NONLINEAR SIGNAL PROCESSING IN A SPEECH RECOGNITION SYSTEM

摘要

NONLINEAR SIGNAL PROCESSING IN A SPEECHRECOGNITION SYSTEMAbstractAn acoustic processor and a method of processingan acoustic wave input which includes a non-linearauditory model of the neural firing rate in theear. The firing rate is determined by thereplenishment of neurotransmitter and loss ofneurotransmitter due to spontaneous decay,spontaneous firing, and acoustic wave inputsdefined, preferably, in sones. The number of freeparameters in the acoustic processor is reducibleto one, namely the ratio R of two steady-statefiring rates each resulting from a differentloudness. Preferably, the free parameter isadjusted to minimize steady-state effects whichare adversely impacted by speaker differences,background noise, distortion, and the like. Thepresent invention addresses the general problem ofprocessing an acoustic wave input in a speechrecognition system and addresses the specificproblem of adjusting acoustic processorperformance to reduce adverse effects.