Particle filtering with enhanced likelihood model for underwater acoustic source DOA tracking

Estimating DOA of an underwater acoustic source is a challenging problem due to low signal-to-noise ratio (SNR) in an ocean environment. This problem is even more challenging when the source is dynamic since the received signal can only be assumed to be stationary for a small number of snapshots. In this paper, a Bayesian framework and its particle filtering (PF) implementation are introduced to cope with this problem. At each time step, the particles are sampled according to a constant velocity model, and then corrected with the corresponding likelihood. Since the likelihood function is usually spread and distorted in the heavy noisy environment, it is exponentially weighted to enhance the weight of particles in high likelihood area. The particles can then be weighted more appropriately and resampled efficiently. Experiments show that the proposed PF tracking algorithm significantly outperforms the traditional localization approaches as well as the existing PF algorithm in challenging environments.