SoC implementation of a modulation classification module for cognitive radios

The Cognitive radios are intelligent Software Defined radios which is aware of its environment by scanning and identifying spectrum holes. The knowledge of the modulation scheme of the received signal is significant to judge the channel and configure the SDR to transmit and receive. The two broad fields of modulation recognition are pattern recognition and decision theoretic approaches. This paper discusses Zynq implementation of an adaptive modulation recognition system which includes decision theoretic approach and higher order cumulants. Zynq is a SoC with Artix-7 FPGA and dual Core ARM Processor. The highly complex computations like FFT, higher order cumulants computations are implemented in FPGA and neural network algorithms are implemented in the ARM processor. The SNR based higher order cumulants computation will differentiate 16-qam, 32-qam, 64-qam, bpsk, qpsk and 8 psk signals. The standard deviation computation of the zero centered instantaneous phase, amplitude and frequency are used to identify the MASK and MFSK signals. The FM signals are identified by the computation of the spectral power density.