Speech enhancement for an in-ear communication system using optimal-filtered accelerometer signals
Kyriakos Papanagiotou, Boaz Rafaely, Christine H. Shadle · ePrints Soton (University of Southampton) · 2002
A method of enhancing speech produced by the wearer of an in-ear communication system is described. The eventual system will use optimal filtering of two signals, from a microphone and an accelerometer, to generate a speech signal that is free of ambient noise and distortion. Current experiments use an accelerometer mounted near the ear (on the temporo-mandibular joint, TMJ) and a reference microphone placed 60 cm in front of the subject. Formal intelligibility tests show the optimally filtered accelerometer signal outperforms high-pass filtering in quiet or noise.