KR20010083267A - Heart sound classification method by intergration of time period and statistical analysis - Google Patents

Abstract

PURPOSE: A method for sorting heart sound by sectional integral and statistical analysis is provided to judge the heart sound efficiently and in a short time and to improve the probability of success in judging the heart sound. CONSTITUTION: In the method for sorting heart sound by sectional integral and statistical analysis, a few stages are included. In the stage of measuring an electrocardiogram, the peak value, R, which corresponds to the beginning period of heart contraction, and the peak value, T, which corresponds to the beginning period of heart relaxation, are measured. In the stage of measuring the heart sound, the periodic heart sound signal over one period is measured. In the stage of pre-treatment, the heart sound signal over one period is formalized. In the stage of detecting the peak point, the main peak points of the first and second heart sound are measured. In the stage of the first judgement, the size of the first and second heart sound and each peak point is compared and it is judged if it is abnormal or not. In the stage of the second judgement, the heart sound signal is integrated and the size of the integrated value is compared with the whole integral value. In the stage of the third judgement, mean, quartiles and skewness about a whole period are measured and the quantities of the noise in the heart sound are judged.

Description

Heart sound classification method by intergration of time period and statistical analysis}

The present invention relates to a heart sound classification method by interval integration and statistical analysis.

Phonocardiography is a convenient tool that provides noninvasive information about heart valne and cardiovascular flow. The lipman capillary electrometer was used by Einthoven in 1894 to record electrocardiogram (ECG) in 1894. (Lippman's capillary electrometer) has been researched and developed as one of the most basic diagnostic tools since the first heart sound graph was developed (Am J Cardiol, vol. 60, pp. 1378-1382, 1987, Acta Cardiologica, vol. XLVIII, pp. 337-344, 1993). However, in recent years, the development of diagnostic technology by heart sound recording method has not been developed compared with other diagnostic technologies. The reason is that the Phonocardiogram (PCG) is a complex signal that cannot be analyzed simply by listening or seeing, and there is still some controversy regarding the source of heart sound. Moreover, tools and methods for measuring heart sound have not been accurately quantified or standardized, and errors in these methods affect the heart sound collection and analysis (CRC Critical Reviews Biomedical Engineering, vol. 23, no. 3, pp. 163-). 219, 1995). The development of ultrasound sound recording method tends to be considered as an assistant diagnostic tool. However, with the recent development of digital signal processing methods, the audible heart analysis method has been developed using modeling, frequency analysis, time-frequency analysis, Research on recognition methods is diversified and active

Kim, Yonsei medical journal, vol. 39, no. 4, pp. 302-308, 1998 and Durand, Medical & Biological Engineering & Computing, vol. 32, pp. 311-316, 1994 Ding, Noninvasive diagnoses of coronary disease by heart sound analysis, Ph.D. Thesis in Biomedical Engineering, The State University of New Jersey, 1991), Akay, Noninvasive detection of coronary artery diseaase, IEEE Engineering in Medicine and Biology, no. 17, pp. 761-764, 1994). A study on the perception of coronary artery disease through wavelet analysis was performed as a result of the frequency analysis of a specific valve or disease, resulting in a very limited recognition method for the presence or absence of a specific disease. D. Barschdor ff, "Automatic phonocardiogram signal analysis in infant based on wavelet transform and artificial neural networks", pp.753-756, IEEE Computers in Cardiology, 1995). In other words, the research on the recognition of heart soundness was not comprehensive compared to other biosignals such as electrocardiogram and electroencephalogram, and it only confirmed the presence of a specific disease. Unlike ECG or EKG, there are many errors due to non-standardization of measurement methods, and even the same pathological symptom causes different heart sounds and somewhat different heart sounds, which makes it more difficult to recognize.

In general, the heart sound is an acoustic wave generated by the movement of the heart, and may be classified into a normal heart sound and a murmur. The actual heart noise is not noise, but is an acoustic signal in contrast to the sound signal in contrast to the normal first and second heart sounds. The heart repeats contraction and relaxation movements, which are caused by the opening and closing of the heart valves, the vibration of the myocardium, and the vibrations of related organs.

1 is a graph of ventricular pressure, heart sound, and electrocardiogram generation time. Heart noise can occur at any location in the heart cycle, depending on the pathological condition of the heart. This murmur is mainly caused by turbulent flow of blood flow and abnormal vibrations of the heart muscle or valves. The normal heart sound is divided into four main elements from the first sound sound (S1) to the fourth heart sound (S4). The first and second heart sounds (S2) Is recognized as a normal signal and the third heart sound (S3) and the fourth heart sound are recognized as normal sounds in children or young people, but are interpreted mainly pathologically in adults.

In auscultation of heart sound, the current method is to assess the human eye by diagnosing by listening to the ear with a stethoscope on the chest part of the subject or by recording the heart sound waveform on a recording sheet. This method is difficult to preserve data and subjective judgment of the subject. As a result, evaluation results may vary. In recent years, with the development of digital signal processing technology, efforts have been made to secure objectivity in diagnosing heart sound through digital signal processing for heart sound, but have not yet developed useful results.

Conventional heart sound classification method can be divided into two categories as follows.

1) US Pat. No. 4,446,873, as shown in FIG. 2, detects the heart sound, generates a square wave window function, searches the heart sound data in the window function, determines the heart sound characteristics, and determines whether the heart sound is abnormal. do.

2) In US Patent 4,378,022, as shown in Fig. 3, after dividing the input data by the window function and Fourier transform, and then determine the abnormality of the heart sound through power spectrum calculation and comparison, in particular in the frequency domain Classify the sound feature by checking the signal magnitude within the range.

However, the above two methods each have the following disadvantages.

In the case of method 1), if there is a signal in a certain time domain, it can be determined whether it is systolic or diastolic during cardiac activity. In addition, it takes time to set a short time window for the whole heart sounding cycle and scan the whole heart sounding cycle with the set window, and if the window size is small, it takes a lot of time to scan the whole cycle and the window size Larger drops noise detection resolution.

In the case of method 2), the boundary between frequency ranges for heart sound, heart noise, and ambient noise is not clearly distinguished, so the efficiency is likely to decrease. In fact, frequency components alone do not distinguish between normal heart sound and heart noise. In addition, the signal processing in the frequency domain requires a lot of calculations, so it is inferior in real time and requires the high performance of the system.

It is a first object of the present invention to provide a method for classifying heart sounds by section integration and statistical analysis, which enables effective determination of heart sounds.

A second object of the present invention is to provide a method for classifying heart sounds by section integration and statistical analysis which can determine the heart sounds within a short time.

A third object of the present invention is to provide a method for classifying heart sounds by interval integration and statistical analysis in which the probability of success of heart sound determination is improved.

1 is a graph of ventricular pressure, heart sound, and ECG generation time.

2 shows a conventional heart sound determination method.

Figure 3 shows another conventional heart sound determination method.

4 illustrates a method for processing a heart sound and an electrocardiogram signal in the heart sound classification method according to the present invention.

5 is a flow chart of an embodiment according to the heart sound classification method by interval integration and statistical analysis of the present invention.

6 is a heart sound waveform diagram applied to the experiment of the present invention, (a) is a normal heart sound (b) is an abnormal heart sound.

7 is a heart sound waveform diagram showing a process of detecting peaks from the normal heart sound shown in FIG.

FIG. 8 is a heart sound waveform diagram illustrating a process of detecting peaks from the abnormal heart sound shown in FIG.

FIG. 9 illustrates a determination process based on the interval integration of the normal heart sound illustrated in FIG. 6A, and is a heart sound waveform diagram showing an integration process for a certain section at each vertex of the normal heart sound.

FIG. 10 is a view illustrating a determination process based on the interval integration of the abnormal heart sound illustrated in FIG. 6B, and is a heart sound waveform diagram showing an integration process for a certain section at each vertex of the normal heart sound.

In order to achieve the above object, according to the present invention,

Obtaining, by an electrocardiogram, a peak value R corresponding to at least the onset of cardiac contraction and a peak value T corresponding to the heart relaxation starter;

A heart sound measurement step of measuring one or more cycles of a periodic heart sound signal including at least a first heart sound and a second heart sound by a heart sound meter;

A preprocessing step of normalizing the measured at least one heart sound signal;

A vertex extraction step of obtaining a first heart sound, a second heart sound, and main peaks from the normalized heart sound signal;

A first determination step of comparing an amplitude of each vertex with a first heart sound, a second heart sound in the normalized heart sound signal to determine whether it is abnormal;

A second determination step of integrating the heart sound signal for a predetermined time interval around the vertices and evaluating the magnitude of each integral value with the integral of one period of the heart sound signal to determine whether each vertex is meaningful as heart noise;

And a third determination step of comparing the mean and quartiles for the entire heart sound period, or obtaining skewness and evaluating the amount of noise included in the heart sound. A method for classifying heart sounds by statistical analysis is provided.

In the method of classifying heart sounds by the interval integration and statistical analysis of the present invention, the third determining step includes all of the mean, quartiles, and skewness of the whole heart sound period during planting. It is desirable to evaluate the amount of noise that is present.

In addition, the heart sound measurement step may be performed so as to normalize the amplitude to filter the input signal to a predetermined frequency band and compensate for the amplitude difference according to the input condition of each heart sound.

In the method of classifying heart sounds by the interval integration and statistical analysis of the present invention, in selecting one heart sound signal, it is preferable to select a signal of several cycles to calculate an average one signal.

In the present invention, the heart sound is classified into seven types according to the occurrence time of the heart noise. Normal heart sound, pre-systolic murmur, early systolic murmur, late systolic murmur, early diastolic murmur, and late relaxation late diastolic murmur and continuous murmur.

In determining the normal and abnormal of the heart sound and classifying the seven types, the following three steps are used to evaluate the position and the amount of heart noise included in the heart sound.

1) Level 1 Judgment

Several peaks of large amplitude are obtained in the time domain, and the vertex size is compared with the first and second heart sounds to determine whether each vertex is abnormal.

2) Level 2 Judgment

Integrate over a certain time interval around each vertex and evaluate the magnitude of each integral value with the integral value of the whole cycle to determine whether each vertex is significant as a murmur.

3) 3-step judgment

The mean, quartiles, and skewness of the entire planting cycle are determined to evaluate the amount of noise contained in the planting plant.

The signal collection condition for implementing the algorithm of the present invention should know the position of the first heart sound among the heart sound components by measuring the heart sound and the electrocardiogram at the same time or by other methods, and the sampling rate when digitizing the signal. However, such a condition is a general method for collecting signals for heart sound analysis, so it is difficult to regard it as a special condition for the present invention. 4 is a block diagram of a general signal acquisition device configuration.

Referring to FIG. 4, a heart sound and an electrocardiogram are measured to amplify, filter, adjust the heart sound / ECG delay, and convert analog and digital (A / D) signals. Then, a few peaks of large amplitude are obtained in the time domain using the digital signal through A / D conversion, and the vertex size is compared with the first heart sound and the second heart sound to determine whether each vertex is abnormal or not. Integrate over a period of time, and compare and evaluate the magnitude of each integral value with the total integral value of one cycle to determine whether each vertex is meaningful as a murmur, and the mean, The quartiles and skewness are calculated to evaluate the amount of noise contained in the planting.

Hereinafter, a preferred embodiment of the present invention will be described in detail with reference to FIG.

1) Perform general preprocessing such as amplification and filtering on the heart sound signal. Here, the input signal is filtered to the 10 ~ 500Hz band and the absolute value is obtained. Normalize the amplitude to compensate for the difference in amplitude according to the input condition of each heart sound.

2) Select the heart sound one cycle signal which is synchronized with ECG. At this time, in order to remove the influence of noise and to select a stable signal, an average one cycle signal may be calculated by selecting a signal of several cycles.

3) The absolute value of the heart sound data is taken.

4) Find the maximum peak in one periodic signal and normalize the rest of the data by using this peak as the maximum value.

5) After the masking operation is performed to set a certain section including the obtained maximum vertex to the minimum value, the maximum vertex is obtained again from the remaining data. The reason for masking is to prevent the next detected vertex from being detected near the previous one.

6) Repeat step 5) N times.

7) Obtain the values for the first heart sound (S1) and the second heart sound (S2). Here, the positions of the first heart sound S1 and the second heart sound S2 are known in advance because the sound is synchronized with the R wave (signal) and the T wave (signal) of the ECG.

8) It is determined whether the maximum peak is the first heart sound or the second heart sound among the heart sound components. If the maximum peak is the first heart sound or the second heart sound, it is normal or abnormal.

9) Perform interval integration on the vertices obtained in step 4) -6).

10) Compare the integral value of each vertex section with the total period integral value, and check whether the integral value of each vertex section is below a certain ratio with respect to the total period integral value. If it is below a certain ratio, it is normal and if it is above, it is abnormal.

11) Comparing the positions of the vertices determined as abnormal in step 10) with the position of the first heart sound and the second heart sound to calculate which position of each period is closest to the period of planting.

12) The mean, quartile, and skewness of the entire planted cycle are calculated by statistical methods.

13) Determine if the average is below a certain percentage of the third quartile. Below is normal and above is abnormal. The greater the murmur, the larger the third quartile is than the average. Also check if the skewness is above a certain number. Above is normal and below is abnormal. The more murmurs, the lower the skewness.

14) Classify heart sounds according to the results in steps 8), 10), and 13) above, and classify the heart sounds into seven types according to the procedure of paragraph 11).

In the above 11), if the position of each vertex is between the first heart sound and the second heart sound, the systolic heart noise is close to the first heart sound side, the systolic initial heart sound is close to the second heart sound side, and the middle systolic heart sound is close to the second heart sound side. In addition, if the position of each vertex is between the second heart sound and the first heart sound, the diastolic heart noise is close to the second heart sound side, the early diastolic heart sound when close to the first heart sound, and the middle diastolic heart sound when the middle heart is close to the first heart sound. In addition, if the peak is after the second heart and very close to the first heart, it is classified as a pre-shrinkage heart noise.

In peak detection, as shown in FIG. Since the first heart sound S1 is correlated with the ECG signal, the first heart sound S1 is determined as the first heart sound S1 by detecting the maximum value within a predetermined section based on the R wave (signal) of the ECG signal. Even if there are heart noises in the surroundings, the first heart sound S1 becomes the maximum peak within a predetermined section based on the R wave. After finding the first heart sound as the first vertex from the first input heart sound data, masking a certain section centered on the first heart sound to the minimum value as shown in FIGS. 7 and 8 (b) and the second heart sound as the next second vertex. After detection, masking is performed as in FIGS. 7 and 8 (c). The reason for masking the predetermined section is to prevent the next maximum vertex from being detected in the section adjacent to the maximum vertex. After detecting the second vertex, the section around the second vertex is masked to a minimum value and the third vertex is detected. The above process is repeated to detect up to the nth vertex. The first vertex corresponds to the first heart sound, and then the second, third, fourth, and fifth vertices in turn become the vertices having the maximum values. In the next vertex evaluation step, the second heart sound is found among these vertices, and the position and size evaluation for each remaining vertex is evaluated to determine normal or abnormal.

In vertex evaluation, the time relationship of each vertex is evaluated based on the previously detected first heart sound. The first and second heartbeats vary slightly from person to person, but are located approximately half of the heart rate cycle. The first heart sound and the second heart sound within a predetermined time range are determined within four vertices except for the first heart sound, which is the first heart sound, among the five vertices previously detected. The remaining three vertices after the second heartbeat are sorted in the order of the vertex values. After the alignment of the five vertices is completed, the judgment on the normal and abnormal is started based on the relationship between each vertex. In the case of an abnormal heart, the interval between the first heart sound and the second heart sound is often shorter than that of a normal heart. Therefore, if the interval between the first heart sound and the second heart sound is shorter than expected, there is a possibility of abnormality. In addition, in the normal case, the values of the remaining vertices should be significantly smaller than the first and second heart sounds. Therefore, the remaining vertex value is evaluated and classified as abnormal if it is above the expected value, and it is determined whether it is systolic or diastolic in consideration of the position of the vertex and the time relationship between the first and second heart sounds.

In the judgment by statistical analysis, the average value, the quartile, and the skewness are obtained as representative values for the entire section of the heart sound. The mean and quartiles are compared relatively and compared with experimental judgments on the number value of the skewness. In this case, the numerical value that divides the values arranged in the order of magnitude of the input signal into quadrants is called the quartile, and the numerical value that determines the degree of left and right symmetry based on the median is called the skewness. If you have a distribution, the skewness is 0.5. In the case of the heart sound, the region of the first heart sound and the second heart sound usually occupies a short time area of about 10 or less than the whole one sound region in one cycle, so the quartile has a value below the average value in the normal case. In the case of no noise, the skewness represents a large value, which means that since most of the signals have low values, the median is very low, so that there are many signals above the median. On the other hand, when the heart noise increases, the median rises and the signal below the median increases, so the skewness becomes a smaller number, and the value becomes extremely 0.5 or less. Judgment classification for heart sounds can be classified into normal, abnormal, early systolic heartbeat, late systolic heartbeat, early diastolic heartbeat, late diastolic heartbeat.

In order to examine the validity of the present invention as described above, the actual test was conducted.

Data collection, ie, heart collection, was performed using Biopac's MP100 system, with a contact microphone (TSD108) as the input microphone and a DA100B as an amplifier. In addition to the normal heart sounds measured directly, we used Frontiers in Bioscience, a non-profit journal, and a heart sound database from Western University. Heart sound data has 16bit resolution with 8KHz sampling. Table 1 below summarizes the heart sound data applied.

division Symptoms and Causes Heart sound database Direct measurement Bioscience Western University Sum normal Normal heart sounds 5 2 2 9 abnormal Early systolic noise (such as mitral regurgitation) - 9 8 17 Late systolic noise (such as mitral valve prolapse) - 2 2 4 Early diastolic noise (such as aortic valve regurgitation) - 3 2 5 Late diastolic heart sounds (tricuspid regurgitation, etc.) - 2 One 3 Continuous heart noise (such as pulmonary branch stenosis) - - 2 2 Sum 5 18 17 40

In Figure 6 (a) represents a normal heart sound, (b) is an abnormal heart sound. In FIG. 6, the horizontal axis represents the time axis, and each number represents the number of sampling data, and the vertical axis represents the size value of each data.

7 and 8 are heart sound wave diagrams showing the process of detecting the peak of the normal heart sound and abnormal heart sound.

In Figures 7 and 8, (a) is a heart sound waveform after the pre-process normalized the heart sound shown in Figure 5, (b) after the detection of the first heart sound (S1), masking the first heart sound to a certain band minimum value (C) is a waveform diagram of the state where the second heart sound is masked after detecting the second heart sound S2, (d) is a waveform diagram of the state where the third heart sound is masked after the detection of the third heart sound S3 Waveform diagram (e) shows the waveform of the fourth heart sound after the detection of the fourth heart sound (S4), and (f) shows the waveform of the fifth heart sound after the detection of the fifth heart sound (S5). It is also.

9 and 10 illustrate a determination process based on interval integration, which is a heart sound waveform diagram showing an integration process for a predetermined section at each vertex of a normal heart sound and an abnormal heart sound. In the determination process by the interval integration, the integral value for the entire interval, and the integral value for each vertex section including the first heart sound and the second heart sound are obtained. Check the abnormality by comparing the integral value of each vertex section with the total integral value.

variable Criteria Normal case One The third vertex value is S! Is it greater than? No (1) 2 Is the third point position before contraction? Yes (0) 3 Is the third apex located early in the systole? Yes (0) 4 Is the third vertex position at the end of the systole? Yes (0) 5 Is the third vertex position early in the diastolic phase? Yes (0) 6 Is the third vertex position at the end of the relaxation period? Yes (0) 7 Is the fourth vertex value larger than S1? No (1) 8 Is the fourth vertex position before contraction? Yes (0) 9 Is the fourth apex located early in the systole? Yes (0) 10 Is the fourth vertex position at the end of the systole? Yes (0) 11 Is the fourth vertex position at the beginning of the diastolic phase? Yes (0) 12 Is the fourth vertex position at the end of the diastolic phase? Yes (0) 13 Is the fifth vertex value greater than S1? No (1) 14 Is the fifth vertex position before contraction? Yes (0) 15 Is the fifth apex located early in the systole? Yes (0) 16 Is the fifth apex located at the end of the systole? Yes (0) 17 Is the fifth vertex position at the beginning of the diastolic phase? Yes (0) 18 Is the fifth vertex position at the end of the relaxation period? Yes (0) 19 Is the integral value of the third vertex interval less than or equal to a certain percentage of the total integral value? Yes (0) 20 Is the integral value of the fourth vertex interval less than or equal to a certain percentage of the total integral value? Yes (0) 21 Is the fifth vertex interval integral value less than or equal to a certain percentage of the total integration values? Yes (0) 22 Is the mean greater than the third quartile? Yes (0) 23 Why is the distortion above a certain value? Yes (0)

Table 2 above shows the determination variables of the normal heart sound and the abnormal heart sound, which is an example of only five vertices including the first heart sound and the second heart sound. In the case of normal heart sound, variables 1, 7, and 13 are set to '1' and the remaining variables are set to '0'. Statistical variables obtained by calculating the heart sounds shown in FIGS. 6 (a) and 6 (b) are as follows.

(a) mean = 0.77, third quartile = 0.63, skewness = 3.7

(b) mean = 0.63, third quartile = 0.81 skewness = 2.8

The final determination result with respect to FIG. 6 is as follows.

(a) Judgment result = normal heart sound

(b) Judgment result = initial systolic noise

division Symptoms and Causes Number of heart sound data Heart sound classification result Correctly classified Recognition rate [] normal Normal heart sounds 9 9 100 abnormal Early contraction 17 13 76 Late contraction 4 4 100 Early relaxation 5 5 100 Late relaxation 3 2 67 Continuous heart noise 2 2 100 Sum 40 35 88

Table 3 shows the results of testing 40 heart sound data with the material. The heart was classified into normal and abnormal. Abnormalities were classified into early systolic noise, late systolic heart noise, early diastolic heart sound, late diastolic heart noise, and continuous heart noise. Recognition rates for each classification were 100 for normal heart sound, 75 early systolic noise, 100 late systolic noise, 100 early diastolic noise, 67 late diastolic noise, and 100 continuous heart noise. In the case of early systolic noise in the abnormal heart sound group, 4 out of 17 samples were mistaken as normal. In addition, one out of three cases of misdiagnosis was late. The case of continuous heart noise was recognized as stable. The total recognition rate was 88 with 5 being misidentified in a total of 40 samples. In the case of normal heart sound, all 9 test data were correctly recognized and the recognition rate was 100. On the other hand, in the abnormal heart sound group, 26 out of 31 heart sounds were correctly recognized and the average recognition rate was 84. Misrecognition in the abnormal heart sound group was found in the early systolic and late diastolic heart sounds. The heart sound was not recognized by the strong tension of the heart in the vicinity of the first heart sound. This is considered to be because the algorithm is designed on the basis of the first heart sound to set the normal range for the detection and integration of the first heart sound wide. Abnormal heart sound was judged as normal heart sound in the middle stage because the abnormal symptoms were weak and the overall signal shape was similar to the normal heart sound. However, even though the signal shape was similar, 28 judgment variables could not be satisfied. Recognition result of abnormality was good. In many clinical applications, the expected problem is also likely to be a perception of the weakness of the heart.

The present invention is not an analysis in the frequency domain where the calculation is complicated, but a small amount of calculation through the relative evaluation through the statistical analysis and the absolute value at the peak with a large amplitude according to the cardiac exercise time in the time domain, and is a strong method against ambient noise. It is effective in diagnosing the heart sound as an evaluation of one cycle of the heart sound, not an evaluation of the constant time portion or the frequency portion of the heart sound. The present invention can be particularly utilized in cardiac sound diagnosers, including small electronic stethoscopes.

Although the present invention has been described with reference to the embodiments shown in the drawings, this is merely illustrative, and those skilled in the art will understand that various modifications and equivalent other embodiments are possible therefrom. Therefore, the true technical protection scope of the present invention should be defined only in the appended claims.

Claims (5)

An electrocardiogram measuring step of obtaining, by an electrocardiogram, a peak value R corresponding to at least the heart contraction start and a peak value T corresponding to the heart relaxation start; A heart sound measurement step of measuring at least one cycle of a periodic heart sound signal including at least a first heart sound and a second heart sound by a heart sound meter; A preprocessing step of normalizing the measured at least one heart sound signal; A vertex extraction step of obtaining a first heart sound, a second heart sound, and main peaks from the normalized heart sound signal; A first determination step of comparing an amplitude of each vertex with a first heart sound, a second heart sound in the normalized heart sound signal to determine whether it is abnormal; A second determination step of integrating the heart sound signal for a predetermined time interval around the vertices and evaluating the magnitude of each integral value with the integral of one period of the heart sound signal to determine whether each vertex is meaningful as heart noise; A third decision step of evaluating the amount of noise included in the planting by comparing the mean and quartiles or calculating the skewness for the entire planting cycle; Heart sound classification method by The method of claim 1, The third determination step is to calculate all the mean, quartiles, skewness for the entire heart period, and to evaluate the amount of noise included in the planting period. Heart sound classification method by analysis. The method according to claim 1 or 2, The heart sound measuring step may include filtering the input signal into a predetermined frequency band, wherein the heart sound classification method using interval integration and statistical analysis. The method according to any one of claims 1 and 2, In selecting a heart sound one cycle signal, the heart sound classification method using interval integration and statistical analysis, characterized in that to select the signal of several cycles to calculate the average one cycle signal. The method of claim 3, In selecting a heart sound one cycle signal, the heart sound classification method using interval integration and statistical analysis, characterized in that to select the signal of several cycles to calculate the average one cycle signal.