Heart Sound Signal Modeling and Segmentation based on Improved Shannon Energy Envelogram using Adaptive Windows

Various segmentation algorithms have been proposed for better classification of the highly nonstationary heart sounds. This paper proposes an improved segmentation technique based on Shannon Energy calculation of the phonocardiogram using adaptive windows. The major focus of the research has been on a simple yet comprehensive signal representation as well as on extracting most information from the Shannon energy envelogram. Earlier algorithms had a major disadvantage of losing the temporal resolution of the signal which can sometimes lead to ausculations unnoticed. First of all sinusoidal modeling of the filtered PCG is done. Zero crossings of the signal are detected and window size for the Shannon Energy calculation is formulated. With variable window size, Shannon Energy envelopes are computed. Sequence analysis is then performed on various features of envelopes and zero segments for simple classification, though aim of the paper is not the classifier design. The algorithm proves to extract details of the signal with high precision.