ADPCM speech coders with backward adaptive predictors for noisy channels.

Bibliographic Details
Main Author:	Reininger, Randall Craig
Other Authors:	Fischer, T. R. (degree committee member.), Halverson, D. R. (degree committee member.), Longnecker, M. T. (degree committee member.)
Format:	Thesis Book
Language:	English
Published:	1983.
Subjects:	Data compression (Telecommunication) Error-correcting codes (Information theory) Pulse-code modulation. Speech processing systems. Electrical Engineering Academic theses
Online Access:	Link to ProQuest Copy Link to OAKTrust copy

Description
Abstract:	In this paper, the use of robust backward adaptive predictors in adaptive differential pulse code modulation (ADPCM) speech coders for a 16 kb/s transmission rate is investigated. The classical adaptive transversal predictors are compared to the more recently developed adaptive lattice predictors. All of the ADPCM systems that are considered use a robust Jayant four level quantizer. Damping schemes are introduced for the adaptive lattice predictors so that the ADPCM system is robust with respect to channel errors. It is shown that the adaptive lattice predictors have an advantage over the transversal predictors for both noise-free and noisy channels. A method for improving the performance of the system for very noisy channels is introduced which involves reducing the sampling rate of the system and adding an error correcting channel code. It is shown that a significant increase in performance for very noisy channels can be attained by using a rate 2/3 code for protecting the magnitude bits of the coded quantizer output and leaving the sign bits unprotected for a system that operates at 6.4 kHz. Only a small loss is incurred in the noise-free case due to the reduction in sampling rate for the system. The ADPCM system with and without channel coding is compared to a 16 kb/s CVSD system (continuously variable slope delta modulation). It is shown that the ADPCM system with channel coding is uniformly better than the CVSD system for all error rates. The dynamic range, tandem, and background noise performance of the systems are also compared.
Item Description:	"Major subject: Electrical Engineering." Typescript (photocopy). Vita.
Physical Description:	ix, 101 leaves : illustrations ; 29 cm
Bibliography:	Includes bibliographical references (leaves 92-94).