Abstract
We address the problem of signal compression, basing on the mathematical model, in which a set of all possible signals is considered as a function space with a metric p. The main attention is focused on the minimization of the size of compressed representation, when function characteristics are not known precisely.