一种有效的语音信号变调算法及其实时实现

Effective Pitch Shifting Algorithm for Sampled Speech Signal and Its RealTime Realization

改变语音信号的音调是许多商用设备的一种功能，如采样合成器、调和器、声码器，也能被许多计算机音乐语言软件实现，然而，许多成熟的算法由于技术的保密性，不公开；一些公开的算法均有各种各样的问题。本文根据音乐的信号的特征，在保持未变调声音的规格特征上，提出了一种有效的变调算法，在具体实现中，为了保证运算的实时快速，采用了ＴＩ公司的ＴＭＳ３２０Ｃ５０ＤＳＰ芯片作为主处理芯片，它的时钟频率是１０ＭＨｚ。本文也介绍了在ＴＭＳ３２０Ｃ５０芯片上实时实现的软、硬件技术。

Shifting the pitch of digitally sampled sounds is an operation performed by a number of commercially available devices (sampling synthesizers, harmonizers and vocoders) as well as by the software of many computer speech languages. However, some ripe algorithms are not open for keeping them secret. Some open algorithms exist a lot of questions. Based on the nature of the musical signal, an effective algorithm for pitch shifting is presented while maintaining formant characteristics of the unshifted sound. In the concrete realization, we used TMS320C50 DSP chip of TI Corporation in U.S.A as the main processing chip for assuring the realtime processing. Its clock frequency is 10 MHz. Also in the paper, software and hardware technologies for its realtime realization in TMS320C50 DSP chip are discussed.