Optimization of an hyperbaric speech transcoder

The aim of this paper is to device a technique to improve the restoration of the spectral properties of the speech of deep-sea divers in operations when breezing mixtures different from air are used. The technique used here is the modified channel vocoder. The intelligibility problems due to hyperbaric speech are described and then a new version of frequency bands is studied. Multirate sampling is proposed that considerably reduces the processing time. The real-time operation is improved (in simulation) to 85%. Hyperbaric speech records from 'Hydra 4' (physiological experiments using pressurised heliox) are processed to validate these optimizations.