No abstract

The analysis and representation of multico tnponent sig,nals, embedcsed in additive Gaussian noise, is oEinterest in many signal processing applicatkns and has been stuclied for years, maidy for the case of stationary signals. However, in the non-st~tionary case only cveral methods aue available, one o f which. k parameter analysis from the hie-frequency distribution (TFD) of the signal. Recently, a new positive distribution free of cross-terms, named MCE-TFD W E B introduced. Based on this TFD, an algorithm for signal recon-CE representation is introduced. The the instantaneous fre:qucncy (IF), the mplitude a ( t ) of each signal compoixe original signal. Together with the pproach, the proposed reconstruction powerful tool for the represent ation and analysis of non-stationary multicomponent signah.

No abstract

We prcxrit ;I. ricw approach to tlic dcsigri of Tirnc-Frequency (TF) filter bariks for riori-stationary rioisy sigrials. Thc input riiiilticornporic~rit sigrial is rcprcscritcd by tho hliriirriurri Cross-Entropy T F distributiori irriti tlic systcrri is hascc1 011 x i array of tirric-varying filtcrs. Each filter processes oric cornporicrit of tho sippi1 ar:cortlirig to its spccifiic TI: support. Tlic output of the proposed T F filtcririg algoritkirii is a set of tlic sigrial cornprierits.Siricc tlic s\iggcstctl rric:tliod is cquivalerit to tu,()-dirricrisiorial riiatchcd filtcririg, iiri algorithrn for riori-stationary sigrial cla.ssification i s also prcscritctl To tlcriionstratc the algorithm pcrforrnance, sirnulatiori results arc givcri.