Abstract:
The electrocardiogram (ECG) is the record of variation of bioelectric potential of human
heart beats with respect to time. ECG represents the electrical activity of heart and is a
well-established tool for the diagnosis of cardiac condition. In order to obtain accurate
information about cardiac condition of a person, correct interpretation of ECG is essential.
The objective of this work is to devise an automatic computer based detection technique
for five particular types of cardiac abnormalities by processing ECG signal with better
accuracy and less computational time.
Five particular types of cardiac phenomena that have been considered are: normal beat,
left bundle branch block (LBBB), right bundle branch block (RBBB), premature
ventricular contraction (PVC) and atrial premature beat (APB). A total set of 19 records
have been taken from MIT- BIH arrhythmia database as test data. For all the test records,
discrete wavelet transform (DWT) with 40,000 samples up to a decomposition level of 4 is
performed in the Matlab 7.4.0 environment. Five different types of mother wavelet
functions are used for this analysis. These are: haar, coiflet2, symlet4, daubechies10 and
biorthogonal6.8. The maximum value of the approximation coefficients of level 4 is
selected as the indicating parameter, which is used to distinguish between different
abnormalities.
A comparison is made among the performances of different types of mother wavelets to
select the best one to differentiate cardiac abnormalities. Among the five wavelets that we
have used in our study, daubechies10 (db10), the only one possessing asymmetric
property, has provided the most optimum result. So as an outcome of this analysis we can
conclude that asymmetric mother wavelet functions are more effective than the symmetric
ones for distinguishing different types of ECG beats. Further analysis with more number
of symmetric and asymmetric wavelet functions should be carried out to generalize the
findings of this study and to detect other types of cardiac abnormalities.
