Analysis of Various DWT Methods for Feature Extracted PCG Signals

Analysis of Various DWT Methods for Feature Extracted PCG Signals

1. Venkata Hari Prasad1,

   1. Research scholar, Andhra University, Visakhapatnam

      Dr. P. Rajesh Kumar2

      Andhra University College of Engineering, Visakhapatnam

      Abstract: One of the main obstacles states that the widespread use of phonocardiogram (PCG) in modern days medicine is the various noise components they invariably contain. Although many advances have been made towards automated heart sound segmentation and heart pathology detection and classification, an efficient method for noise handling would come as a major aid for further development in this field, especially when it comes to working with PCGs collected in realistic environments such as hospitals and clinics. The feature extraction has been gone through 10 levels on PCG recorded signals using transformation techniques. Analyzing PCG signals with calculating parameters Energy, Standard deviation, Variance, Mean square error (MSE), Peak Signal to Noise Ratio (PSNR), Root Mean Square Error (RMSE) and Maximum Entropy (ME) values of human heart signal which were extracted from Phonocardiogram were calculated. These calculations are based on the filtration process. Wavelets considered as filtration technique as well as under goes 10 leveling factors. Different wavelets compared for analysis part such as Haar, Daubechies, Orthogonal, Coiflets and Biorthogonal and also finding histograms and denoising the signal were part in this proposed scheme using wave menu analysis.

      Keywords:- ECG-Electrocardiogram , PCG-Phonocardiogram, AS- Aortic Stenosis, AR- Aortic Regurgitation , MS- Mitral Stenosis, MR- Mitral Regurgitation , DWT- Discrete Wavelet Transform, ENER-Energy ,STD-Standard Deviation, VAR- Variance, NHS- Normal Heart Sound

      1. INTRODUCTION

         The heart is a hollow muscular organ that pumps blood throughout the blood vessels to various parts of the body by repeated, rhythmic contractions. It is found in all animals with a circulatory system, which includes the vertebrates and heart is divided into chambers namely atrium and ventricles. The upper two chambers are known as atria while the lower two chambers are known as ventricles. Heart muscles squeeze the blood from chamber to chamber. During this squeezing process, the valves help the blood to keep flowing smoothly in and out of the heart. This is done by automatically opening of valves to let blood from chamber to chamber and closing to prevent the backflow of blood [1]. Heart sounds are the composite sounds produced by myocardial systolic and diastolic, hoist valve, blood flow and cardiovascular vibration impact, and contain a great deal of physiological and pathological information regarding human heart and vascular.

         The average human heart, beating at 72 beats per minute, will beat approximately 2.5 billion times during an average 66 year lifespan, and pumps approximately 4.7-5.7 litres of blood per minute. It weighs approximately 250 to 300 grams (9 to 11 oz) in females and 300 to 350 grams (11 to 12 oz) in males.

         Research on diagnosis of cardiac abnormalities using wavelet techniques has been carried out from the past few years, due to its good performance in analyzing the signals that present non stationary characteristics, this technique has eventually become a powerful alternative when compared to the traditional Fourier Transform (FT) [1] [2]. Fig.(1) shows the normal heart sounds, composed of four different sounds, namely S1, S2, S3 and S4.The pumping action of a normal heart is audible by the 1st heart sound (S1) and 2nd heart

         Figure (1): Heart Sounds for S1and S2 Signals.

         Sound (S2). During systole, the AV valves are closed and blood tries to flow back to the atrium, causing back bulging of the AV valves. But the taut chordatetendineae (cord-like tendons that connect the papillary muscles to the tricuspid valve and the mitral valve in the heart) stop the back bulging and causes the blood to flow forward. This leads to vibration of the valves, blood and the walls of the ventricles which is presented as the 1st heart sound. During diastole, blood in the blood vessels tries to flow back to the ventricles causing the semi lunar valves to bulge. But the elastic recoil of the arteries cause the blood to bounce forward which vibrates the blood, the walls and the ventricular valves which is presented as the 2nd heart sound . The 3rd heart sound (S3) is heard in the mid diastole due to the blood that fills the ventricles. The 4th heart sound (S4), also known as atrial heart sound, occurs when the atrium contracts and pumps blood to the ventricles. S4 appears with a low energy and is almost never heard by the stethoscope [3].

      2. Methodology

         1. Input Heart signal Acquisition:

            Input heart signals for investigation are downloaded from standard biomedical website, these signals are converted into (.wav) format. Phonocardiogram is a graphic method of recording noises during his heart activity. Heart and vessels sounds are composed by audible and inaudible oscillations, but recordable. If a microphone specially designed to detect low-frequency sound is placed on the chest, the heart sounds can be amplified and recorded by a high-speed recording apparatus.

            Figure 2. Recording A is an example of normal heart sounds, showing the vibrations of the first, second and third heart sounds and even the very weak atrial sound. Note specifically that the third and atrial heart sounds are each a very low rumble. The third heart sound can be recorded in only one third to one half of all people, and the atrial heart sound can be recorded in perhaps one fourth of all people.

         2. Wavelet families:

            Figure 2: Phonocardiograms for normal (A) and abnormal heart sounds

            Wavelet analysis has practically become a ubiquitous tool in signal processing. Two basic properties, space and frequency localization and multi-resolution analysis, make this a very attractive tool in signal analysis. The wavelet transform method processes perfect local property in both time space and frequency space and it use widely in the region of vehicle Faults detection and identification.

            Several families of wavelets that have proven to be especially useful. Some wavelet Families are: Haar, Daubechies, Bi- orthogonal, Coiflets, Sym lets, Morlet's, Mexican hat, Meyer, Other real wavelets, complex wavelets

            • Haar:

              Any discussion of wavelets begins with Haar wavelet, the first and simplest. Haar wavelet is discontinuous, and resembles a step function. It represents the same wavelet as Daubechies db1.

            • Daubechies:

              Figure 3: Step Response for Haar Wavelet

              Ingrid Daubechies, one of the brightest stars in the world of wavelet research, invented what are called compactly supported orthonormal wavelets thus making discrete wavelet analysis practicable. The names of the Daubechies family wavelets are written dbN, where N is the order, and db the surname of the wavelet. The db1 wavelet, as mentioned above, is the same as Haar wavelet. Here is the wavelet functions psi of the next nine members of the family:

              Figure 4: Different Daubechies Wavelets

              • Bi-orthogonal Wavelet: This family of wavelets exhibits the property of linear phase, which is needed for signal and image reconstruction. By using two wavelets, one for decomposition (on the left side) and the other for reconstruction (on the right side) instead of the same single one, interesting properties are derived.

                Figure 5: Different Biorhogonal Wavelets

              • Coiflets: Built by I. Daubechies at the request of R. Coifman. he wavelet function has 2N moments equal to 0 and the scaling function has 2N-1 moments equal to 0. The two functions have a support of length 6N-1. You can obtain a survey of the main properties of this family by typing wave info ('coif') from the MATLAB command line

                .

                Figure 6: Various coiflets

              • Symlets: The symlets are nearly symmetrical wavelets proposed by Daubechies as modifications to the db family. The properties of the two wavelet families are similar. Here is the wavelet functions psi.

                Figure 7: Representing the Symlets

         3. PARAMETERS:

            • Variance:

              The variance of a random variable X is its second central moment, the expected value of the squared deviation from the mean = E[X]:

              This definition encompasses random variables that are discrete, continuous, neither, or mixed. The variance can also be thought of as the covariance of a random variable with itself:

              The expression for the variance can be expanded:

            • MAX ENTROPY:

              The statistics and information theory, a maximum entropy probability distribution is a probability distribution whose entropy is at least as great as that of all other members of a specified class of distributions.

            • Standard Deviation:

              The standard deviation (SD) measures the amount of variation or dispersion from the average. A low standard deviation indicates that the data points tend to be very close to the mean (also called expected value); a high standard deviation indicates that the data points are spread out over a large range of values.

            • ENERGY:

              Energy is a word with more than one meaning. Energy means something has the ability to cause change.

              Mostly it is used in science to describe how much potential a physical system has to change.

              It may also be used in economics to describe the part of the market where energy itself is harnessed and sold to consumers. It can sometimes refer to the ability for someone to act or speak in a lively and vigorous way.

            • Peak signal-to noise-ratio:

            Peak signal-to-noise ratio, often abbreviated PSNR, is an engineering term for the ratio between the maximum possible power of a signal and the power of corrupting noise that affects the fidelity of its representation. Because many signals have a very wide dynamic range, PSNR is usually expressed in terms of the logarithmic decibel scale.

            Signal-to-noise ratio (often abbreviated SNR or S/N) is a measure used in science and engineering that compares the level of a desired signal to the level of background noise. It is defined as the ratio of signal power to the noise power, often expressed in decibels. A ratio higher than 1:1 (greater than 0 dB) indicates more signal than noise. While SNR is commonly quoted for electrical signals, it can be applied to any form of signal (such as isotope levels in an ice core or biochemical signalling between cells).

      3. EXPERIMENTAL RESULTS

         Figure 8: Block Diagram for Proposed Scheme.

         1. SIGNAL ACQUISITION:

            Acquiring the heart beat signal from the database which is extracted from heart biometrics. DWT based biomedical systems has been developed by using the heart Sounds, obtained from a total of 5 heart sounds. Stages of feature extraction have been realized by using MATLAB R2012R software package. Using wave read function converting the heart signal into a graphical representation.

            Fig 9. Converting heart signal in to Graphical form

         2. Wave menu: By accessing wave menu comparing different wavelets along denoising the signal.

            Fig 10. Wave menu

            DWT has been used to derive feature Vectors from the heart sound signals. Selecting proper level of Wavelet and determining decomposition at best level play a significant role in the analysis of Heart Sound using discrete wavelet transform method. This decomposition levels is figured out based on the sampling frequency components of the signal. Using DWT as a filter at several levels helps in finding the best level of decomposition. There are apparent Differences between the graphics of the heart sound of a Normal subject and those from patients with Aortic Stenosis, Mitral Stenosis, Aortic Regurgitation & Mitral Regurgitation Diseases. These differences being reflected to heart sound Graphics are also reflected largely to DWT graphics. Therefore, such a classification system, established by taking such variances in DWT graphics into consideration enables to Decide on respective diseases.

         3. Calculation: After wave menu and denoising signal PSNR, Variance, Energy, Entropy were calculated.

         4. FLOW CHART

            START

            Acquiring Signals

            Converting Signals into Graphs

            Denoising using Wave menu

            Calculating PSNR, Variance, Energy and Entropy

            STOP

            1. A. Selecting Haar Wavelet:

      4. ANALYSIS OF DIFFERENT WAVELETS:

         Figure 11. Haar wavelet applying the input signal

         1. Comparison of original and denoised signal:

            Figure 12. Comparison of original and denoised signal

         2. Analyzing Signal with decomposition levels:

            Figure 13. Analyzing the signal for all decomposition levels

            Decomposition levels s =a10+d9+d8+d7+d6+d5+d5+d4+d3+d2+d1

         3. Calculation of Parameters:

         1. A. Selecting Biorthogonal Wavelets:

            Figure 14. Different parameters

            Figure 15. Biorthogonal wavelet applying the input signal

            1. Comparison of original and denoised signal:

               Figure 16. Comparison of denoised and original signal

            2. Analyzing Signal with decomposition levels:

               Figure 17. Analyzing the signal for all decomposition levels

               Decomposition Levels S =a10+d9+d8+d7+d6+d5+d5+d4+d3+d2+d1

            3. Calculation of parameters:

         2. A. Selecting Coiflet Wavelet:

            Figure 18. Different parameters

            Figure 19. Coiflet wavelet applying the input signal

            B. Comparison of denoised and original signal:

            Figure 20. Comparison of original and denoised signal

            Figure 21. Analyzing the signal for all decomposition levels

            Decomposition Levels S =a10+d9+d8+d7+d6+d5+d5+d4+d3+d2+d1

            D. Calculation of parameters:

         3. A. Selecting Symlet Wavelet:

         Figure 22. Different parameters

         Figure 23. Symlet wavelet applying the input signal

         B. Comparison of denoised and original signal:

         Figure 24. Comparison of original and denoised signals

         Figure 25. Analyzing the signal for all decomposition levels

         Decomposition Levels S =a10+d9+d8+d7+d6+d5+d5+d4+d3+d2+d1

         D. Calculation of parameters:

         Figure 26. Different parameters

      5. COMPARISON TABLES

         BIORTHOGONAL: PARAMETERS
         LEVEL	ENERGY	STD.DEVIATION	VARIANCE	RMS	MEAN	ENTROPY
         1	1.3333	72.32757	0.0552	102.2412	71.7038	1.0000
         2	1.2222	99.0206	0.0981	120.6622	70.5855	1.1345
         3	0.2222	129.1677	0.1668	148.8588	76.0548	1.1038
         4	0	167.4848	0.2805	186.8588	85.0539	1.0868
         5	1.7778	219.2796	0.4808	239.7027	99.4809	1.2263
         6	1.7778	289.9728	0.8408	312.5573	119.9432	1.5095
         7	0.8889	386.4620	1.4935	412.3109	147.8239	1.8041
         8	0	518.2618	2.6860	548.6257	185.2386	2.1766
         9	0.6667	698.5166	4.8793	374.9157	235.1667	2.3310
         10	0.6667	945.3923	8.9377	989.7441	301.7005	2.4729
         HAAR WAVELET:PARAMETERS
         LEVEL		ENERGY	STD.DEVIATION	VARIANCE	RMS	MEAN	ENTROPY
         1		0.6667	77.5479	0.0601	103.3729	72.6199	1.0000
         2		0.5556	96.3941	0.0929	108.9710	58.5455	0.9457
         3		0.6667	114.2607	0.1306	121.8419	53.6462	0.9183
         4		0.6667	120.9712	0.1463	121.8419	37.8276	0.8113
         5		0.7778	164.4227	0.2703	165.1935	48.3080	0.7793
         6		0.8889	224.1892	0.5026	224.8771	62.4415	0.7496
         7		1.2222	306.4556	0.9392	307.0739	81.4885	0.7219

         8	1.1111	419.8046	1.7624	420.3642	107.1675	0.6962
         9	1.4444	576.1471	3.3195	576.6566	141.8220	0.6723
         10	1.5556	792.0186	6.2739	792.4848	188.6489	0.6500
         COIFLET WAVELET:PARAMETERS
         LEVEL	ENERGY	STD.DEVIATION	VARIANCE	RMS		MEAN	ENTROPY
         1	0.8889	76.8953	0.0591	103.8852		72.8121	1.0000
         2	0.5556	100.1194	0.1002	117.7363		66.0721	0.9980
         3	0.3333	126.0577	0.1589	139.9195		66.1651	1.1916
         4	0.2222	164.7530	0.2714	178.9175		76.4024	1.1393
         5	0.1111	217.4429	0.4728	232.7706		91.2881	1.1546
         6	0	289.2748	0.8368	306.5477		111.7750	1.2784
         7	1.0000	387.3072	1.5001	407.3683		139.3936	1.3991
         8	0.2222	521.2832	2.7174	545.1303		176.3294	1.4917
         9	1.0000	704.6552	4.9654	733.5294		225.5993	1.5744
         10	0.4444	986.0221	9.1398	991.5075		291.3196	1.5374
         SYMLET WAVELET:PARAMETERS
         LEVEL	ENERGY	STD.DEVIATION	VARIANCE	RMS	MEAN	ENTROPY
         1	0.8889	78.2069	0.0612	105.4748	74.2853	1.0000
         2	0.4444	100.6584	0.1013	115.0163	61.8091	0.9852
         3	0.6667	124.5850	0.1552	133.6927	57.6402	0.9544
         4	1.0000	163.3963	0.2670	173.2281	68.6646	0.9819
         5	1.1111	216.4295	0.4684	227.4835	83.8819	0.9940
         6	1.0000	288.8590	0.8344	301.6666	104.4068	0.9988
         7	1.7778	387.8446	1.5042	403.0321	131.8471	1.0000
         8	1.6667	523.2856	2.7383	541.6296	168.4277	0.9992
         9	1.5556	708.8686	5.0249	731.3587	217.1863	0.9975
         10	1.0000	963.5249	8.2838	991.4417	282.2549	0.9953

         Table 1: Comparison of different wavelets

      6. CONCLUSION

         Different wavelets are considered for the analysis of heart sounds such as HAAR, DB, Bior, Ortho, Coiflets and Symmlets. In this paper PSNR, Variance, Mean, Entropy, Energy and denoising the values along level 10decomposition has been calculated and shown the graphs above section. The decomposition and Denoising are calculated and plotted under the basis of wave menu from MATLAB. The factors are calculated and were shown their analysis under different wavelets. So, the maximum probable output has been shown and plotted as bar graphs even histogram levels were calculated and plotted in this paper. After completing all dwt methods it is found that coiflet is the best method among the various wavelet methods

      7. REFERENCES:

1. Todd R. Reed, Nancy E. Reed, Peter Fritzson, Heart sound analysis for symptom detection and computer-aided diagnosis, Simulation Modeling Practice and Theory 12 (2004) 129146

2. Guy Amit, Noam Gavriely, Nathan Intrator, Cluster analysis and classification of heart sounds, Biomedical Signal Processing and Control 4 (2009) 2636

3. Faizan Javed, P A Venkatachalam, Ahmad Fadzil M H, A Signal Processing Module for the Analysis of Heart Sounds and Heart Murmurs, Journal of Physics: Conference Series 34 (2006) 10981105

4. Sumeth Yuenyong, Akinori Nishihara, Waree Kongprawechnon, Kanokvate Tungpimolrut, A framework for automatic heart sound analysis without segmentation, Biomed Eng Online. 2011; 10: 13 Published online 2011 February 9

5. Dr. H. B. Kekre, Tanuja Sarode, Shachi Natu & Prachi Natu Performance Comparison Of 2-D DCT On Full/BlockSpectrogram And 1- D DCT On Row Mean Of Spectrogram For Speaker Identification, International Journal of Biometrics and Bioinformatics (IJBB), Vol (4): Issue 3 [6] Dr. H B Kekre , Vaishali Kulkarni, Speaker Identification using Row Mean of Haar and Kekres Transform on Spectrograms of Different Frame Sizes, (IJACSA)

   International Journal of Advanced Computer Science and Applications Special Issue on Artificial Intelligence

6. http://www.peterjbentley.com/heartchallenge/index.html