Tracking a moving target with multiple doppler sensors using an artificial neural network

Tracking of human movements using radar in a high-clutter environment or even through walls is a problem of current interest. Example applications include law enforcement, disaster search-and-rescue and urban military operations. One approach to human tracking is to use wideband waveforms. Such system is capable of high range localization, but the cost of the hardware tends to be high. Doppler-based sensors, on the other hand, offer an inexpensive way to detect moving targets in the presence of stationary clutter. However, target location information is not possible, unless frequency or spatial diversity is incorporated. In this work, we investigate the use of a collection of spatially diverse Doppler sensors to derive the location information of a moving target. This problem has been investigated previously by Armstrong and Holeman in the context of tracking a baseball in 3D using a number of speed guns. In that work, a local search method was employed for the maximum-likelihood estimation of target parameters (position and velocity) from the measured Doppler shifts. However, the results can be very dependent on the initial guess so that the parameter estimation may not be robust. In this paper, an artificial neural network is proposed to estimate the target parameters using Doppler information measured by a set of spatially distributed sensors. The neural network is trained to relate the nonlinear relationship between the observed Doppler information and the target parameters. For the training, point scatterer data generated by simulation are used. Some preliminary measurement data are collected using a toy car that runs a round track. Its trajectory and velocity are estimated by the neural network. The simulation and measurement results are reported.