Non-stationary decomposition using the Discrete Linear Chirp transform (DLCT) for FM demodulation

In this paper, we consider FM demodulation as an application of the decomposition of non-stationary signals. Non-stationary signal decomposition can be done using either the empirical mode decomposition (EMD) or the Discrete Linear Chirp Decomposition (DLCT) methods. These methods decompose non-stationary signals using local time-scale signal characteristics. While the EMD decomposes the signal into a number of intrinsic mode functions (IMFs), the DLCT obtains a parametric model based on a local linear chirp model. Analytically the DLCT considers localized zero-mean linear chirps as special IMFs. The DLCT is a joint frequency instantaneous-frequency orthogonal transformation that extends the discrete Fourier transform (DFT) for processing of non-stationary signals. FM demodulation is commonly done by computing the signal derivative to convert it into an amplitude demodulation. We will show that the demodulation can be approached with the EMD and the DLCT and that the second method provides better results. The performance of the DLCT and the EMD are illustrated and compared when used as an FM demodulation scheme in software defined radio.