Preliminary Adaptation of Speech Source Localization Algorithm for Reduced Bandwidth of Interest in Tornadic Infrasound Signals

摘要

Recent evidence indicates that acoustic waves at frequencies below human hearing (i.e. infrasound) are produced during tornado genesis and continue through the life of a tornado. Oklahoma State University is working in part with local weather stations and NOAA to design tornado detection and localization equipment based on infrasound measurements from GPS-localized static and mobile microphone arrays. This paper focuses on adapting a speech-based acoustic source localization algorithm demonstrated previously on a mobile robot from literature to the current infrasound source tracking problem through implementation of the Chirp Z-Transform (CZT) and recently published inverse Chrip Z-Transform (iCZT). The usage of CZT and iCZT provides higher resolution over relatively small bands of interest at low frequency where spectra generated by the currently used Fast Fourier Transform (FFT) su˙er from few contour points in this region. Preliminary results of localization tests are presented for CZT and FFT implementations with discussion of runtime and secondary source-rejection for frequencies outside the bandwidths of interest.