Spectral Envelope Modelling for Full-Band Speech Coding

Loading...
Thumbnail Image
Journal Title
Journal ISSN
Volume Title
Sähkötekniikan korkeakoulu | Master's thesis
Date
2016-12-12
Department
Major/Subject
Radio Communications
Mcode
S3019
Degree programme
CCIS - Master’s Programme in Computer, Communication and Information Sciences (TS2013)
Language
en
Pages
9+61
Series
Abstract
Speech coding considering historically narrow-band was in the latest years significantly improved by widening the coded audio bandwidth. However, existing speech coders still employ a limited band source-filter model extended by parametric coding of the higher band. In this thesis, a full-band source-filter model is considered and especially its spectral magnitude envelope modelling. To match full-band operating mode, we modified, tuned and compared two methods, Linear Predictive Coding (LPC) and Distribution Quantization (DQ). LPC uses autoregressive modeling, while DQ quantifies the energy ratios between parts of the spectrum. Parameters of both methods were quantized with multi-stage vector quantization. Objective and subjective evaluations indicate the two methods used in a full-band source-filter coding scheme perform on the same range and are competitive against conventional speech coders requiring an extra bandwidth extension.
Description
Supervisor
Alku, Paavo
Thesis advisor
Bäckström, Tom
Fuchs, Guillaume
Keywords
spectral envelope modelling, linear predictive coding, distribution quantization, moving average multi-stage vector quantization, full-band speech coding
Other note
Citation