The problem considered here is that of detecting events from the analysis of sensing signals. Our approach is based on the continuous wavelet transform and defines pseudo power signatures, as functions of the resolution and characterizing the power distribution of an event. Detection of their presence can be used to identify an event. These pseudo power signatures are, ideally, independent of the duration of the event and can be therefore used to provide fast detection of changes as required, for example, in fault detection problems.
The paper gives an overviee of the concept of pseudo power signatures and their application to signal classification, focusing in fault detection. The paper also introduces new computational algorithms that are much more efficient than those previously available.