One of the approaches proposed for Non-Intrusive Load Monitoring (NILM) of residential appliances is based on the analysis of switching transient voltages. The transients induced when and appliance is operated can be learned or classified and thus it can be determined when appliances are turned on and off via analysis of the transients. Acquisition of the transient requires two steps: first, the detection of a transient in the ongoing power line signal, and second, the filtering of the electric power signal itself (e.g., in the US, the 60 Hz signal and its harmonics) to provide a clean transient signal to analyze. Here, a detection method using a masked (envelope trigger) oscilloscope-type instrument is adopted. That allows no computation to take place until a transient is actually detected, thus saving greatly on computational requirements. Then, the filtering is performed mathematically. A novel recursive filtering algorithm is presented here, allowing removal of the natural response of the mathematical filter. By using the methods presented in this article, very clean transient signals from household appliances, with full removal of the power signal and its harmonics, can be obtained, thus greatly facilitating the posterior analysis and classification of these transients.