It is well known that the production of speech involves non linear phenomena. Classical algorithms of speech analysis, nevertheless, are based on the assumption that speech is generated by a linear system. In this paper we will describe how, using signal prediction based on a quadratic Volterra operator, the classical linear techniques can be extended to include non-linear modeling of speech. An adaptive algorithm will de described. Application of this novel analysis approach in speech coding will be described and compared with the classical linear approaches. It will be shown that the non-linear approach yields better performances with respect to the linear one and therefore it is of interest for telecommunication application.
ADPCM with non linear predictors
MUMOLO, ENZO;A. Carini
1994-01-01
Abstract
It is well known that the production of speech involves non linear phenomena. Classical algorithms of speech analysis, nevertheless, are based on the assumption that speech is generated by a linear system. In this paper we will describe how, using signal prediction based on a quadratic Volterra operator, the classical linear techniques can be extended to include non-linear modeling of speech. An adaptive algorithm will de described. Application of this novel analysis approach in speech coding will be described and compared with the classical linear approaches. It will be shown that the non-linear approach yields better performances with respect to the linear one and therefore it is of interest for telecommunication application.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
