Aiming at the drawbacks of manual judgement reliance and low efficiency in power cable fault locating technologies，a novel algorithm using fault soundwave auto-recognition and initial point auto-marking is proposed. Four general features are defined by analyzing the fault soundwave characteristics，by performing feature extractions on fault and non-fault soundwaves，training set and test set are forged. An AdaBoost-support vector machine （SVM） based algorithm for fault soundwave recognition is proposed，which is used to train the four proposed features in spatial distributing difference of discharge and noise signals. A fault soundwave initial point marking algorithm is proposed by combining discrete wavelet transformation and Gaussian distribution regularities. Experiment verifies that the proposed algorithms increase its efficiency in fault pinpointing while maintaining the accuracy.