A One-Dimensional Discrete-Time CNN Chip for Audio Signal Processing

Abstract

Although most of the CNN applications and corresponding VLSI implementations involve two-dimensional arrays, one-dimensional cellular neural networks have recently attracted increasing attention. For the latter, quite different applications have been reported, ranging from 1-D signal processing [153,185,186] to instrumentation and control [187]. In particular a 1-D CNN architecture able to emulate the behavior of FIR filters and to perform the Daubechies Wavelet Transform [188] has been thoroughly discussed in [185,186].