116
|
4 Measurement of Biosignals and Analog Signal Processing
especially if the reliability of the diagnosis depends on a original representation of the
biosignal. One may imagine the fatal consequences, if, as a result of signal filtering,
an ECG change typical of a heart attack would not be apparent, although it is visible
in the raw signal. In the following, the different types of interference and their fre-
quency characteristics shall be considered in principle. A simplified representation of
the spectra of the respective disturbance can be found in Figure 4.24.
Fig. 4.24: Simplified overview representation of the spectra: (a) arbitrary biosignal, (b) power
line hum, (c) high-frequency harmonic disturbance (without sidebands of frequency modulation),
(d) noise or impulsive disturbance, and (e) transient disturbance. The decay of the curve (d) would be
assumed in the case of 1/f- or 1/f 2-noise or a real impulse with temporal expansion. Any harmonics
are not shown for clarity.
1.
Periodic disturbances: The most important periodic disturbances include 50 or
60 Hz mains hum and coupling from high-frequency transmitters. If the inter-
ference signal is purely harmonic (sinusoidal), the corresponding spectrum is
pulse-shaped at the respective frequency. The 50 or 60 Hz mains hum lies within
the spectrum of most biosignals. Carrier signals of transmission channels (e.g.
WLAN, mobile radio) lie in their frequency far above the spectrum of biosignals.
Frequency-modulated transmission channels produce sidebands in the spectrum
near the transmission frequency. In the case of amplitude-modulated signals,
the transmission frequency is also above the biosignal spectrum. However, amp-
litude modulation provides low-frequency spectral components that can overlay
the biosignal.
2.
Noise: Noise is based on stochastic processes. Accordingly, the noise has an ar-
bitrary time course with a very broad spectrum. In the case of thermal noise (cf.
subsection 4.1.2) and shot noise⁹ the power density spectrum is constant and ex-
tended over the entire frequency range. This is referred to as white noise. In the
case of 1/f- or 1/f 2-noise¹⁰ the power spectral density decreases with increasing
9 Shot noise arises from the random fluctuation in the number of charge carriers passing through a
potential barrier (e.g. pn junction).
10 The 1/f 2-noise arises, e.g., from the recharge of surface and interface states.