A method of eliminating millimeter wave bioradar noise in voice
[technical field]
The present invention relates to make an uproar in radar field of voice signal more particularly to a kind of elimination millimeter wave bioradar voice
The method of sound.
[background technique]
As one with the speech detection method non-contact, remote, acoustic resistive noise jamming ability is strong, bioradar
Speech detections that speech detection technological break-through microphone is interfered vulnerable to acoustic noise and other need to contact with human skin
The limitation of device is gradually applied and is developed in speech detection field.Currently, although millimeter wave bioradar can have
Effect obtain 20 meters outside human body voice signal, however obtain voice signal often by electromagnetism harmonic noise, circuit noise and
Ambient noise etc. is interfered, and the presence of these noises reduces the quality of radar voice to a certain extent, or even affects language
The intelligibility of sound, therefore, how the effective noise content eliminated in radar voice exists to millimeter wave bioradar technology
Further application in speech detection field is of great significance.
[summary of the invention]
It is an object of the invention to: a kind of method for eliminating noise in millimeter wave bioradar is provided, this method is according to warp
Radar voice noise characteristic distributions after empirical mode decomposition, under the premise of not causing voice signal to be distorted and excessive damage,
Effectively remove the noise content in radar voice.For bioradar speech detection technology development provide speech de-noising in terms of
Technical support.
Effectively to realize above-mentioned purpose, the technical solution adopted by the present invention is that:
A method of eliminating millimeter wave bioradar noise in voice, comprising the following steps:
1) bioradar voice is sampled, obtains grandfather tape and makes an uproar radar voice signal;
2) to the grandfather tape of acquisition make an uproar radar voice signal carry out empirical mode decomposition be intrinsic mode function IMF;
3) noise content analysis is carried out to the intrinsic mode function after empirical mode decomposition;
4) according to radar voice noise characteristic distributions, the selection of noise component(s) number is carried out using Mutual information entropy, to distinguish
High frequency and intermediate frequency separation distinguish intermediate frequency and low frequency separation using given threshold;
5) intrinsic mode function after self-adaptive solution is reconstructed to obtain enhanced radar voice.
As a further improvement of the present invention, step 2) specifically includes the following steps:
Assuming that be x (t) to speech signal, empirical mode decomposition method by screening process it is adaptive by the original letter
Number it is decomposed into intrinsic mode function, and each intrinsic mode function IMF is a sub- frequency content of original signal, tool
The screening process of body is as follows:
2.1) all maximum of signal x (t) and minimum point are found out;
2.2) maximum and minimum point are fitted by sample interpolation three times respectively, respectively obtain maximum envelope eu(t)
With minimum envelope ed(t);
2.3) the mean value m of maximum envelope and minimum envelope is calculated1(t)=(eu(t)+ed(t))/2, and with original
Signal, which subtracts the mean value, can obtain h1(t)=x (t)-m1(t);
2.4) judge h1(t) whether meet two screening primary conditions of intrinsic mode function IMF:
(a) in all signal lengths, the zero passage points of each intrinsic mode function IMF must be with extreme point number phase
Same, or both number is not much different in 1;
(b) at any sampled point, the mean value of maximum envelope and minimum envelope is zero, that is, meets IMF about the time
Axis is geometrically symmetric;
At this point, if h1(t) meet (a) and (b) two primary conditions of intrinsic mode function IMF, then IMF1(t)=h1
(t);
If 2.5) h1(t) two primary conditions of IMF are unsatisfactory for, then make h1(t) step is repeated as new original signal
2.1), 2.2), 2.3) obtain h2(t)=h1(t)-m2(t);If h2(t) meet IMF and sieve condition then IMF substantially1(t)=h2
(t), at this time if h2(t) it is unsatisfactory for IMF screening condition and then executes stopping screening threshold value, the stopping criterion such as following formula:
Stopping screening threshold value SD value range should meet between 0.2 and 0.3;
If h2(t) meet stopping criterion, then IMF1(t)=h2(t);
If h2(t) it is unsatisfactory for, then by h2(t) step 2.1) -2.4 is repeated), get h under continuingi(t), until hi(t)
Meet two primary conditions or meet stopping criterion, then IMF1(t)=hi(t);
2.6) when getting IMF1(t) after, it is removed from original signal and obtains residual signals r1(t):r1(t)=x
(t)-IMF1(t), the original signal that residual signals are new as one repeats step 2.1) -2.5), until obtaining next residual
Difference signal, then decomposing finally, then residual signals are indicated with following formula: rn(t)=rn-1(t)-IMFn(t);
At this point, rnIt (t) is a monotonic sequence, after the completion of screening, original signal is broken down into multiple intrinsic mode function IMF1
(t)、IMF2(t)…IMFn(t) and residual sequence rn(t);N is positive integer;Therefore, original signal may be expressed as:
As a further improvement of the present invention, original radar voice signal is decomposed into radio-frequency head through empirical mode decomposition
Divide, intermediate-frequency section, low frequency part, noise concentrates on high frequency section, and intermediate-frequency section is useful primary speech signal, and low frequency is language
The detailed information of sound signal.
As a further improvement of the present invention, in step 4), bioradar voice intrinsic mode function calculates Mutual information entropy
Method is as follows:
The Mutual information entropy of two variable Xs and Y can calculate according to the following formula:
Wherein, p (x, y) is variable X and the Joint Distribution of Y, and p (x) and p (y) are marginal distribution or Mutual information entropy respectively
Following formula is expressed as with the form of entropy:
I(X;Y)=H (X)-H (X | Y) (4)
Wherein:
H (X) is the intrinsic entropy of variable X, and H (X | Y) it is the entropy of variable X under the premise of variable Y occurs.
As a further improvement of the present invention, in step 4), high frequency, intermediate frequency separation, intermediate frequency, low frequency separation according to
Lower step determines:
The mutual information entropy between adjacent IMF is calculated, according to Mutual information entropy theory, adjacent Mutual information entropy there will be following rule
Rule: first reduce and increase afterwards, then increasing:
IMF is as distinguishing the separation of high frequency and intermediate frequency when selecting information entropy minimum value according to formula (7), and distinguishes intermediate frequency
With the separation of low frequency by carrying out empirical mode decomposition to radar voice signal after, according to noise in each intrinsic mode function point
Cloth situation finally determines that fixed threshold FT is 10-1;If the amplitude of the maximum value of IMFs is less than threshold value FT, IMFs is as low
Frequency part.
As a further improvement of the present invention, in step 5), to noise-containing intrinsic mode function through adaptive threshold
Denoising is using empirical mode decomposition method to radar speech de-noising, wherein the function of adaptive threshold are as follows:
Wherein, N is the length of signal, and σ is the noise variance of each rank IMF estimation:
The threshold function table that formula (8) are used to high frequency section, the collapse threshold that intermediate-frequency section is used are as follows:
Soft threshold method is used to eliminate the noise signal of high frequency and intermediate-frequency section:
Wherein, sign is sign function, and i represents the i-th rank intrinsic mode function.
As a further improvement of the present invention, in step 5), with remaining low frequency component and the high frequency after threshold denoising
Voice signal is reconstructed with intermediate frequency component, reconstruction signal is indicated with following formula:
Wherein, k is determining high fdrequency component number, and l is the intrinsic mode number of components of determining differentiation high frequency and intermediate frequency, n
It is the intrinsic mode number after empirical modal boundary.
The present invention is allowed to the good effect having compared with existing radar speech enhancement technique due to using above-mentioned technology
It is:
The present invention is through the millimeter radar voice signal to acquisition after empirical mode decomposition, in intrinsic mode function
On the basis of noise content is analyzed, according in intrinsic mode function the characteristics of noise profile, using adaptive threshold to height
Frequently, intermediate frequency, low frequency part are denoised, so as to according to noise content number it is targeted eliminate voice in noise
Content has stronger adaptability and validity, particularly can be good at the detailed information for retaining voice, facilitates in radar
Voice quality significantly improves.Using the example of this method show it is this have targetedly empirical mode decomposition method can have
Effect eliminates the noise content in millimetre-wave radar voice, compared with traditional sound enhancement method, has stronger adaptability, and
The intelligibility of voice can be significantly improved on the basis of eliminating radar noise in voice content.Do not causing radar voice signal
Under the premise of distortion, radar noise in voice is efficiently removed.The present invention can detect human body language using bioradar to be later
Effective technical support is provided in terms of sound.Therefore, the present invention has stronger use value in terms of eliminating radar voice noise
And application prospect.
Further, high frequency, intermediate frequency and low frequency three parts are divided into the intrinsic mode function of decomposition, by observing eigen mode
Noise profile situation in state function, by radar noise in voice characteristic distributions are as follows: noise focuses primarily upon high frequency section, intermediate frequency portion
Divide predominantly useful primary speech signal, but still contain a small amount of noise signal, low frequency part is mainly that voice signal is thin
Save signal.High frequency and intermediate frequency component number are selected by Mutual information entropy on this basis, wherein high frequency section contains
Higher threshold value is arranged in much noise.And the separation for distinguishing intermediate frequency and low frequency finds out fixed threshold 10 according to experiment-1.If
The amplitude of the maximum value of IMFs is less than threshold value FT, then these IMFs are low frequency part.Higher-frequency is arranged for intermediate-frequency section
The lesser adaptive threshold in part, low frequency part do not do denoising.
[Detailed description of the invention]
Fig. 1 is a kind of method flow diagram for eliminating millimeter wave bioradar noise in voice;
Fig. 2 is one section of radar phonetic material after empirical mode decomposition, wherein (a) is noise-containing original radar language
Sound signal;(b) it is each rank intrinsic mode function after empirical mode decomposition;
Fig. 3 is the original radar voice signal of radar system acquisition, wherein when (a) is millimetre-wave radar voice signal
Domain waveform;(b) be millimetre-wave radar voice signal sound spectrograph;
Fig. 4 is that traditional algorithm spectrum-subtraction carries out the voice signal after de-noising, wherein (a) is millimetre-wave radar voice signal
Time domain waveform;(b) be millimetre-wave radar voice signal sound spectrograph;
Fig. 5 is that empirical mode decomposition described in this patent and Mutual information entropy algorithm eliminate the radar voice signal after noise,
Wherein, (a) is the time domain waveform of millimetre-wave radar voice signal;(b) be millimetre-wave radar voice signal sound spectrograph.
[specific embodiment]
With reference to the accompanying drawing, a specific embodiment of the invention is described in detail, but the present invention is not limited to the implementations
Example.In order to make the public have thorough understanding to the present invention, is preferably applied in following present invention and concrete details is described in detail in example.
Referring to Fig. 1, the present invention eliminates the basic principle of the method for millimeter wave bioradar noise in voice are as follows: first to original
Beginning radar voice signal carries out empirical mode decomposition;Noise profile in intrinsic mode function after empirical mode decomposition is carried out
It divides;It is divided using separation of the Mutual information entropy to high frequency section and intermediate-frequency section, each rank intrinsic mode function is carried out
Adaptive Wavelet Thrinkage;To treated, radar voice signal is reconstructed.
A kind of method of elimination millimeter wave bioradar noise in voice of the invention, the specific steps of which are as follows:
Firstly, sampled to bioradar voice, obtains grandfather tape and make an uproar radar voice signal;
Secondly, the original radar voice signal to acquisition carries out empirical mode decomposition;
Making an uproar to described band, steps are as follows for millimetre-wave radar voice signal progress empirical mode decomposition:
Assuming that being x (t) to speech signal, empirical mode decomposition method is original by this by " screening " process adaptive
Signal decomposition is intrinsic mode function (IMFs), and each IMF is a sub- frequency content of original signal.It is specific
Screening process is as follows:
1) all maximum of signal x (t) and minimum point are found out.
2) maximum and minimum point are fitted by sample interpolation three times respectively, respectively obtain maximum envelope eu(t) and
Minimum envelope ed(t)。
3) the mean value m of maximum envelope and minimum envelope is calculated1(t)=(eu(t)+ed(t))/2, and with original letter
H can be obtained by number subtracting the mean value1(t)=x (t)-m1(t).
4) judge h1(t) whether meet two screening primary conditions of IMF:
(a) in all signal lengths, the zero passage points of each IMF must be identical as extreme point number, or at most
Difference one.
(b) at any sampled point, the mean value of maximum envelope and minimum envelope is necessary for zero, that is to say, that IMF is
It is geometrically symmetric about time shaft.
At this point, if h1(t) meet two primary conditions of IMF, then IMF1(t)=h1(t)
If 5) h1(t) two primary conditions of IMF are unsatisfactory for, then h1(t) step 1) is repeated as new original signal,
2), 3) h is obtained2(t)=h1(t)-m2(t), if h2(t) meet IMF and sieve condition then IMF substantially1(t)=h2(t).At this time such as
Fruit h2(t) being unsatisfactory for IMF screening condition, then a new stopping screening threshold value is performed.The stopping criterion such as following formula:
Under normal circumstances, SD value range is between 0.2 and 0.3.If h2(t) meet stopping criterion, then IMF1(t)=
h2(t).If h2(t) it is unsatisfactory for, then by h2(t) step 1) -4 is repeated), get h under continuingi(t), until hi(t) meet two
A stop condition meets stopping criterion.Then IMF1(t)=hi(t)。
6) when getting IMF1(t) after, it is removed from original signal and obtains residual signals r1(t):r1(t)=x (t)-
IMF1(t) is at this point, the residual signals original signal new as one, repeats step 1-5, until obtaining next residual signals.
It so decomposes finally, then residual signals can be indicated with following formula: rn(t)=rn-1(t)-IMFn(t)。
At this point, rnIt (t) is a monotonic sequence, after the completion of screening, original signal is broken down into multiple intrinsic mode function IMF1
(t),IMF2(t),…IMFn(t) and residual sequence rn(t) therefore, original signal may be expressed as:
By original radar voice signal through empirical mode decomposition be intrinsic mode function.Fig. 2 is one section through empirical modal point
Radar phonetic material after solution.Fig. 2 (a) is noise-containing original radar voice signal.After Fig. 2 (b) is empirical mode decomposition
Each rank intrinsic mode function, from top to bottom, the frequency of the intrinsic mode function gradually decreases.Since radar voice is main
Noise source be white Gaussian noise, therefore, in general, noise can be distributed in entire intrinsic mode function, can from figure (b)
To find out, it is substantially not visible voice signal waveform, predominantly noise signal in first three rank intrinsic mode function, and from 4-9 rank
Intrinsic mode starts, mainly voice signal waveform, and there are partial noise signals.From the 10th rank single order to the end, eigen mode
The frequency of state function is low and amplitude very little, however wherein includes the detailed information of some voices, and therefore, we can make such
Assuming that: by original radar voice signal through empirical mode decomposition, it can be decomposed into high frequency section, intermediate-frequency section, low frequency part is made an uproar
Sound focuses primarily upon high frequency section, and intermediate-frequency section is mainly useful primary speech signal, and low frequency is that the details of voice signal is believed
Breath.
Again, noise content analysis is carried out to the intrinsic mode function after empirical mode decomposition.
Then, according to radar voice noise characteristic distributions, the selection of noise component(s) number is carried out, using Mutual information entropy with area
Divide high frequency and intermediate frequency separation, given threshold is utilized to distinguish intermediate frequency and low frequency separation.
It is as follows that bioradar voice intrinsic mode function calculates Mutual information entropy method:
Mutual information entropy is all nonnegative number under normal circumstances, and the Mutual information entropy of two variable Xs and Y can calculate according to the following formula:
Wherein, p (x, y) is variable X and the Joint Distribution of Y, and p (x) and p (y) are marginal distribution respectively.Or Mutual information entropy
Following formula is expressed as with the form of entropy:
I(X;Y)=H (X)-H (X | Y) (4)
Wherein:
H (X) is the intrinsic entropy of variable X, and H (X | Y) it is the entropy of variable X under the premise of variable Y occurs.
When event X is more uncertain, in general, the correlation between two variables is stronger, then mutual between variable by the bigger of entropy
Information entropy is bigger.According to the above feature, the mutual information entropy between adjacent IMF is calculated, according to Mutual information entropy theory, adjacent mutual trust
Breath entropy will have following rule: first reduces and increases afterwards, then increasing:
IMF is as distinguishing the separation of high frequency and intermediate frequency when selecting information entropy minimum value according to formula (7), and distinguishes intermediate frequency
With the separation of low frequency by carrying out empirical mode decomposition to radar voice signal after, according to noise in each intrinsic mode function point
Cloth situation finally determines that fixed threshold (FT) is 10-1;If the amplitude of the maximum value of IMFs is less than threshold value FT, IMFs is
Low frequency part.
Finally, the intrinsic mode function after self-adaptive solution is reconstructed to obtain enhanced radar voice.
To noise-containing intrinsic mode function through Adaptive Wavelet Thrinkage, its step are as follows:
Using empirical mode decomposition method in radar speech de-noising, the selection of threshold value is played a crucial role.It is logical
The generic function of normal threshold value by:
Herein, N is the length of signal, and σ is the noise variance of each rank IMF estimation:
Herein, the threshold function table of formula (8) is used to high frequency section, it is relatively small for intermediate-frequency section noisiness,
To avoid voice signal from being distorted, to the collapse threshold of intermediate-frequency section use are as follows:
Soft threshold method is used to eliminate the noise signal of high frequency and intermediate-frequency section:
Then, weight is carried out to voice signal with remaining low frequency component and the high frequency after threshold denoising and intermediate frequency component
Structure, reconstruction signal are indicated with following formula:
Wherein, k is determining high fdrequency component number, and l is the intrinsic mode number of components of determining differentiation high frequency and intermediate frequency, n
It is the intrinsic mode number after empirical modal boundary.
Generally speaking, the radar voice enhancement algorithm that the application proposes is mainly following steps,
1. original radar voice signal x (t) is sieved using empirical mode decomposition.
2. calculating the Energy-Entropy of each component by formula (5) and (6).
3. the Mutual information entropy for calculating each adjacent component presses formula (4).
4. determining the separation of high frequency and intermediate frequency component according to formula (7).
5. determining the separation of intermediate frequency and low frequency component using IMF fixed threshold.
6. being filtered out using formula (8)-(11) to the noise in high frequency and intermediate frequency voice signal.
7. voice signal is reconstructed with the height after low frequency component and formula (12) denoising not processed, intermediate frequency component.
Embodiment
In order to guarantee the consistency of sound source, the recording materials " 1-2-3-4-5-6 " of a women experimenter is selected, and by sound
Case plays at away from 5 meters of radar system.
Radar voice is enhanced with following step:
1) bioradar voice is sampled, obtains grandfather tape and makes an uproar radar voice signal;
2) empirical mode decomposition is carried out to the original radar voice signal of acquisition;
3) noise content analysis is carried out to the intrinsic mode function after empirical mode decomposition;
4) according to radar voice noise characteristic distributions, the selection of noise component(s) number is carried out using Mutual information entropy;
5) intrinsic mode function after self-adaptive solution is reconstructed to obtain enhanced radar voice.
Processing result comparison, please refers to shown in Fig. 3, Fig. 4, Fig. 5, wherein (a) is millimetre-wave radar voice signal in each figure
Time domain waveform, figure (b) be millimetre-wave radar voice signal sound spectrograph.
Fig. 2 is the original radar voice signal acquired by radar system, the either time domain from original radar voice signal
In waveform or sound spectrograph it can be seen that, voice signal by noise jamming, these noise sources mainly have electromagnetism harmonic noise,
Circuit noise and ambient noise.The presence of these noises often affects the quality of radar voice signal.
Fig. 4 is that the voice signal after de-noising is carried out by traditional algorithm spectrum-subtraction, as can be seen that voice letter from time domain waveform
Noise in number is significantly eliminated, the noise in especially silent section, from Fig. 4 (b) sound spectrograph again it can be seen that through
After spectrum-subtraction processing, the noise in silent section is significantly eliminated, but in high frequency section, some non-primary radar voice letters
Ingredient in number is introduced, referred to as " music noise ", this is acoustically often influencing voice quality.
Fig. 5 is that empirical mode decomposition described in this patent and Mutual information entropy algorithm eliminate the radar voice signal after noise.
It can be seen from the figure that the noise in radar voice signal has also obtained significantly inhibiting, although can see in Fig. 5 (b)
It arrives, in non-speech segment, however it remains part residual noise, this is that this method is to avoid voice signal distortion, and use is adaptive
Answer the result of threshold value.For the validity for further verifying this method, synchronizes and carried out Subjective audiometry experiment, experimental result table
Bright, compared with spectrum cuts algorithm, the method for proposition effectively eliminates radar voice letter under the premise of not causing voice signal to be distorted
Noise in number, improves voice quality.
Although the discussed method for eliminating the noise in millimetre-wave radar voice of the invention has specific aim, it is especially suitable for
Millimetre-wave radar voice signal, but the range that the present invention uses is not limited to millimetre-wave radar voice signal, for centimetre
Wave radar voice signal and some acoustic speech signals acquired under identical environment equally have important directive significance and
Reference value.Embodiments of the present invention are explained in detail above in conjunction with attached drawing, is not intended to limit the invention in any way, all
It is according to the technical essence of the invention to any simple modification, change and equivalent structural changes made by the above embodiment,
It still falls in the protection scope of technical solution of the present invention.