Article details

Research area
Speech recognition

Proceedings ICASSP 1997, Munich, Germany, pp. 3537-3540


Christophe Couvreur, Yoram Bresler

Doppler-Based Movement Estimation for Wide Band Sources From Single Sensor Measurements


We address the problem of estimating the motion of a wide-band source from single passive sensor measurements, for example, estimation of the speed and position of a car moving on a road from the recording of its acoustic signature at a microphone located next to the road. We present a new computationally efficient method based on a time-varying ARMA model for Doppler-shifted random processes. Unlike previously proposed approaches which rely on a “local” periodicity hypothesis for the signal source, or a cyclostationary assumption, our method assumes only that the source is stationary and admits a rational (ARMA) model. The method is tested on synthetic and real acoustic data.

Read/download now