The present paper concerns the estimation of probability density functions using the particular parameterized class of distribution functions implemented by a single non-linear neuron, introduced in the previous contribution [12]. The estimation procedure is applied to the statistical characterization of sorne electrical and mechanical phenomena.