Nonstationary signal modeling with application to bat echo location

Two classes of parametric estimators for the relief of a nonstationary signal were tested in an application to the analysis of bat echo location calls. The first class consists of autoregressive adaptive models. The second is characterized by the prior assumption of a time-variation of the model coefficient restricted to the space spanned by a known set of functions. These evolutionary models are split into autoregressive models, lattice filters, and log area models. Application of the models to three typical bat echo location calls shows the great accuracy of both classes.