Velocity estimation and detection of moving objects by using extended complex Kalman filters in the mixed domain

In this paper, a transform/spatiotemporal mixed-domain filter is proposed that extracts a moving object with its trajectory varying in time (arbitrary trajectory signal). The trajectory signal in. a mixed domain can be considered as a bunch of one-dimensional complex signals. In the case of an arbitrary trajectory signal, this is a bunch of quasi-periodic signals in which the periods of the one-dimensional complex signals change in time. Hence, in the proposed method, the arbitrary trajectory signal is extracted by designing a parallel bank of extended complex Kalman filters for estimating the quasi-periodic signals. The effectiveness of the proposed method is clarified by simulation. (C) 2003 Wiley Periodicals, Inc.