论文部分内容阅读
本文主要阐述了语音信号声源模型的结构,其中主要包括代表一激励参数的脉冲及噪声激励源,以及代表声道特性的线性时变系统。指出声源编码(声码器)即用传送模型参数代替传送语音波形的一种获得极低比特率的编码方式。分别讨论了三种典型声码器即CV、HV及LPC的原理,主要结构及性能。它们都用某种方式提取激励参数,而分别用短时谱幅度、倒谱系数及线性预测参数来描述声道特性,控制短时频谱的包络线,以获得具有一定可懂度的综合语音输出。
This paper mainly elaborates the structure of the speech signal sound source model, which mainly includes a pulse and noise excitation source representing an excitation parameter, and a linear time-varying system representing the channel characteristics. It is pointed out that the source code (vocoder) is a very low bit-rate coding method that uses transmission model parameters instead of transmitting voice waveforms. The principle, main structure and performance of three typical vocoder namely CV, HV and LPC are respectively discussed. They both extract the excitation parameters in some way and describe the channel characteristics with short-time spectral amplitude, cepstral coefficients and linear prediction parameters, respectively, and control the envelope of the short-time spectrum to obtain a comprehensive speech with a certain degree of intelligibility Output.