This contribution presents a novel two microphone speech enhancement system and its optimization using objective measures of speech quality. The speech enhancement system was designed for the voice communication system of a Computed Tomography system (CT scanner) which is used for the observation of the patient during CT examinations. The purpose of the system is to reduce noise and thus to enhance the speech signal of the patient which is transmitted to the control desk. The enhancement is based on the spatial coherence of noise and speech signals and is accomplished in two separate frequency bands by means of adaptive filters, scalar adaptive weighting and a highpass filter. For the optimization of the algorithms we used objective criteria which include the distortion of the clean speech signal and the reduction of noise during speech activity and during speech pause. The real time version of the speech enhancement system is implemented on a single Motorola DSP 56001.
Bibliographic reference. Martin, Rainer (1995): "Design and optimization of a two microphone speech enhancement system", In EUROSPEECH-1995, 2009-2012.