Compression of atomic decompositions using R-D optimum dictionary selection

Atomic decompositions have been increasingly used as signal compression tools. In general, these decompositions are obtained using a single dictionary. One may use instead several dictionaries to decompose the signal, and transmit side information in order to indicate the dictionary employed. This allows the selection of the dictionary leading to the best rate-distortion compromise. Such a scenario is encountered when decompositions that use dictionaries composed of parameterized atoms are to be encoded. In such framework, distinct quantizers applied to parameters of the atoms lead to different dictionaries. In this work, we propose a strategy, based on a training stage, to select the parameter quantizers that give near-optimum rate- distortion performance. The proposed strategy is assessed in the framework of electric power system disturbance signals compression. Simulation results show that the proposed scheme indeed achieves near-optimum R-D performance with low computational complexity.