Speech-pitch detection using maximum likelihood algorithm

Presents a robust algorithm for pitch detection and estimation. The algorithm is based on implementation of the maximum-likelihood estimation. The speech signal is digitized with 8 kHz, 10-bit Analog-to-Digital Converter (ADC) and the digitized data are stored in a PC where it is segmented into frames each of 32 msec. A theoretical analysis has been conducted in this paper to find the maximum-likelihood epoch determination signal that represent the Glottal Closure Instant (GCI) from which the period can be determined. The Hilbert transform is incorporated into the algorithm to improve the overall performance of the algorithm. Utterances from three male and three female speakers have been used to test the algorithm. The algorithm has performed reliably and detected correctly voiced, unvoiced, and mixed modes of speech