Classification of degraded signals by the method of invariants

A novel approach to the recognition of the signals degraded by a linear time-inwtriant system with an unknown impulse response is proposed. It consists of describing the signals by the features which are invariant to the degradation and recognizing signals in the feature space. Unlike the blind-deconvolution techniques, neither the impulse response identilication nor the signal restor,'ttion is performed. Two sets of appropriate blur-invariant features (the first one dctincd in time domain and the other one in spectral domain) arc introduced in this paper and the optimal algorithm for a robust signal classification is proposed. R~um6 Cet article propose une nouvelle :tpproche de lit rcconmtissance de signaux d~:grad~3s par un syst/:mc invariant dans Ic temps lin6aire fi r6ponse impulsionelle inconnue. Cette approche eonsiste fi d~crire Its signaux ~i raide de caract6ristiques invariantcs ~i llt d/:gradation ct fi rcconnaitre cos signaux dans respace des caract/:ristiqucs. A la diff/:rence des techniques de d,2convolution a raveugle, ni ridcntilication de llt r/:ponsc impulsionelle ni la rcstauration du signal ne sont effectu6es. Deux ensembles de caract/:ristiqucs appropri/:cs invariantcs au flou (le premier d/:lini darts le domaine temporel, le second dans Ic domaine spectral) sont introduits, et nous proposons ralgorithme optimal pour une classilication robuste des signaux. C 1997 Elsevier Science B.V.