基于双基频多带激励分离模型的元音分离

听觉场景分析(AuditorySceneAnalysis,ASA)是一种模仿人的听觉特性实现对混叠声音信号进行分离的方法。作为ASA的基础研究,论文针对各种ASA系统存在的相近频带信号无法有效分离问题,提出了一种新的基于双基频多带激励分离模型的元音分离系统,该系统利用两语音信号的基音轨迹特性提取多带激励分离模型中两基频对应的语音的各次谐波参数,将两组参数代入多带激励合成模型从而得到两个分离的语音信号。文中给出了算法的原理与具体描述。仿真实验结果表明,系统对存在基音频率差异的元音信号能实现有效的分离。

Concurrent Vowels Separation Based on Double-pitch Multi-band Excitation Separation Model

Auditory scene analysis(CASA)is systems that can separate mixed sound sources by modify the auditory characteristics of human being.Traditional ASA systems ignore the problem that there are band of two speech mix to-gether which can not be separated block by block,a concurrent vowels separation system based on double-pitch multi-band excitation separation model is presented as a foundation part of ASA.Guided by the two pitches of the speeches,parameters of the two speeches are obtained from the double-pitch MBE separation model,when the parameters are used in the synthesis model,the separated sounds are obtained.The theory of the algorithm and its detail are presented.Experimental results show that concurrent speeches with different pitches are separated effectively.