Wideband speech coding using MELP model
Weisi Lin, Soo Ngee Koh, Xiao Lin · 2003
In this paper, we consider the use of the mixed excitation linear prediction (MELP) model for coding wideband speech signals. In order to achieve a reasonably good subjective quality for the decoded speech and to maintain a low operating bit rate, some modifications to the LP analysis/synthesis stage, pitch estimation, and post filtering of the standard MELP model are needed. With the proposed coder operating at a bit rate below 8 kbit/s, subjective quality of the decoded speech is found to be close to that of G.722 operating at 48 kbit/s.