Acoustic Geo-sensing using sequential Monte Carlo Filtering

We present a Sequential Monte Carlo (SMC) based approach for performing Acoustic Geo-Sensing, i.e., the derivation of geoinformation from audio data, in context of the Graz Cell-phone Cycle Corpus (GC3). In particular, we deal with the process of determining the geographical position of a mobile device in motion, based solely on the acoustic data recorded by its built-in microphone. Our approach estimates the mobile device’s most probable geoposition on-line, improving its accuracy over time by continuously incorporating the most current acoustic information captured from the microphone. In this study, we introduce our algorithm and present the recognition results achieved for the one-dimensional problem of position tracking along a known route. Finally, we will point out our method’s theoretical potential of being generalized to solving position tracking problems on a two dimensional map.