In this paper, a hybrid path tracking system is presented, which exploits the efficiency of a Kalman filter in conjunction with the power of compressive sensing to represent accurately sparse signals and a region-based multivariate Gaussian model. The experimental evaluation reveals an increased localization performance, while maintaining a low computational complexity.