Study and development of MELP vocoder

摘要

In mixed-excitation linear prediction (MELP) coding, a new speech synthesis model that approaches actual human pronunciation mechanism is adopted to improve the performance of low bit-rate speech coding. A penetrating analysis of the MELP algorithm is presented in this paper. Improvements in multi-stage vector searching method and in pitch estimation are suggested. A 2.4 kbit/s MELP vocoder has been simulated on PC and realized real time using one single TMS320C5402 DSP.