Compressão de sinais biomédicos

• Ano de defesa: 2016
• Autor(a) principal: Oliveira, Hugo Neves de
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal da Paraíba; Brasil; Informática; Programa de Pós-Graduação em Informática; UFPB
• Palavras-chave em Português: Compressão de dados; Sinais biomédicos; Transformadas de domínio; Quantização vetorial; Codificadores de entropia; Data compression; Biomedical signals; Domain transforms; Vector quantization; Entropy coders; Entropia - Codificadores; Teoria dos sinais; Sinais biomédicos - Compressão de dados; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: https://repositorio.ufpb.br/jspui/handle/123456789/16919

Compressors for electrocardiograms and electroencephalograms have been reported in the literature over the last decades, but there is a lack of works describing general solutions for all biomedical signals. Aiming to fill this gap, this work discusses compression methods that work well for all biomedical signals. The lossless compression methods are based in wavelet transforms and linear predictors associated with several entropy coders. The lossy compaction strategies use trigonometric and wavelet transforms, followed by vector quantizers based in dead-zone quantization by Lagrangian Minimization and Successive Approximation Quantizations. The entropy coders are based in Prediction by Partial Matching, Run Length Encoding and Set Partitioning in Hierarchical Trees. The methods were tested usign the signals of the MIT/BIH Polysomnographic Database. Lossless compressors achieved compression ratios of at most 4.818:1, while the lossy methods achieved compression ratios of at most 818.055:1. Smoother signals, as respiration and oxygen saturation records, presented better reconstructions with wavelets and Successive Approximation Quantization, despite the lower compression performances. Contrastively, for the same quality of visual reconstruction, trigonometric transforms and Lagrangian Minimization achieved better compression performances for rougher signals – as electroencephalograms and electromyograms.