This paper presents a set of filters with globally exponentially stable error dynamics for source localization and navigation, in 3-D, based on direction measurements from the agent (or vehicle) to the source, in addition to relative velocity readings of the agent. Both the source and the agent are allowed to have constant unknown drift velocities and the relative drift velocity is also explicitly estimated. The observability of the system is studied and realistic simulation results are presented, in the presence of measurement noise, that illustrate the performance of the achieved solutions. Comparison results with the Extended Kalman Filter are also provided and similar performances are achieved.