Background: Phonocardiography is the study of human cardiac sounds. Phonocardiograms (cardiac sounds) as they are called, represent the most vital physiological and pathological information about the human body.
Objective: This paper presents an automatic method of
segmentation of heart sounds using the occurrence of cardiac rhythmic events.
Methods: Noisy heart sound is filtered using the 6th order
Chebyshev type I low pass filter to remove the redundant noise. The Bark
Spectrogram is calculated from the cardiac signal by converting the spectrogram
to the Bark scale. The bark spectrogram is smoothened and the loudness index is
calculated by averaging the amplitude across all frequency bands. The loudness
index is smoothened and differentiated to obtain the event detection function.
The smoothened event detection function gives the occurrence of the cardiac
events namely the first and the second heart sounds.
Result: This method is highly effective in identifying peaks
S1 and S2 with a segmentation accuracy of 96.98% giving an F1 measure of
97.09%.
Conclusion and significance: This method does not require
the setting up of any type of noise threshold. So, it is a highly effective
type of segmentation of phonocardiogram corrupted with noise. To reduce the effect
of noise the noisy phonocardiogram is heavily filtered using the time-frequency
block thresholding method.
Author (s) Details
Vishwanath Madhava
Shervegar
Department of Electronics & Communication Engineering, Moodlakatte
Institute of Technology, Kundapura, Visvesvaraya Technological University,
Belagavi, India.
Please see the book here:- https://doi.org/10.9734/bpi/mmrnp/v12/3241
