CN107463956A - A kind of method and device of the heart and lung sounds separation based on Non-negative Matrix Factorization - Google Patents
A kind of method and device of the heart and lung sounds separation based on Non-negative Matrix Factorization Download PDFInfo
- Publication number
- CN107463956A CN107463956A CN201710651856.9A CN201710651856A CN107463956A CN 107463956 A CN107463956 A CN 107463956A CN 201710651856 A CN201710651856 A CN 201710651856A CN 107463956 A CN107463956 A CN 107463956A
- Authority
- CN
- China
- Prior art keywords
- matrix
- lung sounds
- cardiechema signals
- negative
- heart
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2134—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on separation criteria, e.g. independent component analysis
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B7/00—Instruments for auscultation
- A61B7/02—Stethoscopes
- A61B7/04—Electric stethoscopes
Abstract
The invention discloses a kind of method of the heart and lung sounds separation based on Non-negative Matrix Factorization, first time domain heart and lung sounds mixed signal can be filtered, to obtain priori cardiechema signals and priori Lung Sounds, and Short Time Fourier Transform and sparse Non-negative Matrix Factorization are carried out to the priori cardiechema signals and priori Lung Sounds, finally give the basic matrix of mixed signal, because the process of the sparse Non-negative Matrix Factorization is the equal of the learning process of basic matrix, effect with cluster itself, effectively basic matrix can be excavated from priori cardiechema signals and priori Lung Sounds, and cardiechema signals and Lung Sounds are accurately separated out from time domain heart and lung sounds mixed signal according to the basic matrix in subsequent step;The invention also discloses a kind of device of the heart and lung sounds separation based on Non-negative Matrix Factorization, equally with above-mentioned beneficial effect.
Description
Technical field
The present invention relates to audio separation field, more particularly to a kind of method of the heart and lung sounds separation based on Non-negative Matrix Factorization
And device.
Background technology
Due to the horizontal continuous improvement of the modern life, the health problem of people increasingly highlights, the heart such as its cardiovascular disease
The morbidity and mortality more and more higher of dirty organ lesion, at present cardiovascular death account for the head of the total cause of death of urban and rural residents
Position.At the same time, the PUD D such as increasingly sharpening recently as atmosphere pollution, asthma, chronic obstructive pulmonary disease, pneumonia
The incidence of disease is also increasingly increasing.The health problem of mankind's cardiorespiratory system has been subjected to the concern of current medical field, to cardiorespiratory system
Accurately diagnosis is to ensure the effective means of patient's rehabilitation as early as possible.
Heart sound, that is, the regular movements of human heart organ is the sound sent, wherein including the life seen on heart, blood vessel
Reason and pathological information.Heart sound also characterizes the partial information of whole body of human body circulation and cardiopulmonary circulatory system.Up to now, to heart sound
Auscultation appoint so be detect heart organ disease basic diagnosis and treatment mode.Lungs sound, that is, people's respiratory system in ventilating action
Caused sound, wherein including information that is a large amount of and changing lung's triumph and pathology.At present, the main of respiratory disease is examined
Treatment means solve again by auscultation lungs sound.
In reality, the cardiechema signals and Lung Sounds that collect are typically both mixed signals, due to human heart
Position and lung be positioned relatively close to, can be disturbed when listening to cardiechema signals by Lung Sounds, and listen to Lung Sounds
When can receive the interference of cardiechema signals again.Therefore cardiechema signals and Lung Sounds separation effectively had into pole from mixed signal
Its important use value and clinical meaning.
Realize easy because Non-negative Matrix Factorization has, take that memory space is few, it is excellent that the result of decomposition occurs without negative value etc.
Point, in the prior art, typically heart and lung sounds mixed signal is separated from the method for Non-negative Matrix Factorization.In existing skill
The either semi-supervised non-negative matrix factorization method of the non-negative matrix factorization method supervised entirely is typically used in art.Above-mentioned full prison
Superintend and direct semi-supervised non-negative matrix factorization method needs the elder generation of very pure heart sound either lungs sound in implementation process
Knowledge is tested, in application process clinically, there is very high dependence to manual operation.
The content of the invention
In view of this, it is a primary object of the present invention to provide a kind of side of the heart and lung sounds separation based on Non-negative Matrix Factorization
Method, it can be separated in the case of without very pure cardiechema signals either Lung Sounds from heart and lung sounds mixed signal
Go out cardiechema signals and Lung Sounds;Another object of the present invention is to provide a kind of heart and lung sounds separation based on Non-negative Matrix Factorization
Device, can effectively reduce when separating heart and lung sounds mixed signal to manually-operated dependence.
In order to solve the above-mentioned technical problem, the invention provides a kind of side of the heart and lung sounds separation based on Non-negative Matrix Factorization
Method, methods described include:
Priori cardiechema signals and priori Lung Sounds are isolated from the time domain heart and lung sounds mixed signal got;
Short Time Fourier Transform and sparse nonnegative matrix are carried out to the priori cardiechema signals and the priori Lung Sounds
Decompose, obtain the basic matrix of time domain heart and lung sounds mixed signal;
Non-negative square is carried out to the time domain heart and lung sounds mixed signal according to the basic matrix of the time domain heart and lung sounds mixed signal
Battle array is decomposed, and reconstructs cardiechema signals and Lung Sounds.
Optionally, the time domain heart and lung sounds mixed signal got carries out Short Time Fourier Transform and sparse nonnegative matrix
Decompose, obtaining the basic matrix of time domain heart and lung sounds mixed signal includes:
Short Time Fourier Transform is carried out to the priori cardiechema signals, to obtain the phasing matrix of cardiechema signals and multiple hearts
The non-negative magnitude matrix of sound signal, and Short Time Fourier Transform is carried out to the priori Lung Sounds, to obtain Lung Sounds
The non-negative magnitude matrix of phasing matrix and multiple Lung Sounds;
Sparse Non-negative Matrix Factorization is carried out to the non-negative magnitude matrix of multiple cardiechema signals, obtains the base of cardiechema signals
Matrix, and sparse Non-negative Matrix Factorization is carried out to the non-negative magnitude matrix of multiple Lung Sounds, obtain the base of Lung Sounds
Matrix, and the basic matrix of the basic matrix of the cardiechema signals and the Lung Sounds is combined into time domain heart and lung sounds mixed signal
Basic matrix.
Optionally, the basic matrix according to the time domain heart and lung sounds mixed signal is to the time domain heart and lung sounds mixed signal
Non-negative Matrix Factorization is carried out, and reconstruct cardiechema signals and Lung Sounds to include:
Non-negative square is carried out to the time domain heart and lung sounds mixed signal according to the basic matrix of the time domain heart and lung sounds mixed signal
Battle array is decomposed, and obtains the coefficient matrix of time domain heart and lung sounds mixed signal, and the basic matrix according to the cardiechema signals and the lungs sound
The basic matrix of signal looks for out the coefficient matrix and lung of cardiechema signals from the coefficient matrix of the time domain heart and lung sounds mixed signal
The coefficient matrix of sound signal;
The coefficient matrix preliminary reconstruction of the cardiechema signals is reduced into the preliminary magnitude matrix of cardiechema signals, and by described in
The coefficient matrix reconstruct of Lung Sounds is reduced into the preliminary magnitude matrix of Lung Sounds;
Preliminary magnitude matrix and the preliminary magnitude matrix into Lung Sounds to the cardiechema signals carry out mask, obtain
To the magnitude matrix of cardiechema signals and the magnitude matrix of Lung Sounds;
Cardiechema signals are reconstructed according to the phasing matrix of the magnitude matrix of the cardiechema signals and the cardiechema signals, and led to
Cross the magnitude matrix of the Lung Sounds and the phasing matrix of the Lung Sounds reconstructs Lung Sounds.
Optionally, after the non-negative magnitude matrix of cardiechema signals, the non-negative magnitude matrix of Lung Sounds is obtained, the side
Method further comprises:
The non-negative magnitude matrix to the cardiechema signals and the non-negative magnitude matrix of the Lung Sounds carry out denoising respectively.
Optionally, the non-negative amplitude square of the non-negative magnitude matrix to the cardiechema signals respectively and the Lung Sounds
Battle array, which carries out denoising, to be included:
The average value of cardiechema signals is calculated according to the non-negative magnitude matrix of the cardiechema signals, and believed according to the lungs sound
Number non-negative magnitude matrix calculate the average value of Lung Sounds;
The advance of average value in the non-negative magnitude matrix less than the cardiechema signals less than the cardiechema signals is removed to set
The value of fixed multiple, and remove the average value for being less than the Lung Sounds in the non-negative magnitude matrix less than the Lung Sounds
The value of multiple set in advance.
Optionally, it is described that the heart is reconstructed according to the magnitude matrix of the cardiechema signals and the phasing matrix of the cardiechema signals
Sound signal, and Lung Sounds bag is reconstructed by the magnitude matrix of the Lung Sounds and the phasing matrix of the Lung Sounds
Include:
Fourier's inversion in short-term is passed through according to the phasing matrix of the magnitude matrix of the cardiechema signals and the cardiechema signals
Get the cardiechema signals in return, and by the magnitude matrix of the Lung Sounds and the phasing matrix of the Lung Sounds through too short
When inverse Fourier transform obtain the Lung Sounds.
Present invention also offers a kind of device of the heart and lung sounds separation based on Non-negative Matrix Factorization, described device includes:
Crude separation module:For isolating priori cardiechema signals and priori from the time domain heart and lung sounds mixed signal got
Lung Sounds;
Training module:For carrying out Short Time Fourier Transform and sparse non-negative to the time domain heart and lung sounds mixed signal that gets
Matrix decomposition, obtain the basic matrix of time domain heart and lung sounds mixed signal;
Decomposing module:The time domain cardiopulmonary mixture of tones is believed for the basic matrix according to the time domain heart and lung sounds mixed signal
Number Non-negative Matrix Factorization is carried out, and reconstruct cardiechema signals and Lung Sounds.
Optionally, the training module includes:
Short Time Fourier Transform unit:For carrying out Short Time Fourier Transform to the priori cardiechema signals, to obtain the heart
The non-negative magnitude matrix of the phasing matrix of sound signal and multiple cardiechema signals, and the priori Lung Sounds are carried out in Fu in short-term
Leaf transformation, to obtain the non-negative magnitude matrix of the phasing matrix of Lung Sounds and multiple Lung Sounds;
Sparse Non-negative Matrix Factorization unit:It is sparse non-negative for being carried out to the non-negative magnitude matrix of multiple cardiechema signals
Matrix decomposition, the basic matrix of cardiechema signals is obtained, and the non-negative magnitude matrix progress to multiple Lung Sounds is sparse non-negative
Matrix decomposition, obtains the basic matrix of Lung Sounds, and by the basic matrix of the cardiechema signals and the basic matrix of the Lung Sounds
It is combined into the basic matrix of time domain heart and lung sounds mixed signal.
Optionally, the decomposing module includes:
Non-negative Matrix Factorization unit:For the basic matrix according to the time domain heart and lung sounds mixed signal to the time domain cardiopulmonary
Mixture of tones signal carries out Non-negative Matrix Factorization, obtains the coefficient matrix of time domain heart and lung sounds mixed signal, and believe according to the heart sound
Number basic matrix and the basic matrixs of the Lung Sounds look for out the heart from the coefficient matrix of the time domain heart and lung sounds mixed signal
The coefficient matrix of sound signal and the coefficient matrix of Lung Sounds;
Preliminary reconstruction unit:For the coefficient matrix preliminary reconstruction of the cardiechema signals to be reduced into the preliminary of cardiechema signals
Magnitude matrix, and the coefficient matrix of the Lung Sounds is reconstructed to the preliminary magnitude matrix for being reduced into Lung Sounds;
Masking unit:For the preliminary magnitude matrix to the cardiechema signals and the preliminary amplitude square into Lung Sounds
Battle array carry out mask, obtains the magnitude matrix of cardiechema signals and the magnitude matrix of Lung Sounds;
Reconfiguration unit:Phasing matrix for the magnitude matrix according to the cardiechema signals and the cardiechema signals reconstructs
Cardiechema signals, and Lung Sounds are reconstructed by the magnitude matrix of the Lung Sounds and the phasing matrix of the Lung Sounds.
Optionally, after Short Time Fourier Transform unit, described device further comprises:
Denoising unit:For the non-negative magnitude matrix to the cardiechema signals respectively and the non-negative amplitude of the Lung Sounds
Matrix carries out denoising.
A kind of method of heart and lung sounds separation based on Non-negative Matrix Factorization provided by the present invention, can be first to time domain cardiopulmonary
Mixture of tones signal is filtered, to obtain priori cardiechema signals and priori Lung Sounds, and to the priori cardiechema signals and elder generation
Test Lung Sounds and carry out Short Time Fourier Transform and sparse Non-negative Matrix Factorization, finally give the basic matrix of mixed signal, due to
The process of the sparse Non-negative Matrix Factorization is the equal of the learning process of basic matrix, effect with cluster itself, can be effective
Excavate basic matrix from priori cardiechema signals and priori Lung Sounds, and in subsequent step according to the basic matrix from when
Cardiechema signals and Lung Sounds are accurately separated out in the heart and lung sounds mixed signal of domain.Present invention also offers one kind to be based on non-negative square
The device for the heart and lung sounds separation that battle array is decomposed, equally with above-mentioned beneficial effect, is no longer repeated herein.
Brief description of the drawings
, below will be to embodiment or existing for the clearer explanation embodiment of the present invention or the technical scheme of prior art
The required accompanying drawing used is briefly described in technology description, it should be apparent that, drawings in the following description are only this hair
Some bright embodiments, for those of ordinary skill in the art, on the premise of not paying creative work, can be with root
Other accompanying drawings are obtained according to these accompanying drawings.
The flow chart for the first heart and lung sounds separation method that Fig. 1 is provided by the embodiment of the present invention;
The flow chart for second of heart and lung sounds separation method that Fig. 2 is provided by the embodiment of the present invention;
Fig. 3 is the structured flowchart of heart and lung sounds separator provided in an embodiment of the present invention.
Embodiment
The core of the present invention is to provide a kind of method of the heart and lung sounds separation based on Non-negative Matrix Factorization, in prior art
In, typically use the either semi-supervised non-negative matrix factorization method of the non-negative matrix factorization method supervised entirely.In existing skill
In art, the use of non-negative matrix factorization method is to need the priori of very pure heart sound either lungs sound, that is, needs to pass through
Pure heart sound and lungs sound obtain the basic matrix of heart and lung sounds mixed signal, in the basic matrix pair according to the heart and lung sounds mixed signal
Heart and lung sounds mixed signal carries out Non-negative Matrix Factorization, finally gives cardiechema signals and Lung Sounds.But should in specific clinic
During, it is difficult to get the very pure heart sound and lungs sound of patient in advance, mixed so as to be difficult to get patient's heart and lung sounds
The basic matrix in signal is closed, from not resulting in the heart sound being finally recovered out and lungs sound is inaccurate.To obtain the pure heart in advance
Sound and lungs sound to manual operation, it is necessary to have very high dependence.
And the method for the heart and lung sounds separation provided by the present invention based on Non-negative Matrix Factorization, can be first to time domain heart and lung sounds
Mixed signal is filtered, to obtain priori cardiechema signals and priori Lung Sounds, and to the priori cardiechema signals and priori
Lung Sounds carry out Short Time Fourier Transform and sparse Non-negative Matrix Factorization, the basic matrix of mixed signal are finally given, by institute
The process for stating sparse Non-negative Matrix Factorization is the equal of the learning process of basic matrix, effect with cluster itself, can be effective
Excavate basic matrix from priori cardiechema signals and priori Lung Sounds, and in subsequent step according to the basic matrix from time domain
Cardiechema signals and Lung Sounds are accurately separated out in heart and lung sounds mixed signal.
In order that those skilled in the art more fully understand the present invention program, with reference to the accompanying drawings and detailed description
The present invention is described in further detail.Obviously, described embodiment is only part of the embodiment of the present invention, rather than
Whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art are not making creative work premise
Lower obtained every other embodiment, belongs to the scope of protection of the invention.
It refer to Fig. 1, the flow chart for the first heart and lung sounds separation method that Fig. 1 is provided by the embodiment of the present invention, the party
Method includes:
S101:Priori cardiechema signals and priori Lung Sounds are isolated from the time domain heart and lung sounds mixed signal got.
Caused cardiechema signals are typically at 1 to 400 hertz in easypro contracting activity for normal heart, and 1 to 200
In this section of hertz, the amplitude of cardiechema signals is larger, and it occupies the energy of very middle ratio.And the frequency of Lung Sounds is led to
It is often at 50 to 3000 hertz, and its Energy distribution is than more uniform, is not very big in low-frequency range energy accounting.
Therefore, in this step, LPF typically is carried out to heart and lung sounds mixed signal from low pass filter, with thick
Rough isolates priori cardiechema signals;Priori cardiechema signals and heart and lung sounds mixed signal are contrasted again afterwards, to obtain priori
Lung Sounds.And the cut-off frequency of low pass filter is usually arranged as 200 hertz, naturally it is also possible to which cut-off frequency is arranged into it
His numerical value, as long as coarse priori cardiechema signals and priori Lung Sounds can be isolated, it is not specifically limited herein.
Except from low pass filter, it is also an option that other wave filters, as long as the purpose of this step can be realized,
It is not specifically limited herein.
In embodiments of the present invention, typically it is from the type low pass filter of Chebyshev II, its passband edge frequency
200Hz, stopband cut-off frequency are 250Hz, and passband maximum attenuation is 1dB, minimum attenuation in stop band 80dB, and sampling pulse is
2000Hz。
S102:Short Time Fourier Transform and sparse nonnegative matrix point are carried out to priori cardiechema signals and priori Lung Sounds
Solution, obtains the basic matrix of time domain heart and lung sounds mixed signal.
It is that the cardiechema signals in heart and lung sounds mixed signal are adopted concretely because heart and lung sounds mixed signal has openness
The amplitude of sampling point is most of in most of scopes to be zero or is close to zero, and the larger sampled point number of amplitude is fewer, and the heart
The openness of lungs sound mixed signal can be with making its l0Norm or l1The mathematic(al) representation of Norm minimum describes.And heart and lung sounds mix
Signal is closed in the time domain generally without openness, now needs to carry out sparse transformation to heart and lung sounds mixed signal so that cardiopulmonary
Mixture of tones signal shows openness.
In this step, can be by Short Time Fourier Transform by the priori cardiechema signals shown in time domain and priori lungs sound
Signal is transformed to priori cardiechema signals and priori Lung Sounds on time-frequency domain, to allow the priori cardiechema signals and priori lung
Sound signal shows openness, then carries out sparse Non-negative Matrix Factorization to it, has obtained heart and lung sounds mixed signal in time domain
Basic matrix.
Specific operation will elaborate in the following embodiments in this relevant step, will not be repeated here.
S103:Nonnegative matrix point is carried out to time domain heart and lung sounds mixed signal according to the basic matrix of time domain heart and lung sounds mixed signal
Solution, and reconstruct cardiechema signals and Lung Sounds.
The Non-negative Matrix Factorization is particularly as being that given n × m nonnegative matrixes V is resolved into the non-negative of n × r
Matrix W and a r × m nonnegative matrix H product:V ≈ WH, wherein r < m, r < n, nonnegative matrix W is referred to as basic matrix, non-
Negative matrix H is referred to as coefficient matrix.
In this step, time domain heart and lung sounds mixed signal can be carried out according to the basic matrix of time domain heart and lung sounds mixed signal non-
Negative matrix decomposes, to obtain the coefficient matrix of time domain heart and lung sounds mixed signal, and according to the time domain heart and lung sounds mixed signal
Coefficient matrix reconstructs cardiechema signals and Lung Sounds.
Specific operation will elaborate in the following embodiments in this relevant step, will not be repeated here.
The method for a kind of heart and lung sounds separation based on Non-negative Matrix Factorization that the embodiment of the present invention is provided, when can first pair
Domain heart and lung sounds mixed signal is filtered, and to obtain priori cardiechema signals and priori Lung Sounds, and the priori heart sound is believed
Number and priori Lung Sounds carry out Short Time Fourier Transform and sparse Non-negative Matrix Factorization, finally give the group moment of mixed signal
Battle array, because the process of the sparse Non-negative Matrix Factorization is the equal of the learning process of basic matrix, effect with cluster itself can
Effectively to excavate basic matrix from priori cardiechema signals and priori Lung Sounds, and according to the group moment in subsequent step
Battle array is accurately separated out cardiechema signals and Lung Sounds from time domain heart and lung sounds mixed signal.
In the present invention, the detail about S102 and S103 will be described in detail in the following embodiments.
It refer to Fig. 2, the flow chart for second of heart and lung sounds separation method that Fig. 2 is provided by the embodiment of the present invention, the party
Method includes:
S201:Priori cardiechema signals and priori Lung Sounds are isolated from the time domain heart and lung sounds mixed signal got.
This step is identical with S101 in above-described embodiment, and details refer to above-described embodiment, no longer be repeated herein.
S202:Short Time Fourier Transform is carried out to priori cardiechema signals, to obtain the phasing matrix of cardiechema signals and multiple
The non-negative magnitude matrix of cardiechema signals, and Short Time Fourier Transform is carried out to priori Lung Sounds, to obtain the phase of Lung Sounds
The non-negative magnitude matrix of bit matrix and multiple Lung Sounds.
The Short Time Fourier Transform is a kind of mathematic(al) manipulation related to Fourier transformation, to determine time varying signal its
The frequency and phase of remaining local sine wave.Short Time Fourier Transform is mainly characterized by setting a window function, usual feelings
Also the window function is managed under condition and is referred to as window;It is assumed that stable during signal in a window function, the mobile window function is to obtain
Stationary signal in different finite time width, Fourier transformation is carried out to the stationary signal in each window function, so as to count
Calculate each power spectrum at different moments, i.e. phasing matrix and magnitude matrix.In actual mechanical process, the width phase of window function
When in the length of the signal intercepted from original signal on time span.When the changes in amplitude of original signal is very fast, window is being set
The width of appropriate reduction window function is needed during function, i.e., for non-stationary signal, when signal intensity is violent, it is desirable to window function
There is higher temporal resolution.
In embodiments of the present invention, due to the present invention, decompose is heart and lung sounds mixed signal, and its amplitude is not in negative value,
So phasing matrix and non-negative magnitude matrix can be obtained by Short Time Fourier Transform.
S203:The non-negative magnitude matrix to cardiechema signals and the non-negative magnitude matrix of Lung Sounds carry out denoising respectively.
In this step, carrying out denoising commonly used approach is:
The average value of cardiechema signals is first calculated according to the non-negative magnitude matrix of cardiechema signals, and according to the non-of Lung Sounds
Negative magnitude matrix calculates the average value of Lung Sounds;
Remove the multiple set in advance for the average value for being less than cardiechema signals in the non-negative magnitude matrix less than cardiechema signals
Value, and remove in the non-negative magnitude matrix less than Lung Sounds less than Lung Sounds average value multiple set in advance
Value.
In embodiments of the present invention, it is that will be less than the average value of cardiechema signals and regarded less than the point of the average value of Lung Sounds
For noise, the numerical value of the noise is forced into zero.In this step, it is to filter out the significant lower point of energy.
Except above-mentioned denoising method, can also be not specifically limited herein by other method denoisings, as long as can remove
Noise in the non-negative magnitude matrix of cardiechema signals and the non-negative magnitude matrix of Lung Sounds.
S204:Sparse Non-negative Matrix Factorization is carried out to the non-negative magnitude matrix of multiple cardiechema signals, obtains cardiechema signals
Basic matrix, and sparse Non-negative Matrix Factorization is carried out to the non-negative magnitude matrix of multiple Lung Sounds, obtain the group moment of Lung Sounds
Battle array, and the basic matrix of the basic matrix of cardiechema signals and Lung Sounds is combined into the basic matrix of time domain heart and lung sounds mixed signal.
In embodiments of the present invention, the sparse Non-negative Matrix Factorization uses l0The sparse nonnegative matrix of norm constraint
Decompose, according to the l0Norm constraint and original Non-negative Matrix Factorization, defining its object function isSo that the object function meets W >=0, H >=0;Wherein X is input
Data matrix, W is basic matrix, and H is coefficient matrix, and L is the element number that allow maximum is not 0, and h is that NNLS is (non-negative
Least square method) function, i is the number circulated in following specific calculating process.For l0The sparse nonnegative matrix of norm constraint
Decompose, the constraints means that data matrix X each row combine expression by L base vector to greatest extent.When group moment
When battle array W is as feature, it is meant that each data sample is represented by L combinations of features to greatest extent.
l0The specific steps of the sparse Non-negative Matrix Factorization of norm constraint include:
Random initializtion basic matrix W;
Sparse coding is carried out to data matrix X with basic matrix W, to generate coefficient matrix H;
The sparsity structure of retention coefficient matrix, strengthen basic matrix W and coefficient matrix H.
In above-mentioned steps, the random initializtion basic matrix W is each numerical value in random generation basic matrix W.
Because without containing some old factor elements more in the multiplication more new formula of the Non-negative Matrix Factorization of any constraint
New rule, so above-mentioned enhancing basic matrix W and coefficient matrix H specific method is:
In above-mentioned steps after random initializtion basic matrix W, iteration is understood about sparse coding and the step of strengthening matrix
Repeatedly, obtaining the basic matrix of the basic matrix of cardiechema signals and Lung Sounds.
It is of course also possible to the basic matrix of cardiechema signals is obtained by the sparse Non-negative Matrix Factorization with other constraintss
With the basic matrix of Lung Sounds, it is not specifically limited herein.
In this step, the basic matrix by the basic matrix of cardiechema signals and Lung Sounds is combined into time domain heart and lung sounds and mixed
The concrete mode for closing the basic matrix of signal is typically that two basic matrixs are laterally disposed, to be combined into a time domain cardiopulmonary mixture of tones
The basic matrix of signal.Other combinations can certainly be selected, for example, it is placed longitudinally etc., it is not specifically limited herein.
S205:Nonnegative matrix point is carried out to time domain heart and lung sounds mixed signal according to the basic matrix of time domain heart and lung sounds mixed signal
Solution, obtains the coefficient matrix of time domain heart and lung sounds mixed signal, and according to the basic matrix of cardiechema signals and the basic matrix of Lung Sounds
The coefficient matrix of cardiechema signals and the coefficient matrix of Lung Sounds are looked for out from the coefficient matrix of time domain heart and lung sounds mixed signal.
In this step, in the coefficient matrix of the time domain heart and lung sounds mixed signal cardiechema signals coefficient matrix and lungs sound
The position relationship of the coefficient matrix of signal and cardiechema signals in the basic matrix of time domain heart and lung sounds mixed signal described in S204
The position relationship of the basic matrix of basic matrix and Lung Sounds is corresponding.If such as the basic matrix of cardiechema signals and the base of Lung Sounds
Matrix is into cross direction profiles, then the coefficient matrix of cardiechema signals and the coefficient matrix of Lung Sounds are into genesis analysis;If cardiechema signals
Basic matrix and Lung Sounds the distribution of basic matrix paired linea angulata, then the coefficient square of the coefficient matrix of cardiechema signals and Lung Sounds
The same linea angulata distribution in pairs of battle array.
S206:The coefficient matrix preliminary reconstruction of cardiechema signals is reduced into the preliminary magnitude matrix of cardiechema signals, and by lung
The coefficient matrix reconstruct of sound signal is reduced into the preliminary magnitude matrix of Lung Sounds.
In this step, it is to pass through reSc'=Wc×Hc, and reSr'=Wr×HrThe first of cardiechema signals is calculated
Walk the preliminary magnitude matrix of magnitude matrix and Lung Sounds.Wherein reSc' be cardiechema signals preliminary magnitude matrix, WcFor the heart
The basic matrix of sound signal, HcFor the coefficient matrix of cardiechema signals;reSr' be Lung Sounds preliminary magnitude matrix, WrBelieve for lungs sound
Number basic matrix, HrFor the coefficient matrix of Lung Sounds.
S207:Preliminary magnitude matrix to cardiechema signals and the preliminary magnitude matrix progress mask into Lung Sounds, are obtained
The magnitude matrix of cardiechema signals and the magnitude matrix of Lung Sounds.
In embodiments of the present invention, heart and lung sounds mixed signal is sparse, i.e., on a small time-frequency domain, only one
Source signal is occupied an leading position, otherwise the source signal is cardiechema signals, otherwise it is Lung Sounds.
In this step, mask code matrix M is definedc,iAnd Mr,i, wherein Mc,iFor the mask code matrix of cardiechema signals, Mr,iLungs sound is believed
Number mask code matrix:
Wherein reSc' for the preliminary magnitude matrix of cardiechema signals that is obtained in S206, reSr' obtained in S206
The preliminary magnitude matrix of Lung Sounds;I represents i-th of element in each matrix.Work as reSc' in corresponding element ratio reSr' in
When corresponding element is big, in mask code matrix Mc,iIn corresponding element be set to 1, be otherwise 0;In mask code matrix Mr,iIn corresponding member
Element is set to 0, is otherwise 1.
S208:Cardiechema signals are reconstructed according to the phasing matrix of the magnitude matrix of cardiechema signals and cardiechema signals, and passed through
The magnitude matrix of Lung Sounds and the phasing matrix of Lung Sounds reconstruct Lung Sounds.
In this step, according to obtained cardiechema signals and the magnitude matrix of Lung Sounds, believe with reference to priori heart sound
Number and priori Lung Sounds phasing matrix, by inverse Fourier transform in short-term, the cardiechema signals and Lung Sounds drawn.
The method for a kind of heart and lung sounds separation based on Non-negative Matrix Factorization that the embodiment of the present invention is provided, when can first pair
Domain heart and lung sounds mixed signal is filtered, and to obtain priori cardiechema signals and priori Lung Sounds, and the priori heart sound is believed
Number and priori Lung Sounds carry out Short Time Fourier Transform and sparse Non-negative Matrix Factorization, finally give the group moment of mixed signal
Battle array, because the process of the sparse Non-negative Matrix Factorization is the equal of the learning process of basic matrix, effect with cluster itself can
Effectively to excavate basic matrix from priori cardiechema signals and priori Lung Sounds, and according to the group moment in subsequent step
Battle array is accurately separated out cardiechema signals and Lung Sounds from time domain heart and lung sounds mixed signal;And by appropriate denoising process,
So that final result is more accurate.
A kind of heart and lung sounds separator provided in an embodiment of the present invention is introduced below, cardiopulmonary cent described below
Can be mutually to should refer to from device and above-described heart and lung sounds separation method.
Fig. 3 be heart and lung sounds separator provided in an embodiment of the present invention structured flowchart, the heart and lung sounds separator of reference picture 3
It can include:
Crude separation module 100:For isolated from the time domain heart and lung sounds mixed signal got priori cardiechema signals and
Priori Lung Sounds.
Training module 200:For carrying out Short Time Fourier Transform and sparse to the time domain heart and lung sounds mixed signal that gets
Non-negative Matrix Factorization, obtain the basic matrix of time domain heart and lung sounds mixed signal.
Decomposing module 300:The time domain heart and lung sounds are mixed for the basic matrix according to the time domain heart and lung sounds mixed signal
Close signal and carry out Non-negative Matrix Factorization, and reconstruct cardiechema signals and Lung Sounds.
In embodiments of the present invention, the training module 200 can include:
Short Time Fourier Transform unit 201:For carrying out Short Time Fourier Transform to the priori cardiechema signals, to obtain
The non-negative magnitude matrix of the phasing matrix of cardiechema signals and multiple cardiechema signals, and Fu in short-term is carried out to the priori Lung Sounds
In leaf transformation, to obtain the non-negative magnitude matrix of the phasing matrix of Lung Sounds and multiple Lung Sounds.
Sparse Non-negative Matrix Factorization unit 202:It is sparse for being carried out to the non-negative magnitude matrix of multiple cardiechema signals
Non-negative Matrix Factorization, the basic matrix of cardiechema signals is obtained, and the non-negative magnitude matrix progress to multiple Lung Sounds is sparse
Non-negative Matrix Factorization, obtains the basic matrix of Lung Sounds, and by the basic matrix of the cardiechema signals and the base of the Lung Sounds
Matrix is combined into the basic matrix of time domain heart and lung sounds mixed signal.
In embodiments of the present invention, the decomposing module 300 can include:
Non-negative Matrix Factorization unit 301:For according to the basic matrix of the time domain heart and lung sounds mixed signal to the time domain
Heart and lung sounds mixed signal carries out Non-negative Matrix Factorization, obtains the coefficient matrix of time domain heart and lung sounds mixed signal, and according to the heart
The basic matrix of the basic matrix of sound signal and the Lung Sounds is looked for from the coefficient matrix of the time domain heart and lung sounds mixed signal
Go out the coefficient matrix of cardiechema signals and the coefficient matrix of Lung Sounds.
Preliminary reconstruction unit 302:For the coefficient matrix preliminary reconstruction of the cardiechema signals to be reduced into cardiechema signals
Preliminary magnitude matrix, and the coefficient matrix of the Lung Sounds is reconstructed to the preliminary magnitude matrix for being reduced into Lung Sounds.
Masking unit 303:For the preliminary magnitude matrix to the cardiechema signals and the first stride into Lung Sounds
Spend matrix and carry out mask, obtain the magnitude matrix of cardiechema signals and the magnitude matrix of Lung Sounds.
Reconfiguration unit 304:For the magnitude matrix according to the cardiechema signals and the phasing matrix weight of the cardiechema signals
Structure goes out cardiechema signals, and reconstructs lungs sound by the magnitude matrix of the Lung Sounds and the phasing matrix of the Lung Sounds and believe
Number.
In embodiments of the present invention, after Short Time Fourier Transform unit 201, the training module 200 can enter one
Step includes:
Denoising unit 203:For the non-negative of the non-negative magnitude matrix to the cardiechema signals respectively and the Lung Sounds
Magnitude matrix carries out denoising.
The heart and lung sounds separator of the present embodiment is used to realize foregoing heart and lung sounds separation method, therefore heart and lung sounds separation dress
The embodiment part of the visible heart and lung sounds separation method hereinbefore of embodiment in putting, for example, crude separation module 100,
Training module 200, decomposing module 300, it is respectively used to realize step S101, S102 and S103 institutes in above-mentioned heart and lung sounds separation method
So that its embodiment is referred to the description of corresponding various pieces embodiment, will not be repeated here.
Each embodiment is described by the way of progressive in this specification, what each embodiment stressed be with it is other
The difference of embodiment, between each embodiment same or similar part mutually referring to.For dress disclosed in embodiment
For putting, because it is corresponded to the method disclosed in Example, so description is fairly simple, related part is referring to method part
Explanation.
Professional further appreciates that, with reference to the unit of each example of the embodiments described herein description
And algorithm steps, can be realized with electronic hardware, computer software or the combination of the two, in order to clearly demonstrate hardware and
The interchangeability of software, the composition and step of each example are generally described according to function in the above description.These
Function is performed with hardware or software mode actually, application-specific and design constraint depending on technical scheme.Specialty
Technical staff can realize described function using distinct methods to each specific application, but this realization should not
Think beyond the scope of this invention.
Directly it can be held with reference to the step of method or algorithm that the embodiments described herein describes with hardware, processor
Capable software module, or the two combination are implemented.Software module can be placed in random access memory (RAM), internal memory, read-only deposit
Reservoir (ROM), electrically programmable ROM, electrically erasable ROM, register, hard disk, moveable magnetic disc, CD-ROM or technology
In any other form of storage medium well known in field.
The method and device separated above to a kind of heart and lung sounds based on Non-negative Matrix Factorization provided by the present invention is entered
Go and be discussed in detail.Specific case used herein is set forth to the principle and embodiment of the present invention, and the above is implemented
The explanation of example is only intended to help the method and its core concept for understanding the present invention.It should be pointed out that for the general of the art
For logical technical staff, under the premise without departing from the principles of the invention, some improvement and modification can also be carried out to the present invention, this
A little improvement and modification are also fallen into the protection domain of the claims in the present invention.
Claims (10)
- A kind of 1. method of the heart and lung sounds separation based on Non-negative Matrix Factorization, it is characterised in that methods described includes:Priori cardiechema signals and priori Lung Sounds are isolated from the time domain heart and lung sounds mixed signal got;Short Time Fourier Transform and sparse Non-negative Matrix Factorization are carried out to the priori cardiechema signals and the priori Lung Sounds, Obtain the basic matrix of time domain heart and lung sounds mixed signal;Nonnegative matrix point is carried out to the time domain heart and lung sounds mixed signal according to the basic matrix of the time domain heart and lung sounds mixed signal Solution, and reconstruct cardiechema signals and Lung Sounds.
- 2. according to the method for claim 1, it is characterised in that described to the priori cardiechema signals and the priori lungs sound Signal carries out Short Time Fourier Transform and sparse Non-negative Matrix Factorization, and obtaining the basic matrix of time domain heart and lung sounds mixed signal includes:Short Time Fourier Transform is carried out to the priori cardiechema signals, believed with obtaining the phasing matrix of cardiechema signals and multiple heart sound Number non-negative magnitude matrix, and to the priori Lung Sounds carry out Short Time Fourier Transform, to obtain the phase of Lung Sounds The non-negative magnitude matrix of matrix and multiple Lung Sounds;Sparse Non-negative Matrix Factorization is carried out to the non-negative magnitude matrix of multiple cardiechema signals, obtains the group moment of cardiechema signals Battle array, and sparse Non-negative Matrix Factorization is carried out to the non-negative magnitude matrix of multiple Lung Sounds, obtain the group moment of Lung Sounds Battle array, and the basic matrix of the basic matrix of the cardiechema signals and the Lung Sounds is combined into the base of time domain heart and lung sounds mixed signal Matrix.
- 3. according to the method for claim 2, it is characterised in that the group moment according to the time domain heart and lung sounds mixed signal Battle array carries out Non-negative Matrix Factorization to the time domain heart and lung sounds mixed signal, and reconstructs cardiechema signals and Lung Sounds and include:Nonnegative matrix point is carried out to the time domain heart and lung sounds mixed signal according to the basic matrix of the time domain heart and lung sounds mixed signal Solution, obtains the coefficient matrix of time domain heart and lung sounds mixed signal, and the basic matrix according to the cardiechema signals and the Lung Sounds Basic matrix looked for out from the coefficient matrix of the time domain heart and lung sounds mixed signal cardiechema signals coefficient matrix and lungs sound letter Number coefficient matrix;The coefficient matrix preliminary reconstruction of the cardiechema signals is reduced into the preliminary magnitude matrix of cardiechema signals, and by the lungs sound The coefficient matrix reconstruct of signal is reduced into the preliminary magnitude matrix of Lung Sounds;Preliminary magnitude matrix and the preliminary magnitude matrix into Lung Sounds to the cardiechema signals carry out mask, obtain the heart The magnitude matrix of sound signal and the magnitude matrix of Lung Sounds;Cardiechema signals are reconstructed according to the phasing matrix of the magnitude matrix of the cardiechema signals and the cardiechema signals, and pass through institute State the magnitude matrix of Lung Sounds and the phasing matrix of the Lung Sounds reconstructs Lung Sounds.
- 4. according to the method for claim 2, it is characterised in that obtaining non-negative magnitude matrix, the lungs sound letter of cardiechema signals Number non-negative magnitude matrix after, methods described further comprises:The non-negative magnitude matrix to the cardiechema signals and the non-negative magnitude matrix of the Lung Sounds carry out denoising respectively.
- 5. according to the method for claim 4, it is characterised in that described respectively to the non-negative magnitude matrix of the cardiechema signals Carrying out denoising with the non-negative magnitude matrix of the Lung Sounds includes:The average value of cardiechema signals is calculated according to the non-negative magnitude matrix of the cardiechema signals, and according to the Lung Sounds Non-negative magnitude matrix calculates the average value of Lung Sounds;Removal is preset again less than the average value for being less than the cardiechema signals in the non-negative magnitude matrix of the cardiechema signals Several values, and the advance of average value removed in the non-negative magnitude matrix less than the Lung Sounds less than the Lung Sounds sets Determine the value of multiple.
- 6. according to the method for claim 3, it is characterised in that the magnitude matrix according to the cardiechema signals and described The phasing matrix of cardiechema signals reconstructs cardiechema signals, and passes through the magnitude matrix of the Lung Sounds and the Lung Sounds Phasing matrix, which reconstructs Lung Sounds, to be included:Obtained according to the phasing matrix of the magnitude matrix of the cardiechema signals and the cardiechema signals by inverse Fourier transform in short-term To the cardiechema signals, and by the magnitude matrix of the Lung Sounds and the phasing matrix of the Lung Sounds by Fu in short-term In leaf inverse transformation obtain the Lung Sounds.
- 7. a kind of device of the heart and lung sounds separation based on Non-negative Matrix Factorization, it is characterised in that described device includes:Crude separation module:For isolating priori cardiechema signals and priori lungs sound from the time domain heart and lung sounds mixed signal got Signal;Training module:For carrying out Short Time Fourier Transform and sparse nonnegative matrix to the time domain heart and lung sounds mixed signal got Decompose, obtain the basic matrix of time domain heart and lung sounds mixed signal;Decomposing module:The time domain heart and lung sounds mixed signal is entered for the basic matrix according to the time domain heart and lung sounds mixed signal Row Non-negative Matrix Factorization, and reconstruct cardiechema signals and Lung Sounds.
- 8. according to the method for claim 7, it is characterised in that the training module includes:Short Time Fourier Transform unit:For carrying out Short Time Fourier Transform to the priori cardiechema signals, to obtain heart sound letter Number phasing matrix and multiple cardiechema signals non-negative magnitude matrix, and to the priori Lung Sounds carry out in short-term Fourier become Change, to obtain the non-negative magnitude matrix of the phasing matrix of Lung Sounds and multiple Lung Sounds;Sparse Non-negative Matrix Factorization unit:For carrying out sparse nonnegative matrix to the non-negative magnitude matrix of multiple cardiechema signals Decompose, obtain the basic matrix of cardiechema signals, and sparse nonnegative matrix is carried out to the non-negative magnitude matrix of multiple Lung Sounds Decompose, obtain the basic matrix of Lung Sounds, and the basic matrix of the basic matrix of the cardiechema signals and the Lung Sounds is combined Into the basic matrix of time domain heart and lung sounds mixed signal.
- 9. device according to claim 8, it is characterised in that the decomposing module includes:Non-negative Matrix Factorization unit:The time domain heart and lung sounds are mixed for the basic matrix according to the time domain heart and lung sounds mixed signal Close signal and carry out Non-negative Matrix Factorization, obtain the coefficient matrix of time domain heart and lung sounds mixed signal, and according to the cardiechema signals The basic matrix of basic matrix and the Lung Sounds looks for out heart sound letter from the coefficient matrix of the time domain heart and lung sounds mixed signal Number coefficient matrix and Lung Sounds coefficient matrix;Preliminary reconstruction unit:For the coefficient matrix preliminary reconstruction of the cardiechema signals to be reduced into the preliminary amplitude of cardiechema signals Matrix, and the coefficient matrix of the Lung Sounds is reconstructed to the preliminary magnitude matrix for being reduced into Lung Sounds;Masking unit:Enter for the preliminary magnitude matrix to the cardiechema signals and the preliminary magnitude matrix into Lung Sounds Row mask, obtain the magnitude matrix of cardiechema signals and the magnitude matrix of Lung Sounds;Reconfiguration unit:Phasing matrix for the magnitude matrix according to the cardiechema signals and the cardiechema signals reconstructs heart sound Signal, and Lung Sounds are reconstructed by the magnitude matrix of the Lung Sounds and the phasing matrix of the Lung Sounds.
- 10. device according to claim 8, it is characterised in that described device further comprises:Denoising unit:For after Short Time Fourier Transform unit, respectively the non-negative magnitude matrix to the cardiechema signals and The non-negative magnitude matrix of the Lung Sounds carries out denoising.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710651856.9A CN107463956B (en) | 2017-08-02 | 2017-08-02 | Cardiopulmonary sound separation method and device based on non-negative matrix decomposition |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710651856.9A CN107463956B (en) | 2017-08-02 | 2017-08-02 | Cardiopulmonary sound separation method and device based on non-negative matrix decomposition |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107463956A true CN107463956A (en) | 2017-12-12 |
CN107463956B CN107463956B (en) | 2020-07-03 |
Family
ID=60548127
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710651856.9A Active CN107463956B (en) | 2017-08-02 | 2017-08-02 | Cardiopulmonary sound separation method and device based on non-negative matrix decomposition |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107463956B (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108133200A (en) * | 2018-01-16 | 2018-06-08 | 广东工业大学 | A kind of heart and lung sounds separation method and system based on autoregression regularization NMF |
CN108764184A (en) * | 2018-06-01 | 2018-11-06 | 广东工业大学 | A kind of separation method of heart and lung sounds signal, device, equipment and storage medium |
CN109711073A (en) * | 2018-12-29 | 2019-05-03 | 广东工业大学 | A kind of heart and lung sounds aliasing signal blind source separation method based on rarefaction representation |
CN110010145A (en) * | 2019-02-28 | 2019-07-12 | 广东工业大学 | A method of eliminating electronic auscultation device grating |
CN111863014A (en) * | 2019-04-26 | 2020-10-30 | 北京嘀嘀无限科技发展有限公司 | Audio processing method and device, electronic equipment and readable storage medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103860167A (en) * | 2014-04-02 | 2014-06-18 | 太原理工大学 | J-wave extraction and analysis method based on non-negative matrix factorization |
CN104323772A (en) * | 2014-12-02 | 2015-02-04 | 太原理工大学 | Sparse component analysis-based atrial fibrillation signal extraction method |
CN105787946A (en) * | 2016-03-17 | 2016-07-20 | 广东工业大学 | Nonnegative blind source separation fetal heart sound analysis method and device |
US20170188869A1 (en) * | 2016-01-04 | 2017-07-06 | AventuSoft, LLC | System and method of extraction of the heart valve signals |
-
2017
- 2017-08-02 CN CN201710651856.9A patent/CN107463956B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103860167A (en) * | 2014-04-02 | 2014-06-18 | 太原理工大学 | J-wave extraction and analysis method based on non-negative matrix factorization |
CN104323772A (en) * | 2014-12-02 | 2015-02-04 | 太原理工大学 | Sparse component analysis-based atrial fibrillation signal extraction method |
US20170188869A1 (en) * | 2016-01-04 | 2017-07-06 | AventuSoft, LLC | System and method of extraction of the heart valve signals |
CN105787946A (en) * | 2016-03-17 | 2016-07-20 | 广东工业大学 | Nonnegative blind source separation fetal heart sound analysis method and device |
Non-Patent Citations (1)
Title |
---|
郭鹿鸣: "基于盲源分离的心肺音信号分离方法研究与应用", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108133200A (en) * | 2018-01-16 | 2018-06-08 | 广东工业大学 | A kind of heart and lung sounds separation method and system based on autoregression regularization NMF |
CN108764184A (en) * | 2018-06-01 | 2018-11-06 | 广东工业大学 | A kind of separation method of heart and lung sounds signal, device, equipment and storage medium |
CN108764184B (en) * | 2018-06-01 | 2022-02-15 | 广东工业大学 | Method, device and equipment for separating heart-lung sound signals and storage medium |
CN109711073A (en) * | 2018-12-29 | 2019-05-03 | 广东工业大学 | A kind of heart and lung sounds aliasing signal blind source separation method based on rarefaction representation |
CN109711073B (en) * | 2018-12-29 | 2022-11-22 | 广东工业大学 | Heart and lung sound aliasing signal blind source separation method based on sparse representation |
CN110010145A (en) * | 2019-02-28 | 2019-07-12 | 广东工业大学 | A method of eliminating electronic auscultation device grating |
CN110010145B (en) * | 2019-02-28 | 2021-05-11 | 广东工业大学 | Method for eliminating friction sound of electronic stethoscope |
CN111863014A (en) * | 2019-04-26 | 2020-10-30 | 北京嘀嘀无限科技发展有限公司 | Audio processing method and device, electronic equipment and readable storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN107463956B (en) | 2020-07-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107463956A (en) | A kind of method and device of the heart and lung sounds separation based on Non-negative Matrix Factorization | |
Guo et al. | A review of wavelet analysis and its applications: Challenges and opportunities | |
CN102697493B (en) | Method for rapidly and automatically identifying and removing ocular artifacts in electroencephalogram signal | |
CN108714026A (en) | The fine granularity electrocardiosignal sorting technique merged based on depth convolutional neural networks and on-line decision | |
CN110680308B (en) | Electrocardiosignal denoising method based on fusion of improved EMD and threshold method | |
Wang et al. | An ECG signal denoising method using conditional generative adversarial net | |
Amann et al. | Reduction of CPR artifacts in the ventricular fibrillation ECG by coherent line removal | |
CN106419898A (en) | Method removing electrocardiosignal baseline drift | |
Liu et al. | Research and implementation of heart sound denoising | |
Li et al. | Application of an EMG interference filtering method to dynamic ECGs based on an adaptive wavelet-Wiener filter and adaptive moving average filter | |
El B’charri et al. | The ECG signal compression using an efficient algorithm based on the DWT | |
Pourazad et al. | Heart sound cancellation from lung sound recordings using adaptive threshold and 2D interpolation in time-frequency domain | |
Behrman et al. | Correlation dimension of electroglottographic data from healthy and pathologic subjects | |
Werther et al. | CPR artifact removal in ventricular fibrillation ECG signals using Gabor multipliers | |
Dey et al. | Deep learning algorithms for efficient analysis of ecg signals to detect heart disorders | |
Lander et al. | Time-frequency plane Wiener filtering of the high-resolution ECG: Background and time-frequency representations | |
CN107392149A (en) | A kind of real-time blind separating method of human body heart and lung sounds and system | |
CN113229842A (en) | Heart and lung sound automatic separation method based on complex deep neural network | |
Rajan et al. | Novel Method for the Segregation of Heart Sounds from Lung Sounds to Extrapolate the Breathing Syndrome | |
Haibin et al. | Heart sound analysis based on autoregressive power spectral density | |
Reddy et al. | A tutorial review on data compression with detection of fetal heart beat from noisy ECG | |
CN105662336B (en) | A kind of signal denoising processing method and processing device | |
Potdar et al. | Removal of Heart Sound from Lung Sound using LabVIEW 8.6 | |
Balasubramanian et al. | Electrocardiogram Signal Denoising Using Optimized Adaptive Hybrid Filter with Empirical Wavelet Transform | |
Torbey et al. | Time-Scale Analysis of SignalsWithout Basis Functions: Application to Sudden Cardiac Arrest Prediction. |
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 |