Pre-Emphasis and Linear Prediction Error Filters for Quality Improvement of Bone-Conducted Speech

Abstract

In this paper, a method to improve the quality of the bone-conducted (BC) speech is presented. Two filters, high-pass filter and linear prediction (LP) error filter, are utilized in a cascaded form. The LP coefficients are calculated from a long frame in an input sentence and used as the coefficients for the LP error filter. The proposed method requires only the BC speech signal as the input, while the conventional methods do both the air-conducted (AC) and BC speeches. Through experiments, the performance of the proposed method is investigated and it is shown that a significant quality improvement for the BC speech is obtained.