Author(s): Chorão, Ricardo Daniel Domingos
Date: 2012
Persistent ID: http://hdl.handle.net/10362/8249
Origin: Repositório Institucional da UNL
Subject(s): Parallel processing; Parallel algorithms; Biosignals; Signal processing; Heart rate variability; Respiration
Dissertação para obtenção do Grau de Mestre em Engenharia Biomédica
Patients with neuromuscular and cardiorespiratory diseases need to be monitored continuously. This constant monitoring gives rise to huge amounts of multivariate data which need to be processed as soon as possible, so that their most relevant features can be extracted. The field of parallel processing, an area from the computational sciences, comes naturally as a way to provide an answer to this problem. For the parallel processing to succeed it is necessary to adapt the pre-existing signal processing algorithms to the modern architectures of computer systems with several processing units. In this work parallel processing techniques are applied to biosignals, connecting the area of computer science to the biomedical domain. Several considerations are made on how to design parallel algorithms for signal processing, following the data parallel paradigm. The emphasis is given to algorithm design, rather than the computing systems that execute these algorithms. Nonetheless, shared memory systems and distributed memory systems are mentioned in the present work. Two signal processing tools integrating some of the parallel programming concepts mentioned throughout this work were developed. These tools allow a fast and efficient analysis of long-term biosignals. The two kinds of analysis are focused on heart rate variability and breath frequency, and aim to the processing of electrocardiograms and respiratory signals, respectively. The proposed tools make use of the several processing units that most of the actual computers include in their architecture, giving the clinician a fast tool without him having to set up a system specifically meant to run parallel programs.
