Waveform interpolation and analysis-by-synthesis-a good match?

Abstract

This paper considers the incorporation of analysis-by-synthesis (A-by-S) techniques in waveform interpolation (WI) coding architectures. It proposes an altered A-by-S mechanism which overcomes the non-synchronous input/output speech which is characteristic of WI coders. The proposed architectures operate on a prototype-by-prototype basis, optimising parameter choices within each frame. Three possible architectures are considered, with varying degrees of A-by-S integration. These low-rate, 2.4 kb/s coders are tailored to the transmission of no prototype phase information. It is also shown that the incorporation of A-by-S in WI allows exploitation of CELP style perceptual weighting techniques. The conclusion is that the use of A-by-S can improve the perceptual performance of WI coders.