Compressed Illunination Sensing英文清晰版.pdfVIP

IEEE SIGNAL PROCESSING LETTERS, VOL. 18, NO. 10, OCTOBER 2011 587 Compressed Illumination Sensing Sandeep Gogineni, Student Member, IEEE, and Arye Nehorai, Fellow, IEEE Abstract—Modern day light emitting diodes (LEDs) are ca- pable of producing high intensity light across a wide spread of frequencies. Hence, they are becoming a common ingredient in many lighting systems. In order to obtain desired lighting effects efficiently, it is important to sense the light received across different target locations and estimate the unknown properties (amplitudes, frequency offsets and phases) of the modulating signals. This facilitates the design of the driving waveforms for the LEDs. This procedure is known as illumination sensing and it enables efficient and effective usage of light energy to achieve the intended effects. We propose a novel two step approach to perform this estimation using sparse modeling which exploits the fact that the measurements at the sensors are sparse in the frequency offset space and the phase space. Further, we employ compressive sensing to reduce the dimensions of the measurement vector, thereby reducing the complexity of the estimation algorithm. This will enable quick estimation which is essential to avoid any lag in attaining the desired illumination effects. Also, we demonstrate the performance of the proposed approach in the presence of a modeling mismatch. Fig. 1. LED illuminating system with sensing at target locations. Index Terms—Compressive sensing, illumination sensing, LED, sparsity.