An electrical pulse fed into an electrical cable, e.g. a power transmission or distribution cable buried underground, produces an acoustic signal, e.g. an arc-over or discharge noise, at a fault in the cable. This acoustic signal is detected and used to locate the fault. But the acoustic signal has interference noise superimposed on it, which makes it difficult to identify the acoustic signal in the noise-burdened received signal, and thus also makes it difficult or impossible to accurately locate the fault. To minimize or avoid the influence of interference noises, the relevant acoustic signal is identified in the received signal by comparing the received signal with predetermined sample data, such as signal patterns, characteristics or characteristic sets.