This research is armed at fundamental understanding of identification methods for high impedance and arcing faults on distribution feeders. The intent of the research is to fully understand and characterize the nature of these faults and to explore new detection techniques using advanced concepts of signal processing and pattern recognition such as spectral analysis, signal detection and signature classification. It is expected that the use of these concepts with knowledge based systems and parallel processing hardware can result in the detection and classification of faults which are currently undetectable.