Cotalker nulling based on multi super directional beamformer

摘要

Speech from either a driver 102 or a passenger 104 in a vehicle 100 is selected directionally via a pair of microphones 108, 110 separated by a distance 114 whose signals are sent to a processor (DSP) which analyses their frequency components in order to identify driver and passenger sounds (by eg. examining their phase as in figs. 7-9) and suppress the drivers frequency components in the passenger sound, or vice versa, thus providing directional selection (ie. beamforming). Components below a certain threshold (fig. 6) may be attenuated or removed after Voice Activity Detection (VAD) and a speech recognizer may also be employed.