Heart Sound With Wavelet

8/4/2019 Heart Sound With Wavelet

http://slidepdf.com/reader/full/heart-sound-with-wavelet 1/13

ORIGINAL PAPER

Software Development for the Analysis of Heartbeat Sounds

with LabVIEW in Diagnosis of Cardiovascular Disease

Taner Topal & Hüseyin Polat & İnan Güler

Received: 31 January 2008 /Accepted: 13 March 2008 /Published online: 15 April 2008# Springer Science + Business Media, LLC 2008

Abstract In this paper, a time-frequency spectral analysis

software (Heart Sound Analyzer) for the computer-aidedanalysis of cardiac sounds has been developed with Lab-

VIEW. Software modules reveal important information for

cardiovascular disorders, it can also assist to general

physicians to come up with more accurate and reliable

diagnosis at early stages. Heart sound analyzer (HSA)

software can overcome the deficiency of expert doctors and

help them in rural as well as urban clinics and hospitals.

HSA has two main blocks: data acquisition and pre-

processing, time – frequency spectral analyses. The heart

sounds are first acquired using a modified stethoscope

which has an electret microphone in it. Then, the signals are

analysed using the time – frequency/scale spectral analysis

techniques such as STFT, Wigner – Ville distribution and

wavelet transforms. HSA modules have been tested with

real heart sounds from 35 volunteers and proved to be quite

efficient and robust while dealing with a large variety of

pathological conditions.

Keywords Phonocardiogram signal . Cardiac sound

analysis . FFT . STFT . Wigner – Ville distribution .

Wavelet transform

Introduction

Auscultation is a technique in which a stethoscope is used

to listen the heart sounds of a body. Classic cardiac

auscultation has been the most marked method used in

diagnosing many cardiovascular diseases since two centu-ries. It is still a fundamental and very important diagnostic

tool in the diagnosis of heart disease, since it is cheap and

noninvasive. The structural abnormalities of the heart are

frequently reflected in the sounds produced by the heart.

Physicians use the stethoscope as a device to listen the

heart sounds and make a proper diagnosis. However, the

auscultation of heart sound signals through either a

conventional acoustic or an electronic stethoscope needs

a long-term practice and experience. Cardiologists are

specifically interested in abnormal sounds, which may

propose the presence of a cardiac pathology and also give

diagnostic information. For instance, a very important type

of abnormal sound is murmur caused by the turbulent

flow of blood in the cardiovascular system. The timing

and pitch of a murmur are of significant importance in the

diagnosis of a heart condition, for example murmur during

diastole are signs of malfunctioning of heart valves but

murmur during systole may correspond to either a

pathological or healthy heart, depending on the acoustic

characteristics of the murmurs [1, 2].

Today, electronic stethoscopes with built-in filters make

auscultation more precise and convenient. Murmurs and

tones are separated easily, but background sound, under the

threshold of audibility, still covers weak murmurs, so that

the human ear is not able to identify them. Often, more than

one valve is insufficient and/or stenotic, which makes the

classification of heart sounds altered by valve diseases very

complex and difficult [1, 2].

A time – frequency distribution (TFD) shows the spectral

composition of a signal at some particular time instant. It is

usually derived from a time – frequency representation (TFR),

such as short time Fourier transform (STFT), Wigner – Ville

distribution (WVD) and Wavelet transform (WT) [3].

J Med Syst (2008) 32:409 – 421

DOI 10.1007/s10916-008-9146-8

T. Topal : H. Polat : İ. Güler (*)

Electronics and Computer Education Department,

Faculty of Technical Education, Gazi University,

Ankara, Teknikokullar 06500, Turkey

e-mail: [email protected]



Some authors have point out that the wavelet based

scalogram analysis of phonocardiograms is more successful

than the spectrogram or Wigner – Ville distribution due to

the strong non-stationarity of that signal [4]. Time –

frequency/scale methods have been applied to characterize

heart sounds [3, 5, 6]. In the literature, the authors have

discussed the characterization of heart murmurs using time –

frequency methods over a number of cardiac cycles [4 – 9].In this study, the STFT, WVD and WT were applied to

heart sound signals obtained from 35 volunteers and collected

from online clinical training website [10 – 12]. To visualize

the frequency band activities depending on time, signals are

processed with time – frequency-based STFT, WVD and

continuous wavelet transform (CWT) method. The applied

methods were compared in terms of their frequency

resolution and the effects in determining abnormality in

heart sound signals. For this purpose, we developed a

software package using LabVIEW. In order to obtain

clinically interpretable results, frequency band activities of

the signals were mapped onto frequency – time axes usingthe STFT, WVD, and WT. The performance analysis by the

wavelet transform over other techniques in analyzing heart

sounds of the phonocardiogram signals is presented.

Materials and methods

Heart sound auscultation

The heart sounds are primarily generated from blood

turbulence. The blood turbulence occurs due to fast

accelerations and decelerations of the blood in the cham-

bers and arteries caused by the contraction or closure of the

heart valves, which in turn produce mechanical vibrations

that propagate through the body tissues up to the surface of

the chest. Basic heart sounds mostly occur in the frequency

range of 20 – 200 Hz. Some heart murmurs produce the

sound around 1 kHz. Shown in Fig. 1, auscultation points

generally correlates with a cardiac valve and thus enables

detection of murmur associated with valvular abnormalities.

Auscultation of the precordium with a stethoscope will

reveal an audible first heart sound (S1) and second heart

sound (S2). These normal heart sounds are generated valve

closures as shown in Fig. 2. Closures of the mitral (M1) and

tricuspid (T1) valves produces S1, which is heard best at

the apex of the heart. A split S1 may be heard along the left

lower sternal border, where the tricuspid component might

also be audible. Abnormal S1 sound occur when there is

disease of the mitral valve. For example, the S1 may be

loud and have a “snapping” quality in patients with mitral

stenosis as a result of rapid closure of the mitral valve;

conversely, S1 may be diminished in the presence of mitral

regurgitation [1, 2].

S2 correlates with closure of the aortic (A2) and

pulmonic valves (P2) and is heard best at the base of the

heart. Physiologic splitting of S2 occurs as a result of aortic

valve closure preceding pulmonic closure. The splitting of S2 is maximal at the end of inspiration and heard best at the

second left interspace (parasternally). Increased splitting at

the end of inspiration is resulted from a delayed pulmonic

closure which is caused by right ventricular filling during

the inspiration. Fixed splitting of S2 can occur in patients

with atrial septal defect, pulmonic stenosis, and right

bundle branch block. True fixed splitting occurs only with

atrial septal defect (i.e., does not vary with respirations); in

pulmonic stenosis and right bundle branch block, there is a

widened split S2 that does vary with inspiration.

S3 is low pitched, is heard best at the apex with the bell

of the stethoscope, and is usually not present in healthy

adults. An S3 can be a normal variant in children and may

persist into young adulthood. S4 is a presystolic low-

pitched sound occurring just before S1 that is heard best at

the apex with the bell of the stethoscope; it is normally not

present in healthy persons. A heart murmur is defined

according to its intensity, frequency, quality, configuration,

timing, duration, and radiation. Murmurs can be either

systolic or diastolic as shown in Fig. 2. The intensity or

loudness is graded on a 6-point scale. Grade 1 murmurs are

very faint generally heard only after focusing on the sound.

Grade 2 murmurs are faint but heard immediately on

placing the stethoscope on the precordium. Grade 3

murmurs are moderately loud, whereas grade 4 murmurs

are loud. Grade 5 murmurs are very loud and can be heard

when the stethoscope is partly off the precordium. Grade 6

murmurs are also very loud and can be heard with

stethoscope completely off the precordium. Grade 4 and

higher murmurs also are associated with a thrill. However,

in most of the heart sound analysis, the information of the

first heart sound and the second heart sound are playing an

important rule in cardiac auscultation [1, 2, 4].

Fig. 1 Point of auscultation, AR Aortic region; TR tricuspid region;

PR pulmonic region; MR mitral region; 1 right second intercostal

space; 2 left second intercostal space; 3 midleft sternal border; 4 fifth

intercostal space, mid clavicular line

410 J Med Syst (2008) 32:409 – 421



Hardware

Modified stethoscope with electret microphone, micro-

phone preamplifier circuit, SCXI (SCXI-1000, SCXI

1125) and DAQ (PCI-MIO-16XE-10) modules are used as

hardware components. As seen in Fig. 3, some of the heart

sounds were recorded by a self-produced electronic

stethoscope from 35 volunteers. Fourteen female and 21

male person in the 18 – 40 ages range are joined this study.

A heart sound of 15 s duration was measured successively

on four different auscultation areas as shown in Fig. 1. As

illustrated in Fig. 3, the measurements were taken with a

modified classical stethoscope (frequency bandwidth

20 Hz – 20 kHz), combined with NI hardware system.

An electret microphone, preamplification and adjustable

bias voltage stages were used for classical stethoscope

modification and placed before NI hardware system as

shown in Fig. 4. Recordings were performed using self-developed measurement and analysis software based on

LABVIEW graphical programming language [13, 14]. The

DAQ card (PCI-MIO-16XE-10) can digitize the signal with

128 kHz sampling frequency/16-bit resolution. To extract

highly noisy signals with any audible disturbance from

heart sounds, dual modified stethoscope were used. One of

them acquired heart sounds and the other acquired medium

sounds (i.e, noise). Then, to achieve noise cancellation,

medium sounds extracted from heart sounds. Some abnor-

mal heart sounds collected from training CD-ROMs and

online clinical training websites are used for testing and

software validation [10 – 12].

Software

In this study, LabVIEW graphical programming language is

used. All self-produced software modules are developed

with LabVIEW development system. LabVIEW is a user

friendly and powerful graphical development environment

used for signal acquisition, measurement analysis, and data

presentation [14]. By incorporating LabVIEW, the results in

this work can be obtained faster, more precise and also

easier. A main component of LabVIEW is its ability tosimulate a virtual instrument (VI). The term “virtual

instrumentation” is used to represent a PC-based control

system which is used to acquire data from physical

transducers and then manipulate them in specific ways

using a very high level graphical environment. In the

Fig. 3 Block diagram of the data acquisition system

Fig. 2 Normal and abnormal

heart sounds representations

and pathophysiology of systolic

and diastolic murmurs

J Med Syst (2008) 32:409 – 421 411411



graphical environment, symbolic icons are used that operate

in the same way as real instruments do. It involves using

specific programming methods and a combination of

different measuring hardware to simulate almost any

instrument in the computer. A virtual instrument is a system

comprising of specially designed software and also a

hardware that will convert the input from measuring to

digital data to be processed by the computer. The user will

employ a computerized test and measurement to collect any

data that will be displayed on the computer screen. VIs also

could influence systems where the processes are controlled

based on data collected and processed by a computerizedinstrumentation system.

Heart sound signal processing techniques

Short-time Fourier transform

Fourier analysis decomposes a signal into its frequency

components and determines their relative strengths. The

Fourier transform is defined as

F wð Þ ¼ R 1À1 f t ð ÞeÀ j wt dt , f t ð Þ

¼ 12π

R 1À1 F wð Þe j wt d w: ð1Þ

This transform is applied to stationary signals, that is,

signals whose properties do not evolve in time. When the

signal is non-stationary we can introduce a local frequency

parameter so that local Fourier transform looks at the signal

through a window over which the signal is approximately

stationary. Therefore, we applied the STFT to the heart

sound signals under study. The STFT positions a window

function Ψ (t ) at τ on the time axis, and calculates the

Fourier transform of the windowed signal as

F w; t ð Þ ¼Z À1

1 f t ð Þy * t À t ð ÞeÀ j wt dt : ð2Þ

When the window y (t ) is a Gaussian function, the STFT

is called a Gabor transform. The basis functions of this

transform are generated by modulation and transformationof the window function y (t ), where w and t are modulation

and translation parameters, respectively. The fixed time

window y (t ) is the limitation of STFT as it causes a fixed

time – frequency resolution [3, 8, 13, 15].

Wigner – Ville distribution (WVD)

Wigner – Ville distribution is a distribution directly related to

the signal correlation function [16, 17]. Hence when, for

instance, two frequency components are present, a term

reflects the correlation of the two signal components will

exist in the middle of the two components in the time –

frequency plane. The amplitude of cross-term can be double

that of the original signal, and hence it will severely distort

the interpretation of the original signal. Thus, WVD does

not fulfill the frequency/time support property. This is the

other shortcoming of WVD. Much works have been carried

out to overcome this artifact. One of the most direct ways is

to convert the real signal into an analytical signal.

Generally, for any complex function x the time-dependent

WVD is defined

WVD x t ;w

ð Þ ¼ Z 1

À1 x t

þt

2 x* t

Àt

2 eÀ j wt d t

ð3

Þwhere * denotes the complex conjugate. WVD maps a one-

dimensional function of time into two-dimensional function

of time and frequency [3, 16, 17]. In real applications,

evaluation from minus infinite to plus infinite is impossible.

To overcome this problem, we can impose a running

window h(t ), such as

PWVD x t ;wð Þ ¼Z 1

À1h t ð Þ x t þ t

2

x* t À t

2

eÀ j wt d t

ð4Þ

which is called the pseudo Wigner – Ville distribution. This

windowing operation will result as frequency smoothing in

time domain. The effect of the smoothing is twofold: it can

substantially suppress the cross-terms, but on the other hand

smoothing will reduce the resolution of the WVD [3, 16 – 18].

Wavelet transform analysis

For a continuous input signal, the time and scale parameters

can be continuous, leading to the continuous wavelet

Fig. 4 Schematic representation of the preamp and bias voltage

circuit for electret mic. used in the system

412 J Med Syst (2008) 32:409 – 421





where j , k ∈Z. Hence we can express any f (t )∈ L2( R) as the

superposition

f t ð Þ ¼X

j

Xk

d j ;k y j ;k t ð Þ ð8Þ

where the wavelet coefficient is defined as

d j ;k ¼ f t ð Þ;y j ;k t ð ÞD E

¼ 1

aÀ j =20

Z f t ð Þy a

À j 0 t À kb0

À Ádt ð9Þ

The following is an analogy: assume that the wavelet

analysis is like a microscope. First, one chooses the

magnification, that is aÀ j 0 . Then one moves to the chosen

location. Now, if one looks at very small details, then the

chosen magnification is large and corresponds to j

negative and large. Then aÀ j 0 b0 corresponds to small steps,

which are used to catch small details. As a result,

particularly in researching cardiovascular abnormalities by

using heart sound signals, WT is very useful in the

Fig. 6 Displaying and record-

ing heart sound module of HSA

(belonging to a 36 year old male

healthy volunteer)

Fig. 7 One cardiac cycle (S1 –

S2 heart sounds) extracted from

normal cardiac sound data is

loaded from wave file

414 J Med Syst (2008) 32:409 – 421



expression of discontinuities caused by recording apparatus

[6 – 8, 13, 15, 20].

Description of software and application results

A packet program, Heart Sound Analyzer (HSA), is

designed and developed upon the hardware with LabVIEW

programming language. The sofware is consist of two

sections: acquiring and recording of the displayed heart

sounds, and data analysis of the recorded heart sounds. All

program modules are developed and written in ourselves,

no LabVIEW addons are used. In the application of

program modules, aortic regurgitation (AR), mitral stenosis

(MS), ventricular septal defect (VSD), pulmonic stenosis

(PS) and normal (N) cardiac sounds are used. All defective

heart sound type arbitrarily chosen. Cardiac sounds data

obtained from 18 – 40 ages range volunteers, internet and

cardiac auscultation training CD-ROMs [10 – 12].

Heart sound analyzer software

In Fig. 5, Software running hierarchy is shown as

flowchart. Acquired heart sound data can be recorded to

Fig. 8 Main HSA screen after

reading cardiac sound data from

file

Fig. 9 PSD and phase spectrum

of the normal heart sound in the

HSA amplitude and phase spec-

trum module

J Med Syst (2008) 32:409 – 421 415415



hardisk and loaded from harddisk for data analysis as well

as other heart sound data. All modules are called from HSA

main module and will be explained following sections

Displaying and Recording Heart sounds With modified

stethoscope, acquired heart sounds from healthy and

unhealthy person, as digitized by DAQ card are displayed

and also can be recorded if it is desired. In this work,sampling frequency and recording interval are conveniently

adjustable. Sampling frequency is selected as 8 kHz which

is sufficient to digitize the heart sound data. In the HSA

module, heart sound data can be recorded both in wav and

txt file formats. In Fig. 6, it is seen the heart sound

recording belonging to a 36 year old male healthy

volunteer.

Analysis of the Cardiac Sounds In these HSA modules,PSD, STFT, WVD and CWT analyses are made as offline.

a S1 S2

normal cardiac sound normal cardiac sound

S1 S1

c

A2

P2

A m p l i t u d e

(Hz)

b

M1

T1

A m p l i t u d e

(Hz)

S2 S2

Fig. 10 a One cardiac cycle of the normal heart sound and its two dominant parts S1 (first) and S2 (second) sounds. b PSD of the first (S1)

cardiac sound. c PSD of the second (S2) cardiac sound

416 J Med Syst (2008) 32:409 – 421



Recorded data are loaded from file and then processed.

Amplitude and phase functions, spectrograms and scalo-

grams of the heart sounds are linearly plotted by HSA.

Read data from file Recorded heart sound data are loaded

as array from wave or text file for time – frequency analysis.

Sound data are divided by the maximum value of the array

thus scaled [−1,1] data range. The reading heart sound data

from file HSA module is shown in Fig. 7. As it is seen, any

portion of data remaining between two cursors can be

extracted from the signal waveform. Both signal waveform

and extracted waveform are transported to main HSA

screen as in Fig. 8. Then, extracted waveform part is used

for analyses.

Fig. 11 STFT spectrogram of one cardiac cycle of the normal heart sound, main screen and different color table plots

(a) (b) (c)

(d) (e)

Fig. 12 One cardiac cycle STFT spectrograms of a N, b AR, c MS, d PS and e VSD heart sounds, respectively

J Med Syst (2008) 32:409 – 421 417417



Amplitude and phase spectrum HSA module application

HSA amplitude and phase spectrum modules are shown in

Fig. 9. Log/linear input specifies linear or log spectrum

output. Display unit selection is the output unit for the

spectrum. Display unit can be set to one of the values listedin the display unit section such as amplitude spectral

density and power spectral density (PSD).

In window section, many well-known window types

such as Hanning, Hamming, Blackman, Blackman – Harris,

Gaussian etc. are listed. Window selection specifies

information about the window and it is used to eliminate

the spectral leakage while computing amplitude and power

spectrum.

The cardiac sound signal is clearly contained two

principal sounds (S1 and S2) as is demonstrated in

Fig. 10a. The FFT can be applied to the normal first and

second sounds to analyse the frequency spectrum as shown

in Fig. 10 b and c. A 1,024 point FFT is applied with

window function. The two internal components for thecardiac sound S1 (M1 and T1) and the two components A2

and P2 of the sound S2 are apparent in Fig. 10 b and c, but

FFT analysis cannot give the time delay between these

internal components. Therefore, the conventional FFT and

PSD are unable to diagnose heart diseases accurately.

Hence, it is necessary to seek a transform which is named

time-varying spectrum. The WVD can give better results

under the same conditions and same sampling rate [8, 18].

Cross-terms

S1 S2 S1 S2

a bFig. 13 a WVD and b PWVD

spectrograms of the normal car-

diac sound (one cardiac cycle)

(a) (b) (c)

(e)(d)

Fig. 14 WVD spectrograms (one cardiac cycle) of a N, b AR, c MS, d PS and e VSD heart sounds, respectively

418 J Med Syst (2008) 32:409 – 421



STFT HSA module application

In Fig. 11, HSA STFT modul front panel is shown. A 1,024

point FFT is applied with window function. In the STFT

calculation, 1,000 point overlap (97%) is used between the

Window frames. Sampling frequency, overlap rate andwindow filter type are appropriately adjustable. The

maximum of Y scale can be changed with slider. As it

shown at the right part of the Fig. 11, color table can also be

changed from options if desired.

Figure 12 shows the STFT spectrograms of the normal

(N), aortic regurgitation (AR), mitral stenosis (MS),

pulmonic stenosis (PS) and ventricular septal defect

(VSD) cardiac sounds.

WVD HSA module application

In the WVD and PWVD analysis, number of data aredecimated to 1,024 point. When the data point below the

1,024, then zero paddind is used. Colour table of the graph

and analysis method are chosen from HSA front panel.

Frequency scale (vertical axis) is also adjustable, as the

others. Figure 13 shows the WVD and PWVD spectro-

grams of the HSA modul with normal heart sound. It can be

(a) (b) (c)

(e)(d)

Fig. 15 PWVD spectrograms (one cardiac cycle) of a N, b AR, c MS, d PS and e VSD heart sounds, respectively

(a) (b)

Fig. 16 HSA CWT modul front panel with normal a S1 sound and b S2 sound scalograms

J Med Syst (2008) 32:409 – 421 419419



noticed that it is difficult to discern the sound S1 and soundS2 as seen in Fig. 13a. because of the cross-terms. With the

Gaussian filtering which can assist to suppress the cross-

terms (PWVD), as it is seen in Fig. 13 b, both main

components S1 and S2 sounds of the cardiac cycle are

appeared clearly. WVD and PWVD methods results may be

improved by increasing the sampling rate of original signal

or using different filtering type, but it may be still possible

to see the cross-term effects because of the nonlinearity of

the WVD. The WVD and PWVD analysis applied to heart

sound signal and they provides high time-and frequency-

resolution in spectral analysis.

In Figs. 14 and 15, applications of the WVD andPWVD methods on the normal (N), aortic regurgitation

(AR), mitral stenosis (MS), pulmonic stenosis (PS) and

ventricular septal defect (VSD) heart sounds are shown

respectively.

CWT HSA module application

HSA CWT module is shown in Fig. 16 with two principal

component S1 and S2 sounds of the normal heart sound. To

achieve CWT, the Morlet function is used as a mother wavelet. This is a smooth, symmetric and having a

compact support properties wave [9, 13, 20]. Five hundred

twelve point data frame has been used in CWT calculations

(W nop N1). One hundred twenty-eight point scale

resolution is applied. Color table of the graph and scale

maximum can be changed conveniently.

As seen in Fig. 16a and b the time delays between M1

and T1 components for the sound S1 and A2 and P2 for the

sound S2 these internals components can be easily

measured by the experts. The wavelet transform permits

us to estimate and determine this time difference easily [8,

9, 13].Figure 17 shows the continuous wavelet transform of the

selected heart sounds ((a) N, (b) AR, (c) MS, (d) PS and (e)

VSD heart sounds, respectively.). In one cardiac cycle of

the normal heart beat sound, the first S1 and the second S2

(Fig. 17a) sounds are clearly observed with CWT method

scale-time representation.

All results were compared, and it was determined that

the STFT was more applicable for real-time processing of

heart sound signals, because of its short process time.

(a) (b) (c)

(d) (e)

Fig. 17 CWT scalograms (one cardiac cycle) of a N, b AR, c MS, d PS and e VSD heart sounds, respectively

420 J Med Syst (2008) 32:409 – 421



However, the WVD and CWT had good resolution and

performance high enough for use in clinical and research

settings. Especially CWT was the best frequency – time

characteristic over the counterparts.

Conclusion

LabVIEW is a very powerful tool for data acquisition.

LabVIEW has benefit specifications such as flexibility,

future improvement, and fast software development. Data

acquired can be easily manipulated and processed to further

gain a more detailed analysis. It is customizable easily to

include any improvements to the virtual instrument as seen

necessary.

In this paper a time – frequency spectral analysis modules

for the computer-aided analysis of heart sounds has been

developed with LabVIEW. The HSA modules exhibits

important information of cardiovascular disorders and can

assist general physicians to come up with more accurate andreliable diagnosis at early stages. It can overcome the

deficiency of expert doctors and help them in rural as well

as urban clinics and hospitals. The software has two main

blocks: data acquisition and pre-processing part and time –

frequency spectral analysis modules. The heart sounds are first

acquired and recorded using a modified stethoscope which has

an electret microphone in it. Then the signals are analysed

using the time – frequency/scale spectral analysis techniques

such as STFT, Wigner – Ville distribution and wavelet trans-

form. All modules have been tested with real heart sounds

from 35 volunteers, heart sounds data form internet and

training CD-ROMs and has proved to be quite efficient and

robust while dealing with a large variety of pathological

conditions. Examples of the normal (N), aortic regurgitation

(AR), mitral stenosis (MS), pulmonic stenosis (PS) and

ventricular septal defect (VSD) cardiac sounds are presented.

Acknowledgements This study has been supported by the Scientific

and Research Projects Department of Gazi University (BAP, Project

no: 07/2007-07).

References

1. Karnath, B., and Thornton, W., Review of clinical signs:

Auscultation of the heart. Hospital Physician. 38(9):39 – 43,

2002, (September).

2. Tavel, M., Cardiac auscultation: A glorious past — but does it have

a future? Circulation. 93:1250 – 1253, 1996.

3. Akay, M., Time frequency and wavelets in biomedical signal

processing. Series in biomedical engineering . IEEE Press, New

York, pp. 271 – 301, 1997.

4. Obaidat, M. S., Phonocardiogram signal analysis: techniques and

performance comparison. Journal of Medical Engineering &

Technology. 17:221 – 227, 1993.

5. Jandre, F. C., & Souza, M. N., Wavelet Analysis of Phonocardio-

grams: Differences between Normal and Abnormal Heart sounds,Proceedings — 19th International Conference — IEEE/EMBS

4:1642 – 1644, Chicago IL, USA, 1997

6. Novak, P., and Novak, V., Time/frequency mapping of the heart

rate, blood pressure and respiratory signals. Medical & Biological

Engineering & Computing . 31:103 – 110, 1993.

7. Debbal, S. M., and Bereksi-Reguig, F., Automatic measure of the

split in the second cardiac sound by using the wavelet transform

Technique. Computers in Biology and Medicine. 37:269 – 276,

2007.

8. Debbal, S. M., and Bereksi-Reguig, F., Time – frequency analysis

of the first and the second heartbeat sounds. Applied Mathematics

and Computation. 184:1041 – 1052, 2007.

9. Debbal, S. M., and Bereksi-Reguig, F., Analysis of the second

heart sound using continuous wavelet transform. Journal of

Medical Engineering & Technology. 28:151 – 156, 2004.

10. Nakao, K., Online bed side learning: Heart sound auscultation,

http://www.medic.mie-u.ac.jp/student/sinnzou.html.

11. Sawayama, T., Auscultation training by CD: Heart Sound .

Nankodo, Japan, 1994.

12. University of Wales College of Medicine (UWCM)’s Database,

http://mentor.uwcm.ac.uk:11280/aspire/ .

13. K ıymık, M. K., Güler, İ., Dizibüyük, A., and Ak ın, M.,

Comparison of STFT and Wavelet transform methods in deter-

mining epileptic seizure activity in EEG signals for real-time

application. Computers in Biology and Medicine. 35:603 – 616,

2005.

14. Johnson, G. W., LabView graphical programming: practical

applications in instrumentation and control , 2nd edition.

McGraw-Hill, New York, 1997.

15. Bulgrin, J. R., Rubal, B. J., Thompson, C. R., and Moody, J. M.,

Comparison of short-time Fourier, Wavelet and time domain

analyses of intercardiac sounds. Biomedical Sciences Instrumen-

tation. 29:465 – 472, 1993.

16. Hlawatsch, F., and Flandrin, P., The interference structure of

the Wigner distribution and related time – frequency signal repre-

sentations. In: MecklenbraukerW., and HlawatschF. (Eds.), The

Wigner distribution — theory and applications in signal processing

Elsevier, New York, pp. 59 – 133, 1997.

17. Qian, S., and Chen, D. P., Joint time – frequency analysis. IEEE

Signal Processing Magazine. 16(2):52 – 67, 1999, Mar.

18. Kudriavtsev, V., Polyshchuk, V., and Roy, D. L., Heart energy

signature spectrogram for cardiovascular diagnosis. BioMedical

Engineering OnLine. 6:16, 2007, (May).

19. Daubechies, I., Ten lectures on wavelets. Society for Industrial

and Applied Mathematics, Philadelphia, PA, USA, 1992.

20. Olivier, R., and Duhamel, P., Fast algorithms for discrete and

continuous wavelet transforms. IEEE Transactions on Information

Theory. 38(2):569 – 586, 1992.

J Med Syst (2008) 32:409 – 421 421421

http://www.medic.mie-u.ac.jp/student/sinnzou.html

http://mentor.uwcm.ac.uk:11280/aspire/

http://mentor.uwcm.ac.uk:11280/aspire/

http://www.medic.mie-u.ac.jp/student/sinnzou.html

Documents

Heart Sound With Wavelet