This paper presents a new algorithm to detect the intended fundamental frequency (F0) in human voice. Accurate detection of F0 is the first step when computing multiple important voice analyses, such as harmonics to noise ratio (HNR), frequency variations within utterances and voice harmonics. These, in turn, allow voice recognition and identification, which have wide applications from voice pathology diagnosis to automatic answering algorithms and robotic voice commanded systems. The proposed algorithm explicitly incorporates the following voice-specific characteristics in frequency domain: presence of harmonics in human voice, continuity of intended F0 over small time intervals and presence of intensity reinforcement in F0. The latter allows the algorithm to remain robust when analyzing voices that present subharmonic (pathological) components.