HIGH-RESOLUTION PURSUIT FOR FEATURE-EXTRACTION

Recently, adaptive approximation techniques have become popular for ob taining parsimonious representations of large classes of signals. Thes e methods include method of Frames, matching pursuit, and, most recent ly, basis pursuit. In this work, high resolution pursuit (HRP) is deve loped as an alternative to existing function approximation techniques. Existing techniques do not always efficiently yield representations w hich are sparse and physically interpretable. HRP is an enhanced Versi on of the matching pursuit algorithm and overcomes the shortcomings of the traditional matching pursuit algorithm by emphasizing local fit o ver global lit at each stage. Further, the HRP algorithm has the same order of complexity as matching pursuit. In this paper, the HRP algori thm is developed and demonstrated on 1D functions. Convergence propert ies of HRP are also examined. HRP is also suitable for extracting feat ures which may then be used in recognition. (C) 1998 Academic Press.