基于数学形态学的语音反蓄意模仿研究

Research of Voice Anti-deliberate Imitation Based on Mathematical Morphology

人的声音虽然具有个性特征,但随着现代高科技的发展,人的语音也可以被模仿;当模仿者蓄意模仿说话人声音相似度极高时,当前语音身份确认系统就有可能会被模仿欺骗;为了防止造成国家和人民财产的重大损失,就此展开了语音反蓄意模仿的研究;为了进一步消除说话人语音中的噪声干扰以增强不同说话人的区分能力,研究了数学形态学及加权形态滤波器在一维语音信号中去噪的应用;研究表明,数学形态学对带噪语音信噪比的改善有较好的效果,进一步增强了区别蓄意模仿说话人的能力。

Although there is the personality in the voice of people, with the development of modern high--tech the human voice can be / imitated. When the voice of deliberate imitator sounds much similar, current voice identification system may be deceived. In order to prevent causing a great property loss of countries and people, the research on anti--deliberate imitation of voice will be carried out. Then the applica- tion of mathematical morphology in denoising of one--dimensional speech signal was introduced. To further eliminate the noise of the speaker＇ s voice for enhancing the ability to distinguish between different speakers, mathematical morphology and weighted morphological filter are researched in the application of one--dimensional speech signal denoising. The results show that, mathematical morphology is better for improving the SNR of noisy speech, further enhancing the ability to distinguish the speakers of deliberate imitation.