Authors:
Tiia Siiskonen
;
Tapio Grönfors
and
Niina Päivinen
Affiliation:
University of Kuopio, Finland
Keyword(s):
Lossy data compression, Electromyography, Discrete Cosine Transform and Medical parameters.
Related
Ontology
Subjects/Areas/Topics:
Informatics in Control, Automation and Robotics
;
Optimization Problems in Signal Processing
;
Signal Processing, Sensors, Systems Modeling and Control
;
Time Series and System Modeling
Abstract:
Typically used simplified error measures, like mean-squared-error (MSE), do not reveal everything about the clinical quality of lossy compressed medical signals. Errors have to be interpreted via essential medical parameters. The medical parameters depend on the type of the signal and only the preservation of essential medical parameters can guarantee the correct clinical quality. In this study, short electromyography (EMG) signals are compressed with DCT transformation -based lossy compression method. The compression is gained with irreversible masking and scalar quantization of the DCT coefficients. The most prominent medical parameters of EMG signal are the mean frequency (MNF) and the median frequency (MDF). The behaviors of these parameters are studied both by fitting a regression line and by examining the mean absolute errors frequency-by-frequency over clinically interesting frequency range. This reveals the frequency dependency of errors of the medical parameters and inspires
the idea that the generated linear model can be used for estimating the correct value of the processed medical parameter.
(More)