Upload
abubuker-sidique
View
216
Download
0
Embed Size (px)
Citation preview
8/2/2019 2nd 4
1/8
i
ABSTRACT
Phonocardiographic signals contain bioacoustic informationreflecting the operation of the heart. Normally there are two heart sounds,
systole and diastole. If a third heart sound is present it could be a sign of
heart failure whereas a murmur indicates defective valves or an orifice in
the septal wall. The primary aim of this thesis is to find the heart diseases
from their heart sounds. In order to improve the automatic diagnosis
capabilities of auscultations, signal processing algorithms are developed.
Acoustic heart signals are generated by the mechanical processes of the
cardiac cycle, carry significant information about the underlying
functioning of the cardiovascular system. We describe a computational
analysis framework for identifying distinct morphologies of heart sounds
and classifying them into physiological states. The analysis framework is
based on hierarchical clustering, compact data representation in the
feature space of cluster distances and a classification algorithm. A novel
method of histogram matching algorithm based on structural and
perceptual features such as Zero-Crossing Rate (ZCR), Mel-Frequency
Cepstral Coefficient (MFCC), Fast Fourier transform (FFT), power
spectrum analysis are discussed. And based on this algorithm the heart
sounds are classified using matlab toolbox. The proposed framework
could be a potential solution for automatic analysis of heart sound that
may be implemented in real time for classification of heart sounds.
8/2/2019 2nd 4
2/8
ii
TABLE OF CONTENTS
CHAPTER CONTENTS PAGE
NO NO
ABSTRACT iLIST OF FIGURE iv
LIST OF TABLES v
ACRONYMS vi
1. 1.1.INTRODUCTION 1
1.1.1.CARDIOVASCULAR ANATOMY AND
PHYSIOLOGY 1
1.1.2.THE HEART VALVES 3
1.1.3.PHONOCARDIOGRAPHY 4
1.1.4.ORIGIN OF HEART SOUND 6
1.1.5.VALVULAR HEART DISEASES 7
1.2.PROBLEM STATEMENT 11
1.3.PROPOSED SOLUTIONS 121.4.IMPLEMENTATION 13
1.5.OBJECTIVES 15
1.6.ORGANISATION OF THE REPORT 15
2. 2.1.LITERATURE REVIEW 16
3 3.1METHODOLOGY 27
3.2SEGMENTATION 28
3.3FEATURE EXTRACTION AND THEIR CHARACTERISTICS 29
3.3.1MEL-FREQ CEPSTRAL COEFFICIENTS (MFCC) 33
3.3.2POWER SPECTRAL DENSITY 35
3.3.3FAST FOURIER TRANSFORM (FFT) 36
3.3.4ZERO-CROSSING RATE 38
8/2/2019 2nd 4
3/8
iii
3.4SUPPORT VECTOR MACHINE CLASSIFIER 39
3.4.1DATA PREPROCESSING 40
3.4.2SCALING 40
3.4.3RBF KERNEL 40
4 4.1MATLAB SIMULATOR 42
4.2SIGNAL PROCESSING TOOLBOX 42
4.3SEGMENTATION IN MATLAB 43
4.4FEATURE EXTRACTION OF PCG SIGNALS 44
4.4.1ZERO CROSSING RATE 45
4.4.2HISTOGRAM MODELING 45
4.4.3CEPSTRAL ANALYSIS 46
4.4.4FFT-BASED TIME-FREQUENCY ANALYSIS 47
4.4.5DISCRETE FOURIER TRANSFORM 48
4.4.6SPECTRAL ANALYSIS 484.5K-MEANS CLUSTERING 50
4.6CLASSIFICATION OF UNKNOWN MODULATED
SIGNAL 51
5 5.1 RESULTS 87
5.2CONCLUSION 98
REFERENCES 100
8/2/2019 2nd 4
4/8
iv
LISTOFFIGURES
FIGURE TITLE PAGE
NO NO
1.1 Anatomy of the heart (left figure) and the blood flowpathways through left and right side of the heart 2
1.2 Illustration of the mitral valve and its associated chordaetendineae and papillary muscles and the heart valves
and the fibrous rings surrounding each valve. 4
1.3 Implementation process of heart sound classification 143.1 Design model of heart sound classification 28
3.2 Flow chart of Mel-Freq Cepstral Coefficients 35
3.3 Representation of s1,s2,s3,s4 using fast fourier transform. 37
3.4 Representation of Heart Sound In terms of ZCR 39
4.1 Segmentation of input signal 444.2 Histogram of Normal and Arotic-regugration heart sound
Using (MFCC) 46
4.3 Example of waveforms and MFCC spectrograms of S1. 47
8/2/2019 2nd 4
5/8
v
LIST OF TABLE
3.1 Feature Extraction And Their Characteristics 30
8/2/2019 2nd 4
6/8
vi
ACRONYMS
PCG - Phonocardiograph
MRI - Magnetic resonance imaging
TFR - Time-Frequency Representation
S1 - First heart second
S2 - Second heart second
S3 - Third heart sound
S4 - Fourth heart sound
A2 - aortic component
P2 - pulmonary component
FHR - Fetal heart rate
KNN - k-nearest neighbor
MLP - Multilayer perceptron
SVM - Support vector machines
STFT - Short-term Fourier transforms
DWT - Discrete wavelet transformNAR-PSD - Normalized autoregressive power
spectral density
HSAS - HeartSound Authentication System
MSE - Mean Square Error
DSK - Digital signal processor starter kit
DSPPCG - Digital signal processorbasedphonocardiogram
PCPCG PC - Based phonocardiogram
CWT - Continuous WaveletTransform
ECG - Electrocardiogram
RPROP - Resilient Back propagation
SDOF - Single degree-of freedom
8/2/2019 2nd 4
7/8
vii
CSCW - Cardiac sound characteristic
waveform
THV - Threshold value
ZCR - zero crossing rate
MFCC - mel frequency cepstral coefficient
FFT - fast fourier transform
DCT - dicrete cosine transform
8/2/2019 2nd 4
8/8
viii