Fast Fourier transform computation using a digital CNN simulator

Abstract

We explore the advantages of more general topology of cellular neural network (CNN) arrays, where cell neighbourhood is defined from the functional, rather than topological, point of view. In this way it is possible to build many new applications, thus extending possibilities of CNN. To illustrate this, we have chosen a fast Fourier transform algorithm, which can be successfully used in many applications. Both fast Fourier and inverse fast Fourier transform (FFT and IFFT) can easily be built using our digital CNN simulator proposed. In contrast to direct Fourier transform, as proposed for CNN by Moreira-Tamayo et al. (1996), FFT is far more economical. This paper also clarifies some computational techniques of the proposed digital CNN simulator and focuses on its timing and accuracy aspects.