1Mar del Plata National University, Mar del Plata 7600. Argentina
2University of Toronto, Toronto, Ontario, Canada
3Ryerson Polytechnic University, Toronto, Ontario, Canada
We propose a probabilistic extension of the Matching Pursuit algorithm introduced by Mallat and others. After a discussion of the Matching Pursuit algorithm a conditional probability model is set up over coherent components. The predicted denoised function is given by an expected value with respect to a probability measure. This expected value is computed by means of an ergodic average. This algorithm improves the original Matching Pursuit when the optimization that drives the latter algorithm proposes coherent components which reproduce details of the noise present in the signal. Several simulated examples are presented.
The research of S. E. Ferrando was supported by NSERC grant OGP 0194624
References:
[1] M. B. Alaya, \emph{On the simulation of expectations of random variables depending on a stopping time}, Stochastic Analysis and Applications, \textbf{11}:2 (1993), 133--153.
[2] R. B. Ash, \emph{Real Analysis and Probability}, Academic Press, New York, 1972.
[3] J. Besag, P.J. Green, D. Higdon and K. Mengersen, \emph{Bayesian computation and stochastic systems} (with discussion)), Statistical Science \textbf{10} (1995), 3--66).
[4] N. Bouleau and D. Lepingle, \emph{Numerical Methods for Stochastic Processes}, Wiley and Sons, New York, 1994.
[5] G. Davis, S. Mallat and M. Avellaneda, \emph{Greedy adaptive approximations}, J.\ of Constr.\ Approx.\ \textbf{13} (1997) 57--98.
[6] L. Devroye, \emph{Non-Uniform Random Variate Generation}, Springer Verlag, 1986.
[7] S. Mallat and Z. Zhang, \emph{Matching pursuit with time-frequency dictionaries}, IEEE Transactions on Signal Processing \textbf{41}:12 (1993), 3397--3415.
[8] L. Tierney, \emph{Markov chains for exploring posterior distributions}, The Annals of Statistics \textbf{22}:4, (1994), 1701--1762.
[9] B. Torresani, \emph{Wavelets associated with representations of the affine Weyl-Heisenberg group}, Journal of Mathematical Physics \textbf{32}:5 (1991), 1273--1279.
[10] E. T. Whittaker and G. N. Watson, \emph{A Course of Modern Analysis}, Cambridge University Press, 1927.