论文部分内容阅读
MBE语音编译码算法的关键是合成语音的连续性.本文在研究MBE语音编译码模型、方法的基础上,提出用一个慢变频正弦信号产生窄带等功率谱信号作激励产生清音,同时采用增衰因子,把清音、浊音的合成统一到时域方法下的新算法,进一步保证了合成语音的幅度及基音连续性.对在4.8Kbps计算机模拟结果的非正式听音表明,新的算法较传统MBE在音质上自然度更好.
The key of MBE speech codec algorithm is the continuity of synthesized speech. On the basis of studying the model and method of MBE speech codec, this paper proposes to use a slow-frequency sinusoidal signal to generate narrow-band equal-power spectrum signal as excitation to generate the unvoiced sound. At the same time, using the increasing factor, the unification of unvoiced speech and voiced speech into the time domain method The new algorithm further ensures the amplitude and pitch continuity of synthesized speech. The informal listening of the results of the computer simulation at 4.8 Kbps shows that the new algorithm is more natural in sound quality than the traditional MBE.