Robust recursive AR speech analysis based on quadratic classifier with sliding training data set and a heuristic decision threshold

A robust recursive procedure for identification of nonstationary AR speech model based on a quadratic classifier with a heuristic decision threshold is proposed and evaluated. A comparative experimental analysis is done through processing natural speech signal with voiced and mixed excitation segments. Obtained results show that the proposed robust procedure based on the quadratic classifier with sliding training data set and the heuristic decision threshold achieves more accurate AR speech parameter estimation and provides improved tracking performance.