Abstract :
A pattern recognition approach to the automatic testing of loaded transmission lines is investigated. The pattern recognition technique selected allows automatic testing of a loaded line requiring, as data, only the magnitude of the input impedance of the line over the frequencies of interest. For standard voice quality lines, these frequencies lie between 200-4000 Hz. Thirty-two features are automatically extracted from the input impedance and used as the input to a minimum-mean-square weighted distance classifier. The classifier determines the nature and location of the line fault, if any, for the line under test. Twenty different line configurations or classes are experimentally investigated. A training set of five lines for each line class is used to obtain statistical data for each class. Thirty-six transmission lines are then tested. All 36 of these transmission lines were correctly identified as to the configuration of the line and the nature and location of any existing fault.
