Enhancing FS-1016 CELP via selective harmonic pre-emphasis

An algorithmic enhancement for FS-1016 CELP, referred to as selective harmonic pre-emphasis, is described. A limitation of CELP is the occasional occurrence of delay estimation errors, particularly in voiced/unvoiced transition areas and at word onsets. This type of error is somewhat similar to pitc...

Full description

Saved in:
Bibliographic Details
Main Author: Teague, K.A.
Format: Conference Proceeding
Language:English
Subjects:
Code standards
Computer errors
Delay estimation
Harmonic analysis
Signal synthesis
Speech analysis
Speech coding
Speech synthesis
Stochastic processes
Synthesizers
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:An algorithmic enhancement for FS-1016 CELP, referred to as selective harmonic pre-emphasis, is described. A limitation of CELP is the occasional occurrence of delay estimation errors, particularly in voiced/unvoiced transition areas and at word onsets. This type of error is somewhat similar to pitch and voicing errors in other types of coders and results in audible roughness and distortion when it occurs. Harmonic pre-emphasis is applied in the analysis by synthesis loop of the analyzer to strengthen selectively the harmonic structure in voiced and mixed areas so that adaptive codebook delay values can be estimated more reliably and smoothly. The procedure is applied adaptively as a function of apparent voicing. The result is an objective improvement of approximately 0.8 (1.5) DRT points and approximately 0.9 (1.2) DAM points in quiet, and office, environments, respectively. Full interoperability with the FS-1016 standard is preserved.
DOI:10.1109/ETWCS.1999.897313