Premium
Quasi‐periodic self‐stabilization of adaptive ARMA predictors
Author(s) -
JaïdaneSaïdane Meriem,
Macchi Odile
Publication year - 1988
Publication title -
international journal of adaptive control and signal processing
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.73
H-Index - 66
eISSN - 1099-1115
pISSN - 0890-6327
DOI - 10.1002/acs.4480020102
Subject(s) - control theory (sociology) , transcoding , encoder , quantization (signal processing) , computer science , stability (learning theory) , algorithm , mathematics , transmission (telecommunications) , artificial intelligence , statistics , telecommunications , control (management) , machine learning
The stability of adaptive ARMA predictors, which present random time‐varying poles, is investigated. The updating algorithm is a suboptimal version of the stochastic gradient. For narrow‐band inputs local and quasi‐periodic instabilities occur. At those times the output error increases exponentially but has essentially the frequency of the unstable pole of the MA predictor part. This unstable error generates in the algorithm a stabilizing drift for the (unstable) MA parameters. This behaviour, called self‐stabilization (SS), is attenuated by inclusion of a (leakage) σ‐factor in the updating algorithm. The SS analysis is applied to the important application of the ADPCM encoding of speech signals. It is shown that in successive encoder/decoder connections the SS phenomenon results in the accumulation of quantization distortions at each transcoder. For a digital transmission chain this is a serious drawback.