TY - JOUR

T1 - Detecting the harmonics of oscillations with time-variable frequencies

AU - Sheppard, Lawrence

AU - Stefanovska, A.

AU - McClintock, P. V. E.

N1 - © 2011 American Physical Society

PY - 2011/1/7

Y1 - 2011/1/7

N2 - A method is introduced for the spectral analysis of complex noisy signals containing several frequency components. It enables components that are independent to be distinguished from the harmonics of nonsinusoidal oscillatory processes of lower frequency. The method is based on mutual information and surrogate testing combined with the wavelet transform, and it is applicable to relatively short time series containing frequencies that are time variable. Where the fundamental frequency and harmonics of a process can be identified, the characteristic shape of the corresponding oscillation can be determined, enabling adaptive filtering to remove other components and nonoscillatory noise from the signal. Thus the total bandwidth of the signal can be correctly partitioned and the power associated with each component then can be quantified more accurately. The method is first demonstrated on numerical examples. It is then used to identify the higher harmonics of oscillations in human skin blood flow, both spontaneous and associated with periodic iontophoresis of a vasodilatory agent. The method should be equally relevant to all situations where signals of comparable complexity are encountered, including applications in astrophysics, engineering, and electrical circuits, as well as in other areas of physiology and biology.

AB - A method is introduced for the spectral analysis of complex noisy signals containing several frequency components. It enables components that are independent to be distinguished from the harmonics of nonsinusoidal oscillatory processes of lower frequency. The method is based on mutual information and surrogate testing combined with the wavelet transform, and it is applicable to relatively short time series containing frequencies that are time variable. Where the fundamental frequency and harmonics of a process can be identified, the characteristic shape of the corresponding oscillation can be determined, enabling adaptive filtering to remove other components and nonoscillatory noise from the signal. Thus the total bandwidth of the signal can be correctly partitioned and the power associated with each component then can be quantified more accurately. The method is first demonstrated on numerical examples. It is then used to identify the higher harmonics of oscillations in human skin blood flow, both spontaneous and associated with periodic iontophoresis of a vasodilatory agent. The method should be equally relevant to all situations where signals of comparable complexity are encountered, including applications in astrophysics, engineering, and electrical circuits, as well as in other areas of physiology and biology.

KW - VENTRICULAR-FIBRILLATION

KW - IDENTIFICATION

KW - TRANSFORM

KW - SERIES

KW - NONLINEARITIES

KW - ALGORITHMS

KW - SIGNALS

KW - COMPLEX

KW - SYSTEMS

KW - SEARCH

UR - http://www.scopus.com/inward/record.url?scp=78751494771&partnerID=8YFLogxK

U2 - 10.1103/PhysRevE.83.016206

DO - 10.1103/PhysRevE.83.016206

M3 - Journal article

VL - 83

SP - -

JO - Physical Review E

JF - Physical Review E

SN - 1539-3755

IS - 1

M1 - 016206

ER -