Statistical models and recursive prediction of speech signals

On the basis of Markovian filtering theory, regression analysis, and a renewal process method, recursive linear and nonlinear algorithms for adaptive prediction are presented. A simplified forecasting algorithm is obtained which can be adapted to the technical realization of digital elements for microcircuit technology. Consideration is given to algorithms for predicting and filtering speech signals as they are mixed with background noise. Prediction algorithms can be applied to the digital representation of signals in systems with adaptive differential PCM and in encoders.