Optimized dictionary learning algorithm for compressive data gathering

To improve the adaptability of compressive data gathering for various classes of sensory data, and to reduce the recovery error caused by environmental noise, an optimized dictionary learning algorithm was proposed to adaptively construct the sparse dictionary in compressive data gathering. Theoretical analysis shows that in compressive data gathering the recovery error caused by environmental noise is positively correlated to the self-coherence of the sparse dictionary. Therefore, in order to alleviate the recovery error caused by environmental noise, the proposed algorithm introduces a penalty term into the dictionary learning procedure to reduce the self-coherence of the learned dictionary. The introduced penalty term can also alleviate the over-fitting on the training data during the dictionary learning procedure, which further improves the sparse representation performance of the learned dictionary. The experimental results verify that the proposed method achieves better sparse representation performance than other dictionary learning methods, and can alleviate the recovery error caused by environmental noise.