Underwater acoustic imaging: sparse models and implementation issues

Projet ANR : ANR-09-EMER-001International audienceWe present recent work on sparse models for underwater acoustic imaging and on implementation of imaging methods with real data. By considering physical issues like non-isotropic scattering and non-optimal calibration, we have designed several structured sparse models. Greedy algorithms are used to estimate the sparse representations. Our work includes the design of real experiments in a tank. Several series of data have been collected and processed. For such a realistic scenario, data and representations live in high-dimensional spaces. We introduce algorithmic adaptations to deal with the resulting computational issues. The imaging results obtained by our methods are finally compared to standard beamforming imaging