CN105249986A - Heart sound signal period parameter estimating method - Google Patents

Abstract

The invention discloses a heart sound signal period parameter estimating method, and belongs to the field of research on biological signal detection and identification, for solving the problem of poor reliability of period parameter estimation. The method comprises the following steps: 1, preprocessing heart voice signals; 2, extracting average Shannon energy envelope signals from preprocessed heart sound signals; 3, smoothing the extracted envelope signals by use of a singular value spectrum analysis method; and 4, performing estimation of period parameters of the heart sound signals on smoothed envelope signals. The method provided by the invention has the advantages of easy realization, low algorithm complexity and good reliability, thereby being suitable for popularization and application.

Description

A kind of cardiechema signals cycle parameter method of estimation

Technical field

The invention belongs to the research field of bio signal detection and indentification.

Background technology

The health status of the various physiological signal reflection people of human body.Cardiechema signals is one of of paramount importance physiological signal of human body, it is the reflection of heart and cardiovascular system mechanical movement situation, be in cardiac cycle due to myocardial contraction diastole, valve opens and closes and blood flow impacts a kind of mechanical vibration of causing such as ventricle wall and large artery trunks.Contain the important information about human health status in hear sounds, objective digitized cardiophony can be realized by extracting this information and carrying out effective identification, thus reliable diagnostic result can be provided for patient.

But cardiechema signals is cycle signal a kind of jiggly paracycle, needs to carry out segment processing to it thus obtains each cardiocirculatory cardiechema signals of reflection.Therefore, an important prerequisite of reliable recognition of heart sound technology needs to carry out segment processing accurately by cardiechema signals, and this just needs to estimate the cycle parameter in cardiechema signals exactly.Cycle parameter in hear sounds comprises: heart cycle, systole and diastole, and the inverse in its center cycle is exactly heart rate.The cycle parameter of cardiechema signals estimates it is one of Major Difficulties in hear sounds gradation study.There is the problem of poor reliability in the hear sounds cycle parameter of the cycle parameter method of estimation acquisition of existing cardiechema signals.

Summary of the invention

There is the problem of poor reliability in the hear sounds cycle parameter that the object of the invention is the cycle parameter method of estimation acquisition in order to solve existing cardiechema signals, the invention provides a kind of cardiechema signals cycle parameter method of estimation.

A kind of cardiechema signals cycle parameter method of estimation of the present invention, described method comprises the steps:

Step 1: pretreatment is carried out to cardiechema signals to be estimated;

Step 2: to the average shannon energy envelope signal of pretreated heart sound signal extraction;

Step 3: adopt singular value spectral analysis method smoothing to the envelope signal extracted;

Step 4: the cycle parameter envelope signal after level and smooth being estimated to cardiechema signals.

In described step 1, pretreated method is carried out to cardiechema signals to be estimated and comprises:

Cardiechema signals to be estimated is carried out energy normalization process, then down-sampled is 2kHz, and adopt 6 rank Butterworth strainer accepter to carry out bandpass filtering with other sound beyond filtering cut-off frequency and noise to the cardiechema signals after down-sampled, obtain pretreated cardiechema signals x (n), wherein n is the sample index of hear sounds.

In described step 2, the method for the average shannon energy envelope signal of pretreated heart sound signal extraction is comprised:

Employing length is W, the frame sliding window moved as W/2 acts on pretreated cardiechema signals x (n) and carries out framing, calculates its average shannon energy envelope to the cardiechema signals in frame window by formula (1):

$e\left(m\right)=-\frac{1}{W}\underset{n=1}{\overset{W}{\Σ}}x{\left(n\right)}^2\log x{\left(n\right)}^2$ Formula (1)

The envelope signal obtained after carrying out envelope extraction to cardiechema signals x (n) is e (m), m is frame index.

Described step 4 comprises the steps:

Step 4.1: the monolateral auto-correlation function calculating level and smooth rear envelope signal:

After level and smooth, envelope signal e (m) calculates its monolateral auto-correlation function r (l) according to formula (2):

$r\left(l\right)=\underset{m=1}{\overset{M-l}{\Σ}}e(m+l)e\left(m\right)l\≥1$ Formula (2)

Wherein: l is the amount of delaying of level and smooth rear envelope signal e (m), M is the length of level and smooth rear envelope signal e (m).

Step 4.2: choose eligible peak value and sort in auto-correlation function r (l):

Choose all K peak values in maximum hear sounds periodic regime and carry out sequence by its size and be;

R ₍₁₎> r ₍₂₎> r _(K)formula (3)

And record and delay position corresponding to each peak value, be denoted as: L=}l ₍₁₎, l ₍₂₎..., l _(K);

Step 4.3: estimate heart cycle l according to formula (4) _t:

formula (4)

Step 4.4: estimate systole l according to formula (5) _syswith diastole l _dia:

formula (5)

Wherein, l _tfor the heart cycle, i, j=1,2 ... K.

Described step 4.4, estimates systole l according to formula (5) _syswith diastole l _diadetailed process be:

Step 4.4.1: make initial distance d=+ ∞, wherein+∞ represents positive infinity;

Step 4.4.2: choose l arbitrarily from L _(i)and l _(j)and meet l _(i)and l _(j)all be less than heart cycle l _t; If the empty set of being chosen for, then perform step 4.4.5, if be chosen for nonvoid set, then perform step 4.4.3;

Step 4.4.3: if | l _i+ l _j-l _t| < d then performs l _sys=l _i, l _dia=l _j, work as l _i≤ l _jor l _sys=l _j, l _dia=l _i, work as l _i> l _j;

Step 4.4.4: remove l from L _(i)and l _(j), forward step 4.4.2 to;

Step 4.4.5: end loop, the current l obtained _sysand l _diathe systole will estimated exactly and diastole.

In described step 3, the method adopting singular value spectral analysis method smoothing to the envelope signal extracted comprises:

Be envelope signal e (m) foundation formula (6) structure track matrix (trajectorymatrix) X of M by the length of extraction:

formula (6)

Wherein: k is previously selected constant, the time span corresponding to it is made to be greater than the maximum heart cycle; Then, by formula (7), Eigenvalues Decomposition is carried out to track matrix X:

X=USV ^tformula (7)

Wherein: U, V are the matrix be made up of as column vector the left and right characteristic vector of singular value decomposition respectively, and S is the diagonal matrix be made up of k eigenvalue; The left and right characteristic vector chosen corresponding to a front q eigenvalue forms matrix U _qand V _qand the diagonal matrix S that a front q eigenvalue is formed _q, then the approximate matrix X of track matrix X is constructed _qcan be expressed as:

X _q=U _qs _qv _q ^tformula (8)

By the approximate matrix X obtained in formula (8) _qgenerate level and smooth rear envelope signal e (m), m is frame index.

Beneficial effect of the present invention is, the present invention includes the parameter estimation that heart cycle, systole and diastole three are important, thinks that cardiechema signals segmentation provides reliable foundation.The present invention has the advantages that to be easy to realization, good reliability, is therefore applicable to promoting the use of.

Accompanying drawing explanation

Fig. 1 is the curve synoptic diagram of pretreated cardiechema signals;

Fig. 2 is the curve synoptic diagram of average shannon energy envelope signal cardiechema signals;

Fig. 3 is the curve synoptic diagram of monolateral auto-correlation function, and abscissa is for delaying position, and the heart cycle is 72, and systole is 30, and diastole is 42.

Detailed description of the invention

Composition graphs 1, Fig. 2 and Fig. 3 illustrate present embodiment, and a kind of cardiechema signals cycle parameter method of estimation described in present embodiment, described method comprises the steps:

Step 1, pretreatment is carried out to cardiechema signals to be estimated:

Cardiechema signals to be estimated is carried out energy normalization process, then down-sampled is 2kHz, and adopt 6 rank Butterworth strainers to lead to 20 ~ 900Hz ripple device to carry out bandpass filtering with other sound beyond filtering cut-off frequency and noise to the cardiechema signals after down-sampled, obtain pretreated cardiechema signals x (n), wherein n is the sample index of hear sounds.The cardiechema signals obtained as shown in Figure 1.

Step 2: to the average shannon energy envelope signal of pretreated heart sound signal extraction:

Employing length is W, the frame sliding window moved as W/2 acts on pretreated cardiechema signals x (n) and carries out framing, calculates its average shannon energy envelope to the cardiechema signals in frame window by formula (1):

$e\left(m\right)=-\frac{1}{W}\underset{n=1}{\overset{W}{\&Sigma;}}x{\left(n\right)}^2\log x{\left(n\right)}^2$ Formula (1)

The envelope signal obtained after carrying out envelope extraction to cardiechema signals x (n) is e (m), m is frame index.Institute obtains average shannon energy envelope signal e (m) as shown in Figure 2.

Step 3: adopt singular value spectral analysis method smoothing to the envelope signal extracted:

Be envelope signal e (m) foundation formula (6) structure track matrix (trajectorymatrix) X of M by the length of extraction:

formula (6)

Wherein: k is previously selected constant, the time span corresponding to it is made to be greater than the maximum heart cycle 1250 milliseconds; Then, by formula (7), Eigenvalues Decomposition is carried out to track matrix X:

X=USV ^tformula (7)

Wherein: U, V are the matrix be made up of as column vector the left and right characteristic vector of singular value decomposition respectively, and S is the diagonal matrix be made up of k eigenvalue; The left and right characteristic vector chosen corresponding to a front q eigenvalue forms matrix U _qand V _qand the diagonal matrix S that a front q eigenvalue is formed _q, then the approximate matrix X of track matrix X is constructed _qcan be expressed as:

X _q=U _qs _qv _q ^tformula (8)

By the approximate matrix X obtained in formula (8) _qgenerate level and smooth rear envelope signal e (m), m is frame index.

Step 4: the cycle parameter envelope signal after level and smooth being estimated to cardiechema signals, comprises the steps:

Step 4.1: the monolateral auto-correlation function calculating level and smooth rear envelope signal:

After level and smooth, envelope signal e (m) calculates its monolateral auto-correlation function r (l) according to formula (2):

$r\left(l\right)=\underset{m=1}{\overset{M-l}{\&Sigma;}}e(m+l)e\left(m\right)l\&GreaterEqual;1$ Formula (2)

Wherein: l is the amount of delaying of level and smooth rear envelope signal e (m), M is the length of level and smooth rear envelope signal e (m).

Step 4.2: choose eligible peak value and sort in auto-correlation function r (l):

Choose all K peak values in maximum hear sounds periodic regime 250 ~ 1250 milliseconds and carry out sequence by its size and be;

R ₍₁₎> r ₍₂₎> r _(K)formula (3)

And record and delay position corresponding to each peak value, be denoted as: L={l ₍₁₎, l ₍₂₎..., l _(K);

Step 4.3: estimate heart cycle l according to formula (4) _t:

formula (4)

Step 4.4: estimate systole l according to formula (5) _syswith diastole l _dia:

formula (5)

Wherein, l _tfor the heart cycle, i, j=1,2 ... K.

According to by step 4.3 obtain heart cycle l _twith the peak value obtained in step 4.2 corresponding to delay location sets L={l ₍₁₎, l ₍₂₎..., l _(K), the specific implementation process of formula (5) comprises the following steps:

Step 4.4.1: make initial distance d=+ ∞, wherein+∞ represents positive infinity;

Step 4.4.2: choose l arbitrarily from L _(i)and l _(j)and meet l _(i)and l _(j)all be less than heart cycle l _t; If the empty set of being chosen for, then perform step 4.4.5, if be chosen for nonvoid set, then perform step 4.4.3;

Step 4.4.3: if | l _i+ l _j-l _t| < d _{then perform}l _sys=l _i, l _dia=l _j, work as l _i≤ l _jor l _sys=l _j, l _dia=l _i, work as l _i> l _j;

Step 4.4.4: remove l from L _(i)and l _(j), forward step 4.4.2 to;

Step 4.4.5: end loop, the current l obtained _sysand l _diathe systole will estimated exactly and diastole.

Position L={30 is delayed, eligible l in 42,59,72,84,102,115,130,146} corresponding to the peak value selected in monolateral auto-correlation function in figure 3 _(i)and l _(j)all be less than heart cycle l _tposition be: { 30,42,59}.By calculating and comparing | l _i+ l _j-l _t| can calculate and work as l ₍₁₎=30, l ₍₂₎when=42 | l _i+ l _j-l _t|=0 is minimum range, therefore can obtain systole l _sys=30 and diastole l _dia=42.

Claims (6)

1. a cardiechema signals cycle parameter method of estimation, is characterized in that, described method comprises the steps:

Step 1: pretreatment is carried out to cardiechema signals to be estimated;

Step 2: to the average shannon energy envelope signal of pretreated heart sound signal extraction;

Step 3: adopt singular value spectral analysis method smoothing to the envelope signal extracted;

Step 4: the cycle parameter envelope signal after level and smooth being estimated to cardiechema signals.

2. a kind of cardiechema signals cycle parameter method of estimation according to claim 1, is characterized in that,

In described step 1, pretreated method is carried out to cardiechema signals to be estimated and comprises:

Cardiechema signals to be estimated is carried out energy normalization process, then down-sampled is 2kHz, and adopt 6 rank Butterworth strainer accepter to carry out bandpass filtering with other sound beyond filtering cut-off frequency and noise to the cardiechema signals after down-sampled, obtain pretreated cardiechema signals x (n), wherein n is the sample index of hear sounds.

3. a kind of cardiechema signals cycle parameter method of estimation according to claim 1 and 2, is characterized in that,

In described step 2, the method for the average shannon energy envelope signal of pretreated heart sound signal extraction is comprised:

Employing length is W, the frame sliding window moved as W/2 acts on pretreated cardiechema signals x (n) and carries out framing, calculates its average shannon energy envelope to the cardiechema signals in frame window by formula (1):

$e\left(m\right)=-\frac{1}{W}\underset{n=1}{\overset{W}{\&Sigma;}}x{\left(n\right)}^2\log x{\left(n\right)}^2$ Formula (1)

The sample index of the envelope signal obtained after carrying out envelope extraction to cardiechema signals x (n) to be e (m), n be hear sounds, m is frame index.

4. a kind of cardiechema signals cycle parameter method of estimation according to claim 3, is characterized in that,

Described step 4 comprises the steps:

Step 4.1: the monolateral auto-correlation function calculating level and smooth rear envelope signal:

After level and smooth, envelope signal e (m) calculates its monolateral auto-correlation function r (l) according to formula (2):

$r\left(l\right)=\underset{m=1}{\overset{M-l}{\&Sigma;}}e(m+l)e\left(m\right),l\&GreaterEqual;1$ Formula (2)

Wherein: l is the amount of delaying of level and smooth rear envelope signal e (m), M is the length of level and smooth rear envelope signal e (m).

Step 4.2: choose eligible peak value and sort in auto-correlation function r (l):

Choose all K peak values in maximum hear sounds periodic regime millisecond and carry out sequence by its size and be;

R ₍₁₎>r ₍₂₎>r _(K)formula (3)

And record and delay position corresponding to each peak value, be denoted as: L={l ₍₁₎, l ₍₂₎..., l _(K);

Step 4.3: estimate heart cycle l according to formula (4) _t:

formula (4)

Step 4.4: estimate systole l according to formula (5) _syswith diastole l _dia:

formula (5)

Wherein, l _tfor the heart cycle, i, j=1,2 ... K.

5. a kind of cardiechema signals cycle parameter method of estimation according to claim 4, is characterized in that,

Described step 4.4, estimates systole l according to formula (5) _syswith diastole l _diadetailed process be:

Step 4.4.1: make initial distance d=+ ∞, wherein+∞ represents positive infinity;

Step 4.4.2: choose l arbitrarily from L _(i)and l _(j)and meet l _(i)and l _(j)all be less than heart cycle l _t; If the empty set of being chosen for, then perform step 4.4.5, if be chosen for nonvoid set, then perform step 4.4.3;

Step 4.4.3: if | l _i+ l _j-l _t| <d then performs l _sys=l _i, l _dia=l _j, work as l _i≤ l _jor l _sys=l _j, l _dia=l _i, work as l _i>l _j;

Step 4.4.4: remove l from L _(i)and l _(j), forward step 4.4.2 to;

Step 4.4.5: end loop, the current l obtained _sysand l _diathe systole will estimated exactly and diastole.

6. a kind of cardiechema signals cycle parameter method of estimation according to claim 3, is characterized in that, in described step 3, the method adopting singular value spectral analysis method smoothing to the envelope signal extracted comprises:

Be envelope signal e (m) foundation formula (6) structure track matrix (trajectorymatrix) X of M by the length of extraction:

formula (6)

Wherein: k is previously selected constant, the time span corresponding to it is made to be greater than the maximum heart cycle; Then, by formula (7), Eigenvalues Decomposition is carried out to track matrix X:

X=USV ^tformula (7)

Wherein: U, V are the matrix be made up of as column vector the left and right characteristic vector of singular value decomposition respectively, and S is the diagonal matrix be made up of k eigenvalue; The left and right characteristic vector chosen corresponding to a front q eigenvalue forms matrix U _qand V _qand the diagonal matrix S that a front q eigenvalue is formed _q, then the approximate matrix X of track matrix X is constructed _qcan be expressed as:

X _q=U _qs _qv _q ^tformula (8)

By the approximate matrix X obtained in formula (8) _qgenerate level and smooth rear envelope signal e (m), m is frame index.