CN105147252A - Heart disease recognition and assessment method - Google Patents

Abstract

The invention relates to signal processing. The invention provides a heart disease recognition and assessment method. The method includes that a system acquires at least one grade of heart sound signals, takes the heart sound signals as training samples, preprocesses the training samples, carries out autocorrelation segmentation to obtain period signals, and calculates Mel-frequency cepstral coefficients of the period signals and a Stockwell transform time-frequency complex matrix; the system acquires the maximum of every plurality of rows in the Stockwell transform time-frequency complex matrix as a characteristic vector, to form a list of corresponding relation about types of heart sound signals, Mel-frequency cepstral coefficients thereof and characteristic vectors; the system acquires heart sound signals to be recognized, preprocesses the heart sound signals to be recognized, carries out autocorrelation segmentation on preprocessed heart sound signals to be recognized, calculates Mel-frequency cepstral coefficients of the heart sound signals to be recognized after autocorrelation segmentation and characteristic vectors of the heart sound signals to be recognized, compares the Mel-frequency cepstral coefficients and the characteristic vectors with the list of corresponding relation to obtain corresponding grades of the heart sound signals to be recognized. The heart disease recognition and assessment method is suitable for assessment of heart states.

Description

Heart disease identification and appraisal procedure

Technical field

The present invention relates to signal processing technology field, particularly astable periodic signal recognition methods and heart state assessment.

Background technology

As the vibration signal that heart and trunk mechanical movement produce, hear sounds is one of most important physiological signal of human body.When cardiovascular disease not yet develop into be enough to produce clinical pathology change before, just there will be some important pathological informations in hear sounds, these pathological informations have characteristic to embody in numerous disease, and this is all very significant to the estimation of the diagnosis of cardiovascular disease and the state of an illness.Therefore, heart sound analysis is the important means of Non-invasive detection cardiovascular disease, has become one of effective ways of such disease of clinical assistant diagnosis.

In prior art, the Classification and Identification function great majority of hear sounds just distinguish normal and abnormal cardiechema signals, seldom finer kinds of Diseases differentiation is carried out to abnormal cardiechema signals, also do not carry out estimation to disease or health degree to quantize, therefore classify still not meticulous, and recognition correct rate needs to be improved further, and the quantitative evaluation of disease or health degree is conducive to being transplanted to the various platforms such as mobile terminal, and with low cost, be adapted at the practical application of health medical treatment electronic device field.

Summary of the invention

Technical problem to be solved by this invention, is just to provide a kind of heart disease identification and appraisal procedure and system to realize carrying out abnormal information signal the quantitative evaluation of finer resume degree, make user convenient learn oneself cardiac health.

The present invention solve the technical problem, and the technical scheme of employing is, heart disease identification and appraisal procedure, comprise the following steps:

The cardiechema signals of step 1, system acquisition at least one grade, carries out pretreatment as training sample to described training sample; Higher grade, and heart is more healthy;

Step 2, system carry out auto-correlation segmentation to pretreated training sample, obtain periodic signal;

The mel-frequency cepstrum coefficient of step 3, system-computed periodic signal and S-transformation time-frequency complex matrix; (Stockwelltransform, hereinafter referred to as S-transformation);

Maximum in step 4, system acquisition S-transformation time-frequency complex matrix in every some row, as characteristic vector, forms the corresponding relation list of cardiechema signals type and its mel-frequency cepstrum coefficient and characteristic vector; Every some row are by user's sets itself;

Step 5, system acquisition cardiechema signals to be identified carries out pretreatment and carries out auto-correlation segmentation to pretreated cardiechema signals to be identified, calculates mel-frequency cepstrum coefficient and the characteristic vector thereof of the cardiechema signals to be identified after auto-correlation segmentation;

The mel-frequency cepstrum coefficient of cardiechema signals to be identified and characteristic vector thereof compare with corresponding relation list by step 6, system respectively, obtain the grade that cardiechema signals to be identified is corresponding.

Concrete, in described step 1, system carries out pretreatment to the training sample collected, and specifically comprises:

Step 11, system carry out resampling according to the sample frequency preset to the training sample collected;

Step 12, system carry out Butterworth low pass ripple to the training sample after resampling;

Step 13, system carry out denoising to the training sample after low-pass filtering.

Concrete, in described step 2, system carries out auto-correlation segmentation to pretreated training sample, specifically comprises:

The amplitude equalizing value of the training sample after step 21, system-computed denoising;

The signal length of training sample and signal period are gathered that counting is divided by and obtain the segments of training sample by step 22, system, and the described signal period gathers to count counts n, 750 points≤n≤2500 point; System, by the combination of two successively from start to end of the training sample after segmentation, if segments is odd number, casts out unnecessary segmentation; Signal period gathers the meter n that counts, according under mankind's usual condition heart beating and frequency acquisition calculate, computational methods are this area conventional techniques means, repeat no more herein.Experiment proves that the most cases lower signal period gathers the n that counts and should be between 750 points≤n≤2500, and 2500 be that signal period collection maximum in experiment is at present counted.When carrying out segmentation, best segmented mode uses maximum signal period collection to count to carry out segmentation calculating, and in other cases, the maximum signal period gathers to count and is likely greater than 2500 points, also should belong to protection scope of the present invention.

The starting point of first heart sound in each combined segment of step 23, system looks, when the cardiechema signals amplitude of continuous 100 collection points after a certain collection point is all less than amplitude equalizing value, then this point is the starting point of first heart sound; Described first heart sound starting point is at least one;

Step 24, calculate the hear sounds cycle at first heart sound starting point place and the autocorrelation coefficient in next hear sounds cycle in every combined segment, relatively and the first heart sound initial point position of the maximum autocorrelation coefficient of preserving in each combined segment and correspondence thereof;

Step 25, the hear sounds cycle at the first heart sound initial point position place that maximum autocorrelation coefficient is corresponding in all combined segment of Systematic selection, as periodic signal.

Further, in described step 3, the mel-frequency cepstrum coefficient of system-computed periodic signal comprises the following steps:

Step 31, system are by periodic signal by a high pass filter, and form is:

H(z)＝1-a*(z-1)；

Wherein, the value of coefficient a is between 0.9 and 1.0;

Step 32, system carry out fast Fourier transform to the periodic signal after high-pass filtering, then the discrete power spectrum of squared computing cycle signal;

Step 33, system are multiplied by one group of L V-belt bandpass filter to the spectrum energy that discrete power is composed, and try to achieve the logarithmic energy that each wave filter exports, L altogether; And bring an above-mentioned L logarithmic energy into discrete cosine transform, obtain cepstral domain parameter:

$C=\underset{j=1}{\overset{L}{\Σ}}log\left(p_j\right)cos\[k(j-\frac{1}{2})\frac{\mathrm{\π}}{L}\]k=1,2,...,P;$

Wherein C is MFCC parameter, and P is the exponent number of MFCC, p _jfor a jth performance number parameter, j is present filter.

Further, in described step 3, the S-transformation time-frequency complex matrix of system-computed periodic signal, comprises the following steps:

Step 311, system carry out fast Fourier transform to periodic signal, and be augmented by transformation results H [m] as H [m+n], m is that the collection of periodic signal is counted;

Step 322, system do fast Fourier transform conversion to Gauss function, obtain G (m, n);

Fourier's inverse operation that step 333, system press stepped-frequency signal calculating Y=H [m+n] G (m, n) namely obtains S-transformation time-frequency complex matrix.

Concrete, the corresponding relation list input libsvm grader of the mel-frequency cepstrum coefficient of cardiechema signals to be identified and characteristic vector and training sample is carried out classification process by system, completes the identification to different brackets cardiechema signals.

Concrete, also comprise:

The MFCC of the cardiechema signals of at least one group of highest ranking and the different feature input SVDD of S-transformation two is set up the model of a suprasphere, i.e. MFCC suprasphere and S-transformation suprasphere by step 7, system respectively;

Step 8, system arrange weight coefficient a and b according to MFCC and S-transformation respectively to the significance degree of Influence on test result;

Step 9, system are by two of test sample book characteristic parameters input SVDD, and the distance relation according to the cardiechema signals of these two characteristic vectors and highest ranking determines two fractional values;

Step 10, system are added after being multiplied with respective weight coefficient by determine two fractional values, and obtain a final mark, as the assessment result of heart, the higher heart of mark is more healthy.

To the invention has the beneficial effects as follows: the cardiechema signals of system acquisition at least one grade, as training sample, pretreatment is carried out to described training sample; Higher grade, and heart is more healthy; Tell according to credit signal the medical science general knowledge that health of heart degree is medical practitioner, then carried out grade distribute be simple possible.Mel-frequency cepstrum coefficient and the calculating of S-transformation time-frequency complex matrix are carried out to the credit signal of each grade, and using the maximum in some row every in S-transformation time-frequency complex matrix as characteristic vector, the corresponding mel-frequency cepstrum coefficient of cardiechema signals of each grade and S-transformation characteristic vector, form corresponding lists; Then the cardiechema signals that system acquisition is to be identified carries out pretreatment and carries out auto-correlation segmentation to pretreated cardiechema signals to be identified, calculates mel-frequency cepstrum coefficient and the characteristic vector thereof of the cardiechema signals to be identified after auto-correlation segmentation; The mel-frequency cepstrum coefficient of cardiechema signals to be identified and characteristic vector thereof are compared with corresponding relation list respectively, obtains the grade that cardiechema signals to be identified is corresponding.System to quantitative evaluation abnormal information signal being carried out to finer resume degree, make user convenient learn oneself cardiac health.

Detailed description of the invention

Technical scheme of the present invention is described in detail below in conjunction with embodiment:

The Classification and Identification function great majority that the present invention is directed to hear sounds in prior art just distinguish normal and abnormal cardiechema signals, seldom finer kinds of Diseases differentiation is carried out to abnormal cardiechema signals, also do not carry out estimation to disease or health degree to quantize, therefore classify still not meticulous, and recognition correct rate needs the problem that improves further, there is provided a kind of heart disease identification and appraisal procedure first, the cardiechema signals of system acquisition at least one grade, carries out pretreatment as training sample to described training sample; Higher grade, and heart is more healthy; Secondly, system carries out auto-correlation segmentation to pretreated training sample, obtains periodic signal and the mel-frequency cepstrum coefficient of computing cycle signal and S-transformation time-frequency complex matrix; Then, the maximum in system acquisition S-transformation time-frequency complex matrix in every some row, as characteristic vector, forms the corresponding relation list of cardiechema signals type and its mel-frequency cepstrum coefficient and characteristic vector; Every some row are by user's sets itself; Subsequently, the cardiechema signals that system acquisition is to be identified carries out pretreatment and carries out auto-correlation segmentation to pretreated cardiechema signals to be identified, the mel-frequency cepstrum coefficient of the cardiechema signals to be identified after the segmentation of calculating auto-correlation and characteristic vector thereof; Finally, the mel-frequency cepstrum coefficient of cardiechema signals to be identified and characteristic vector thereof compare with corresponding relation list by system respectively, obtain the grade that cardiechema signals to be identified is corresponding.The cardiechema signals of system acquisition at least one grade, carries out pretreatment as training sample to described training sample; Higher grade, and heart is more healthy; Tell according to credit signal the medical science general knowledge that health of heart degree is medical practitioner, then carried out grade distribute be simple possible.Mel-frequency cepstrum coefficient and the calculating of S-transformation time-frequency complex matrix are carried out to the credit signal of each grade, and using the maximum in some row every in S-transformation time-frequency complex matrix as characteristic vector, the corresponding mel-frequency cepstrum coefficient of cardiechema signals of each grade and S-transformation characteristic vector, form corresponding lists; Then the cardiechema signals that system acquisition is to be identified carries out pretreatment and carries out auto-correlation segmentation to pretreated cardiechema signals to be identified, calculates mel-frequency cepstrum coefficient and the characteristic vector thereof of the cardiechema signals to be identified after auto-correlation segmentation; The mel-frequency cepstrum coefficient of cardiechema signals to be identified and characteristic vector thereof are compared with corresponding relation list respectively, obtains the grade that cardiechema signals to be identified is corresponding.System to quantitative evaluation abnormal information signal being carried out to finer resume degree, make user convenient learn oneself cardiac health.

Embodiment

First, the cardiechema signals of system acquisition at least one grade, carries out pretreatment as training sample to described training sample for the heart disease identification of this example and appraisal procedure; Higher grade, and heart is more healthy; Tell according to credit signal the medical science general knowledge that health of heart degree is medical practitioner, then carried out grade distribute be simple possible, repeat no more herein.

Secondly, system carries out auto-correlation segmentation to pretreated training sample, obtains periodic signal.

In order to obtain periodic signal more accurately, need to be handled as follows step to training sample: first, the sample frequency according to presetting carries out resampling to the training sample collected; Again Butterworth low pass ripple is carried out to the training sample after resampling; Finally, denoising is carried out to the training sample after low-pass filtering.

System carries out auto-correlation segmentation to pretreated training sample, specifically comprises:

The amplitude equalizing value of the training sample after steps A, system-computed denoising;

The signal length of training sample and signal period are gathered that counting is divided by and obtain the segments of training sample by step B, system, and the described signal period gathers to count counts n.Usually according to the cardiechema signals of different people, the general cardiechema signals cycle gathers the scope of counting and is: 750 points≤n≤2500 point; System, by the combination of two successively from start to end of the training sample after segmentation, if segments is odd number, casts out unnecessary segmentation;

The starting point of first heart sound in each combined segment of step C, system looks, when the cardiechema signals amplitude of continuous 100 collection points after a certain collection point is all less than amplitude equalizing value, then this point is the starting point of first heart sound; Described first heart sound starting point is at least one;

Step D, calculate the hear sounds cycle at first heart sound starting point place and the autocorrelation coefficient in next hear sounds cycle in every combined segment, relatively and the first heart sound initial point position of the maximum autocorrelation coefficient of preserving in each combined segment and correspondence thereof;

Step e, the hear sounds cycle at the first heart sound initial point position place that maximum autocorrelation coefficient is corresponding in all combined segment of Systematic selection, as periodic signal.

During concrete operations, a little stroke of window must be established when doing autocorrelative within each cycle to guarantee the hear sounds cycle found truly, this autocorrelation coefficient is exactly to find this point.Also the positional information of this point is saved, aspect use below while saving autocorrelation coefficient.Complete hear sounds cycle is at least comprised in each combined segment, also be likely be greater than one, at this moment segmentation is more rough, which position from which position is terminated the hear sounds cycle actually, and auto-correlation that Here it is draws the meaning of window, draws 2500 points from minimum hear sounds Cycle Length (750 points) always, every standardized point just calculates an autocorrelation coefficient, from these coefficients, choose maximum coefficient, the signal in a cycle before saving.Each segmentation is not always the case.The previous signal combination of packing up of preserving in each combined segment becomes periodic signal.

Finally, the mel-frequency cepstrum coefficient of computing cycle signal and S-transformation time-frequency complex matrix.

Wherein, the mel-frequency cepstrum coefficient of system-computed periodic signal comprises the following steps:

Step one, system are by periodic signal by a high pass filter, and form is:

H(z)＝1-a*(z-1)；

Wherein, the value of coefficient a is between 0.9 and 1.0;

Step 2, system carry out fast Fourier transform to the periodic signal after high-pass filtering, then the discrete power spectrum of squared computing cycle signal;

Step 3, system are multiplied by one group of L V-belt bandpass filter to the spectrum energy that discrete power is composed, and try to achieve the logarithmic energy that each wave filter exports, L altogether; And bring an above-mentioned L logarithmic energy into discrete cosine transform, obtain cepstral domain parameter:

$C=\underset{j=1}{\overset{L}{\Σ}}log\left(p_j\right)cos\[k(j-\frac{1}{2})\frac{\mathrm{\π}}{L}\]k=1,2,...,P;$

Wherein C is MFCC parameter, and P is the exponent number of MFCC, p _jfor a jth performance number parameter, j is present filter.

Secondly, the S-transformation time-frequency complex matrix of system-computed periodic signal, comprises the following steps:

Step 4, system carry out fast Fourier transform to periodic signal, are augmented by transformation results H [m] as H [m+n], and n gathers to count the signal period; M is that the collection in periodic signal is counted;

Step 5, system do fast Fourier transform conversion to Gauss function, obtain G (m, n);

Fourier's inverse operation that step 6, system press stepped-frequency signal calculating Y=H [m+n] G (m, n) namely obtains S-transformation time-frequency complex matrix.

Then, the maximum in system acquisition S-transformation time-frequency complex matrix in every some row, as characteristic vector, forms the corresponding relation list of cardiechema signals type and its mel-frequency cepstrum coefficient and characteristic vector; The value of every some row is by user's sets itself.

Need to treat thought-read tone signal when classifying, cardiechema signals to be measured is done the mel-frequency cepstrum coefficient that calculate to be measured cardiechema signals same with training sample and characteristic vector by system, and the corresponding relation list of training sample and the mel-frequency cepstrum coefficient of cardiechema signals to be measured and characteristic vector input libsvm grader are carried out classification process, complete the identification to different brackets hear sounds type.Training sample is more, and grade is clearer respectively.

Subsequently, the cardiechema signals that system acquisition is to be identified carries out pretreatment and carries out auto-correlation segmentation to pretreated cardiechema signals to be identified, the mel-frequency cepstrum coefficient of the cardiechema signals to be identified after the segmentation of calculating auto-correlation and characteristic vector thereof; Finally, the mel-frequency cepstrum coefficient of cardiechema signals to be identified and characteristic vector thereof compare with corresponding relation list by system respectively, obtain the grade that cardiechema signals to be identified is corresponding.The cardiechema signals of system acquisition at least one grade, carries out pretreatment as training sample to described training sample; Higher grade, and heart is more healthy; Tell according to credit signal the medical science general knowledge that health of heart degree is medical practitioner, then carried out grade distribute be simple possible.Mel-frequency cepstrum coefficient and the calculating of S-transformation time-frequency complex matrix are carried out to the credit signal of each grade, and using the maximum in some row every in S-transformation time-frequency complex matrix as characteristic vector, the corresponding mel-frequency cepstrum coefficient of cardiechema signals of each grade and S-transformation characteristic vector, form corresponding lists; Then the cardiechema signals that system acquisition is to be identified carries out pretreatment and carries out auto-correlation segmentation to pretreated cardiechema signals to be identified, calculates mel-frequency cepstrum coefficient and the characteristic vector thereof of the cardiechema signals to be identified after auto-correlation segmentation; The mel-frequency cepstrum coefficient of cardiechema signals to be identified and characteristic vector thereof are compared with corresponding relation list respectively, obtains the grade that cardiechema signals to be identified is corresponding.System to quantitative evaluation abnormal information signal being carried out to finer resume degree, make user convenient learn oneself cardiac health.

Further, in order to provide health of heart scoring, the MFCC of the cardiechema signals of at least one group of highest ranking and the different feature input SVDD of S-transformation two is set up the model of a suprasphere, i.e. MFCC suprasphere and S-transformation suprasphere by system respectively; Secondly, system arranges weight coefficient a and b according to MFCC and S-transformation respectively to the significance degree of Influence on test result; Then, system is by two of test sample book characteristic parameters input SVDD, and the distance relation according to the cardiechema signals of these two characteristic vectors and highest ranking determines two fractional values; Finally, system is added after being multiplied with respective weight coefficient by determine two fractional values, and obtain a final mark, as the assessment result of heart, the higher heart of mark is more healthy.

In sum, be the feature of periodic signal for hear sounds, the invention provides an auto-correlation segmentation algorithm and combine the MFCC algorithm and S-transformation that improve, realize the identification of all kinds of heart disease, adjuvant clinical diagnosis of cardiovascular diseases.And the characteristic parameter extracted using MFCC algorithm and S-transformation realizes the quantitative evaluation of cardiechema signals disease or health degree as quantitative basis, with the heart helping the user in terminal to understand oneself easily.

Claims (7)

1. heart disease identification and appraisal procedure, is characterized in that, comprises the following steps:

The cardiechema signals of step 1, system acquisition at least one grade, carries out pretreatment as training sample to described training sample; Higher grade, and heart is more healthy;

Step 2, system carry out auto-correlation segmentation to pretreated training sample, obtain periodic signal;

The mel-frequency cepstrum coefficient of step 3, system-computed periodic signal and S-transformation time-frequency complex matrix;

Maximum in step 4, system acquisition S-transformation time-frequency complex matrix in every some row, as characteristic vector, forms the corresponding relation list of cardiechema signals type and its mel-frequency cepstrum coefficient and characteristic vector; Every some row are by user's sets itself;

Step 5, system acquisition cardiechema signals to be identified carries out pretreatment and carries out auto-correlation segmentation to pretreated cardiechema signals to be identified, calculates mel-frequency cepstrum coefficient and the characteristic vector thereof of the cardiechema signals to be identified after auto-correlation segmentation;

The mel-frequency cepstrum coefficient of cardiechema signals to be identified and characteristic vector thereof compare with corresponding relation list by step 6, system respectively, obtain the grade that cardiechema signals to be identified is corresponding.

2. heart disease identification according to claim 1 and appraisal procedure, is characterized in that, in described step 1, system carries out pretreatment to the training sample collected, and specifically comprises:

Step 11, system carry out resampling according to the sample frequency preset to the training sample collected;

Step 12, system carry out Butterworth low pass ripple to the training sample after resampling;

Step 13, system carry out denoising to the training sample after low-pass filtering.

3. heart disease identification according to claim 2 and appraisal procedure, is characterized in that, in described step 2, system carries out auto-correlation segmentation to pretreated training sample, specifically comprises:

The amplitude equalizing value of the training sample after step 21, system-computed denoising;

The signal length of training sample and signal period are gathered that counting is divided by and obtain the segments of training sample by step 22, system, and the described signal period gathers to count counts n, 750 points≤n≤2500 point; System, by the combination of two successively from start to end of the training sample after segmentation, if segments is odd number, casts out unnecessary segmentation;

The starting point of first heart sound in each combined segment of step 23, system looks, when the cardiechema signals amplitude of continuous 100 collection points after a certain collection point is all less than amplitude equalizing value, then this point is the starting point of first heart sound; Described first heart sound starting point is at least one;

Step 24, calculate the hear sounds cycle at first heart sound starting point place and the autocorrelation coefficient in next hear sounds cycle in every combined segment, relatively and the first heart sound initial point position of the maximum autocorrelation coefficient of preserving in each combined segment and correspondence thereof;

Step 25, the hear sounds cycle at the first heart sound initial point position place that maximum autocorrelation coefficient is corresponding in all combined segment of Systematic selection, as periodic signal.

4. heart disease identification according to claim 3 and appraisal procedure, is characterized in that, in described step 3, the mel-frequency cepstrum coefficient of system-computed periodic signal comprises the following steps:

Step 31, system are by periodic signal by a high pass filter, and form is:

H(z)＝1-a*(z-1)；

Wherein, the value of coefficient a is between 0.9 and 1.0;

Step 32, system carry out fast Fourier transform to the periodic signal after high-pass filtering, then the discrete power spectrum of squared computing cycle signal;

Step 33, system are multiplied by one group of L V-belt bandpass filter to the spectrum energy that discrete power is composed, and try to achieve the logarithmic energy that each wave filter exports, L altogether; And bring an above-mentioned L logarithmic energy into discrete cosine transform, obtain cepstral domain parameter:

$C=\underset{j=1}{\overset{L}{\Σ}}log\left(p_j\right)cos\[k(j-\frac{1}{2})\frac{\mathrm{\π}}{L}\]k=1,2,...,P;$

Wherein C is MFCC parameter, and P is the exponent number of MFCC, p _jfor a jth performance number parameter, j is present filter.

5. heart disease identification according to claim 4 and appraisal procedure, is characterized in that, in described step 3, the S-transformation time-frequency complex matrix of system-computed periodic signal, comprises the following steps:

Step 311, system carry out fast Fourier transform to periodic signal, and be augmented by transformation results H [m] as H [m+n], m is that the collection of periodic signal is counted;

Step 322, system do fast Fourier transform conversion to Gauss function, obtain G (m, n);

Fourier's inverse operation that step 333, system press stepped-frequency signal calculating Y=H [m+n] G (m, n) namely obtains S-transformation time-frequency complex matrix.

6. heart disease identification according to claim 5 and appraisal procedure, it is characterized in that, the corresponding relation list input libsvm grader of the mel-frequency cepstrum coefficient of cardiechema signals to be identified and characteristic vector and training sample is carried out classification process by system, completes the identification to different brackets cardiechema signals.

7. heart disease identification according to claim 1 and appraisal procedure, is characterized in that, also comprise

The MFCC of the cardiechema signals of at least one group of highest ranking and the different feature input SVDD of S-transformation two is set up the model of a suprasphere, i.e. MFCC suprasphere and S-transformation suprasphere by step 7, system respectively;

Step 8, system arrange weight coefficient a and b according to MFCC and S-transformation respectively to the significance degree of Influence on test result;

Step 9, system are by two of test sample book characteristic parameters input SVDD, and the distance relation according to the cardiechema signals of these two characteristic vectors and highest ranking determines two fractional values;

Step 10, system are added after being multiplied with respective weight coefficient by determine two fractional values, and obtain a final mark, as the assessment result of heart, the higher heart of mark is more healthy.