Logarithmic cyclic frequency domain profile for automatic modulation recognition

Cyclostationary techniques have been applied widely to the problem of recognising communication modulation schemes. As these techniques are processing intensive, much effort has been invested in researching algorithms that can reduce the number of computational steps required, with fast Fourier transform approaches predominating. A novel approach to improve the extent of the cyclic frequency (a) is proposed. By using the constant Q transform (CQT), a logarithmic form of the spectral correlation function (SCF) can be produced. This allows the a-axis to be extended, which can be advantageous when the receiver bandwidth cannot be well matched to the signal frequency and bandwidth using a priori knowledge of spectrum allocation. It is found that a CQT-based SCF can form the basis of a logarithmic cyclic frequency domain profile algorithm without loss of sensitivity compared with the conventional, linear form.