Partial discharges in high-voltage power cables are a focus of concern due to their harmful effects regarding insulation degradation. Therefore, their fast and accurate identification is of paramount importance. This paper deals with the diagnosis and identification of cable defects in medium-voltage XLPE insulated cables. To this end partial discharge pulses are acquired by means of a standard partial discharge detector. Data acquired are further processed by means of the fast Fourier transform and by applying suitable multivariate feature extraction and classification methods, namely principal component analysis, canonical variate analysis and k nearest neighbors. Experimental results show that the proposed identification methodology provides improved classification accuracy, simplicity and very low time response to classify a new sample object.