论文部分内容阅读
对现有变调方法作了分类分析,分析了3种典型方法的变调原理和特点,即基于时域同步叠加固定合成变调法(synchronized overlap-add fixed synthesis,SOLA-FS)、频域插值法和基于相位声码器法,重点给出改进的时域SOLA-FS实现方法;通过仿真实验对比3种变调方法的效果:3种变调方法均能在保持音频播放时间不变的前提下,实现音调的改变,但在语音自然度的感知上有差别;通过主观测听实验评估了各种变调方法的音效。结果表明:不论对语音音高的提升还是降低,在相同变调系数下,时域SOLA-FS方法均具有最好的变调效果。
Based on the classification and analysis of the existing transposition methods, the principles and characteristics of the transposition of the three typical methods are analyzed, that is, based on the synchronized overlap-add fixed synthesis (SOLA-FS) method, the frequency domain interpolation method and Based on the phase vocoder method, the improved time-domain SOLA-FS implementation method is given. The simulation results show that the three kinds of modulation methods can achieve the same effect without changing the audio playback time However, there is a difference in perceived voice naturalness; the sound effects of various tone-changing methods are evaluated through subjective listening experiments. The results show that the time-domain SOLA-FS method has the best pitch-shifting effect under the same pitch coefficient, regardless of whether the pitch of the voice is increased or decreased.