Abnormality analysis of pcg signal using vmd and mlp neural network

Phonocardiogram signal includes heart sound with murmurs gives valuable information for the detection of cardiac diseases. This paper focus for the detection of all the peaks of 300 Heart sound from using Variational Mode Decomposition. The starting and end of each heart sound is detected using the normalized envelogram of Shannon energy, the extraction of heart murmurs is thereafter accomplished by setting a threshold level for them and finding the peaks using Variational Mode Decomposition method. Finally, 250 peaks data are trained using Multi-Layer perceptron neural network with two and three hidden layers by changing the weightage of the hidden layer neuron and all the 300 peaks data are randomly tested for best results. The Multi-Layer Perceptron based neuron network has shown a best correct prediction rate of 93.685%. The technique indicates that a combination of signal processing, MLP classification and mathematical modelling can be used as a precise method for abnormality analysis of heart.