Speech Quality Preservation with Bone-Conducted Speech in Highly Noisy Environments and Its Application to Speech and Speaker Recognition

提出了两种骨导语音高频成分估计方法，并将其应用于说话人识别。通过硬件和软件的识别实验表明，该方法可以为说话人提供一个非常有用的特征。

Two methods to estimate the high frequency components of bone-conducted speech from its low frequency counterparts were derived, and applied to speaker recognition. Recognition experiments through hardware and software showed that the proposed methods could provide a very useful feature for speakers.