FILTERING RANDOM NOISE FROM DETERMINISTIC SIGNALS VIA DATA-COMPRESSION

We present a novel technique for the design of filters for random nois e, leading to a class of filters called Occam filters. The essence of the technique is that when a lossy data compression algorithm is appli ed to a noisy signal with the allowed loss set equal to the noise stre ngth, the loss and the noise tend to cancel rather than add. We give t wo illustrative applications of the technique to univariate signals. W e also prove asymptotic convergence bounds on the effectiveness of Occ am filters.