论文部分内容阅读
文中提出了一种基于离散余弦变换(DCT)和基音同步叠接相加(PSOLA)的语音变换方法。此方法可以自由调整源语音的基音频率、能量分布和时长,以达到变换要求,并且变换后的语音具有较高的质量。变换方法首先对基音标记过的语音段通过DCT进行基音频率和能量分布的调整,之后再通过PSOLA法进一步对基音频率进行修正。实验表明,此方法在男女声变换中,能够使变换前、后语音的性别感觉明显变化,并且保持了较高的语音质量。
In this paper, a speech transform based on discrete cosine transform (DCT) and pitch-synchronous splicing (PSOLA) is proposed. This method can freely adjust the pitch frequency, energy distribution and duration of the source speech to meet the transformation requirement, and the transformed speech has higher quality. The transform method first adjusts the pitch frequency and energy distribution of pitch-marked speech segments through DCT, and then further modifies the pitch frequency by PSOLA method. Experiments show that this method can significantly change the gender perception of the voice before and after the transformation, and maintain a high voice quality.