Efficient FFT algorithms for mobile devices

Abstract

Increased traffic on wireless communication infrastructure has exacerbated the limited availability of radio frequency ({RF}) resources. Spectrum sharing is a possible solution to this problem that requires devices equipped with Cognitive Radio ({CR}) capabilities. A widely employed technique to enable {CR} is real-time {RF} spectrum analysis by applying the Fast Fourier Transform ({FFT}). Today’s mobile devices actually provide enough computing resources to perform not only the {FFT} but also wireless communication functions and protocols by software according to the software-defined radios paradigm. In addition to that, the pervasive availability of mobile devices make them powerful computing platform for new services. This thesis studies the feasibility of using mobile devices as a novel spectrum sensing platform with focus on {FFT}-based spectrum sensing algorithms. We benchmark several open-source {FFT} libraries on an Android smartphone. We relate the efficiency of calculating the {FFT} to both algorithmic and implementation-related aspects. The benchmark results also show the clear potential of special {FFT} algorithms that are tailored for sparse spectrum detection.