A comparison of quadratic TFDs for entropy based detection of components time supports in multicomponent nonstationary signal mixtures

Abstract

Separation of different signal components, produced by one or more sources, is a problem encountered in many signal processing applications. This paper proposes a fully automatic undetermined blind source separation method, based on a peak detection and extraction technique from a signal time-frequency distribution (TFD). Information on the local number of components is obtained from the TFD Short-term Rényi entropy. It also allows to detect components time supports in the time-frequency plane, with no need for predefined thresholds on the components amplitude. This approach allows to extract different signal components without prior knowledge about the signal. The method is also used as a quality criterion to compare Quadratic TFDs. Results for synthetic and real data are reported for different TFDs, including the recently introduced Extended Modified B distribution.