Avtor/Urednik		Korošec, Dean
Naslov		Parametric time-varying spectrum and its application to SEMG signals
Tip		članek
Vir		In: Kokol P, Zupan B, Stare J, et al, editors. Medical informatics Europe '99. Amsterdam: IOS press,
Leto izdaje		1999
Obseg		str. 385-90
Jezik		eng
Abstrakt		Frequency spectrum of surface electromyographic signals (SEMGs) exhibits a non-stationary nature even in case of constant level isometric muscle contractions. This changes are related to muscle fatigue processes and can be evaluated by methods for estimation of timevarying (TV) spectrum. We describe a two estimation techniques: sliding window approach and decompostion of linear model coeficients trajectories. Drawback of the first is its large variance which causes unrealistic abrupt changes in the curve of Qverall dynamics, calculated either as the second derivative of the MF or as the ARMA distance between subsequent linear models forming the TV parametric spectrum. We demonstrate, that the second method is more appropriate for description of a slowly changing spec- trum of surface EMG signals. Dynamics measures obtained from this estimation are proposed as an indication of the fatigue process character.
Deskriptorji		ELECTROCARDIOGRAPHY ISOMETRIC CONTRACTION MUSCLE FATIGUE LINEAR MODELS