This thesis presents research on the radar micro-Doppler drone characterization using features from the long windowed short-time Fourier transform (STFT) representation referred to as HElicopter Rotation Modulation (HERM) line signatures. The long windowed STFT representation enables drone identification at longer ranges using a low pulse repetition frequency (PRF) radars. A simple mathematical model is proposed to describe the HERM line signature using three parameters or features, namely the number of HERM lines, the fundamental frequencies and the number of fundamental frequencies. This model motivates research into parametric spectral estimation techniques for micro-Doppler feature extraction which is lacking in the current state of the art. Subspace-based spectral estimation techniques are applied to real and simulated data and compared with the Fourier-based techniques from the literature. These proposed techniques are shown to perform better for smaller windows of data. An application of these features for drone detection/classification is demonstrated.