Non-intrusive measurement of impulsive noise in telephone-type networks

An in-service, non-intrusive technique for the characterization of stationary and impulsive noise in a telecommunication system is presented. It is based on the segmentation of the signal observed during a telephone call in speech and noise periods by means of a Bayesian classifier. The signal observed during noise periods is further processed by means of a hypothesis testing procedure in order to detect noise-impulses. Experimental results that show the effectiveness of the proposed technique are discussed