Source localization using Over-Complete Dictionary

Abstract

Direction of Arrival (DOA) estimation or source localization by listening to the radiated acoustic signal from the target is of primal importance in underwater surveillance system. In general, the sensor array data vector is projected on a Complete Dictionary which is made up of a set of linearly independent basis vectors. In this work, the notion of Over-Complete Dictionary (OCD) is built for underwater source localization application. This is implemented by stacking multiple set of basis vectors. Each set of basis vectors are selected for a specific statistical properties of the signal and the ambient noise into consideration. In a dynamically changing ocean scenario, OCD implementation gives better results in comparison with the conventional or adaptive beamforming techniques. As expected, the simulation results also show significant performance improvement with OCD. Also, it is observed that the accuracy and bearing resolution improves as the dictionary size grows. The improved dictionary performance is demonstrated by Monte-Carlo simulations for various ocean scenarios which shows the effectiveness of the proposed technique.