In this paper we present a new method for pitch estimation using a system based on phase-locked-loop devices. Three main blocks define our system. The aim of the first one is to make an harmonic decomposition of the speech signal. This stage is implemented using a band-pass filter bank and phase-locked-loops cascaded to the output of each filter. A second block enhances the harmonic corresponding to the fundamental frequency and attenuates all other harmonics. Finally a third stage re-synthesizes a new signal with high energy at the fundamental frequency and extracts pitch contour from that signal using another phase locked-loop. Performance is evaluated over two databases of laryngograph-labeled speech and compared to various well known pitch estimation algorithms.