An adaptive non-parametric short-time Fourier transform: Application to echolocation

•An excellent time frequency representation method has been proposed.•An echolocation signal has been well analyzed by the adaptive STFT.•An iterative approximation algorithm has been proposed for reducing computation.•Some numerical experiments demonstrate the efficiency of the novel method.

This paper studies a novel time–frequency representation method—adaptive non-parametric short-time Fourier transform (ANSTFT), together with its application to the echolocation signal analysis. By rotating the signal in the analysis window, the local instantaneous frequency has been reset to parallel to the time axis. Then the high frequency and time resolution have been achieved with the mono-frequency signal simultaneously. To find the optimal rotating angle of the local signal, an iterative approximation algorithm has been utilized, which makes the ANSTFT a non-parametric data driving method and have the better generalization ability than the conventional adaptive STFT. Moreover, several numerical examples are presented to illustrate the aforementioned characteristics and the application of ANSTFT to echolocation signal analysis demonstrates its validity.