Pitch and spectral estimation of speech based on auditory synchrony model

摘要: This paper describes a system for processing sonorant regions of speech, motivated by knowledge the human auditory system. The spectral representation is intended to reflect proposed model which takes advantage synchrony in nerve firing patterns enhance formant peaks. also applied pitch extraction, and thus temporal processor envisioned. spectrum derived from outputs set linear fillers with critical bandwidths. Saturation adaptation are incorporated each filter independently. Each "spectral" coefficient determined weighting amplitude response at that frequency measure center filter. Pitch front waveform generated adding rectified across dimension. estimator illustrated pure tones natural speech.