech power k successivelyth- 1] each value, k values successively
It is [Pth,Pm] each value；
Using formula uT=w0uv+w1unThe mean power of each frame main channel signal of default frame number is drawn, wherein, uTRepresent
The mean power of each frame main channel signal of default frame number；
By formula σ2=w0(uv-uT)2+w1(un-uT)2The inter-class variance between phonetic speech power and noise power is obtained, and
Must be sent as an envoy to inter-class variance σ2The maximum performance number of value, the performance number is the power demarcation threshold P of voice and noisethbest；
Further, the initial value that the determination module sets state value VAD first is 0；For certain frame main channel signal, when
State value VAD is 0, while the performance number P of the frame main channel signalxCompare PnBig second preset value, then the determination module judgement should
Frame main channel signal is speech frame, and state value VAD is updated into 1, when state value VAD is 1, while the frame main channel signal
Performance number PxCompare PnSmall 3rd preset value, then the determination module judge that the frame main channel signal is noise frame, and by state value
VAD is updated to 0；
Further,
The centre frequency of the time varing filter is 0.8kHz, amplitude-frequency response
Difference equation isWherein, r is represented
Bandwidth control variables, excursion are [0.005,0.995], specifically, when a frame main channel signal is speech frame, bandwidth control
Variable r processed is changed to 0.005 by 0.995 according to step-length step1, and when a frame main channel signal is noise frame, bandwidth control becomes
Amount r is changed to 0.995 by 0.005 according to step-length step2, wherein：fsRepresent that sample frequency, y (n) represent the frame main channel signal
After time-varying bandpass filtering the corresponding output result of n-th sampled point, y (n-1) represent previous frame main channel signal through when variable speed
The 1st corresponding output result of sampled point reciprocal, y (n-2) represent that previous frame main channel signal is filtered through time-varying band logical after pass filter
The corresponding output result of second-to-last sampled point, j represent imaginary unit, j after ripple2=-1, ω represents circumference angular frequency；
When a frame main channel signal is speech frame, it is secondary logical that the masking unit adjusts certain frame corresponding with the speech frame
Road signal power gradually decreases to the speech frame by the power after time-varying bandpass filtering by previous frame subaisle signal power
1 percent, when a frame main channel signal is noise frame, the masking unit adjusts certain frame pair corresponding with the noise frame
Channel signal power is gradually increased to the power of present frame subaisle signal by previous frame subaisle signal power；It is described to shelter list
Unit passes through formulaDraw the work(of certain the frame subaisle signal corresponding with a frame main channel signal
Rate Py, in formula：LEN represents sampling number, in_y (n) expressions and the frame main channel signal phase that a frame main channel signal is included
Corresponding certain frame subaisle signal is in n-th sampled value of sampled point.
A kind of sound mask method, voice signal is received by main channel, and receives background signal by subaisle, described
Sound mask method comprises the following steps：
Step 1：Signal on the main channel is carried out into sub-frame processing and obtains each frame main channel signal, and by the pair
Signal on passage carries out sub-frame processing and obtains each frame subaisle signal, performs step 2；
Step 2：Each frame main channel signal is carried out into bandpass filtering, step 3 is performed；
Step 3：Each frame main channel signal after to bandpass filtering is that speech frame or noise frame judge, performs step 4；
Step 4：Time-varying bandpass filtering treatment is carried out to each frame main channel signal, while being covered to each frame subaisle signal
Cover treatment；When a frame main channel signal is speech frame, filtering bandwidth gradually increases, while certain frame corresponding with the speech frame
Subaisle signal is gradually reduced, when a frame main channel signal be noise frame when, filtering bandwidth is gradually reduced, at the same with the noise frame
Certain corresponding frame subaisle signal gradually increases, and performs step 5；
Step 5：By by the frame main channel signal after time-varying bandpass filtering treatment with by sheltering treatment and the frame
The corresponding frame subaisle signal correspondence superposition of main channel signal obtains a frame output signal；
Further, the step 3 specifically includes following steps：
Step 31：The power by each frame main channel signal after bandpass filtering is calculated, step 32 is performed；
Step 32：Voice work(is obtained according to the power by each frame main channel signal after bandpass filtering for calculating
The power demarcation threshold of rate set, noise power set and voice and noise, performs step 33；
Step 33：Probability distribution to each noise power of noise power set in each frame main channel signal power
Sued for peace, and the maximum in each noise power is defined as P when obtain and value is more than the first preset valuen, hold
Row step 34；
Step 34：Work as PnMeet pre-conditioned, and each phonetic speech power average in the phonetic speech power set with it is described
When meeting preset relation between each noise power average in noise power set, P is updatedn, otherwise keep PnIt is constant, perform step
35；
Step 35：According to the power and P by the frame main channel signal after bandpass filtering that calculatenSize compare
As a result come determine the frame main channel signal be speech frame or noise frame.
By adopting the above-described technical solution, the sound mask device and sound mask method of present invention offer, can avoid
It is weaker in voice signal, and noise signal it is stronger when, use fixed decision level to cause the court verdict of voice and noise
Inaccurate technical problem, and can realize that, when a certain frame in main channel is speech frame, it should not have lossy whole to lead to
Cross, but then filtered as far as possible when the frame is noise frame, so as to the influence of noise be preferably minimized；This kind of sound is covered simultaneously
Covering device and sound mask method can realize when it is speech frame to judge a certain frame in main channel, the subaisle signal in respective frame
Power quickly reduces with phonetic speech power but remains to be identified, at the end of voice is judged, the subaisle signal power in respective frame
Slow increase, until the frame end；The present invention shelters that result is good, and flexibility is high.
Brief description of the drawings
Fig. 1 is the structured flowchart of sound mask device of the present invention；
Fig. 2 is the structured flowchart of speech detection unit of the present invention；
The power schematic diagram of each frame main channel signal when Fig. 3 is default frame number value 200 of the invention；
Fig. 4 is the voice signal as sound mask device input signal of the present invention；
Fig. 5 is the background signal as sound mask device input signal of the present invention；
Fig. 6 is the change curve schematic diagram of state value VAD of the present invention；
Fig. 7 be background signal of the present invention in the presence of voice signal whether there is, the process schematic of its change in gain；
Fig. 8 is by the background signal after sound mask device of the present invention treatment；
Fig. 9 is the output signal of sound mask device of the present invention；
Figure 10 is the probability distribution histogram of the performance number of each frame main channel signal of the default frame number of the present invention；
Figure 11 is the flow chart of sound mask method of the present invention.
Specific embodiment
A kind of sound mask device as depicted in figs. 1 and 2, including：Receive the main channel of voice signal；Receive background letter
Number subaisle；Framing unit, each frame main channel signal is obtained for the signal on the main channel to be carried out into sub-frame processing,
And the signal on the subaisle is carried out into sub-frame processing obtain each frame subaisle signal；It is connected with the framing unit
Bandpass filter；The bandpass filter is used to for each frame main channel signal to carry out bandpass filtering；With the band logical
The speech detection unit that wave filter is connected；The speech detection unit be used for bandpass filtering after each frame main channel signal
It is that speech frame or noise frame are judged；The time varing filter being connected with the speech detection unit；The time-variable filtering
Device is used to carry out time-varying bandpass filtering treatment to each frame main channel signal after carrying out speech frame or noise frame judgement；With institute
State the masking unit that framing unit is connected；Time-varying bandpass filtering is carried out to each frame main channel signal in the time varing filter
While treatment, the masking unit is used to that each frame subaisle signal to be carried out sheltering treatment；When a frame main channel signal is language
During sound frame, the filtering bandwidth of the time varing filter gradually increases, while masking unit adjustment is corresponding with the speech frame
Certain frame subaisle signal be gradually reduced, when a frame main channel signal be noise frame when, the filtering bandwidth of the time varing filter
It is gradually reduced, while the masking unit adjusts certain frame subaisle signal corresponding with the noise frame gradually increasing；With time-varying
The sum unit that wave filter is connected with masking unit；The sum unit is used for by after time-varying bandpass filtering treatment
Frame subaisle signal is corresponding is superimposed with certain corresponding with the frame main channel signal by sheltering treatment for frame main channel signal
To a frame output signal；Further, if the frame main channel signal carried out before bandpass filtering is corresponding in n-th sampled point
Sampled value is in_x (n), then the corresponding sampled value of n-th sampled point is by the output result after bandpass filterWherein, IIR_A (k) is k-th filtering
Coefficient, IIR_B (r) are that r-th filter factor, u (n-k) represent the n-th-k sampling after bandpass filtering of this frame main channel signal
The corresponding output result (as n >=k) of point or previous frame main channel signal k-th sampled point reciprocal after bandpass filtering are corresponding
Output result (works as n<During k), in_x (n-r) represents before this frame main channel signal bandpass filtering the n-th-r sampled point is corresponding and adopt
The corresponding sampled value of sample value (as n >=r) or previous frame main channel signal r-th sampled point reciprocal before bandpass filtering (works as n<r
When)；Further, the speech detection unit includes：Power computation module, for calculating by each after bandpass filter
The power of frame main channel signal；The threshold value acquisition module being connected with the power computation module；The threshold value acquisition module is used
Voice work(is obtained in the power by each frame main channel signal after bandpass filter calculated according to power computation module
The power demarcation threshold of rate set, noise power set and voice and noise；The P being connected with threshold value acquisition modulenKnow
Module；The PnKnow module for each noise power to noise power set in the power of each frame main channel signal
Probability distribution sued for peace, and when obtain and value more than the first preset value when the maximum in each noise power is determined
Justice is Pn；Work as PnMeet pre-conditioned, and each phonetic speech power average and the noise power collection in the phonetic speech power set
When meeting preset relation between each noise power average in conjunction, the PnKnow that module updates Pn, otherwise keep PnIt is constant；With threshold
Value acquisition module and PnKnow the determination module that module is connected；The determination module is filtered for what basis was calculated by band logical
The power and P of the frame main channel signal after ripple devicenSize comparative result determining the frame main channel signal for speech frame or
Noise frame；Further, the power computation module passes through formula Px=-10 × log10(Preal/327682) pass through band calculating
The power of the frame main channel signal after pass filter, in formula：PxRepresent the work(by the frame main channel signal after bandpass filtering
Rate,LEN represents that the corresponding sampling number of the frame length of main channel signal one, u (n) represent that frame master is led to
Road signal carries out after bandpass filtering the corresponding sampled value of n-th sampled point；Further, the threshold value acquisition module obtains voice
The detailed process of the power demarcation threshold of power set, noise power set and voice and noise is：To power computation module
The frequency that the performance number of each frame main channel signal of the default frame number for calculating each occurs is counted；Set up default frame number
The probability distribution histogram of each frame main channel signal performance number, phonetic speech power distributed areas are obtained according to the probability distribution histogram
With noise power distributed areas；The power demarcation threshold of voice and noise is calculated using maximum variance between clusters；Further, if
Phonetic speech power collection in attribute sound power distribution area is combined into [0, Pth- 1], the noise power set in noise power distributed areas
It is [Pth,Pm], wherein, Pth- 1 represents phonetic speech power minimum value, the P in phonetic speech power distributed areasthRepresent noise power distribution
Noise power maximum in region, PmRepresent the noise power minimum value in noise power distributed areas；The threshold value is obtained
Module calculates the power demarcation threshold P of voice and noise using maximum variance between clustersthbestDetailed process be：By formulaCalculate each phonetic speech power average in phonetic speech power set, by formulaCalculate and make an uproar
Each noise power average in acoustical power set, wherein,uvIn expression phonetic speech power set
Each phonetic speech power average, unRepresent each noise power average, p in noise power setiRepresent distribution probability, the p of phonetic speech power ik
Value is [0, P to distribution probability, the i of expression phonetic speech power k successivelyth- 1] value is [P to each value, k successivelyth,Pm] each value；Profit
Use formula uT=w0uv+w1unThe mean power of each frame main channel signal of default frame number is drawn, wherein, uTRepresent default frame number
The mean power of each frame main channel signal；By formula σ2=w0(uv-uT)2+w1(un-uT)2Obtain phonetic speech power and noise power
Between inter-class variance, and the inter-class variance σ that must send as an envoy to2The maximum performance number of value, the performance number is the power of voice and noise
Demarcation threshold Pthbest；Further, the initial value that the determination module sets state value VAD first is 0；For certain frame main channel
Signal, when state value VAD is 0, while the performance number P of the frame main channel signalxCompare PnBig second preset value, then the judgement mould
Block judges that the frame main channel signal is speech frame, and state value VAD is updated into 1, when state value VAD is 1, while frame master is logical
The performance number P of road signalxCompare PnSmall 3rd preset value, then the determination module judges that the frame main channel signal is noise frame, and incites somebody to action
State value VAD is updated to 0；Further, the centre frequency of the time varing filter is 0.8kHz, amplitude-frequency responseDifference equation is Wherein, r represent bandwidth control variables, excursion for [0.005,
0.995], specifically, when a frame main channel signal is speech frame, bandwidth control variables r is become by 0.995 according to step-length step1
Change to 0.005, when a frame main channel signal is noise frame, bandwidth control variables r is changed to by 0.005 according to step-length step2
0.995, wherein：fsRepresent that sample frequency, y (n) represent the frame main channel signal n-th sampled point pair after time-varying bandpass filtering
The output result answered, y (n-1) represent that previous frame main channel signal the 1st sampled point reciprocal after time-varying bandpass filtering is corresponding
Output result, y (n-2) represent the corresponding output of second-to-last sampled point after time-varying bandpass filtering of previous frame main channel signal
As a result, j represents imaginary unit, j2=-1, ω represents circumference angular frequency；It is described to cover when a frame main channel signal is speech frame
Unit adjustment certain frame subaisle signal power corresponding with the speech frame is covered to be gradually reduced by previous frame subaisle signal power
It is described when a frame main channel signal is noise frame to the speech frame by 1 percent of the power after time-varying bandpass filtering
Masking unit adjusts certain frame subaisle signal power corresponding with the noise frame and is gradually increased by previous frame subaisle signal power
The big power to present frame subaisle signal；The masking unit passes through formulaDraw and one
The power P of certain corresponding frame subaisle signal of frame main channel signaly, in formula：LEN represents what a frame main channel signal was included
Sampling number, in_y (n) expressions certain frame subaisle signal corresponding with a frame main channel signal are in n-th sampling of sampled point
Value.
A kind of sound mask method as shown in figure 11, receives voice signal, and receive by subaisle by main channel
Background signal, the sound mask method comprises the following steps：
Step 1：Signal on the main channel is carried out into sub-frame processing and obtains each frame main channel signal, and by the pair
Signal on passage carries out sub-frame processing and obtains each frame subaisle signal, performs step 2；
Step 2：Each frame main channel signal is carried out into bandpass filtering, step 3 is performed；
Step 3：Each frame main channel signal after to bandpass filtering is that speech frame or noise frame judge, performs step 4；
Step 4：Time-varying bandpass filtering treatment is carried out to each frame main channel signal, while being covered to each frame subaisle signal
Cover treatment；When a frame main channel signal is speech frame, filtering bandwidth gradually increases, while certain frame corresponding with the speech frame
Subaisle signal is gradually reduced, when a frame main channel signal be noise frame when, filtering bandwidth is gradually reduced, at the same with the noise frame
Certain corresponding frame subaisle signal gradually increases, and performs step 5；
Step 5：By by the frame main channel signal after time-varying bandpass filtering treatment with by sheltering treatment and the frame
The corresponding frame subaisle signal correspondence superposition of main channel signal obtains a frame output signal；
Further, the step 3 specifically includes following steps：
Step 31：The power by each frame main channel signal after bandpass filtering is calculated, step 32 is performed；
Step 32：Voice work(is obtained according to the power by each frame main channel signal after bandpass filtering for calculating
The power demarcation threshold of rate set, noise power set and voice and noise, performs step 33；
Step 33：Probability distribution to each noise power of noise power set in each frame main channel signal power
Sued for peace, and the maximum in each noise power is defined as P when obtain and value is more than the first preset valuen, hold
Row step 34；
Step 34：Work as PnMeet pre-conditioned, and each phonetic speech power average in the phonetic speech power set with it is described
When meeting preset relation between each noise power average in noise power set, P is updatedn, otherwise keep PnIt is constant, perform step
35；
Step 35：According to the power and P by the frame main channel signal after bandpass filtering that calculatenSize compare
As a result come determine the frame main channel signal be speech frame or noise frame.
Further, if carrying out a frame main channel signal before bandpass filtering in the corresponding sampled value of n-th sampled point
It is in_x (n), then the corresponding sampled value of n-th sampled point is by the output result after bandpass filteringWherein, IIR_A (k) is k-th filtering
Coefficient, IIR_B (r) are that r-th filter factor, u (n-k) represent the n-th-k sampling after bandpass filtering of this frame main channel signal
The corresponding output result (as n >=k) of point or previous frame main channel signal k-th sampled point reciprocal after bandpass filtering are corresponding
Output result (works as n<During k), in_x (n-r) represents before this frame main channel signal bandpass filtering the n-th-r sampled point is corresponding and adopt
The corresponding sampled value of sample value (as n >=r) or previous frame main channel signal r-th sampled point reciprocal before bandpass filtering (works as n<r
When)；
Further, the step 31 is specially：By formula Px=-10 × log10(Preal/327682) calculate process
The power of the frame main channel signal after bandpass filtering, in formula：PxRepresent by the frame main channel signal after bandpass filtering
Power,LEN represents that the corresponding sampling number of the frame length of main channel signal one, u (n) represent frame master
Channel signal carries out after bandpass filtering the corresponding sampled value of n-th sampled point；
Further, the step 32 specifically includes following steps：
Step 321：The frequency that the performance number for presetting each frame main channel signal of frame number each occurs is counted；
Step 322：The probability distribution histogram of each frame main channel signal performance number of default frame number is set up, according to the probability
Distribution histogram obtains phonetic speech power distributed areas and noise power distributed areas；
Step 323：The power demarcation threshold P of voice and noise is calculated using maximum variance between clustersthbest；
Further, the phonetic speech power collection in setting phonetic speech power distributed areas is combined into [0, Pth- 1], noise power distribution
Noise power collection in region is combined into [Pth,Pm], wherein, Pth- 1 represents that the phonetic speech power in phonetic speech power distributed areas is minimum
Value, PthRepresent noise power maximum, the P in noise power distributed areasmRepresent the noise work(in noise power distributed areas
Rate minimum value；The power demarcation threshold P of voice and noise is calculated using maximum variance between clustersthbestDetailed process be：
By formulaCalculate each phonetic speech power average in phonetic speech power set, by formulaEach noise power average in noise power set is calculated, wherein, uvTable
Show each phonetic speech power average, the u in phonetic speech power setnRepresent each noise power average, p in noise power setiRepresent voice
The distribution probability of power i, pkValue is [0, P to distribution probability, the i of expression phonetic speech power k successivelyth- 1] each value, k values successively
It is [Pth,Pm] each value；
Using formula uT=w0uv+w1unThe mean power of each frame main channel signal of default frame number is drawn, wherein, uTRepresent
The mean power of each frame main channel signal of default frame number；
By formula σ2=w0(uv-uT)2+w1(un-uT)2The inter-class variance between phonetic speech power and noise power is obtained, and
Must be sent as an envoy to inter-class variance σ2The maximum performance number of value, the performance number is the power demarcation threshold P of voice and noisethbest；
Further, the step 35 is specially：The initial value for setting state value VAD first is 0；For certain frame main channel letter
Number, when state value VAD is 0, while the performance number P of the frame main channel signalxCompare PnBig second preset value, then judge that frame master is led to
Road signal is speech frame, and state value VAD is updated into 1, when state value VAD is 1, while the performance number of the frame main channel signal
PxCompare PnSmall 3rd preset value, then judge that the frame main channel signal is noise frame, and state value VAD is updated into 0；
Further, time-varying bandpass filtering treatment is carried out to each frame main channel signal using time varing filter；The time-varying
The centre frequency of wave filter is 0.8kHz, amplitude-frequency responseDifference side
Cheng WeiWherein,
R represents bandwidth control variables, excursion for [0.005,0.995], specifically, when a frame main channel signal is speech frame,
Bandwidth control variables r is changed to 0.005 by 0.995 according to step-length step1, when a frame main channel signal is noise frame, bandwidth
Control variables r is changed to 0.995 by 0.005 according to step-length step2, wherein：fsRepresent that sample frequency, y (n) represent that frame master is led to
The corresponding output result of road signal n-th sampled point after time-varying bandpass filtering, y (n-1) represent previous frame main channel signal warp
The 1st corresponding output result of sampled point reciprocal, y (n-2) represent previous frame main channel signal through time-varying after time-varying bandpass filtering
The corresponding output result of second-to-last sampled point, j represent imaginary unit, j after bandpass filtering2=-1, ω represents circumference angular frequency
Rate；
When a frame main channel signal is speech frame, adjust certain frame subaisle signal power corresponding with the speech frame by
Previous frame subaisle signal power gradually decreases to the speech frame by 1 percent of the power after time-varying bandpass filtering, when one
When frame main channel signal is noise frame, certain frame subaisle signal power corresponding with the noise frame is adjusted by previous frame subaisle
Signal power is gradually increased to the power of present frame subaisle signal；By formulaDraw with
The power P of certain corresponding frame subaisle signal of one frame main channel signaly, in formula：LEN represents that a frame main channel signal is included
Sampling number, in_y (n) represent certain frame subaisle signal corresponding with a frame main channel signal adopting in n-th sampled point
Sample value.
It is specific that the present invention is counted to the frequency that the performance number for presetting each frame main channel signal of frame number each occurs
Process is：Floor operation is carried out respectively to the performance number for presetting each frame main channel signal of frame number first, then from the first frame master
The performance number of channel signal starts successively using the performance number of each frame main channel signal as accumulator subscript, until by default frame number
The performance number of each frame main channel signal all travel through, if existing in the performance number of each frame main channel signal of default frame number a few
Individual value is equal, then plus one to that should be worth lower target accumulator, and then draws the probability that the performance number of the i-th frame main channel signal occursThe default frame number can be with value 200.
Each frame main channel signal is carried out bandpass filtering by the present invention, can reduce influence of the noise to voice detection results；
A width of 0.3kHz~the 3.4kHz of band of bandpass filter；The Reference Design index of the bandpass filter is：Centre frequency is
1.55kHz, passband frequency range are 0.3kHz~3.4kHz, lower stopband cut-off frequency is 1Hz, upper stopband cut-off frequency is
4kHz, stopband attenuation are more than 60dB；The Reference Design result of the bandpass filter is a 7 rank iir filters, its filter factor
IIR_B (7)={ 1.012205768830948 × 10-3,-4.911110647449132×10-4,7.807184279553245
×10-5,-1.198332967805191×10-3,7.807184279553245×10-5,-4.911110647449132e×
10-4,1.012205768830948×10-3}；Filter factor IIR_A (7)=1.0, -5.380875974547,
12.23643655587558,-15.06088779864848,10.58294743567488,-4.024466821830663,
0.6468480167658538}；Work as n<During k, previous frame main channel signal k-th sampled point reciprocal after bandpass filtering is corresponding
It is 0 that output result u (n-k) assigns initial value, works as n<During r, reciprocal r-th sampled point of the previous frame main channel signal before bandpass filtering
It is 0 that corresponding sampled value in_x (n-r) also assigns initial value.
First preset value of the present invention is 0.1；Second preset value is 12dB；3rd preset value is 6dB；Institute
It refers to P to state pre-conditionednIn [Pthbest,Pm] scope, the preset relation refers to | uv-un| ＞ 10；LEN can be with value 960；
The present invention passes through formula Px=-10 × log10(Preal/327682) calculate by the frame main channel signal after bandpass filtering
Power, wherein,It is an intermediate variable, specifically, PrealRepresent by the frame master after bandpass filtering
The actual power of channel signal, correspondingly, PxIt is to represent the relative power value by the frame main channel signal after bandpass filtering,
It is 32768 that the present invention defines corresponding actual power value at 0dBoV2, therefore PxIt is a relative power value of relative 0dBoV；This hair
The bright envelope that main channel signal and subaisle signal can also be obtained using envelope detection or peak detection, is recycled per frame envelope
Obtain power.
The power schematic diagram of each frame main channel signal when Fig. 3 shows the present invention default frame number value 200, such as Fig. 3 institutes
Show, PxThe default frame number i.e. power of 200 frame main channel signals is collectively constituted with the power of the 199 frame voice signals tried to achieve before.
Figure 10 shows the probability distribution histogram of each frame main channel signal performance number of the default frame number of the present invention, such as Figure 10
It is shown, after the probability distribution histogram foundation of the performance number for presetting each frame main channel signal of frame number, can be according to the probability
Distribution histogram obtains phonetic speech power distributed areas and noise power distributed areas, and obtain the phonetic speech power distributed areas and
The power demarcation threshold P of the border region between noise power distributed areas, voice and noisethbestJust in the border region
Calculating acquisition is carried out by maximum variance between clusters.
When a frame main channel signal is speech frame, the filtering bandwidth of the time varing filter gradually increases the present invention, while
The masking unit adjusts certain frame subaisle signal corresponding with the speech frame and is gradually reduced, specifically, by controlling certain frame pair
The power of channel signal realizes that the amplitude of the frame subaisle signal reduces, when a frame main channel signal is noise frame, the time-varying
The filtering bandwidth of wave filter is gradually reduced, while the masking unit adjusts certain frame subaisle signal corresponding with the noise frame
Gradually increase, specifically, realize that the amplitude of the frame subaisle signal increases by controlling the power of certain frame subaisle signal；This hair
Bright y (n-1), y (n-2) initialization are assigned initial value and are 0；When one
When frame main channel signal is speech frame, certain frame subaisle signal power corresponding with the speech frame is by previous frame subaisle signal
Power gradually decreases to the speech frame by 1 percent of the power after time-varying bandpass filtering, when a frame main channel signal is to make an uproar
During acoustic frame, certain frame subaisle signal power corresponding with the noise frame is gradually increased to work as by previous frame subaisle signal power
The power of previous frame subaisle signal, specifically, when state value VAD is 1, subaisle signal gain curr_gain is in a frame in
Quickly reduce according to every point step size 0.01, until reaching minimum valuePxxRepresent by after time-varying bandpass filtering
A frame main channel signal power, when state value VAD be 0 when, subaisle signal gain curr_gain is in a frame according to every
Point step size 0.000005 slowly increases, until reaching default maximum gain 1；Frame length can be 20ms, sampling frequency during practical application
Rate is 48kHz, and then sampling number corresponding to a frame length is 960, and every point step size here then refers to 960 steps of sampled point
It is long；Output outputm (n)=curr_gain*in_y (n) of the subaisle signal in n-th sampled point of a frame in；The present invention is most
Afterwards by by the frame main channel signal after time-varying bandpass filtering treatment with by sheltering treatment and the frame main channel signal phase
Corresponding certain frame subaisle signal correspondence superposition obtains a frame output signal, specifically, in n-th sample point, output signal
It is output (n)=y (n)+curr_gain*in_y (n).
The excursion of bandwidth control variables r of the present invention is [0.005,0.995], specifically, when state value VAD is 1,
Bandwidth control variables r is changed to 0.005 by 0.995 according to step-length step1, when state value VAD be 0 when, bandwidth control variables r by
0.005 is changed to 0.995 according to step-length step2, wherein：fs
Represent sample frequency, y (n) represent the corresponding output result of the frame main channel signal n-th sampled point after time-varying bandpass filtering,
Y (n-1) represents previous frame main channel signal the 1st the corresponding output result of sampled point, y (n- of inverse after time-varying bandpass filtering
2) represent previous frame main channel signal after time-varying bandpass filtering the corresponding output result of second-to-last sampled point (when y (n) is
During the 1st sampled point, y (n-1) and y (n-2) need to assign initial value for 0), j represent imaginary unit, j2=-1, w represents circumference angular frequency
Rate, specifically, works as fsDuring value 48kHz, step1=0.0006875, step2=0.000006875, due to step-length step1 ratios
Step-length step2 is much larger, therefore, when r is changed by step-length step1,0.005 quickly can be changed in the short time, when r is pressed
When step2 changes, then need the long period slowly varying to 0.995.
The specific effect tested with the subaisle signal as background music signal below after the present invention is implemented, Fig. 4 shows
The reception signal (voice signal) gone out on sound mask device main channel of the present invention, Fig. 5 is shown as of the present invention
Reception signal (background music signal) on sound mask device subaisle；During test sound mask device, maximum kind is used
Between variance method (Ostu) seek the power demarcation threshold of voice and noise, and combine PnJudge that a frame main channel signal is speech frame
(state value VAD is 1) or noise frame (state value VAD is 0), Fig. 6 shows the change curve schematic diagram of state value VAD, such as
Shown in Fig. 6, it can be seen that having at voice, state value VAD is 1, when voice disappears, state value VAD is equal to 0；Fig. 7 shows
Background music signal of the present invention in the presence of voice signal whether there is, the process of its change in gain, as shown in Figure 7, it can be seen that
In the presence of voice signal, background music signal gain quickly reduces, and after voice signal terminates, background music signal gain is then
Slow increase；Fig. 8 shows that, by the background music signal after sound mask device of the present invention treatment, Fig. 9 shows this hair
The output signal of the bright sound mask device；Test result shows, in the presence of voice signal, can both protrude voice signal or
Background music can be picked out；When not having voice signal, background music smootherly gradually increases, human auditory system experience effect
Well.
The present invention can avoid it is weaker in voice signal, and noise signal it is stronger when, can be made using fixed decision level
Voice and noise the inaccurate technical problem of court verdict, and can realize when a certain frame in main channel is voice signal
When, it should not have lossy whole to pass through, but then be filtered as far as possible when the frame is noise, so as to by the influence of noise
It is preferably minimized；This kind of sound mask device and sound mask method can be realized judging that a certain frame in main channel is voice letter simultaneously
Number when, the subaisle signal power in respective frame quickly reduces with phonetic speech power but remains to be identified, at the end of voice is judged,
Subaisle signal power in respective frame slowly increases, until the frame end；Further, the present invention is obtained using Ostu algorithms
The power demarcation threshold of voice and noise is obtained, the accuracy of power demarcation threshold is improve；When using speech frame and noise frame
Becoming wave filter carries out time-varying bandpass filtering treatment, both ensure that speech energy is as much as possible and has passed through, and also reduces noise to letter
Number influence；Subaisle signal is to follow main channel signal size conversion by frame when carrying out and sheltering treatment simultaneously.
The above, the only present invention preferably specific embodiment, but protection scope of the present invention is not limited thereto,
Any one skilled in the art the invention discloses technical scope in, technology according to the present invention scheme and its
Inventive concept is subject to equivalent or change, should all be included within the scope of the present invention.
Claims (10)
1. a kind of sound mask device, it is characterised in that the sound mask device includes：
Receive the main channel of voice signal；
Receive the subaisle of background signal；
Framing unit, each frame main channel signal is obtained for the signal on the main channel to be carried out into sub-frame processing, and by institute
State the signal on subaisle and carry out sub-frame processing and obtain each frame subaisle signal；
The bandpass filter being connected with the framing unit；The bandpass filter is used to enter each frame main channel signal
Row bandpass filtering；
The speech detection unit being connected with the bandpass filter；The speech detection unit be used for bandpass filtering after it is each
Frame main channel signal is that speech frame or noise frame are judged；
The time varing filter being connected with the speech detection unit；The time varing filter is used for carrying out speech frame or noise
Each frame main channel signal after frame judges carries out time-varying bandpass filtering treatment；
The masking unit being connected with the framing unit；The variable speed when the time varing filter is carried out to each frame main channel signal
While pass filter treatment, the masking unit is used to that each frame subaisle signal to be carried out sheltering treatment；When frame main channel letter
Number for speech frame when, the filtering bandwidth of the time varing filter gradually increases, at the same the masking unit adjustment with the speech frame
Certain corresponding frame subaisle signal is gradually reduced, when a frame main channel signal is noise frame, the filter of the time varing filter
Wavestrip is wide to be gradually reduced, while the masking unit adjusts certain frame subaisle signal corresponding with the noise frame gradually increasing；
The sum unit being connected with time varing filter and masking unit；The sum unit is used for will be by time-varying bandpass filtering
A frame main channel signal after treatment is with certain the frame subaisle signal corresponding with the frame main channel signal by sheltering treatment
Correspondence superposition obtains a frame output signal.
2. sound mask device according to claim 1, it is characterised in that
If it is in_x (n), institute to carry out a frame main channel signal before bandpass filtering in the corresponding sampled value of n-th sampled point
The corresponding sampled value of n-th sampled point is stated by the output result after bandpass filter Wherein, IIR_A (k) is that k-th filter factor, IIR_B (r) are r-th filtering system
Number, u (n-k) represent this frame main channel signal the n-th-k corresponding output result of sampled point after bandpass filtering (as n >=k)
Or the corresponding output result of previous frame main channel signal k-th sampled point reciprocal after bandpass filtering (works as n<During k), in_x (n-
R) represent that the n-th-r corresponding sampled value of sampled point (as n >=r) or previous frame master are logical before this frame main channel signal bandpass filtering
The corresponding sampled value of road signal r-th sampled point reciprocal before bandpass filtering (works as n<During r).
3. sound mask device according to claim 1, it is characterised in that the speech detection unit includes：
Power computation module, for calculating the power by each frame main channel signal after bandpass filter；
The threshold value acquisition module being connected with the power computation module；The threshold value acquisition module is used for according to power calculation mould
The power by each frame main channel signal after bandpass filter that block is calculated obtains phonetic speech power set, noise power
Set and the power demarcation threshold of voice and noise；
The P being connected with threshold value acquisition modulenKnow module；The PnKnow module for each noise to noise power set
Probability distribution of the power in the power of each frame main channel signal is sued for peace, and when obtain and value is preset more than first
The maximum in each noise power is defined as P during valuen；Work as PnMeet pre-conditioned, and in the phonetic speech power set
Each phonetic speech power average and the noise power set in when meeting preset relation between each noise power average, the PnObtain
Know that module updates Pn, otherwise keep PnIt is constant；
With threshold value acquisition module and PnKnow the determination module that module is connected；The determination module is used for according to the warp for calculating
The power and P of the frame main channel signal crossed after bandpass filternSize comparative result be determining the frame main channel signal
Speech frame or noise frame.
4. sound mask device according to claim 3, it is characterised in that the power computation module passes through formula Px=-10
×log10(Preal/327682) calculate the power by the frame main channel signal after bandpass filtering, in formula：PxRepresent and pass through
The power of the frame main channel signal after bandpass filtering,LEN represents the frame length of main channel signal one
Corresponding sampling number, u (n) represent that the frame main channel signal carries out after bandpass filtering the corresponding sampled value of n-th sampled point.
5. sound mask device according to claim 3, it is characterised in that the threshold value acquisition module obtains phonetic speech power collection
Close, the detailed process of the power demarcation threshold of noise power set and voice and noise is：
The frequency that the performance number of each frame main channel signal of the default frame number calculated to power computation module each occurs is carried out
Statistics；
The probability distribution histogram of each frame main channel signal performance number of default frame number is set up, is obtained according to the probability distribution histogram
To phonetic speech power distributed areas and noise power distributed areas；
The power demarcation threshold of voice and noise is calculated using maximum variance between clusters.
6. sound mask device according to claim 5, it is characterised in that the voice work(in setting phonetic speech power distributed areas
Rate collection is combined into [0, Pth- 1], the noise power collection in noise power distributed areas is combined into [Pth,Pm], wherein, Pth- 1 represents voice
Phonetic speech power minimum value in power distribution area, PthRepresent noise power maximum, the P in noise power distributed areasmTable
Show the noise power minimum value in noise power distributed areas；The threshold value acquisition module calculates language using maximum variance between clusters
The power demarcation threshold P of sound and noisethbestDetailed process be：
By formulaCalculate each phonetic speech power average in phonetic speech power set, by formulaEach noise power average in noise power set is calculated, wherein, uvTable
Show each phonetic speech power average, the u in phonetic speech power setnRepresent each noise power average, p in noise power setiRepresent voice
The distribution probability of power i, pkValue is [0, P to distribution probability, the i of expression phonetic speech power k successivelyth- 1] each value, k values successively
It is [Pth,Pm] each value；
Using formula uT=w0uv+w1unThe mean power of each frame main channel signal of default frame number is drawn, wherein, uTRepresent default
The mean power of each frame main channel signal of frame number；
By formula σ2=w0(uv-uT)2+w1(un-uT)2The inter-class variance between phonetic speech power and noise power is obtained, and is drawn
Make inter-class variance σ2The maximum performance number of value, the performance number is the power demarcation threshold P of voice and noisethbest。
7. sound mask device according to claim 2, it is characterised in that the determination module sets state value VAD's first
Initial value is 0；For certain frame main channel signal, when state value VAD is 0, while the performance number P of the frame main channel signalxCompare PnIt is big by
Two preset values, then the determination module judges that the frame main channel signal is speech frame, and state value VAD is updated into 1, works as state
Value VAD is 1, while the performance number P of the frame main channel signalxCompare PnSmall 3rd preset value, then the determination module judge frame master
Channel signal is noise frame, and state value VAD is updated into 0.
8. sound mask device according to claim 1, it is characterised in that
The centre frequency of the time varing filter is 0.8kHz, amplitude-frequency response
Difference equation isIts
In, r represents bandwidth control variables, excursion for [0.005,0.995], specifically, when a frame main channel signal is speech frame
When, bandwidth control variables r is changed to 0.005 by 0.995 according to step-length step1, when a frame main channel signal is noise frame, band
Control variables r wide is changed to 0.995 by 0.005 according to step-length step2, wherein：fsRepresent that sample frequency, y (n) represent frame master
The corresponding output result of channel signal n-th sampled point after time-varying bandpass filtering, y (n-1) represent previous frame main channel signal
After time-varying bandpass filtering the 1st corresponding output result of sampled point reciprocal, y (n-2) represent previous frame main channel signal through when
The corresponding output result of second-to-last sampled point, j represent imaginary unit, j after change bandpass filtering2=-1, ω represents circumference angular frequency
Rate；
When a frame main channel signal is speech frame, the masking unit adjusts certain the frame subaisle letter corresponding with the speech frame
Number power gradually decreases to percentage of the speech frame by the power after time-varying bandpass filtering by previous frame subaisle signal power
One of, when a frame main channel signal is noise frame, the masking unit adjusts certain frame subaisle corresponding with the noise frame
Signal power is gradually increased to the power of present frame subaisle signal by previous frame subaisle signal power；The masking unit leads to
Cross formulaDraw the power of certain the frame subaisle signal corresponding with a frame main channel signal
Py, in formula：LEN represents that the sampling number that a frame main channel signal included, in_y (n) represent relative with a frame main channel signal
Certain the frame subaisle signal answered is in n-th sampled value of sampled point.
9. a kind of sound mask method, it is characterised in that the sound mask method receives voice signal by main channel, and leads to
Cross subaisle and receive background signal, the sound mask method comprises the following steps：
Step 1：Signal on the main channel is carried out into sub-frame processing and obtains each frame main channel signal, and by the subaisle
On signal carry out sub-frame processing and obtain each frame subaisle signal, perform step 2；
Step 2：Each frame main channel signal is carried out into bandpass filtering, step 3 is performed；
Step 3：Each frame main channel signal after to bandpass filtering is that speech frame or noise frame judge, performs step 4；
Step 4：Time-varying bandpass filtering treatment is carried out to each frame main channel signal, while carrying out cover to each frame subaisle signal
Reason；When a frame main channel signal is speech frame, filtering bandwidth gradually increases, while certain frame corresponding with the speech frame is secondary logical
Road signal is gradually reduced, and when a frame main channel signal is noise frame, filtering bandwidth is gradually reduced, while relative with the noise frame
Certain the frame subaisle signal answered gradually increases, and performs step 5；
Step 5：Will be same by sheltering leading to frame master for treatment by the frame main channel signal after time-varying bandpass filtering treatment
Signal corresponding frame subaisle signal correspondence superposition in road obtains a frame output signal.
10. sound mask method according to claim 9, it is characterised in that the step 3 specifically includes following steps：
Step 31：The power by each frame main channel signal after bandpass filtering is calculated, step 32 is performed；
Step 32：Phonetic speech power collection is obtained according to the power by each frame main channel signal after bandpass filtering for calculating
The power demarcation threshold of conjunction, noise power set and voice and noise, performs step 33；
Step 33：Probability distribution to each noise power of noise power set in each frame main channel signal power is carried out
Summation, and the maximum in each noise power is defined as P when obtain and value is more than the first preset valuen, perform step
Rapid 34；
Step 34：Work as PnMeet pre-conditioned, and each phonetic speech power average and the noise work(in the phonetic speech power set
When meeting preset relation between each noise power average in rate set, P is updatedn, otherwise keep PnIt is constant, perform step 35；
Step 35：According to the power and P by the frame main channel signal after bandpass filtering that calculatenSize comparative result,
To determine that the frame main channel signal is speech frame or noise frame.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611029084.7A CN106782587B (en) | 2016-11-20 | 2016-11-20 | Sound masking device and sound masking method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611029084.7A CN106782587B (en) | 2016-11-20 | 2016-11-20 | Sound masking device and sound masking method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106782587A true CN106782587A (en) | 2017-05-31 |
| CN106782587B CN106782587B (en) | 2020-04-28 |
Family
ID=58971532
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611029084.7A Active CN106782587B (en) | 2016-11-20 | 2016-11-20 | Sound masking device and sound masking method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106782587B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102099851A (en) * | 2008-07-18 | 2011-06-15 | 皇家飞利浦电子股份有限公司 | Method and system for preventing overhearing of private conversations in public places |
| JP2012194415A (en) * | 2011-03-17 | 2012-10-11 | Yamaha Corp | Masker sound measurement instrument and sound masking device |
| CN104156578A (en) * | 2014-07-31 | 2014-11-19 | 南京工程学院 | Recording time identification method |
| CN204495996U (en) * | 2011-10-26 | 2015-07-22 | 菲力尔系统公司 | broadband sonar receiver |
| CN205004029U (en) * | 2015-09-29 | 2016-01-27 | 苏州一天声学科技有限公司 | Ware is sheltered to array sound |
-
2016
- 2016-11-20 CN CN201611029084.7A patent/CN106782587B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102099851A (en) * | 2008-07-18 | 2011-06-15 | 皇家飞利浦电子股份有限公司 | Method and system for preventing overhearing of private conversations in public places |
| JP2012194415A (en) * | 2011-03-17 | 2012-10-11 | Yamaha Corp | Masker sound measurement instrument and sound masking device |
| CN204495996U (en) * | 2011-10-26 | 2015-07-22 | 菲力尔系统公司 | broadband sonar receiver |
| CN104156578A (en) * | 2014-07-31 | 2014-11-19 | 南京工程学院 | Recording time identification method |
| CN205004029U (en) * | 2015-09-29 | 2016-01-27 | 苏州一天声学科技有限公司 | Ware is sheltered to array sound |
Non-Patent Citations (6)
| Title |
|---|
| AJAY NATARAJAN: ""An Auditory-Masking-Threshold-Based Noise Suppression Algorithm GMMSE-AMT For Listeners"", 《EURASIP》 * |
| JUGURTA MONTALVAO: ""IMPROVED SIGNAL REPRESENTATION FOR EVENT DETECTION IN REMOTE HEALTH CARE THROUGH PSYCHOANALYICAL MASKING"", 《RESEARCHGATE》 * |
| 吕勇: ""基于最小统计和人耳掩蔽特性的语音增强算法"", 《电声技术》 * |
| 张勇: ""非平稳噪声环境下结合听觉掩蔽的语音增强"", 《计算机工程与设计》 * |
| 曹亮: ""基于听觉掩蔽效应的多频带谱减语音增强算法"", 《计算机工程与设计》 * |
| 曹龙涛: ""基于噪声估计的二值掩蔽语音增强算法"", 《计算机工程与应用》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106782587B (en) | 2020-04-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107393542B (en) | Bird species identification method based on two-channel neural network | |
| CN102282867B (en) | Hearing aid and a method of detecting and attenuating transients | |
| DE60108401T2 (en) | SYSTEM FOR INCREASING LANGUAGE QUALITY | |
| KR101676211B1 (en) | Reduction of transient sounds in hearing implants | |
| CN106340303B (en) | A kind of voice de-noising method based on temporal frequency domain | |
| CN106504765B (en) | A kind of auto gain control method and device of audio signal | |
| CN102984634A (en) | Digital hearing-aid unequal-width sub-band automatic gain control method | |
| CN106448712B (en) | A kind of auto gain control method and device of audio signal | |
| CN110265065B (en) | Method for constructing voice endpoint detection model and voice endpoint detection system | |
| DE112011105908B4 (en) | Method and device for adaptive control of the sound effect | |
| CN110248300B (en) | Howling suppression method based on autonomous learning and sound amplification system | |
| CN112242147A (en) | Voice gain control method and computer storage medium | |
| CN103812462B (en) | Volume control method and device | |
| CN107274913A (en) | A kind of sound identification method and device | |
| CN106409309A (en) | Tone quality enhancement method and microphone | |
| CN110012331A (en) | A kind of far field diamylose far field audio recognition method of infrared triggering | |
| CN102610232B (en) | Method for adjusting self-adaptive audio sensing loudness | |
| CN106448690A (en) | Automatic gain control method and apparatus of audio signals | |
| CN106782592A (en) | A kind of echo and the system and method uttered long and high-pitched sounds for eliminating network sound transmission | |
| US7646912B2 (en) | Method and device for ascertaining feature vectors from a signal | |
| CN106782587A (en) | Sound mask device and sound mask method | |
| CN1355916A (en) | Signal noise reduction by time-domain spectral substraction | |
| CN110010150A (en) | Auditory Perception speech characteristic parameter extracting method based on multiresolution | |
| CN112564655A (en) | Audio signal gain control method, device, equipment and storage medium | |
| US20060087380A1 (en) | Method for limiting the dynamic range of audio signals, and circuit arrangement for this purpose |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |