基于情感基音模板的情感语音合成

Synthesis of emotional speech based on emotional pitch template

为了合成能够模拟表达说话人的情感状态的语音,提出一种基于情感基音模板的情感语音合成方法。该方法分别建立高兴、愤怒、悲伤和中立4种不同情感下的韵母基音模板库,建立4种声调模型,统计分析语音库中情感语音的韵律特征参数,运用基音同步叠加算法(PSOLA)合成含情感色彩的语音。实验以音节为合成单位,根据情感特征参数的统计分析结果调节合成语音的韵律特征,合成各种情感的语音。仿真实验结果表明:用情感基音模板合成的目标情感语音具有目标情感的音质色彩,再通过韵律参数调节,可合成较理想的情感语音。该方法可用于增加语音合成系统的智能化,提高人机交互的能力。

In order to synthesize the speech which can express the speaker＇s emotional state,a method of emotional speech synthesis based on the emotional pitch template was presented.By the method,happy,angry,sad and neutral vowel pitch template libraries were established,and four kinds of tone model were also established,the prosody characteristic parameters of the emotional speech were analyzed,and pitch synchronous overlap algorithm（PSOLA） to synthesis speech with emotional colors was used.Using the syllable as the synthetic unit,the prosodic parameters of the synthetic speech were adjusted according to the statistical analysis of the prosodic parameters to synthesize various emotional speech.Simulation results show that with the same prosodic parameters,the emotional speech synthesized with the targeted emotional pitch template has the tone color of the targeted emotion.After the adjustment of prosodic parameters,the ideal emotional speech can be gotten.The method can be used to increase the intelligence of speech synthesis system and improve the capabilities of human-computer interaction.