Abstract:
In this paper, a Phonocardiography (PCG)- based comparative study for cardiovascular anomalies’ early detection system
is proposed. Some of the main signal processing and Artificial Intelligence methods applied in the literature to PCG signals are
contrasted to differentiate normal heartbeats from abnormal ones and classify five of the most common murmurs. The results of this
comparative study show an average of 92.14 % for heart anomaly detection and 71.02 % for classification rates. This is achieved by
using Deep Neural Network (DNN) classification with Hyperbolic Tangent (tanh) activation function, a 5-layer with 100 neurons in
each layer. The Discrete Wavelet Transform (DWT) was found to be the best denoising algorithm and the Heart Sound Envelogram
(HSE) was the best segmentation method for the PCG signal. Mel Frequency Cepstral Coefficients (MFCC) Features outperformed
their Time and Frequency Domain counterparts. This work proved to be useful in the framework of intelligent and preventative health
care systems, offering a convenient early warning home-care tool that should help to direct potentially ill individuals to cardiologists
for more precise diagnoses.