Abstract

Parameterization of bats' echolocation signal is essentially based on determination of spectral power density by means of the classic Fourier transform FFT. This study presents an alternative solution in this area of research, that is parametric and non-parametric modelling of short-time signals. The above mentioned methods are based on modelling of white noise with the use of digital filters the transmission of which was set in a way that allows the output signal to be as close to the modelled signal as possible. Proper selection of parameterization method - MA ( Moving Average), AR (Autoregressive), ARMA ( Autoregressive Moving Average), in respect of the character of signal spectrum ( line spectrum, noise) maximally reduces the number of filter coefficients and improves the accuracy of bat's signal modelling. The work also presents the possibility of using the suggested parameterization methods in automatic species identification.