qualitative evaluation of algorithmic approaches to biomedical signal compression

Abstract

In the digital era, the study of the biomedical signal compression is of prime importance. The newlineincreasing data size demands more storage space, higher bandwidth and the extra transmission newlinetime of these signals over different communication channels for applications like telemedicine. newlineAmong these signals, ECG is a vital physiological signal that helps in diagnosis of heart diseases. newlineThe size of ECG data grows with channels, sampling frequency, recording time and sample newlineresolution. The ECG data requires 40 to 50 MB per channel space for 24-hour recording. newlineThus, limitations of storage size and communication channels, force to study efficient compression newlinealgorithms. The wavelet-based signal compression algorithms provide better compression newlineperformance. Hence, these algorithms are reviewed in detail. These algorithms involve transformation, newlinethresholding, progressive coding, and encoding. The thesis mainly contributes in newlinethresholding and progressive coding algorithms. newlineThresholding provides simple and fast compression but is a lossy compression technique. newlineIn biomedical signal compression, the selection of a proper threshold is the challenging task, newlineas it is required to preserve all clinically important information in these signals. Progressive newlinecoder allows required quality retrieval of the signal and is useful for transmission of the signal newlineeven at a low bandwidth of the channel. SPIHT and EZW are the popular progressive coders newlinethat require variable size data dependent working memory which enforces constraints on the newlinehardware implementation of the algorithm. Evaluation parameters play an important role in the newlinequality assessment of retrieved signals, but one cannot rely on a single evaluation parameter as newlinethey are not 100% perfect. newlineHypothesis testing is closely related to thresholding. Taking advantage of the same multiple newlinehypothesis methods are proposed that uses Benjami-Hochbergs procedure for error rate newlinecontrol. This thesis proposed error rate control algorithms, to determine the signal adaptive newlinethreshold that