The endomorphic model: An approach to adaptive prediction

This paper describes a method by which a set of piece-wise constant autoregressive parameters could be obtained when the source signal is subject to a multistage analysis. It also illustrates an FIR prediction scheme with application to a speech process. It describes a method where the prediction is carried out on a sample by sample basis. This prediction scheme demonstrates the possibility of having control over the prediction error by constraining it to be within a predetermined limit.