A study of acoustic array signal processing applied to tornado tracking

Abstract

In low signal-to-noise ratio (SNR) applications it is often necessary to integrate array measurement data over an extended temporal period to achieve satisfactory estimates of source bearings-of-arrival (BOA), especially in the case of multiple sources. For example, in popular frequency-domain methods such as MUSIC and its variants, the number of so-called snapshots (temporal windows) can be increased to improve performance. However, this improvement is achieved under the assumption that the bearing-of-arrival is not changing during the integration period. In this presentation, a study of the estimation performance of two algorithms (a MUSIC variant and a maximum-likelihood based method) is performed as a function of the number of snapshots, and the result of application of these algorithms to infrasound data recorded during the presence of tornadic storms (moving sources) is presented. Additionally, a method for integration over time using a moving source model is introduced.