The Implementation and Analysis of the CELP Coder with a Dsp Processor

• Main Authors: Chiu, Ming-Fung, 邱明鋒
• Other Authors: Chou, Chun-Hsien
• Format: Others
• Language: zh-TW
• Published: 1997
• Online Access: http://ndltd.ncl.edu.tw/handle/84560912385965372785

碩士 === 大同大學 === 電機工程研究所 === 85 === Nowadays, in the popular internet and personal communication network, the applications of digital speech signal will be promising and will be applied widely, so there is no doubt that the coder with low bit rates and high synthetic speech quality is important in this field. Code Excited Linear Prediction (CELP) is developed base on linear prediction and code excitation model, the scheme with potential to synthesize high speech quality at low bit rates. For example, the Federal Standard 1016 speech CODEC (encoder, decoder) can achieve high synthetic speech quality under 4.8k bps. Federal Standard 1016 speech encoder has three basic features: (1) Short-term linear prediction analysis, (2) Long-term adaptive codebook search and (3) Innovation stochastic codebook search. The objective of linear prediction analysis is to model physical shape of human vocal tract. Adaptive codebook is used to emulate the vibration of human vocal cords in order to produce period phenomenon of speech waveform. Stochastic codebook represents the air flow from lungs to lips. In this thesis, we will implement the Federal Standard 1016 speech CODEC with C language and discuss the problems we met in detail. Meanwhile, we will use TMS320C30 development system to implement the Federal Standard 1016 speech CODEC and analyze the implementation process.