An Exercise in Signal Processing Techniques
Theodora Ananidou, Theo S. Sarris, I. Lerche · Journal of Geoscience Education · 2002
An exercise in a course on signal processing techniques for students was the motivation to report some of the procedures on signal recovery capability so that other students can, perhaps, use them as an introduction to data analysis. This paper illustrates some of the procedures that can be used easily in order to extract “signal” and “noise” from a set of data. The rather unconventional dataset of the publication record with time of one of the authors has been used, and under specific assumptions for the set of data, smoothing routines were applied using 5, 7, 9, and 11-point running average smoothers. For the original and the smoothed data, the power spectra were calculated. Significant peaks at certain periods are observed irrespective of the smoothing procedure and presumably record historical event periods of some importance. In this example, the Blackman-Tukey theorem is employed regarding signal and noise resolution. The estimates of signal and noise power were integrated over the total frequency range to obtain a statistically stable estimate of the “signal”, indicating that about 85% of the data are robust to variations in the processing techniques used.