| ABSTRACT: | THE RESEARCH FOR THE MELP VOCODER AND ITS REAL-TIME REPLEMENTATION
W. Jun, Z. Jiyin, Y. Yusheng
Jilin University, Changchun, Jilin, China.
The MELP coder is based on the LPC model with additional features that include mixed excitation.aperiodic
pulses, adaptive spectral enhancement, pulse dispersion filtering, and Fourier magnitude modeling. this
allow the MELP coder to simulate accurately natural speech. This algorithm has been implement on
TMS320VC5416 DSP processor. Objective speech quality testing and in-Formal listening testing results
show that the intelligibility, naturalness and anti-noise performance of the 1.2 kb/s speech codinf algorithm
is similar to or better than that of the Federal Standard at 4.8 kbps(FD 1016).
Key Words: MELP, mixed excitation aperiodic pulses,TMS320VC5416 DSP processor
|