LP-Based Quality Improvement of Noisy Bone Conducted Speech

Abstract

Bone-conducted (BC) speech seems to be more advantageous than normal air conducted (AC) speech in noisy environments because of its robustness against ambient noise. However, induction of noise to BC speech can not be completely ignored in highly noisy environments. We propose a quality improvement method for noisy BC speech without exploiting any spectral characteristics of AC speech. In the method, the BC speech signal is inputted to the linear prediction (LP) error filter and followed by the speech enhancement process. Through experiments we show that the proposed method provides a notable quality improvement for noisy BC speech.