Efficient 2D Processing of 1D Sensor Signals

Abstract

Signal processing had been the flagship technology behind the intelligent systems for applications ranging from multimedia to biomedicine, from renewable energy systems to telecommunications. It is customary to apply processing tools dedicated for the natural sensor output. For instance, audio signals are processed with 1D techniques, whereas captured images are processed via 2D methods. On the other hand, many 1D sensor outputs exhibit an intrinsically cyclic behavior. Solar radiation recordings, captured line voltage values, cardiac potential, electric consumption, etc. are all fine examples to 1D signals which already have the quasi-periodicity. Recent research efforts of the authors have shown that the cyclic behavior of such signals may help a 2D rendition of the same information, provided that the natural period is accurately determined and assigned as the “width” of the 2D matrix. Experimental results indicate improved efficiency of 2D representation in terms of modelling, prediction and error detection. This work aims to provide a mathematical reasoning to the efficiency of such 2D rendition over 1D processing in terms of reduced autocorrelation orders. © 2022, ICST Institute for Computer Sciences, Social Informatics and Telecommunications Engineering.