Searching for Similarities in Nearly Periodic Signals With Application to ECG Data Compression



This paper proposes a new methodology to identify and correlate patterns on nearly periodic signal, based on signal simplification and clustering approaches. Using cubic Bezier curves some significant signal samples (control points), enabling to segment adequately the original signal, are extracted in a first step. Next, given the correlation among extracted control points, the detection of similarities within the overall signal is then performed through a clustering technique.
Although the approach is useful for many types of signals, the compression of Electrocardiogram signals (ECG) is here investigated. Results with standard MIT-BIH databases show promising compression ratios, in particular, high compression ratios are found for long duration signals, when the signal presents strong regularities.


Intelligent Signal Processing; ECG compression


Intelligent Signal Processing;


International Conference on Pattern Recognition - ICPR2006, August 2006

PDF File

Cited by

Year 2008 : 1 citations

 • MVM Figueredo, JC Nievola, SR Rogal Jr, AB Neto, Compression of Electrocardiogram Using Neural Networks and Wavelets, Springer, Book Series Studies in Computational Intelligence, pp 27-40, 2008. ISBN 978-3-540-79186-7