On the use of auditory models' elements to enhance a sawtooth waveform inspired pitch estimator on telephone-quality signals

Abstract

An enhancement to an existing sawtooth waveform inspired pitch estimator is proposed in order to increase its performance on telephone speech, characterized for being bandlimited. The enhancement consists in recovering the missing harmonics through the use of an auditory-based preprocessing stage, that includes a spectral envelope flattener, a time-aligned gammatone filterbank, and half-wave rectifiers followed by weighted low-pass filters. This enhancement produces an improvement in the pitch estimator's performance on telephone-quality speech, as well as on musical instruments.