Speech Signal Autoregression Modeling Based on the Discrete Fourier Transform and Scale-Invariant Measure of Information Discrimination