Estimation of densities on freeways and arterials is critical to traffic control and management. Most previous work, however, focused on freeway density estimation based merely on detector data. This study attempts to estimate traffic density along a signalized arterial by using data from both detectors and a global positioning system (GPS). The approximation to the previously developed MARCOM (Markov compartment) model is adapted to describe arterial traffic states. A hybrid extended Kalman filter is then implemented to integrate the approximated MARCOM with detector and GPS measurements. The proposed model is tested on a single signal link simulated by using VisSim. Test results show that the hybrid extended Kalman filter with GPS data can significantly improve density estimation.