Clustering Methods for Extremely Low Frequency Subsurface Signals

Extremely Low Frequency (ELF) electromagnetic signals are of great interest in a variety of applications, including the transmission of communication signals over long distances in seawater and below the Earth’s surface, deep geophysical sounding, and the study of electrical atmospheric phenomena known as sferics. In subsurface signaling applications, wireless electromagnetic methods generally require far less infrastructure than wireline approaches and can provide effective signaling over much longer distances than acoustic methods. However, the ELF band is highly susceptible to interfering signals emanating from electronic and electro-mechanical equipment that is often associated with geophysical exploration. This interference can significantly degrade the demodulation and subsequent analysis of the received signals. To improve the demodulation of ELF communication signals, data can be segmented via clustering methods into sequences representing different signal states. In this work, several clustering approaches are compared on their ability to segment ELF signals in the presence of severe interference. K-Means and ‘k-Means like’ clustering methods are found to be superior overall.