Parallel Matching Pursuit Algorithm to Decomposition of Heart Sound Using Gabor Dictionaries

Abstract

In order to improve the performance of matching pursuit algorithm, we propose a Parallel Matching Pursuit algorithm to decompose phonocardiogram sounds of long duration. The main goal is to demonstrate the performance of the Parallel Matching Pursuit algorithm (PMP) compared with traditional iterative matching pursuit algorithm to decompose normal and pathological heart sounds of Phonocardiogram (PCG) using a Gabor dictionary. This architecture is implemented in open source Java SE 8 using a concurrency library, which is able to reduce computational cost using multi-threading until 83 % compared with traditional Matching Pursuit. Java language is widely used in the creation of web components and enterprise solutions so based on this point the main idea of this research is to set the base to implement Parallel Matching Pursuit algorithm (PMP) on web platforms focused on the monitoring of heart to sounds. This implementation allows exploring and applying iterative algorithms or sparse approximation which require processing long audio signals with low processing time.