ABSTRACT
Nearly all physiological signals that can be recorded from the body contain signicant amounts of noise that aects their resolution. In some cases, the eect of the noise can be minimized by simple linear, analog, band-pass ltering; in other cases, a nonlinear amplitude transformation can be eective in improving the signal-to-noise ratio (SNR). Signal averaging is used to extract periodically evoked potentials from noisy electroretinogram (ERG), electroencephalogram (EEG), and magnetoencephalogram recordings, and the lock-in amplier (LIA) can extract a low-frequency signal that is amplitude-modulating a coherent, constant-frequency carrier buried in noise (Northrop, 2004, Section 9.8.6).
