Spectral Enhancement of Whispered Speech Based on Probability Mass Function

Abstract

Whispered speech can be effectively used for quiet and private communications over mobile phones and is also the communication means for ENT patients under a regime of voice rest. The reconstruction of natural sounding speech from such whispers can be useful for several types of application across different scientific fields ranging from communications to biomedical engineering. Despite the useful applications for a such technology, the reconstruction of natural speech from whispers has received relatively little research effort to date. This paper presents novel methods for spectral enhancement and formant smoothing with the aim of attaining more natural sounding speech within the reconstruction process. The proposed approach uses a probability mass-density function to identify a reliable formant trajectory through whispers and apply vocal modifications accordingly. Subjective evaluation experiments were performed, and are reported, to assess the performance of the techniques. A method for the near real-time conversion of whispers to normal phonated speech through a modified CELP codec has been discussed in our previously published work which, the proposed formant modification approach in this paper builds upon.