Transmission lines (TL) are one of the most importante component of na electricl system and, due to their extension, They are most vulnerable to faults. Fast and acurate fault location on transmission lines is necessary to reduce repair costs, maintein system reability and cause the least demage to costumers. In this context, this work proposes as implementation of a method for locating faults in TL using Mel-Frequency Cepstral Coefficients (MFCC) for processing voltage and current signals to extract principal componentes using principal component analysis to obtaing the best parameters to be used in a cuve fitting algorithm to determine the fault location. A general relative mean error of 0.074483837 was found. The highest relative errors were found near the meassurement terminals, removing the i nitial and final distances od faults Where the erro ris greater and less critical in practice, na average of 0.06621711 is obtained. Our proposal was to find a polynomial expression, diferent numbers of coefficients and sample rate were tested. The best result was found using 20 coefficients, 400khz and 16 principal componentes as imput to 4 th degree polynomial. The proposal proved to be insensitive to the fault insertion angle, to the fault resistence and proved to be stable for difefferent ranges of fault distances.