Blind separation of signals with mixed kurtosis signs using threshold activation functions

A parameterized activation function in the form of an adaptive threshold fo r a single-layer neural network, which separates a mixture of signals with any distribution (except for Gaussian), is introduced. This activation func tion is particularly simple to implement, since it neither uses hyperbolic nor polynomial functions, unlike most other nonlinear functions used for bl ind separation. For some specific distributions, the stable region of the t hreshold parameter is derived, and optimal values for best separation perfo rmance are given. If the threshold parameter is made adaptive during the se paration process, the successful separation of signals whose distribution i s unknown is demonstrated and compared against other known methods.